...
Hmm.. but where is CUDA_V2.3 ?
...
Page hasnt been updated but if you manually copy the link and pastes into a new link changing 2_2 to 2_3 everywhere you can get it to download..
In the meantime, my 9800-series cards are also compute capability 1.1, and the new 190.38 drivers plus CUDA 2.3 DLLs have just speeded them up by ~30%.
On a GTX260 I just did a shorty in 1minute 20 seconds wall time... an improvement of... 2.1x faster!!
Now the question is... are they valid??
It seems almost too good to be true
-Bob
I didn't see any speed up after my update to 190 and 2.3 dll (sample size of 3). I'm still getting about 45 minutes (wall clock time) for the GPU and 1:20 hours for the CPU (Q6600). I suspect my 8600GTS is doing all it can. It may have changed a minute or so, but I don't watch it that closely so I can't tell.
Where can I download the nanobots to upgrade my 8600GTS to infinite speed? ;D
Rick
On a GTX260 I just did a shorty in 1minute 20 seconds wall time... an improvement of... 2.1x faster!!
Now the question is... are they valid??
It seems almost too good to be true
-Bob
Get only the needed 2.3 dlls in beta downloads here (http://lunatics.kwsn.net/index.php?module=Downloads;catd=29), and get 190.38 driver from nVidia. Some small speedup at high angle ranges is observed with V12, will be experimenting to see if new compiler makes a better build, or not, for a while.I took the CUDA_V2.3 (.dll's) from the 'public DL area' here.
O.K., one more question to the opt. crew.. ;)
For only CUDA crunching and best performance, PhysX ON/OFF ?
Over the system tray icon from nVIDIA.. right click -> performance and quality setting / system performance -> I can choose:* high performance
Stock is 'quality'.
* performance
* quality
* high quality
How should be the setting for best CUDA performance?
Now that the task list is turned on again...
They are valid!
Bob
Did you update the cufft and cudart.dll's in the projects\setiathome folder? If not then you'd still be running the older cuda version (whichever one you had previously).Drat! I put them in the BOINC folder. It's fixed now. I wasn't paying attention again :)
Get only the needed 2.3 dlls in beta downloads here (http://lunatics.kwsn.net/index.php?module=Downloads;catd=29), and get 190.38 driver from nVidia. Some small speedup at high angle ranges is observed with V12, will be experimenting to see if new compiler makes a better build, or not, for a while.
Are these the same dll's that are in the \CUDA\bin folder after you install the CUDA 2.3 Toolkit?Yes.
correct, the same old fftw library remains, and has nothing to do with cuda, but other parts of the application. Since the files have the same name, there are no app_info changes. Do make sure your driver is 190.38+ though.
N275GTX @666MHz, 896MB GDDR3 ram@2322MHz, 240 stream processors for $200I'd have to say the 275 looks like the better deal to me, since there's extra processing (~10% more processors & some extra clock speed) grunt all around. I'm not exactly sure what would need that much RAM as is on the 260, but the apps in their current (& likely foreseeable future) form wouldn't use it. So, ignoring processing, the choice will come down to power requirements and perhaps cooling etc..., but the 275 looks better to me.
N260GTX @655MHz, 1792MB DDR3 ram@2100MHz, 216 stream processors for $295
So the questions are, are stream processors important to CUDA? Is less but faster memory better than more but slower memory? Is there something important to CUDA in the 275 vs 260?
I'm getting a video card, one of the less expensive types. It will be a MSI card, so my MSI mobo monitor will cover the GPU as well. I have two in mind - which will provide best CUDA performance?
N275GTX @666MHz, 896MB GDDR3 ram@2322MHz, 240 stream processors for $200
N260GTX @655MHz, 1792MB DDR3 ram@2100MHz, 216 stream processors for $295
So the questions are, are stream processors important to CUDA? Is less but faster memory better than more but slower memory? Is there something important to CUDA in the 275 vs 260?
N275GTX @666MHz, 896MB GDDR3 ram@2322MHz, 240 stream processors for $200
N260GTX @655MHz, 1792MB DDR3 ram@2100MHz, 216 stream processors for $295
So the questions are, are stream processors important to CUDA? Is less but faster memory better than more but slower memory? Is there something important to CUDA in the 275 vs 260?
I'd definitely go with the 275. See this page for relative performance (GFLOPS column) of each at their default speeds: http://en.wikipedia.org/wiki/GeForce_200_Series . These GFLOPS numbers are not the same you see in BOINC, but they do scale well IME (8800GT, G94 9600 GSO, and G92 9600 GSO). I'm now running dual-GPU in linux with some of these and all works great!
I'm running boinc currently on the default settings. At this point to get up and running on the unified installer, I believe I have to perform the following correct?0. Make sure stock CUDA app running OK on your host, it's required step.
1. Download Lunatics_Win64v0.2_(SSE3+)_AP505r168_AKv8bx64_CudaV12.exe
2. Install it (I only want to run cuda units)
- Selecting only: Enable processing Multibeam tasks on nvidia cuda gpu (6.08) option
- Default destination folder
3. Download the 2.3 Cuda DLL's from this site (cudart.dll & cufft.dll) and place them in the C:\ProgramData\BOINC\projects\setiathome.berkeley.edu folder
4. Restart BOINC and done?
... Cuda 3.0 dll's to try. ;D
released CUDA 3.0beta time to recompilePrev CUDAs didn't increase speed much from recompile. Most speedup came just for using new CUFFT library with older build.
You do know that CUDA Toolkit 3.0 beta released (http://forums.nvidia.com/index.php?showtopic=149959) for registered developers?released CUDA 3.0beta time to recompilePrev CUDAs didn't increase speed much from recompile. Most speedup came just for using new CUFFT library with older build.
That is, if we will have ability to get CUFFT 3.0 some speedup could be recived "for free" I hope ;)
nVidia has too "corporation-oriented" registration form for beta program...You do know that CUDA Toolkit 3.0 beta released (http://forums.nvidia.com/index.php?showtopic=149959) for registered developers?released CUDA 3.0beta time to recompilePrev CUDAs didn't increase speed much from recompile. Most speedup came just for using new CUFFT library with older build.
That is, if we will have ability to get CUFFT 3.0 some speedup could be recived "for free" I hope ;)
Claggy
nVidia has too "corporation-oriented" registration form for beta program...
nVidia has too "corporation-oriented" registration form for beta program...You do know that CUDA Toolkit 3.0 beta released (http://forums.nvidia.com/index.php?showtopic=149959) for registered developers?released CUDA 3.0beta time to recompilePrev CUDAs didn't increase speed much from recompile. Most speedup came just for using new CUFFT library with older build.
That is, if we will have ability to get CUFFT 3.0 some speedup could be recived "for free" I hope ;)
Claggy
Ok, will try to fill that huge form again :o
i am not in corporation and i am there ... nobody was verifified my given infos .... ;)
it is for free download for now ....where's that ? (I only see it in registered developer area, they launching in the forums again ? ::) )
...Ooh .. them's fighting moves, LoL.
btw have u know that intel has bought rapidmind ? (http://software.intel.com/en-us/data-parallel/)
Note that for CUFFT 3.0, the layout of batched data must be side-by-side
and not interleaved.
Well AFAICT without bothering to look much further, they've simply added stride parameters, which would just require examining the data arrangement & setting the strides for the appropriate format you want. No doubt that makes some more advanced multidimensional FFTs easier to implement. For our purposes it'll probably just be a minor annoyance, to be overcome with macro definitions or wrapper functions.QuoteNote that for CUFFT 3.0, the layout of batched data must be side-by-side
and not interleaved.
so it seems that we must rearrange data before fft to real parts and imaginary parts in rhe same array and after this doing it back ....
....Good. No Cpus probably would be good for the test I have in mind.
I don't know if it would be well (any profit) also for the GPU cruncher (4x OCed GTX260-216), for to let run 4x MB + 4x CUDA. (Because the GPUs would lose performance, and the CPU wouldn't have much for to compensate
* You would use the current CUDA_V3.0_BETA ?First initial alpha test would be an existing 2.3 build , susbsequent tweaking would likely be v3.0 Beta.
Because I see lot of people in the forum, which let run this.
Speed difference? Well for the opt._MB_6.08_V12_app ?
,,,
If ~ 2 % speed up, this would mean + ~ 1,000 RAC at my GPU cruncher. ;)I dispute that the changes would amount to 2% consistently ... since Cuda tasks take such short time this 2% can be other natural variation. The tests were done with short synthetic tests, which typically vary as much as +/- 5% depending on machine state & how long windows has been running. There just aren;t any changes in cuda libraries we would use until fermi boards are available.
* I would see a speed difference between:
nVIDIA_driver_190.38 + CUDA_V2.3
and
nVIDIA_driver_195.62 + CUDA_V2.3
Normally nVIDIA release new driver for the gaming community.
The only CUDA community profit also from new driver versions?
(I don't mean, new driver for new CUDA version - this is clear.. ;))
I think we have determined that the 195.62 nvidia driver is not the best thing to run on pre-Fermi cards.
Has anybody determined if there is any performance difference between the 190.62 (which is currently working very well for me) and the last pre-Fermi update 191.07?
I hate to break anything that is working, unless somebody has noticed an advantage in doing so.
look here:Thank you so much for the link and the test, Raistmer.
http://setiathome.berkeley.edu/forum_thread.php?id=56399&nowrap=true#953054
(in short, I see slowdown dor 195 but no difference for prev drivers)
I used to get the crashes as well especially on the 195 drivers.are you running quad sli or higher? i've got three gtx 295's, one in one rig and two in another, and am only able to get two optimized apps to run without what i call thrashing (invalid results and nvidia driver restarts).
Went back to the 191.07 drivers and problem solved
Thought I'd try thee 196.75 drivers as there seems to have been some stability work done by nVidia on this set - had them since thrusday and no crash yet - also reported problem of fan speed issues hasn't affected me
No improvement or reduction of speed either
Why?
OK, here you are. These are the 32-bit v3.0.14 DLLs from the public toolkit, renamed and ready to drop into a project folder (didn't take long to download and unpack them, did it?)Thanks a lot! :)
You can test and report back. If your test report is convincing enough, I might even update my drivers to the newer, slower version! ;D :P
Starting xxxxxxxxxxxxxxxx.26_1
Starting task xxxxxxxxxxxxxxxx.26_1 using setiathome_enhanced version 608
.. ~ 2 sec. after..
Computation for task xxxxxxxxxxxxxxxx.26_1 finished
Output file xxxxxxxxxxxxxxxx.26_1_0 for task xxxxxxxxxxxxxxxx.26_1 absent
Exit status -1073741515 (0xffffffffc0000135)
<core_client_version>6.10.18</core_client_version>
<![CDATA[
<message>
- exit code -1073741515 (0xc0000135)
</message>
]]>
Yes I have two crunch machines both with 64 bit Win 7
One have 9800 green and another is GT 240 and both computing via cuda 3 dlls without problem
Yes I have two crunch machines both with 64 bit Win 7
One have 9800 green and another is GT 240 and both computing via cuda 3 dlls without problem
What's your experience on the speed question?
Edit - and did you try it with only one of the files doubled-up? I can see you would need a named copy of cudart32_30_14.dll, because cufft32_30_14.dll depends on it: but there's no reverse dependency. Does it matter that cufft32_30_14.dll depends on itself (???)
.
.
<file_info>
<name>MB_6.08_CUDA_V12_VLARKill_FPLim2048.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart.dll</name>
<executable/>
</file_info>
<file_info>
<name>cudafft.dll</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3-1-1a_upx.dll</name>
<executable/>
</file_info>
.
.
The same in the <file_ref> part?Because the application (SETI MB CUDA) only knows the names "cudart.dll" and "cufft.dll".
But the new v3 cufft.dll expects to have "cudart32_30_14.dll" available.
[...]
Here's some good stuff. Some improvement (1%-2%) with 197.13 compared to my last test.
driver : 197.13
CPU: 100%
GPU: gtx 295 core 0
Which nVIDIA driver you tested with CUDA V2.3?
I guess you used only 197.13 with CUDA V2.3 and CUDA V3.0.14 in your test, right?
Which nVIDIA driver you tested with CUDA V2.3?
I guess you used only 197.13 with CUDA V2.3 and CUDA V3.0.14 in your test, right?
My test was CUDA v2.3 vs CUDA v3.0.14 on MB_6.08_CUDA_V12_VLARKill_FPLim2048.exe
Nvidia 257.21 drivers released today.Nice, already used the beta drivers with these features, but maybe this release driver is even better. Was still a lot slower than to XP 64 version.
I havn't done any any benchmarks yet but for multiple gpu users (vista/7), you don't need to extend your desktop (using dummy plugs or real monitors). ALL cuda gpus will be detected.
Nvidia 257.21 drivers released today.Well blow me......seems like nVidia have finally got their act sorted with these drivers
I havn't done any any benchmarks yet but for multiple gpu users (vista/7), you don't need to extend your desktop (using dummy plugs or real monitors). ALL cuda gpus will be detected.
Trying out 258.19 now (Developer Driver)... Seems Good, time will tell.
Trying too ;)See what you reckon. 'Feels' good, but slower Cuda (by quite a bit) I reckon.
Downloading 257.21 now... wonder what the chances are that this will allow service-mode installs on Win7/Vista, seeing as how they must have decoupled CUDA from the user mode drivers... (or I suppose maybe not)Not very likely.
Have anyone been able to run the CUDA version on a Nvidia Quadro NVS 3100m with 256M memory?
Been trying to run the optimised app (via the Unified Installer) and both Cuda 2.3 and 2.2 versions tend to error out before completion. Machine is a Lenovo Thinkpad T410, using driver 257.21 and OS is Windows 7 - I have switched it to classic mode in order to reduce video memory consumption (now round about 43M)
If anyone have any insights please let me know.
I think you'll have problems. The only time I got a 256Mb GFX card to work was on Windows XP 32. It was also running at 1280x1024 as the screen I had was only a 17" CRT at the time. The screen modes for Vista and Win 7 seem to require more memory, and not leave enough for the app to use...Thanks. Guess there's nothing much that can be done about it, after playing around with various settings and Cuda version, it is a pity that the GPU have to remain idle
Giz.
Hi,
its no problem to running 4 tasks parallel on GTX470,
with latest unified installer.
change 0.5 to 0.25 does it
mine GTX470 in the V8-Xeon runs it already sucessful.
heinz
OS Vista 64Bit Ultimate.Hi,
its no problem to running 4 tasks parallel on GTX470,
with latest unified installer.
change 0.5 to 0.25 does it
mine GTX470 in the V8-Xeon runs it already sucessful.
heinz
What O/S are you using? I am using Win XP 32 & 64. I tried 3 tasks but the more you run in parallel the longer each one takes so it was less productive.
John.
Can someone compare how fast is GTX460 comparing to GTX275 and GTS250 for SETI crunching? Is it worth upgrading?
I'm thinking of upgrading GTX275 (power hungry monster, over 220W) to faster and more power efficient GTX460 (around 160W), which I also expect to be faster (in games, it should be around 50% faster, but how much is it in SETI does anyone know?)It's not just Seti that is poorly optimised for Fermi, it's brand new so all developers need time to get used to what the cards can do and where their limits are.
As I understand, SETI application is currently poorly optimized for Fermi architecture, am I right?
Thanks.
Can someone compare how fast is GTX460 comparing to GTX275 and GTS250 for SETI crunching? Is it worth upgrading?
I'm thinking of upgrading GTX275 (power hungry monster, over 220W) to faster and more power efficient GTX460 (around 160W), which I also expect to be faster (in games, it should be around 50% faster, but how much is it in SETI does anyone know?)
As I understand, SETI application is currently poorly optimized for Fermi architecture, am I right?
Thanks.
Here's a link to how nVidia rate their current cards
http://www.nvidia.co.uk/object/graphics_cards_buy_now_uk.html
Not sure how accurate it is but may give you an idea
Here's a link to how nVidia rate their current cardsInteresting, I'm running a GTX285 and a GTX470 in the same box using V0.37. with no overclock I'm finding that the 285 is quicker by roughly a minute on a "standard" 0.4AR unit.
http://www.nvidia.co.uk/object/graphics_cards_buy_now_uk.html
Not sure how accurate it is but may give you an idea
new Beta driver is released http://www.nvidia.co.uk/object/win7-winvista-64bit-260.63-beta-driver-uk.html version 260.63Will be interested to see what you are able to find out.
Gonna give it a go and see what happens ;D
My bet is the main reason for the release was the 450s.
... and possibly sneaking in Cuda 3.2 support before the programming tools are released.
Number of platforms: 1
Plaform Profile: FULL_PROFILE
Plaform Version: OpenCL 1.0 CUDA 3.2.1
Plaform Name: NVIDIA CUDA
Plaform Vendor: NVIDIA Corporation
Plaform Extensions: cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_d3d9_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll
Looks like ...
G:\Apps\ATIStream\bin\x86_64>CLInfo.exe
Number of platforms: 2
Platform Profile: FULL_PROFILE
Platform Version: OpenCL 1.0 CUDA 3.2.1
Platform Name: NVIDIA CUDA
Platform Vendor: NVIDIA Corporation
Sneaky, nothing in the release notes or published about this going in that I've been able to find
New Releases of Parallel Nsight and CUDA Toolkit
NVIDIA today announced new versions of the industry-leading developer tools Parallel Nsight and CUDA Toolkit.
– Parallel Nsight v1.5 includes support for Microsoft Visual Studio 2010, Tesla Compute
Cluster (TCC) debugging and CUDA Toolkit 3.2
– CUDA Toolkit v3.2 includes new math libraries, performance improvements and support
for the new 6GB Tesla and Quadro products
– For an overview of new features, see: http://is.gd/f9hKn
And some bad news - analyser (profiler?) will be included only in professional ($349) version of NSight , standart (free) version has only debugger :/
Any sightings of 3.2 yet?
Ok, is there any good reason for us non-fermi types to go for this new driver before it comes out for everybody? Do you guys need any testing from us on this or can we just go back to sleep and wait for it?
Hi Mark,Short response: OK, mate.
Short answer: No.
Long answer:
Cuda SDKs & Libraries from 3.0 onwards are 'strongly versioned'. That means apps built with a certain SDK are meant to be used with certain DLLs. The New SDK with New DLLs will require a new App.
I would not advise playing swippy swappy with DLLs, as was possible in the past. You could get lucky and it works, but if you break it you own both parts ;) Also, the 3.2 SDK is so far accessible to registered developers only, and is under NDA, so no-one will be posting Cuda 3.2 DLLs around here while I'm on watch ;D
Interesting, with the 268.63 beta driver I'm seeing a change in the stable overclock (a tad higher than before) and now it seems to prefer 3 tasks at a time with the x32f app (instead of 2 before).looks promising to start seeing Fermi coming into it's own then and delivering on its expectations ;)
have just checked my GPU usage after changing to the 260.63 drivers - seems I'm only getting about 60% usage running 2 tasks at a time
Have just changed to 3 at a time - hitting 95% now using the x32f build and times seem to be about the same as running 2 at a time
have just checked my GPU usage after changing to the 260.63 drivers - seems I'm only getting about 60% usage running 2 tasks at a timeDo you accounting for GPU idle startup time in some way?
Have just changed to 3 at a time - hitting 95% now using the x32f build and times seem to be about the same as running 2 at a time
Do you accounting for GPU idle startup time in some way?Hadn't thought of that
This delay can be hided by running few tasks per GPU and the more tasks running simultaneously the better (until no performance degradation in GPU part per se).
I have upgraded to 260.63 drivers, including nvcuda.dll (CUDA 3.2.1), supposed to be much better optimized for fermi then existing CUDA 3.0.
<snip>
Since the features that Cuda 3.2 adds require further development to make use of, it'll be 'a while' before any new release uses it. Not a lot of point making Cuda 3.2 builds that are the same speed as teh Cuda 3.0 builds & don't use the new features ;) Using those will take time.
Jason
Would a cuda 3.2 version without any additional code changes make sense for the Fermi guys? Given the supposed speed increases on Fermi perhaps it might be worthwhile.
That's what I saw, but the search result doesn't match the download page.None that I saw, but checking again anyway...
They don't usually keep the same version number for Beta and WHQL, do they? Just wonder if there were any clues from your developer area access.
Is Cuda 3.2 officially out then?
Is Cuda 3.2 officially out then?
Not, according to http://developer.nvidia.com/object/cuda_archive.html - still the RC from last month. But maybe they're planning to sort that out after coffee, too. ;D
[Edit:] handiest page in NVida's site is ....
... the one I linked? :P
... any really good reason for me to grab it for my little 9500GT?258.96 were good drivers, 260 range seem improved to me in some general aspects, so IMO worth a try.
260.89. I think it may be pre-coffee post weekend hangover time in the UK, for UK nVidia guys, and possibly for Richard's fingers.
Okay, I've downloaded it, now for the dummy questions. One, will it interfere with my app_info in any way that I would have to change? Two, it says on the download page that it now has an option to remove older drivers, should I do this? I've always just installed over top of my old drivers and not had any problem. I do know to exit boinc completely before I do this. Three, I'm running 6.10.58, would there be any reason to go to 6.12.2 while I'm at it? I'm kinda avoiding that right now until they figure out what they are going to do with the messages and notices.
Keep in mind that the 'project backoff' times in the newer versions of Boinc are INSANE.Okay, I've downloaded it, now for the dummy questions. One, will it interfere with my app_info in any way that I would have to change? Two, it says on the download page that it now has an option to remove older drivers, should I do this? I've always just installed over top of my old drivers and not had any problem. I do know to exit boinc completely before I do this. Three, I'm running 6.10.58, would there be any reason to go to 6.12.2 while I'm at it? I'm kinda avoiding that right now until they figure out what they are going to do with the messages and notices.
Hi Perryjay,This is what I found when I installed these drivers:
Didn't have to modify the app_info and I didn't bother removing the old drivers (couldn't be bothered to setup my ntune profile again) and it runs as good as the old drivers did.
I'm still on 6.10.58 at the moment as I'm testing something, and apart from the benefit of the backup project change can see no reason to swap over to 6.12.2 just yet
NVidia UK (GBR) is offering 160.89 WHQL, released today, as recommended for GeForce 9, XP.
NVidia UK (GBR) is offering 160.89 WHQL, released today, as recommended for GeForce 9, XP.
Do y'all mean 260.89 rather than 160.89?
I wouldn't ask other than I'm having SETI computation issues with my 9500GT W2K rig at 257.21 with the 0.37 installer. :(
(The non-optimized Collatz is still running fine on the GPU tho.)
I haven't seen a lot of computation errors with the new 0.37 installer, would be suprised if the installer was causing this.
But for the computation errors, have you used 0.36 installer? I would think that's the problem more than the Nvidia drivers.
I haven't seen a lot of computation errors with the new 0.37 installer, would be suprised if the installer was causing this.
But for the computation errors, have you used 0.36 installer? I would think that's the problem more than the Nvidia drivers.
The x32f build is a lot more stable and produced less errors than the previous cuda build in Beta testing - what errors are you getting?
Didn't have errors when I tried 0.37 (I do use stock 6.09 now since it's faster on my PC). The reason I suggested trying 0.36 was more be course I just thought it's being more likely, solely based on a feeling :)
After constant random conflicts/stops with a GTX470 & 480, I removed, for now, the 470. Maybe PSU issue or Mobo related. (Chipset=X38)
(Driver related?) 258.96? Works OK now, 480 is doing 2 at a time. No stops.
interesting :)Downloaded and installed the APP version yesterday - so far haven't seen a big improvement with the MB build - maybe a couple of % on my 5670
What AP and MB crunch times with mew Cat ?
interesting :)Haven't tried it yet, just downloading it now, but first i've got to try Cat 10.9 (without my fix) and see if ATI computation works with Nvidia GPU fitted,
What AP and MB crunch times with mew Cat ?
Could be possible of course, if app_info is incorrect or file missing/corrupt from project folder - all really depends on what errors the cuda tasks are giving
I'm running the x32f build and the 260.89 drivers so may be worth BANZAI56 upgrading the drivers and re-running the installer to double check everything is correct
FWIW, the rig in question has Win2K Pro for an OS and I thought I had read where 257.21 was going to be one of the last to have support for that OS.?.
The underlying thing here is that it’s a case of ‘if it isn’t broken, fix it until it is’. :)
The rig had an ample quantity of ghosts, but no actual work during one of the early server outages. A quick detach/reattach seemed like a great time to also update the client and the opti app. It is looking like it might become just another reason to simply replace the rig… :/
Any 256+ series driver installed ( 260.99 seems good ) will be enough for 0.37 . Cuda 3.2 builds won't be released until sometime after nVidia relase one witnout the RC tag, and some worthwhile imrovements can be added.
For CUDA 2.3 and prior there were specific dll's etc to install. I didn't see that for V3.x. Has this changed while I was busy ranching?
Tnx Art
OpenCL 1.1 conformance tests have not been run with this version of the ATI Catalyst Driver Suite in conjunction with the ATI Stream SDK v2.2. If you require an OpenCL 1.1 conformant driver, we recommend that you install the ATI Catalyst 10.10 Driver Suite
Somebody at GPUGrid has noticed that the Cuda 3.2 toolkit is also available for general download - http://developer.nvidia.com/object/cuda_3_2_downloads.html
What a coincidence ;D. Haven't gone looking for any matching drivers yet.
Somebody at GPUGrid has noticed that the Cuda 3.2 toolkit is also available for general download - http://developer.nvidia.com/object/cuda_3_2_downloads.html
What a coincidence ;D. Haven't gone looking for any matching drivers yet.
must have been in the last 36h. I checked yesterday. 260.99 are 3.2 compatible, 258.96 looks like it isn't.
Are you sure about those drivers? The developer drivers linked with the toolkit are up to 263.06 for desktops, though still at 260.99 for notebooks. Probably best to wait until the dust has settled in the morning before making any definitive recommendations.
I'd say that's 3.2 ;)
16-Nov-2010 14:54:53 [---] NVIDIA GPU 0: Quadro FX 570M (driver version 25896, CUDA version 3010, compute capability 1.1, 242MB, 61 GFLOPS peak)
17/11/2010 17:41:35 | | NVIDIA GPU 0: Quadro FX 570M (driver version 26099, CUDA version 3020, compute capability 1.1, 242MB, 61 GFLOPS peak)
The developer drivers linked with the toolkit are up to 263.06 for desktops, though still at 260.99 for notebooks. Probably best to wait until the dust has settled in the morning before making any definitive recommendations.
As is all too often the case i acted in haste and was less thanJust install a 260.xx or later driver,
clear as a result.
i downloaded the 3.2 toolkit because that was the first
reference i found to CUDA 3.2. Developing apps and needing
specific versions of rt and fft dll's is beyond me.
i am looking to get and install 3.2 CUDA on my machine as
it seems likely that with the newest drivers and my GTX 460 i
would see some significant performane increases.
What is the best way to do this ?
Hello Claggy:
i downloaded and installed the 266.35 driver. it seems i messed up something though
as it seems that i still have a 3.0 CUDA version
1/15/2011 9:26:10 AM NVIDIA GPU 0: GeForce GTX 460 (driver version 26635, CUDA version 3020, compute capability 2.1, 739MB, 684 GFLOPS peak)
1/15/2011 9:26:10 AM NVIDIA GPU 1: GeForce GTS 250 (driver version 26635, CUDA version 3020, compute capability 1.1, 998MB, 470 GFLOPS peak)
1/15/2011 9:26:10 AM SETI@home Found app_info.xml; using anonymous platform
1/15/2011 9:26:11 AM SETI@home URL http://setiathome.berkeley.edu/; Computer ID 5738837; resource share 1000
I just tried to install the 266.35 driver and suffered a nice BSOD. Went back to 263.06 and back up.
I just tried to install the 266.35 driver and suffered a nice BSOD. Went back to 263.06 and back up.
I put it on four machines last night (one Windows 7, three XP, all 32 bit). No problems with any of them.
266.58 WHQL drivers are out...
Btw, does anyone know how come that GTX460 is compute capability 2.1, and newer few month newer GTX570/GTX580 only compute capability 2.0? Is nVidia going backwards???
Any benefit noticed in anything beyond 260.63?I noticed a small speed improvement somewhere around 263, but since then it has mostly been adding support for new cards.
My rigs are thumping along nicely on 260.63 and 260.99, except of course the rigs with Win2K, which I think stopped being supported at 257.21.
266.58 WHQL drivers are out...
Btw, does anyone know how come that GTX460 is compute capability 2.1, and newer few month newer GTX570/GTX580 only compute capability 2.0? Is nVidia going backwards???
Compute capability mainly differentiates 2 things, The instruction set & compute geometry. Although they have the same instructions, 2.1 as used in the 460 is different to the cuda core configuration used in the 470/480/570/580 , which is 2.0. Basically the 460 has extra warp schedulers & a few minor differences, while 2.0 covered the complete original Fermi design.
Thanks. I knew that 470/480 were "older" design then 460, however I (wrongly) assumed that 570/580 would be at least same or newer design then 460...
For Win 7 64, the new driver is working smoothly. Like Mark said, there is a slight increase in power usage. I migrated up from 260.99. I have no ill side effects.Thanks Steve....
Steve
orion@anubis:~$ strings /usr/lib/nvidia-current/libcuda.so|grep "CUDA "
OpenCL 1.0 CUDA 4.0.1
(Unfortunately, this makes the cuda profiler fail for me, as the new drivers output an extra column: gridZ.)Aha, great! 2D grid limitation was very annoying. OpenCL has 3D for both local and global thread blocks. This made OpenCL->CUDa port quite cumbersome...
What’s New in AMD APP SDK V2.4
Improved OpenCL runtime performance:
Zero copy transfer of data between CPU and GPU for APUs (Windows only)
Improved PCIe transfer speed
Improved kernel launch times
Support for AMD A-series devices
Support for:
AMD C-30 Accelerated Processor with AMD Radeon™ HD 6250 graphics card
AMD C-50 Accelerated Processor with AMD Radeon™ HD 6250 graphics card
AMD E-240 Accelerated Processor with AMD Radeon™ HD 6310 graphics card
AMD E-350 Accelerated Processor with AMD Radeon™ HD 6310 graphics card
AMD Radeon™ HD 6670 graphics card
AMD Radeon™ HD 6570 graphics card
AMD Radeon™ HD 6450 graphics card
Support for CPU/x86 images (Enables the support for image formats, as described in the Khronos OpenCL Specification, to be run on the x86 CPU)
Support for cl_khr_fp64 on the CPU.
Cached reads (-fno-alias)
Binary image format improvements
DXVA buffer sharing
FFT performance improvements
Uninstallation of SDK 2.4 or its components is now done through the Catalyst Install Manager (CIM). See the Installation Notes for a description of the procedure.
For the AMD APP Profiler v2.2:
Support for OpenCL™ application summary
Support for profiling with AMD Radeon™ HD6900 series GPUs (requires AMD APP SDK v2.4 and AMD Catalyst 11.2 or newer)
Improved Session List Explorer
Improved user documentation (now also accessible online in Documentation)
For the AMD APP KernelAnalyzer v1.8:
Support for AMD Catalyst 11.1 – 11.3
Support for setting the OpenCL compiler build options
Improved analysis for AMD Radeon™ HD 6970 graphics cards
Support for displaying x86 assembly code
Commencing with Catalyst 11.3, the AMD Accelerated Parallel Processing (APP) OpenCL runtime is included in the GPU drivers. More frequent updates to the run-time may be obtained by updating the drivers.
yes, in true AMD way as usual.... compatible driver 11.4 listed... but they offer only 11.3 from official page still :-\ :-\ :-\Tested drivers are Cat 11.4, but Cat 11.3 is additionally supported by AMD APP SDK 2.4,
This is the first WHQL-certified release from the Release 270 family of drivers (versions 270.xx to 274.xx).
I tried 270.51 when it was in Beta, and ended up with significant down clocking that could only be cured by a reboot. I'll give this version a go and see what happens.
Steve
I tried 270.51 when it was in Beta, and ended up with significant down clocking that could only be cured by a reboot. I'll give this version a go and see what happens.
Steve
I was getting this with the beta as well, everytime I had to restart Boinc the card would downclock.
So far though it looks like it's been resolved in this release
Now, how do I get rid of this thing? Can I just load the older driver over top of this one or do I have to remove it first then load the older driver?
One more little note, I tried installing 266.58 over the 270 before I uninstalled and that didn't work either.
275.27 Beta has been released today,
Claggy
It took a bit, but it downclocked once again. Back to 266.58
Yep, same observation here. Version 266.58 seems to be rock solidIt took a bit, but it downclocked once again. Back to 266.58
Message from your messenger pidgeon: is there any way we can help to work out if the new drivers do help with the down-clock problem? If that can be established, I can start doing the rounds again....
Key phrase there was I tried the newer drivers before.... That was when they were first released. I was saying I was not going to update drivers again until this mess is sorted out. That's my round about way of agreeing with you "Grumpy" !! ;D
Great idea Pepi but my cheapo machine disabled onboard GPU when I put the extra in there. Can't find any way to get it running again with the 450 in there.And your motherboard is?
Catalyst 11.6 was released yesterday,
Claggy
Catalyst 11.6 was released yesterday,
Claggy
New release notes for APP as well.
(http://www.planet3dnow.de/photoplog/file.php?n=15219&w=o)
Mmmm is this a anything to do with Nvidia or Cuda ??? :-\
Mmmm is this a anything to do with Nvidia or Cuda ??? :-\
Maybe not, but this thread was started almost two years ago, in the early days of CUDA development, and before ATI work had really got underway. Now we're working on both platforms in parallel, it makes sense (to me at least) to have a shared reference point for things like this. Maybe a mod could change the title to reflect the changed development circumstances ("Latest nVIDIA and ATI driver releases"?).
Nvidia 275.5 BETA drivers have been released, with support for CUDA 4.1
Nvidia 275.5 BETA drivers have been released, with support for CUDA 4.1
Anybody got any positive or negative experience with this release ??
Question: Is rescheduling still necessary? Some posts imply changes at SETI have reduced the need. Others refer to Fred's reschedule 2.0 but Search Site didn't find me a link.See the CPU <-> GPU rebranding (http://lunatics.kwsn.net/12-gpu-crunching/cpu-gpu-rebranding-perl-script.360.html) thread for a link to Fred's tool.
and of course VLAR's aren't sent to Nvidia GPU's anymore,
Claggy
VLAR's process reasonably well on ATI GPU's, while there is a slowdown compared to running them on a CPU, it's not nearly as bad as running them on Nvidia GPUs,
and of course VLAR's aren't sent to Nvidia GPU's anymore,
Claggy
Tnx info! The new question is, do VLAR process reasonably on ATI GPU's? Or not process at all? I have two ATI equipped PCs.
Catalyst 11.7 is out,Only 118. worth to try
Claggy
Catalyst 11.7 is out,
Claggy
Only 118. worth to try
Catalyst 11.7 is out,Only 118. worth to try
Claggy
Catalyst 11.7 is out,Only 118. worth to try
Claggy
Win 7 64 only
CAT 11.8. preview (http://support.amd.com/us/kbarticles/Pages/Catalyst118DriverPreview.aspx)
Thanks Mike, got it downloaded too,Catalyst 11.7 is out,Only 118. worth to try
Claggy
Win 7 64 only
CAT 11.8. preview (http://support.amd.com/us/kbarticles/Pages/Catalyst118DriverPreview.aspx)
Working with AMD APU E-350 (HD 6350) and ver 11.6 having so much problems (flickering screen, restarted GPU drivers, not responsive computer, etc.)From the ATI MB release Notes:
with ATI MB optimized, only with this application.
with Collatz no issue or really seldom.
From the ATI MB release Notes:
-period_iterations_num 1 (default value)
If you see lags in GUI or even driver restarts - add this parameter with value >1 (integer numbers).
Claggy
From the ATI MB release Notes:
-period_iterations_num 1 (default value)
If you see lags in GUI or even driver restarts - add this parameter with value >1 (integer numbers).
Claggy
Finally a release notes that does not say this about HD 5970: "supported only in single mode" - which means that now multi gpu 5970 is finally supported. Hopefully.
I will hold my horses until final GA on SDK and CAL before I test, as it's a real hazzle to update/change on a Windows server 2008...
In all previous SDK release notes it was clearly stated in footnote that 5970 was only supported in single GPU mode. Now there is no "single" in release notes, or info stating multi gpu is not supported on 5970.
As for HD 6990 - it's not on the supported list, so I guess this card it yet to be supported fully.
AMD APP SDK 2.5 RC2 is available for download
http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx
Claggy
Still RC2. Release link gives not found 404 errorAMD APP SDK 2.5 RC2 is available for download
http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx
Claggy
Its full version now.
If your get the URL for the 32bit release, and edit it to the 64bit filename, then the download works,Still RC2. Release link gives not found 404 errorAMD APP SDK 2.5 RC2 is available for download
http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx
Claggy
Its full version now.
If your get the URL for the 32bit release, and edit it to the 64bit filename, then the download works,Still RC2. Release link gives not found 404 errorAMD APP SDK 2.5 RC2 is available for download
http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx
Claggy
Its full version now.
Claggy
If your get the URL for the 32bit release, and edit it to the 64bit filename, then the download works,Still RC2. Release link gives not found 404 errorAMD APP SDK 2.5 RC2 is available for download
http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx
Claggy
Its full version now.
Claggy
ROFL, if such tricks required only to get this SDK I could imagine how it will work after that.... better stay with older one for now ;)
I have the 64-bit version downloading now.
It will not do any good at the moment as I am running only Nvidia currently.
concurrent kernel execution is not supported in SDK 2.4 and not planned for 2.5 either.
-------------------------
Micah Villmow
Advanced Micro Devices Inc.
--------------------------------
The information presented in this document is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors. Links to third party sites are for convenience only, and no endorsement is implied.
I also have a NVIDIA Fermi card, and there is no way to get concurrent kernels on that using their OpenCL (driver v280.19 w/ OpenCL 1.1), with either single out-of-order queue or multiple queues. You can get concurrent kernels using CUDA Runtime API, but then it's not in OpenCL.
Ken
Let me eat lunch first, then I'll download it and run chirp test on the 420M (we have temps for that, too). Just 3d, or would any of the others be useful?
For the best experience, NVIDIA strongly recommends installing the following DELL SBIOS update before installing the NVIDIA notebook reference driver.
Installing this driver without the latest Dell SBIOS may cause the system fan noise to become more noticeable.
...Temperatures seemed lower during the ChirpTest, perhaps because of that. Now I'm back to normal running, the GPU temperature is stable at 71C, much as it was before the upgrade.
...Only problem so far is that she chose all good gear...
She didn't have any, er, assistance in making that choice, did she - by any chance?
...
I am coming back up from 270.48 that I installed trying to find the desktop bug I am still seeing.
How about 280?considerably better I'm told
I might end up loading that one once I finish rebuilding my AMD machine, I think I will end up with a mixed GPU environment for testing purposes.When i tested the preview version of Cat 11.8, OpenCL apps consumed lots of CPU time, just like Cat 11.7, and the Nvidia 27x.xx and later drivers,
Primary will be my ATI HD5830 and the GTX560 will go in the second slot.
Installed 280.26 on two of my Win 7 64 machines with GTX 295.
One produced a BSOD or hang ever couple of hours and produced Seti errors. Back to 280.26 no problems yet.
Other one produced no errors but only about 16 second runtime before it finished (Seti enhanced), with an extremely high CPU load of 100%, more than normal.
I already noticed the RAC going down on that machine.....
Did a restart, that didn't solve a thing.
Back to 208.26 and it works just fine.
Using the latest Lunatics installer.
Other one produced no errors but only about 16 second runtime before it finished (Seti enhanced), with an extremely high CPU load of 100%, more than normal.
I already noticed the RAC going down on that machine.....
When i tested the preview version of Cat 11.8, OpenCL apps consumed lots of CPU time, just like Cat 11.7, and the Nvidia 27x.xx and later drivers,
Claggy
installed the latest Nvidia driver and I can confirm the memory mamangement was improvement compared to 277.How about 280?considerably better I'm told
... NVIDIA GPU 0: GeForce 8600 GTS (driver version unknown, CUDA version 4000, compute capability 1.1, 240MB, 96 GFLOPS peak)
Fred, you're getting me really confused here.
Me too ;D.
I tried 280.26 on 2 computers and had to back to 275.33. 275.33 works just fine.
But this is a driver pack and drivers are probably specific to certain cards.
No it is a desktop PC. I'm still running Boinc client version 6.10.60 if it matters.... NVIDIA GPU 0: GeForce 8600 GTS (driver version unknown, CUDA version 4000, compute capability 1.1, 240MB, 96 GFLOPS peak)
is that an Optimus notebook/platform ? I've provided a fix for this to Berkeley via messenger pigeon. It's cosmetic only really, but very annoying. That should be fixed in some future Boinc (It's not a driver problem).
Jason
is there any improvement with last ATI 11.8??
Out of curiosity, how many entries are there in the 'display adapters' section of device manager - and what else is listed, apart from your 8600 GTS? (i'm thinking motherboard video here).It is a mixed host having an ATI 4550 (soon to be replaced by a 6670) along with the 8600 GTS.
OK, you must have just found some other weird configuration that Boinc driver version detection (cosmetic) fails under. Just checking :)
Yup, I reckon that qualifies as 'one David didn't think of' ::)
i notice that you are only linking to Throttle and BoincTasks.A bit off topic.
Do you no longer support/reccomend your rescheduler ?
Thanks for the verification. Yeah it was a relatively trivial cosmetic, but oddly annoying, problem. As mentioned the updated code was passed onto Berkeley, and I see no reason it shouldn't function the same way there when they get around to a build with the change.Hi Jason,
Jason
Could one of our resident ATI driver gurus have a look at http://boinc.berkeley.edu/dev/forum_thread.php?id=6846, please?
........
The posters in that thread seem to be implying that ATI drivers up to and including 11.6 could be used in service mode, but 11.7 and onwards have moved towards the NVidia behaviour. Is that a change which the people who monitor such things can corroborate?
NVidia behaviour
Could one of our resident ATI driver gurus have a look at http://boinc.berkeley.edu/dev/forum_thread.php?id=6846, please?
It's always been the case that NVidia drivers are inaccessible to BOINC installed in 'Protected Application Execution' (aka 'service') mode under Windows Vista and Windows 7.
The posters in that thread seem to be implying that ATI drivers up to and including 11.6 could be used in service mode, but 11.7 and onwards have moved towards the NVidia behaviour. Is that a change which the people who monitor such things can corroborate?
BUT (!) ATi claims that starting from SDK 2.5 OpenCL can be used in service mode...
BUT (!) ATi claims that starting from SDK 2.5 OpenCL can be used in service mode...
'can' and whether Boinc fully implements the required new procedures are two different things. Cuda apps 'can' too, but there is required a protected part & a user mode part for those. If AMD found a way to make it work completely transparently then good show, but I do expect some code gymnastics would be required even then.
Jason
Is that like ATI saying they will resolve the high CPU usage in Cat 11.8 & SDK2.5 ::) :D
I've just tested this out of idle curiosity, and with Boinc installed in protected execution mode, and Cat 11.8 SDK2.5 installed, Boinc mngr can't connect to the client let alone see the GPU
I'd love to be able to run Boinc as a service, but think there's a way to go before it gets there
With 11.8 I would get constant and continuous 100% CPU usage (of one core) as soon as a GPU task was running.
This seems to be fixed with 11.9RC. But not for multi-GPU setups. ... %$!?@# AMD/ATI ::)
This seems to be fixed with 11.9RC. But not for multi-GPU setups. ... %$!?@# AMD/ATI ::)
Thanks for this pointer Frizz. Got the 11.9RC drivers working on my 5670, with only a couple of % CPU usage on r365 ;D Finally it looks like ATI are getting there. Will test later that it's the same with ATI AP tasks
Anyone can confirm CPU usage reduction with 11.9RC for HD6xxx GPUs ?
Also, link to drivers would be nice...
Anyone can confirm CPU usage reduction with 11.9RC for HD6xxx GPUs ?
Also, link to drivers would be nice...
Here's a link to Guru3d who have the drivers http://downloads.guru3d.com/AMD-Catalyst-11.9-RCx-Win7-Vista-download-2783.html
285.27 has been released in betaAnyone know if CPU usage has been reduced on OpenCL apps with this driver?
http://www.nvidia.com/object/win7-winvista-64bit-285.27-beta-driver.html
285.27 has been released in betaAnyone know if CPU usage has been reduced on OpenCL apps with this driver?
http://www.nvidia.com/object/win7-winvista-64bit-285.27-beta-driver.html
Claggy
Looks like AMD ahead then ;)
Looks like AMD ahead then ;)
or not as far back ;D ;D ;D
Right at the end of a task ? Fancy that! LoL. ::)
I think part of the problem is he is running CUDA 3.1 and 285.xx is CUDA 4.1.
I don't know if I am posting in the correct thread but I will take a chance.1) Older drivers show low CPU consumption.
Referencing the problem of having excessively high cpu run time on the OpenCL science apps when using the newer video drivers (Version 270+). Is this a video driver fault or an Opencl app fault? I am using NVIDIA so specially interested in this fault on this product.
Thanks.
O, no... benchmarking again... ;D
Cat 11.10 preview released.
http://support.amd.com/us/Pages/AMDCatalyst1110Previewdriver.aspx
...The system ran fine without any such problems when I was still using the latest WHQL drivers (280.26). Have other people here encountered similar errors with the latest BETA drivers?...
...The system ran fine without any such problems when I was still using the latest WHQL drivers (280.26). Have other people here encountered similar errors with the latest BETA drivers?...
Have made the transition myelf yesterday on my main machine (with GTX 480), and not had such incidents (yet) but there's still time :) ... Something rings a bell with what you describe though, though I can't place it. Did you use the 'clean install' advanced option ?
Yes, I used the Advanced -> Clean Install option. (I always do ever since NVidia introduced this option...)
Cat 11.10 preview V2 released.
http://support.amd.com/us/kbarticles/Pages/GPU122AMDCat1110PreDriverV2.aspx
Time to burn AMD... http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=155039&startpage=1
with Cat 11.10 preview 2 I getting invalids on each and every test WU.... >:( >:( >:( >:( >:(
I checked MB app.Time to burn AMD... http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=155039&startpage=1
with Cat 11.10 preview 2 I getting invalids on each and every test WU.... >:( >:( >:( >:( >:(
Still waiting for any of my AP tasks to be validated using the 11.10v2 drivers all tasks reported on Oct. 10th (http://setiathome.berkeley.edu/results.php?hostid=6195918&offset=0&show_names=0&state=2&appid=5) were completed with the 11.10v2 drivers after clean Windows install so we'll see what happens with them :o
Edit: Just ran a quick AP Bench with 11.10v2 SDK2.5 & then 11.2 SDK2.4 results (attached) look fairly similar to me
Edit2: Previous ATI build was ran against CAL 1.4.1523 (can't remember what Cat version that is :()
I checked MB app.
Just downloaded: http://www.nvidia.com/object/win7-winvista-64bit-285.62-whql-driver.html
almost afraid to install as big crunch here has been doing so well on 285.38 Beta.
Just downloaded: http://www.nvidia.com/object/win7-winvista-64bit-285.62-whql-driver.html
almost afraid to install as big crunch here has been doing so well on 285.38 Beta.
6.13.9 does not work, something is not getting saved properly and it loses the client state.
I did a full backup of my BOINC folder and am back up and running with 6.12.39.
Cat 11.10 (final) is available on AMDs download servers for Windows and Linux. (They just have to update their websites.) ;)
...And some other nasty bugs for Linux.Using still SDK 2.4 on some openSuse 11.3, 64bit.
I wonder why AMD drivers are of such bad quality (recently). Maybe because they are internally already adding support for the new GCN architecture? And break support for "older" cards?More that they never reached a "clean level" of having a minimum number of problems in a driver. Instead carrying this payload of problems with them makes it harder to implement something new(er) a way that it works. :-\
We need to locate place of failure. I've seen the same with windows Cat preview.Added my findings over there.
still unknown where exactly data start to differ between driver versions.
Please, post your findings in corresponding thread on AMD forum too.
http://forums.amd.com/devforum/messageview.cfm?catid=390&threadid=155591&enterthread=y
Anyone here found any improvements for CUDA?
BTW, what computation interfaces NV Tegra2-based devices are support?
There are tabletPCs running Android OS equipped with Tegra2. AFAIK no CUDA or OpenCL support, right?
I keep my GTX460 on 266.58 for that reason.
I keep my GTX460 on 266.58 for that reason.
My GTX460 likes the 285.38 Beta better than anything before or since.
I keep my GTX460 on 266.58 for that reason.
My GTX460 likes the 285.38 Beta better than anything before or since.
Are you running OpenCL AP work on yours?
I keep my GTX460 on 266.58 for that reason.
My GTX460 likes the 285.38 Beta better than anything before or since.
Are you running OpenCL AP work on yours?
i haven't tried that yet. Would you suggest it and if so
what do i need to do ?
[edit]Not seeing any problems on my computer, Milkyway is using 0.05CPU and about 1.1% or so.MW never was OpenCL app.... Though, with deprecation of CAL it would be in some future....
290.36 Beta is out.
Claggy
I will stick with 285.79 for now as it supports the newer stuff.
I have to wonder if that is going to matter since the new ATI apps are incompatible with cat 11.10 and cat 11.11 drivers. Will this solve the incompatibility problem ?
AMD SDK 2.6 will be released december 13th.
Some key changes pronounced.
http://blogs.amd.com/developer/2011/11/28/amd-opencl%E2%84%A2-app-sdk-preview/
Mike
I have to wonder if that is going to matter since the new ATI apps are incompatible with cat 11.10 and cat 11.11 drivers. Will this solve the incompatibility problem ?
AMD SDK 2.6 will be released december 13th.
Some key changes pronounced.
http://blogs.amd.com/developer/2011/11/28/amd-opencl%E2%84%A2-app-sdk-preview/
Mike
Would an 'older card' (i.e. 9800 GT-series) beta bench be useful at this stage? (I don't usually upgrade drivers until the WHQL stage). If so, XP or Win7 (it'll be 32bit without the option).
From Open CL 1.2 ... Device partitioning this is what I was waiting for ...years :P ...any plans for the future?I finally realized that I can see the maximum power and GFLOPS from my cpu and gpu with sisoft sandra.
It would be great to assign just one CU to windows, and the rest to boinc :P. It's better than temp throttling :)
OpenCL is only available on HD4*** and later GPU's, HD3850 is too early, Supported Devices:
11.11 drivers are incompatible with the OpenCL apps.
You can try 11.9 for example.
Did you check workgroup size of the 3850 ?
If its only 128 you might want to try LHD4k version.
Mike
Has anyone tried crunching now a days with SLI connector attached and enabled? Does it work now?Not with with my cards, My 590 got real upset with SLI enabled.
For all ATI AGP(like mine HD4670) cards driver 11.11 is availableIt should work, Treasurer has an AGP HD4650, and has found a combination of drivers/SDK that work (http://setiathome.berkeley.edu/forum_thread.php?id=64196&nowrap=true#1175333):
CatalystAGPHotfix (http://support.amd.com/us/kbarticles/Pages/CatalystAGPHotfix.aspx)
Last Updated
11/15/2011
hoping this works with ap_5.06_win_x86_SSE2_OpenCL_ATI_r521.exe
did anyone tried with AGP ?
heinz
11.11 drivers are incompatible with the OpenCL apps.
You can try 11.9 for example.
Did you check workgroup size of the 3850 ?
If its only 128 those apps won´t work.
Mike
Cat 11.12 drivers have just been releasedSDK_2.6 is out as well, the MD5 checksums have been updated, but not the links, so changing the url from v2.5 to v2.6 gets a download started,
and finally it seems that we're going to get decent drivers from AMD
Bench is looking good against stock at the moment - Q's ~95% and upwards - will post when its finished
With Catalyst 11.12 (tested: Win7x64) AMD finally managed to solve the 100% CPU load bug (even for systems with more than one GPU) :)
for me still no crunching with my HD3850 Cat 11.12 already set-up supposed that should be another problem except for driver some kind of ghost inside computer.With Catalyst 11.12 (tested: Win7x64) AMD finally managed to solve the 100% CPU load bug (even for systems with more than one GPU) :)
Sounds great ! What's average CPU usage at the moment ?
Sounds great ! What's average CPU usage at the moment ?
With Catalyst 11.12 (tested: Win7x64) AMD finally managed to solve the 100% CPU load bug (even for systems with more than one GPU) :)Meanwhile have tested this and it seems to be solved for single GPU :) , but NOT for multi-GPU hosts on 64 bit linux. CPU load with Cat 11.12 is around 80% when crunching MBv7r390_HD5 (Example: http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=3710311 ) :(
Sounds great ! What's average CPU usage at the moment ?
Almost zero. For non-blanked APs its between 2-4 %.
(for bitcoin mining it's between 0-1%).
Sounds great ! What's average CPU usage at the moment ?
Almost zero. For non-blanked APs its between 2-4 %.
(for bitcoin mining it's between 0-1%).
Thats real good info. ;D
Dump File | Crash Time | Bug Check String | Bug Check Code | Parameter 1 | Parameter 2 | Parameter 3 | Parameter 4 | Caused By Driver | Caused By Address | File Description | Product Name | Company | File Version | Processor | Crash Address | Stack Address 1 | Stack Address 2 | Stack Address 3 | Computer Name | Full Path | Processors Count | Major Version | Minor Version | Dump File Size |
Mini121511-01.dmp | 15.12.2011 02:00:54 | THREAD_STUCK_IN_DEVICE_DRIVER | 0x000000ea | 0x89989948 | 0x89a68e08 | 0x88f0d248 | 0x00000001 | ati2cqag.dll | ati2cqag.dll+1aeb4 | Central Memory Manager / Queue Server Module | ATI Radeon Family | ATI Technologies Inc. | 6.14.10.0534 | 32-bit | ati2cqag.dll+1ad92 | | | | | C:\WINDOWS\Minidump\Mini121511-01.dmp | 2 | 15 | 2600 | 98.352 |
Found some freeware too called "BlueScreenView" to decipher bluescreens (http://www.nirsoft.net/utils/blue_screen_view.html) on windows XP and later. Hopefully that can help AMD to ease bug search a little.
Dump File Crash Time Bug Check String Bug Check Code Parameter 1 Parameter 2 Parameter 3 Parameter 4 Caused By Driver Caused By Address File Description Product Name Company File Version Processor Crash Address Stack Address 1 Stack Address 2 Stack Address 3 Computer Name Full Path Processors Count Major Version Minor Version Dump File Size Mini121511-01.dmp 15.12.2011 02:00:54 THREAD_STUCK_IN_DEVICE_DRIVER 0x000000ea 0x89989948 0x89a68e08 0x88f0d248 0x00000001 ati2cqag.dll ati2cqag.dll+1aeb4 Central Memory Manager / Queue Server Module ATI Radeon Family ATI Technologies Inc. 6.14.10.0534 32-bit ati2cqag.dll+1ad92 C:\WINDOWS\Minidump\Mini121511-01.dmp 2 15 2600 98.352
Hi corsair,
I had have similar issue when I changed from my Sapphire HD3870 to NVidia.
I could resolve it by using Revo Uninstaller Pro (http://www.revouninstallerpro.com/) to clean the computer from all its ATI stuff.
Perhaps this can halp you too.
heinz
Only if this will be posted not here but on AMD forums and even better, bug ticket will be opened.I would have done so, but last night when i was ready to do so, AMDs developer fora were not accessible.
Can you install Vista or Win7 on that host ?Only if this will be posted not here but on AMD forums and even better, bug ticket will be opened.I would have done so, but last night when i was ready to do so, AMDs developer fora were not accessible.
Win7x64 if necessary.Can you install Vista or Win7 on that host ?Only if this will be posted not here but on AMD forums and even better, bug ticket will be opened.I would have done so, but last night when i was ready to do so, AMDs developer fora were not accessible.
Try this, please.Reinstall in progress.
WinXP support ceased so we never get this bug fixed if it's WinXP only and we will not have evidences that Win7 involved too.
Received repaired the NVidia GTX460, un-installed the ATI HD 3850 and drivers (with un-installing application + driversweeper and set-up with NVidia ver 285.62 now crunching properly, why is not crunching with ATI ?? ??Which projects couldn't you get the HD3850 to crunch on?, and what drivers did you have installed?, did try rolling back to earlier drivers?
Which projects couldn't you get the HD3850 to crunch on?, and what drivers did you have installed?, did try rolling back to earlier drivers?
Claggy
O.K, with Seti ATI MB and PrimeGrid you're never going to get them to work, as both those projects ATI apps are OpenCL apps, and the HD3850 isn't OpenCL capable, you need an HD4*** or later ATI GPU for OpenCL capability.Which projects couldn't you get the HD3850 to crunch on?, and what drivers did you have installed?, did try rolling back to earlier drivers?
Claggy
with HD3850 milky-way (this card is DP), Collatz, SETI MB ATI and Prime grid and no success with any of then, when I built the computer I was able to crunch at least in milky-way and Collatz, at that time there was not MB for ATI and I was not signed for Prime grid.
I've used several version the last 11.11 & 11.12 don't know what happens but in another computer it works but not in this since the first time I exchanged the ATI for a NVidia 9800 and because a problem I switched back to ATI again and in those times no success too, well I suppose I'll should have to give retirement to this GPU and start thinking in fdisk, format and re-install OS in computer (win7-pro-x64-sp1-ES)
Need a tip please : How do i turn that watchdog system in win7 off ?
I'm thinking on what has become my yearly upgrade. This year I think it will be just a new GPU. What's the advice on the most cost effective, and project effective, GPU right now?
Thanks Art
Need a tip please : How do i turn that watchdog system in win7 off ?
Urs if you're referring to User account control go here:
Control Panel\All Control Panel Items\User Accounts
Choose "Change User Account Control Settings". Set it to minimum for no nagging, or a position of your choice.
If you meant the so-called "god mode" here's one link of many with instructions:
http://howto.cnet.com/8301-11310_39-10423985-285/understanding-windows-7s-godmode/
Gives instant access to all those things now hidden in W7 that were more accessible in XP.
Regards
Art
I expect Urs was looking for control of the 2 seconds a GPU kernel can run before Windows thinks it has hung and forces a reinitialization. He's probably already found it, but the MSDN Timeout Detection and Recovery of GPUs through WDDM (http://msdn.microsoft.com/en-us/windows/hardware/gg487368.aspx) article defines the registry keys and how they interact. A web search for TdrDelay or TdrLevel turns up many pertinent discussions with advice.Actually with Win7 i feel much like a newby, so didn't even know the correct terms to look for. In Linux and XP this is or was called "VPUrecover" and it does or did restart possibly hanging drivers. Actually i wanted the "old" behaviour back to extract the place of failure in the driver and derive a dump containing that info. This dump would help AMDs developers to look up where they need to improve their drivers.
Joe
And 560 TI if you have the skills to control it.
regards
Mike
Mike can you elaborate on "...skills to control it." ? What is it that needs controlling? Maybe there's a link to a discussion here? Today I just about decided on 560Ti or, if I can't "control it" a 460.
Has anyone tried crunching now a days with SLI connector attached and enabled? Does it work now?To answer my own question, the answer is "yes"
What PSU do you have running that host?, I'd expect at least 800 Watt PSU is required,
Claggy
I agree, Asus's Recommended Power Supply Wattage Calculator (http://support.asus.com/powersupply.aspx) says 800W when you put in 2*560Ti's and 900Watts when you put in 2*465's,What PSU do you have running that host?, I'd expect at least 800 Watt PSU is required,
Claggy
For a 465 + 560ti factory OC'd model, it would have to be at least a very good 850W, or preferably larger IMO.
Which application Art ? x41g Cuda 3.2 release ? How many tasks per card ? What are the temperatures like & did you do a 'Clean Install' with the driver installtion advanced options?
Sorry for the questions, but they'll help narrow down the possibilities.
Offthe top of my head, One possibility could be other wacky drivers in the system, so a check with DPC latency Checker (http://www.thesycon.de/deu/latency_check.shtml) might be in order as an easy pre-check before troubleshooting.
[Edit:] for reference, her's what my 480 system (main dev machine) looks like at full crunch while browsing with Google Chrome, which took a lot of work to find and force installation of better drivers for chipset & Wifi etc.
Jason
OK, there's good news & there's bad news. Sounds mostly good, but no way is a 750 watt PSU enough for both those cards at once, I have a 480 & a 560ti in separate machines. The 560ti by itself seems happy with a Corsair AX 750 Gold... The 480 took about 4 months or so by itself to hose a seasonic gold 750. I'd suggest pulling the 465 for now, before your PSU blows up. I'm not being silly.
Jason
I hear ya. Ran to get my VOM. THe PS outlet air didn't feel very hot and it isn't, about 37 deg C. Doesn't seem like its very stressed -- but the point is a good one. I'll get a 1kW in there. My latency picture is not so sweet with that program you suggested, got work to do there.
Thanks ARt
Report at GPUGrid (http://www.gpugrid.net/forum_thread.php?id=2818) that the CUDA 4.1 developer toolkit went onto public production release today - downloadable from http://developer.nvidia.com/cuda-downloads
Presumably there'll be consumer-grade drivers to follow, but I can't find anything newer than 290.53 beta yet. WHQL drivers are still 285.58
Anything useful in cuda 4.1 for us? Does the optimizing compiler make any difference to earlier ones?I saw OpenCL 1.1 support on FERMI (not sure it was added only now though).
Anything useful in cuda 4.1 for us? Does the optimizing compiler make any difference to earlier ones?
The kitties were rather hoping for another driver release beyond 290.53 to go along with 4.1 too. Hope they haven't given up on it.
Anything useful in cuda 4.1 for us? Does the optimizing compiler make any difference to earlier ones?
Decent performance gain on Fermis. Older GPUs are seeing a slowdown over Cuda 3.2, which has been reported & apparently reproduced by nVidia. I was hoping they'd fix that by release, as it means will probably have to support 2 builds next installer.
Jason
It was Cuda41rc2 and Cuda41 apps that were quite a bit slower than Cuda32 apps on Pre-Fermi GPUs (like my 9800GTX+) with the Cuda41 drivers, Fermi's have a performance increase with Cuda41rc2 and Cuda41 apps,Anything useful in cuda 4.1 for us? Does the optimizing compiler make any difference to earlier ones?
Decent performance gain on Fermis. Older GPUs are seeing a slowdown over Cuda 3.2, which has been reported & apparently reproduced by nVidia. I was hoping they'd fix that by release, as it means will probably have to support 2 builds next installer.
Jason
Was that cuda 3.2 apps on the beta driver?
Have you compiled a cuda 4.1 app to see how it goes?
MarkJ
295.73 WHQL is out, downloading now.
Claggy
295.73 WHQL is out, downloading now.
Edit: the disappearing cuda device bug is still apparent in these drivers too.
Claggy
My E8500/9800GTX+ machine had no problem with the 295.51 drivers when connected through the VGA port, not tried 295.73 on it yet,295.73 WHQL is out, downloading now.
Edit: the disappearing cuda device bug is still apparent in these drivers too.
Claggy
Damn. :( The performance increases for several of the games I play (or want to play) are very tempting.
IIRC the disappearing cuda device bug happens when the card switches off its output to blank the screen? So that's something that would happen to me after ~15 minutes of no activity. Is there a way to get around this?
Regards, Patrick.
My E8500/9800GTX+ machine had no problem with the 295.51 drivers when connected through the VGA port, not tried 295.73 on it yet,295.73 WHQL is out, downloading now.
Edit: the disappearing cuda device bug is still apparent in these drivers too.
Claggy
Damn. :( The performance increases for several of the games I play (or want to play) are very tempting.
IIRC the disappearing cuda device bug happens when the card switches off its output to blank the screen? So that's something that would happen to me after ~15 minutes of no activity. Is there a way to get around this?
Regards, Patrick.
Claggy
1/31/2012 15:20:30 | SETI@home | Restarting task 04oc11ad.507.107123.6.10.252_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:20:38 | SETI@home | Restarting task 04oc11ad.507.90813.6.10.118_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:20:58 | SETI@home | Restarting task 04oc11ad.13433.885.8.10.85_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:21:06 | SETI@home | Restarting task 04oc11ad.13433.885.8.10.185_0 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:21:26 | SETI@home | Restarting task 02oc11ag.7971.6806.5.10.28_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:21:34 | SETI@home | Restarting task 04oc11ad.507.4566.6.10.54_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:21:54 | SETI@home | Restarting task 02oc11ag.7971.6806.5.10.26_0 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:22:03 | SETI@home | Restarting task 02oc11ag.7971.13303.5.10.138_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:22:22 | SETI@home | Restarting task 04oc11ad.507.87132.6.10.207_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:22:32 | SETI@home | Restarting task 02oc11ag.7971.13303.5.10.149_0 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:22:50 | SETI@home | Restarting task 02oc11ag.19404.4569.3.10.79_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:23:00 | SETI@home | Restarting task 02oc11ag.19404.4569.3.10.119_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:23:18 | SETI@home | Restarting task 25oc11ab.4355.134274.16.10.91_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:23:29 | SETI@home | Restarting task 02oc11ag.7971.13303.5.10.147_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:23:47 | SETI@home | Restarting task 25oc11ab.4355.140409.16.10.114_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:23:57 | SETI@home | Restarting task 04oc11ad.507.107123.6.10.254_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:24:14 | SETI@home | Restarting task 04oc11ad.507.90813.6.10.118_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:24:26 | SETI@home | Restarting task 04oc11ad.507.107123.6.10.252_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:24:43 | SETI@home | Restarting task 04oc11ad.13433.885.8.10.85_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:24:55 | SETI@home | Restarting task 04oc11ad.13433.885.8.10.185_0 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:25:11 | SETI@home | Restarting task 04oc11ad.507.4566.6.10.54_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:25:23 | SETI@home | Restarting task 02oc11ag.7971.6806.5.10.28_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:25:39 | SETI@home | Restarting task 02oc11ag.7971.6806.5.10.26_0 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:25:51 | SETI@home | Restarting task 04oc11ad.507.87132.6.10.207_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:26:07 | SETI@home | Restarting task 02oc11ag.7971.13303.5.10.149_0 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:26:23 | SETI@home | Restarting task 02oc11ag.19404.4569.3.10.79_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:26:36 | SETI@home | Restarting task 02oc11ag.19404.4569.3.10.119_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:26:52 | SETI@home | Restarting task 25oc11ab.4355.134274.16.10.91_1 using setiathome_enhanced version 610 (cuda_fermi)
1/31/2012 15:27:04 | SETI@home | Restarting task 02oc11ag.7971.13303.5.10.147_1 using setiathome_enhanced version 610 (cuda_fermi)
setiathome_CUDA: cudaGetDeviceCount() call failed.
setiathome_CUDA: No CUDA devices found
setiathome_CUDA: Found 0 CUDA device(s):
setiathome_CUDA: No SETI@home capabale CUDA GPU found...
Is there any improvement for ATI with new 12??Milkyway doesn't appear to like the drivers. I get a message through the eventlog to back down from12.1
I can't download for the moment I'm away home with roaming mobile tariffs and is too expensive to check it.
I'll try in next April with my shore leaves.
Primegrid just posted a message to either downgrade from 295.xx or not install it at all.GPUgrid also did a notice to everyone about 295.x so just stay away from it. Hopefully Nvidia will push a fixed one out for the GTX600's launch some time this month.
I had a PM off someone at Nvidia on friday morning who was trying to reproduce the issue, he had attached to GPUGRID but hadn't managed to reproduce it,Primegrid just posted a message to either downgrade from 295.xx or not install it at all.GPUgrid also did a notice to everyone about 295.x so just stay away from it. Hopefully Nvidia will push a fixed one out for the GTX600's launch some time this month.
I had a PM off someone at Nvidia on friday morning who was trying to reproduce the issue, he had attached to GPUGRID but hadn't managed to reproduce it,
i suggested he try different projects with shorter workunits (Seti & PrimeGrid) and i gave him the steps i used with the MBBench program to reproduce the issue,
Claggy
ATI Catalyst 12.2. pre-certified drivers released.Couldn't get that one to work with a HD7750, but 12.3 preview works ok. (look for it at guru3d)
http://support.amd.com/us/kbarticles/Pages/catalyst122ppecertifieddriver.aspx
Official HD 7.x support.
Mike
NVidia 296.10 WHQL drivers are out.
Update: this bug is considered as release critical (show-stopper) for the next NVIDIA driver release that's due in 2-4 weeks. Thus a fix will be available by that time.
More news from Einstein (http://einstein.phys.uwm.edu/forum_thread.php?id=9307&nowrap=true#116415):Quote from: Oliver BockUpdate: this bug is considered as release critical (show-stopper) for the next NVIDIA driver release that's due in 2-4 weeks. Thus a fix will be available by that time.
You have got to wonder why this was not picked up at Beta stage !!!More news from Einstein (http://einstein.phys.uwm.edu/forum_thread.php?id=9307&nowrap=true#116415):Quote from: Oliver BockUpdate: this bug is considered as release critical (show-stopper) for the next NVIDIA driver release that's due in 2-4 weeks. Thus a fix will be available by that time.
More news from Einstein (http://einstein.phys.uwm.edu/forum_thread.php?id=9307&nowrap=true#116415):Quote from: Oliver BockUpdate: this bug is considered as release critical (show-stopper) for the next NVIDIA driver release that's due in 2-4 weeks. Thus a fix will be available by that time.
Capeverde - is it HD7xxx GPU ?
Good to see OpenCL 1.2 available. It definitely has some opportunities for better app performance with proper optimization done.
Capeverde - is it HD7xxx GPU ?I use a pair of HD7750 in my Sandy Bridge host since late February and was sitting without a working driver before 12.4 preview was reaching public mid of march. :D
Good to see OpenCL 1.2 available. It definitely has some opportunities for better app performance with proper optimization done.
301.24 Beta drivers are available from Nvidia nowThis one works too:http://www.nvidia.co.uk/Download/Find.aspx?lang=en-uk
Edit links broken this one works http://www.geforce.com/drivers/results/43622
Does the CUDA sleep bug exist for this driver release?301.24 Beta drivers are available from Nvidia nowThis one works too:http://www.nvidia.co.uk/Download/Find.aspx?lang=en-uk
Edit links broken this one works http://www.geforce.com/drivers/results/43622
http://uk.download.nvidia.com/Windows/301.24/301.24-desktop-win7-winvista-64bit-international-beta.exe
Claggy
I don't know, there's no mention of it in the Release Notes as being fixed, or being an outstanding Bug,Does the CUDA sleep bug exist for this driver release?301.24 Beta drivers are available from Nvidia nowThis one works too:http://www.nvidia.co.uk/Download/Find.aspx?lang=en-uk
Edit links broken this one works http://www.geforce.com/drivers/results/43622
http://uk.download.nvidia.com/Windows/301.24/301.24-desktop-win7-winvista-64bit-international-beta.exe
Claggy
I don't know, there's no mention of it in the Release Notes as being fixed, or being an outstanding Bug,Does the CUDA sleep bug exist for this driver release?301.24 Beta drivers are available from Nvidia nowThis one works too:http://www.nvidia.co.uk/Download/Find.aspx?lang=en-uk
Edit links broken this one works http://www.geforce.com/drivers/results/43622
http://uk.download.nvidia.com/Windows/301.24/301.24-desktop-win7-winvista-64bit-international-beta.exe
Claggy
It'll be a few hours before my i7-2600K gets to a point where i can install 301.24, I'll let this thread know then, unless someone else gets it installed before hand,
Claggy
Does the CUDA sleep bug exist for this driver release?301.24 Beta drivers are available from Nvidia nowThis one works too:http://www.nvidia.co.uk/Download/Find.aspx?lang=en-uk
Edit links broken this one works http://www.geforce.com/drivers/results/43622
http://uk.download.nvidia.com/Windows/301.24/301.24-desktop-win7-winvista-64bit-international-beta.exe
Claggy
And... What is the verdict? Is the CUDA bug fixed, or still there?Yes, it is fixed, i even downgraded to 295.73 to double check, now crunching on 301.24 again, it did a few Einstein BRP4Cuda tasks this morning no problem,
I'm very tempted to install it, but since I use DVI, and blank my screen, I will probably very susceptible to the bug occurring (and ruining Seti work at the same time).
Regards, Patrick.
And... What is the verdict? Is the CUDA bug fixed, or still there?Yes, it is fixed, i even downgraded to 295.73 to double check, now crunching on 301.24 again, it did a few Einstein BRP4Cuda tasks this morning no problem,
I'm very tempted to install it, but since I use DVI, and blank my screen, I will probably very susceptible to the bug occurring (and ruining Seti work at the same time).
Regards, Patrick.
and has done ~40 odd Seti Cuda tasks late this afternoon with the monitor asleep no problem,
Claggy
Any improvement in performance ?Not good here. On a 3 GPU rig, the 3rd always down clock. Overall its about 2mins slower. I've reverted back 296.35.
when arrived home just updated to the latest official NVidida (296.10 x64) driver and really a nightmare no work could be done to the end in Collatz and Primegrid all CUDA work was ruined I revert to (285.62 x64) and work is running smoothly on a GTX460.As has mentioned repeatilily across various forums, 295.xx and 296.xx drivers have a Monitor Sleep Bug, if a monitor connected to the DVI or HDMI ports goes to sleep the cuda device becomes unavailable,
I am starting to see a few users in 'top computers' on Seti using 301.10, 301.24, and 301.32....
Has anybody had the time to run any valid speed comparisons against 290.53, which has been my trusty standby so far?
thanks for the update supposed that 301.xx still beta?when arrived home just updated to the latest official NVidida (296.10 x64) driver and really a nightmare no work could be done to the end in Collatz and Primegrid all CUDA work was ruined I revert to (285.62 x64) and work is running smoothly on a GTX460.As has mentioned repeatilily across various forums, 295.xx and 296.xx drivers have a Monitor Sleep Bug, if a monitor connected to the DVI or HDMI ports goes to sleep the cuda device becomes unavailable,
eithier set the monitor to never sleep, and physically turn it off, or revert to 290.xx or earlier drivers, or 301.xx or later drivers.
Claggy
Yes, it was when I checked a couple of hours ago.thanks for the update supposed that 301.xx still beta?when arrived home just updated to the latest official NVidida (296.10 x64) driver and really a nightmare no work could be done to the end in Collatz and Primegrid all CUDA work was ruined I revert to (285.62 x64) and work is running smoothly on a GTX460.As has mentioned repeatilily across various forums, 295.xx and 296.xx drivers have a Monitor Sleep Bug, if a monitor connected to the DVI or HDMI ports goes to sleep the cuda device becomes unavailable,
eithier set the monitor to never sleep, and physically turn it off, or revert to 290.xx or earlier drivers, or 301.xx or later drivers.
Claggy
thanks for the update supposed that 301.xx still beta?QuoteYes, it was when I checked a couple of hours ago.
New NVIDIA WHQL 301.42 drivers are available.
New NVIDIA WHQL 301.42 drivers are available.
Any noticable difference from 301.24 beta ?
So from those testing it, is 301.42 a good choice?LoL, offtopic: Diablo 3 has problems here still - no available copies in shops ;D
I'm getting what appears to be thrashing from Diablo 3 & SETI/Collatz, using 280.26.
I ran the cache out and Diablo now has no problems...
304.48 Beta is out,Finally NVidia have fixed the issue where my 3rd card runs at 50Mhz.
Claggy
304.48 Beta is out,
Claggy
Catalyst 12.8 has been released.
Claggy
Catalyst 12.8 has been released.
Claggy
Do the video game increases in performance they claim for this new driver also relate to increased CUDA performance, or are the two not necessarily related?
I just replaced cuda 3.2 with cuda 4.1 dll's in x41g (in app_info.xml). Not sure this is the right way, but it seems working ok for now...Check using Process Explorer (http://technet.microsoft.com/en-us/sysinternals/bb896653.aspx) exactly which DLLs are being used, and where they are being loaded from. You may be surprised - I certainly was, the first time I loaded a CUDA application which could not possibly have worked (as yours can't) - but found that it did.
If this is not a right way, then I probably need a new build...
According to Process Explorer, it seems to be using cuda from C:\Windows\SysWOW64\nvcuda.dll, which is Cuda 5.0.1 from 305.68 drivers.The ones which matter are CUDART and CUFFT.
NVIDIA just released new drivers: 306.97 - WHQL.Hmm ..
(...)
New Driver 310.54 now available
http://www.nvidia.com/Download/Find.aspx?lang=en-us
New Driver 310.54 now available
http://www.nvidia.com/Download/Find.aspx?lang=en-us
Says 'Beta' for me.
GeForce 310.61 drivers are a minor update from GeForce 310.54 drivers with improved compatibility for PlanetSide 2 and Hitman: Absolution.
and now 310.61 Beta is available:QuoteGeForce 310.61 drivers are a minor update from GeForce 310.54 drivers with improved compatibility for PlanetSide 2 and Hitman: Absolution.
I wouldn't be surprised if other things have changed too through
Claggy
and now 310.61 Beta is available:QuoteGeForce 310.61 drivers are a minor update from GeForce 310.54 drivers with improved compatibility for PlanetSide 2 and Hitman: Absolution.
I wouldn't be surprised if other things have changed too through
Claggy
310.70 Beta now available
http://www.nvidia.com/object/win8-win7-winvista-64bit-310.70-driver.html
NVIDIA 310.70 drivers have been released as WHQLdescription:
Driver support for 7xx generation GPUs is EOL. AMD drivers no longer support 7xx generation
GPUs; this includes support for ATI Radeon™ HD and ATI Mobility Radeon™ HD 4000 series
devices, ATI FirePro™ V8750, V8700, V7750, V5700, V2750, ATI Mobility FirePro™ M7740,
and AMD FireStream™ 9270, 9250.
It is necessary to install Catalyst 12.10 graphics drivers prior to installation of SDK 2.8. Vital
components of AMD's OpenCL solution are now contained within the drivers.
:-\ AMDs catalyst-uninstall-utility (http://sites.amd.com/us/game/downloads/Pages/catalyst-uninstall-utility.aspx) for windows was updated, too.Maybe because of Claggy's post in AMD forums? ;) He provided quite big list of leftovers... hope AMD will try to fix that.
:-\ AMDs catalyst-uninstall-utility (http://sites.amd.com/us/game/downloads/Pages/catalyst-uninstall-utility.aspx) for windows was updated, too.Downloaded, I'll try Cat 13.1 tomorrow, If i need to go back to Cat 12.8 i'll try the uninstaller,
Report if problem with broken compiler solved, please.Delayed response: No, Cat 13.1 produced compilations still produce driver restarts with latest AMD/ATI MB apps,
Catalyst 13.2beta3 has been releasedReport if problem with broken compiler solved, please.Delayed response: No, Cat 13.1 produced compilations still produce driver restarts with latest AMD/ATI MB apps,
Claggy
The Windows Cat 13.2 Beta 3 was released two days ago, that version won't even do the compilations,Catalyst 13.2beta3 has been released today. Another one for testing, i guess. Downloading linux driver ....Report if problem with broken compiler solved, please.Delayed response: No, Cat 13.1 produced compilations still produce driver restarts with latest AMD/ATI MB apps,
Claggy
The Windows Cat 13.2 Beta 3 was released two days ago, that version won't even do the compilations,Couldn't get it the linux driver to download yesterday, but today download works properly. Will test it anyway, if it still not works then maybe the next beta will.
Claggy
Just finished test with Catalyst 13.2 beta3 on 64bit openSuSE Linux : The driver crashes have stopped, no more rebooting of the host is necessary now. But the OpenCL compiler still refuses to make binaries from our kernels. It finishes on my SBi5 with a HD7750 right away with the error message : "Internal error: Compilation failed."Tested Catalyst 13.2 beta7 today (on i7IB host), but can't report an progress. Same issue like before!! >:(
So, lets hope for something better ... :-\
And now v314.22 WHQL. That should be the official fixed version for OpenCL on TITAN.
OpenCL-kernels filename : MultiBeam_Kernels_r1779.cl
INFO: can't open binary kernel file: .\\MultiBeam_Kernels_r1779.clHD5_Capeverde.bin_V7, continue with recompile...
Error : Building Program (source, clBuildProgram):main kernels: not OK code -11
Internal error: Compilation failed.
Nvidia has just released Geforce 320.00 Beta GPU driver. I will install tomorrow and see what gives !!!!I'd lay off trying that version just yet, I'm just getting the following on all the x41zc apps:
Cuda error 'cudaMalloc((void**) &dev_cx_DataArray' in file 'c:/[Projects]/__Sources/sah_v7_opt/Xbranch/client/cuda/cudaAcceleration.cu' in line 328 : all CUDA-capable devices are busy or unavailable.
setiathome_CUDA: CUDA runtime ERROR in device memory allocation, attempt 1 of 6
cudaAcc_free() called...
waiting 5 seconds...
Reinitialising Cuda Device...
setiathome_CUDA: Found 1 CUDA device(s):
Device 1: GeForce GTX 460, 1023 MiB, regsPerBlock 32768
computeCap 2.1, multiProcs 7
pciBusID = 1, pciSlotID = 0
clockRate = 1600 MHz
setiathome_CUDA: No device specified, determined to use CUDA device 1: GeForce GTX 460
Cuda error 'cudaMalloc((void**) &dev_cx_DataArray' in file 'c:/[Projects]/__Sources/sah_v7_opt/Xbranch/client/cuda/cudaAcceleration.cu' in line 328 : all CUDA-capable devices are busy or unavailable.
setiathome_CUDA: CUDA runtime ERROR in device memory allocation, attempt 2 of 6
Nvidia has just released Geforce 320.00 Beta GPU driver. I will install tomorrow and see what gives !!!!
http://support.amd.com/us/kbarticles/Pages/AMDCatalyst13-4WINReleaseNotes.aspxAs is Cat 13.5 Beta 2
Cat 13.4 out.
http://support.amd.com/us/kbarticles/Pages/AMDCatalyst13-4WINReleaseNotes.aspxGetting major driver crashes with these on a 6670 / Win7 64bits.
Cat 13.4 out.
While doing what exactly? running what apps? Link to the host?http://support.amd.com/us/kbarticles/Pages/AMDCatalyst13-4WINReleaseNotes.aspxGetting major driver crashes with these on a 6670 / Win7 64bits.
Cat 13.4 out.
Trying 13.1 now.
I somehow have a feeling I'll be going back to 12.6 in the next days :'(
Nothing particular namely being on the desktop, watching a youtube video etc could already lead to driver restart/BSODs.Why are you still running that ancient app, we frequently release updated apps, like the recent ones to fix the Cat 13.2 and later problems, Cat 13.1 is still a driver to avoid,
Leaving SETI running overnight, I found the host had rebooted again.
Tested Cat 13.1 and it seems to have broken OpenCL altogether.
And.......back to 12.6 where all works fine as before.
Host is http://setiathome.berkeley.edu/show_host_detail.php?hostid=4876884
Windows optimized S@H Enhanced application by Alex Kan
Version info: SSE3x (AMD/Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan
SSE3x Win32 Build 426 , Ported by : Raistmer, JDWhale
Nothing particular namely being on the desktop, watching a youtube video etc could already lead to driver restart/BSODs.Newer app versions for enhanced and astropulse are in that post for download (http://lunatics.kwsn.net/12-gpu-crunching/update-for-ati-mb-gpu-app.msg52234.html#msg52234).
Leaving SETI running overnight, I found the host had rebooted again.
Tested Cat 13.1 and it seems to have broken OpenCL altogether.
And.......back to 12.6 where all works fine as before.
Host is http://setiathome.berkeley.edu/show_host_detail.php?hostid=4876884
Nothing particular namely being on the desktop, watching a youtube video etc could already lead to driver restart/BSODs.
Leaving SETI running overnight, I found the host had rebooted again.
Tested Cat 13.1 and it seems to have broken OpenCL altogether.
And.......back to 12.6 where all works fine as before.
Host is http://setiathome.berkeley.edu/show_host_detail.php?hostid=4876884
NVIDIA 320.18 (WHQL) drivers released.The 320.00 drivers had CUDA 5.5 Support too.
"Includes support for applications built using CUDA 5.5"
NVIDIA 320.18 (WHQL) drivers released.The 320.00 drivers had CUDA 5.5 Support too.
"Includes support for applications built using CUDA 5.5"
Claggy
Yep,NVIDIA 320.18 (WHQL) drivers released.The 320.00 drivers had CUDA 5.5 Support too.
"Includes support for applications built using CUDA 5.5"
Claggy
They were the Betas for this public release, I presume?
320.18 has been causing some bsod's here.
Rolling back to 320.00
My question: Is ATI 13.4 Beta working, or should I revert to an earlier version, 12.6 if I can find the link?There should be no problem with running Cat 13.2 or later with the current released apps, (Not had a chance to run the Cat 13.8 Beta driver yet)
Last week I had two MSI twin frozr cards whose fans froz up. Only an ATI47504850 replacement so I installed that. Reading thru the posts I see ATI 13.1 isn't fully functional.
My question: Is ATI 13.4 Beta working, or should I revert to an earlier version, 12.6 if I can find the link?
Thanks
Art
PS MSI wants $35 per card just to look into it. Tech support says the fans are not available separately. There's a Chinese company on Ebay selling MSI branded fans for these cards for about $25 so I ordered replacements. Probably the same company that makes them for MSI.
Fans seem to be a weak point for MSI Frozr cards, I had two MSI Nvidia cards crap out on me last year after the fan started rattling and then slowing and ceasing up. Returned to MSI under warranty and they repaired, but same happened again within 90 days.Last week I had two MSI twin frozr cards whose fans froz up. Only an ATI47504850 replacement so I installed that. Reading thru the posts I see ATI 13.1 isn't fully functional.
My question: Is ATI 13.4 Beta working, or should I revert to an earlier version, 12.6 if I can find the link?
Thanks
Art
PS MSI wants $35 per card just to look into it. Tech support says the fans are not available separately. There's a Chinese company on Ebay selling MSI branded fans for these cards for about $25 so I ordered replacements. Probably the same company that makes them for MSI.
The fans work, we'll see about longevity. GPU 4850 is working fine after updating to latest BOINC and KWSN apps.Several - link to host?
New issue, this PC is getting no CPU work, only GPU. Hard to believe SETI and Milkyway both ran out of CPU work at the same time. How many ways are there for me to screw up and make this happen??
Thanks
Art
MW have lots of CPU apps - if boinc thinks they haven't then something is wrong. try and check their project setting page if you've allowed CPU work and the kind of apps that have CPU. sometimes boinc will choke.
I see you have anon for MW too - did you install CPU apps?
Released AMD Catalistyc 13.9.
will see if is there any improvement.
from my side have to do double install as for the first time the display driver was not updated and after re-boot and done again the set-up the drivers was updated.
GeForce 327.23 WHQL was also released today.
It seems Windows Vista is no longer supported. :-\
It seems Windows Vista is no longer supported. :-\
that's the first thing I noticed before download the bundle that there was no "Vista" naming anywhere in the pack.
time runs too fast. :'(
Yes, i`m still very happy with Vista.
But i will thinking about upgrading to Win 7 since i dont like Win 8 or 8.1.
GeForce 327.23 WHQL was also released today.
already updated this morning too in my GeForce GTX 650 Ti, will see how is the improvement.
GeForce 327.23 WHQL was also released today.
already updated this morning too in my GeForce GTX 650 Ti, will see how is the improvement.
this is the first time that the set-up process goes smoothly as always in previous versions the process stuck at the end of the install of 3D software and had to kill the installation but there was not problem because the system didn't notice any wrong setup.
this time the setup was fine and after installing the 3D feature the program closes itself.
Getting good reports of the BETA GeForce 331.40 Driver at GPUGrid, in particular for high-end (TITAN and GTX 780) cards.I use it on my Titans
Boinc 7.2.5 did not use the RAM correct, it Shows 4GB but we have 6GB
@Richard, can you make a error-report ?
Is Windows 8.1 any better as I found the original Windows 8 to be an abortion and I ended installing my trusty Windows 7 DVD.
Intel 15.33.8.64.3345 (10.18.10.3345) drivers are out for Windows 7/8/8.1, up from the last revision of 15.31.17.64.3257 (9.18.10.3257) as of the 20th Nov.
Claggy
New GeForce v334.89 drivers available as of 1 hour ago.
New GeForce v334.89 drivers available as of 1 hour ago.
New driver for the GTX750 & 750Ti most likely.
New GeForce v334.89 drivers available as of 1 hour ago.
New driver for the GTX750 & 750Ti most likely.
The drivers do include the new cards but also fix various glitches and notably boost performance in Hitman Absolution (Yee Haaa !!), but its to soon to see if this will filter through to Seti performance.
downloading 340.52 now, and i'm back to crunching
I upgraded to the 344.11 drivers on my GTX 660 superclocked and have noticed better crunching times by a few seconds.
Everything seems fine but I did have to boost my fan to 60% to keep the temps that I like.
Are you saying that the driver upgrade has effected your temps ?Ya, But it's not the first time that has happened.I upgraded to the 344.11 drivers on my GTX 660 superclocked and have noticed better crunching times by a few seconds.
Everything seems fine but I did have to boost my fan to 60% to keep the temps that I like.
http://support.amd.com/en-us/kb-articles/Pages/OpenCL2-Driver.aspxAlso consider that first Intel hardware to support OpenCL 2.0 features will be based on Broadwell cores/GPUs (so intel HD5300 and up)
Only Windows supported for now is Win8.1 x64. And only latest GPUs. Looks not too promizing for SETI project where much diversity occurs. OpenCL 2.0 has some nice features worth to be used in SETI apps code though.
And from Intel:
https://software.intel.com/en-us/forums/topic/531074
nVidia out of consideration :-\
https://forums.geforce.com/default/topic/671157/geforce-drivers/open-cl-support-in-nvidia-drivers/
Anyone still crunching on an ATI 6670 card ? What drivers are you running?
I'm having issues with recent Catalyst and Seti@home beta that insists on sending me tasks labeled with _cat132 (like opencl_ati5_cat132).
These tasks usually crash video card driver in a matter of minutes and causes the BOINC job to just sit there and wait forever.
Some of them do complete successfully but it's far too unpredictable.
Lunactic's very own optimized apps run just fine on this card. For now I have disable AMD task in my SETI@home beta preferences.
Host ID: http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=71860
(and yes this is host mixing intel iGPU, Nvidia and AMD ;) )
Anyone still crunching on an ATI 6670 card ? What drivers are you running?Try deleting the compilations for the opencl_ati_cat132 app, and let the app generate new ones.
I'm having issues with recent Catalyst and Seti@home beta that insists on sending me tasks labeled with _cat132 (like opencl_ati5_cat132).
These tasks usually crash video card driver in a matter of minutes and causes the BOINC job to just sit there and wait forever.
Some of them do complete successfully but it's far too unpredictable.
Lunactic's very own optimized apps run just fine on this card. For now I have disable AMD task in my SETI@home beta preferences.
Host ID: http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=71860
(and yes this is host mixing intel iGPU, Nvidia and AMD ;) )
Try Cat 13.12.I had them installed and situation was the same.
Try deleting the compilations for the opencl_ati_cat132 app, and let the app generate new ones.You mean all the .bin files have a name like MB_clFFTplan_Turks_512_r1831.bin ?
Have you tried restarting the host? It was working with that app version two weeks ago.Yes host is set to auto-start and shutdown everyday giving it about 10 hours of computation time per day.
Have you freed any CPU cores?Yes, host is a dual core with BOINC set to use only 50% of avail CPU cores.
Добрый день,
13 января для зарегистрированных разработчиков станет доступен релиз-кандидат CUDA 7.
Вебинар, посвященный обзору CUDA 7, пройдет 21 января в 21:00 по Московскому времени (на английском языке). Для регистрации пройдите по ссылке https://cc.readytalk.com/r/1vhzxvcxwnq8&eom
Ключевые особенности CUDA 7:
Поддержка С++11 позволяет разработчикам С++ проще ускорять свои приложения
· Сократить код с использованием ‘auto’ и ‘lambda’, особенно при задействовании библиотеки шаблонов Thrust
Новая библиотека решателей cuSOLVER для плотных и разреженных матриц обеспечивает значительное ускорение приложениям компьютерного зрения, вычислительной гидродинамики, вычислительной химии и приложениям линейной оптимизации.
· Ключевые решатели LAPACK для плотных матриц быстрее, чем MKL в 3-6 раз
o Решатели для плотных матриц включают в себя Cholesky, LU, SVD и QR
· Прямые решатели для разреженные матриц в 2-14 раз быстрее, чем их CPU-эквиваленты
o Решатели для разреженных матриц включают в себя прямые решатели и собственные решатели
Рантайм компиляция позволяет генерировать оптимизированные ядра во время запуска приложения.
· Повышение производительности путем удаления условной логики и оценки только особых случаев при необходимости.
Catalyst 15.5 was released for Linux:
http://support.amd.com/en-us/kb-articles/Pages/AMDCatalyst15-5LINReleaseNotes.aspx
Windows still not available.
Catalyst 15.5 was released for Linux:Sadly Cat 15.5 took over all the problems that were reported two month ago with 15.3 beta (on openSuSE 13.2, which is not supported by this driver(?)).
http://support.amd.com/en-us/kb-articles/Pages/AMDCatalyst15-5LINReleaseNotes.aspx
Windows still not available.
Catalyst 15.7 for all platforms (Windows and Linux) was released yesterday.
Release notes Windows (http://support.amd.com/en-us/kb-articles/Pages/AMDCatalyst15-7WINReleaseNotes.aspx)
Release notes Linux (http://support.amd.com/en-us/kb-articles/Pages/AMDCatalyst15-7LINReleaseNotes.aspx)