Is it possible to do something to prevent this? Do you mind adding VLAR-killing ability to this mod?
Finally cleared the last of my VLARs at Beta, installed this mod - and guess what, I got a 20-full of VLAR at Main!It's really possible. Especially on quad!
Like http://setiathome.berkeley.edu/result.php?resultid=1134023275 - AR=0.009331
This rig will do that sort of AR quicker on the CPU than the GPU, so I'd be interested in option (1) - hand over to CPU at VLAR - rather than autokill.
Thanks, will be waiting for these modifications :)No sense to wait - use current version right now instead. Even with slow VLAR processing the use GPU is free addon to CPU now (of course if you don't participate in other GPU-related projects.)
Algorithm:
1) Suspend all tasks.
2) resume non-VLAR task - it will be assigned to CUDA app. (I suppose SETI is single active project - you could enhance algorithm for more complex cases)
3) resume 4 (for quad) VLAR tasks - they will be assigned to CPU. Resume some non-VLARS for GPU. (GPU will crunch much faster so while CPU finish VLAR, GPU will finish few non-VLARs).
One possible drawback of your mod, and the way it bypasses BOINC's knowledge of CUDA: it doesn't do task-switching very well - I've already found it trying to run 5 x CPU and 0 x GPU a couple of times. Easy to manage with task suspension, but it still means babysitting.
AMD SSE3 (X2 Athlon ) working on this build ?No, will do SSE3 soon.
@ Raistmer
Your hard work is much appreciated - thanks :bow:
regards
Kim Vater, Team Norway.
So it really means that only 3 CPU are used even if I put 6 cores (4 CPU +1 GPU +1) in the website prefs...
Thanks for your answer and for the clue,
Would yit be possible to have a link to such a template, and also in which folder to put it ?
<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>
As stated on that page, it goes at the root level of your BOINC data folder.
Once in place, you can load and activate it from the advanced menu in BOINC Manager, 'Read config file'. When you change the number of CPUs this way, expect BOINC to re-run benchmarks.
As expected it only uses 1 GPU on multi GPU setup...
So far so good though...
SETI@home error -12 Unknown errorIn this result: http://setiathome.berkeley.edu/result.php?resultid=1138810950
cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel
File: d:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu
Line: 236
Yes.As expected it only uses 1 GPU on multi GPU setup...
So far so good though...
Anyone got any ideas on how to get 2 GPU going? It pains me to leave one idle...
I'm getting lots of computation errors. Cuda app running fine + 4x AK V8 running fine, but fifth AK V8 will start and stop at 8-9 seconds with computation error. Fifth AK V8 starts again and repeats. Computation errors stack up quick! I'm doing SETI and SETI Beta, but only installed package to setiathome.berkeley.edu folder not setiathome.berkeley.edu_beta folder.
Specs:
i7 920
GTX 260 216SP w/178.xx
Hi there.
Some errors from me:QuoteSETI@home error -12 Unknown errorIn this result: http://setiathome.berkeley.edu/result.php?resultid=1138810950
cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel
File: d:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu
Line: 236
Computer specs here: http://setiathome.berkeley.edu/show_host_detail.php?hostid=3503812
No overclock in CPU or videocard, video driver is 181.22 WHQL.
I tried to install your version 8aon my quad core, but all the units that went to the gpu all were getting computation errors, so I removed the opt app to just let the defalut process for now.You didn't install it correctly, reinstall. And try to describe sequence of operations you did first time.
Here are two of the errored out tasks:
http://setiathome.berkeley.edu/result.php?resultid=1140067400
http://setiathome.berkeley.edu/result.php?resultid=1140067408
http://setiathome.berkeley.edu/result.php?resultid=1140502904I'm getting lots of computation errors. Cuda app running fine + 4x AK V8 running fine, but fifth AK V8 will start and stop at 8-9 seconds with computation error. Fifth AK V8 starts again and repeats. Computation errors stack up quick! I'm doing SETI and SETI Beta, but only installed package to setiathome.berkeley.edu folder not setiathome.berkeley.edu_beta folder.
Specs:
i7 920
GTX 260 216SP w/178.xx
Could you provide link on that host, please?
5th AK v8 (first one, actually) shouldn't take 8-9 seconds. It should take 0 seconds to pass control to CUDA MB....
Looks all done right.
But, it seems you use service BOINc install (net stop boinc - it's for stopping service). and pre-V6 BOINC, V6 BOINC use BOINCdata folder instead BOINC folder. IS it true?
http://setiathome.berkeley.edu/result.php?resultid=1140502904I'm getting lots of computation errors. Cuda app running fine + 4x AK V8 running fine, but fifth AK V8 will start and stop at 8-9 seconds with computation error. Fifth AK V8 starts again and repeats. Computation errors stack up quick! I'm doing SETI and SETI Beta, but only installed package to setiathome.berkeley.edu folder not setiathome.berkeley.edu_beta folder.
Specs:
i7 920
GTX 260 216SP w/178.xx
Could you provide link on that host, please?
5th AK v8 (first one, actually) shouldn't take 8-9 seconds. It should take 0 seconds to pass control to CUDA MB....
protected execution=service execution (almost).
But under Vista service BOINC can't get access to CUDA-capable GPU. It seems you should try to reinstall BOINC first.
Yes, this is known inherently error of CUDA MB (but was unknown if this bug still with us or not).
Probably you can do nothing to avoid it, just kepp crunching - I hope this error prety rare. Thanks for report (I passed bug report to developer).
Stock app should experience the same difficulties with server installation. So if stock works the reason in something else.
What driver do you use? 180.x required. I looked on yor link to result - "invalid device symbol" there.
This one: http://setiathome.berkeley.edu/result.php?resultid=1140067400
Hmmm, but GPU WU's aren't the problem, it's the AK V8 WU's. :-\http://setiathome.berkeley.edu/result.php?resultid=1140502904I'm getting lots of computation errors. Cuda app running fine + 4x AK V8 running fine, but fifth AK V8 will start and stop at 8-9 seconds with computation error. Fifth AK V8 starts again and repeats. Computation errors stack up quick! I'm doing SETI and SETI Beta, but only installed package to setiathome.berkeley.edu folder not setiathome.berkeley.edu_beta folder.
Specs:
i7 920
GTX 260 216SP w/178.xx
Could you provide link on that host, please?
5th AK v8 (first one, actually) shouldn't take 8-9 seconds. It should take 0 seconds to pass control to CUDA MB....
Ok, 180.xx drivers are required.
Look at link you provided. It points on CUDA processed task.Hmmm, but GPU WU's aren't the problem, it's the AK V8 WU's. :-\http://setiathome.berkeley.edu/result.php?resultid=1140502904I'm getting lots of computation errors. Cuda app running fine + 4x AK V8 running fine, but fifth AK V8 will start and stop at 8-9 seconds with computation error. Fifth AK V8 starts again and repeats. Computation errors stack up quick! I'm doing SETI and SETI Beta, but only installed package to setiathome.berkeley.edu folder not setiathome.berkeley.edu_beta folder.
Specs:
i7 920
GTX 260 216SP w/178.xx
Could you provide link on that host, please?
5th AK v8 (first one, actually) shouldn't take 8-9 seconds. It should take 0 seconds to pass control to CUDA MB....
Ok, 180.xx drivers are required.
Working very Great on my Amd X2 5000+ with an old 8800 Gs Card!
But I cant use my SLI 9800Gtx+ on my Q6600. :(
Working very Great on my Amd X2 5000+ with an old 8800 Gs Card!
But I cant use my SLI 9800Gtx+ on my Q6600. :(
SLI is not supported by CUDA. It will see cards in SLI mode as one device (if will work at all).
Yes, it shouldn't, read again first post.
For dual GPU configs V7x of combo is more suitable, V8x for single GPU only.
Yeah, unfortunately both approaches have their own flaws. Hope sooner or later BOINC will do shceduling work by itself as it should be.Yes, it shouldn't, read again first post.
For dual GPU configs V7x of combo is more suitable, V8x for single GPU only.
Yup I read it, and i used it! But I cant get any more AP Wu since the last 3-4days, So i switched to V8x for that reason ;D
Look at link you provided. It points on CUDA processed task.I noticed that, but I didn't think anything of it because it was the AK V8 processes failing; however, I updated to 180.xx and I now see how everything works. MB works but CUDA doesn't show up in Boinc Manager. This is normal, correct? Can I suspend CUDA without suspending anything else for when I play games, or do I have to cancel the whole project?
ADDON: form stderr of your result:
Cuda error 'cudaMemcpyToSymbol(cudaAcc_GaussFit_settings, (void*) &settings, sizeof(settings))' in file 'd:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_gaussfit.cu' in line 454 : invalid device symbol.
The answer is the same - update to 18x.xx drivers.
...
AP work is now ceased cause new AP version in testing...
It's offtopic to current thread but too curious not to ask :)...
AP work is now ceased cause new AP version in testing...
The ap_splitters at main finished splitting all the data the project had on hand, a hoped-for shipment from Arecibo didn't arrive Friday, and the HPSS which has the reserve stock of unsplit data was being upgraded so Matt couldn't get more data. If they've also decided to not split AP work when there's some available, that's good too.
Joe
I noticed that, but I didn't think anything of it because it was the AK V8 processes failing; however, I updated to 180.xx and I now see how everything works. MB works but CUDA doesn't show up in Boinc Manager. This is normal, correct? Can I suspend CUDA without suspending anything else for when I play games, or do I have to cancel the whole project?
1. When MB_6.08_mod_CPU_team_CUDA.exe starts uses 100% of one core for about 1 minute.
3. I was confused when I saw 3 threads , 2 astropulse and one seti MB running on my 2 cores (AMD x2 5400+) and CUDA dissappeared from my task manager. the 2 AP were using 25% and seti_MB_cpu 50%.
I really would like to see which of three SETI_MB_wu's is using my GPU. Before it was written like this: (CUDA) in boinc manager.
"
How it works:
for BOINC it looks as usual CPU opt app installed. BOINC will call CPU app (AK_v8b_win_SSSE3x_GPU_CPU_team.exe in our case) and assign one of SETI MB tasks for it. But this app aware of possibility to use GPU for computations. It will check if another instance (it knows only itself and its clones, so - no other GPU-related projects please) already use GPU and if not - will start GPU-related app (MB_6.08_mod_CPU_team_CUDA.exe in our case) and suspend itself until GPU app finish. This CUDA app will do all work as usually but will do it on GPU leave CPU almost free.
"
So on single core you should see one working AK_v8, one suspended AK_v8 (0 CPU time and 0 % CPU usage) and one CUDA app.
1. When MB_6.08_mod_CPU_team_CUDA.exe starts uses 100% of one core for about 1 minute.1. It's OK, but not for min but ~half min (maybe for your system 60 secons is OK too).
2. I was surprised when I saw BOINC was crunching astropulse units I had stored in my pc even if I had installed correctly your mod. It used optimised ap_5.00r103_SSE3.exe so it is good..
3. I was confused when I saw 3 threads , 2 astropulse and one seti MB running on my 2 cores (AMD x2 5400+) and CUDA dissappeared from my task manager. the 2 AP were using 25% and seti_MB_cpu 50%.
I really would like to see which of three SETI_MB_wu's is using my GPU. Before it was written like this: (CUDA) in boinc manager.
Hi,Try to use some beta driver.... If you can;t upgrade to 18x.x it seems this modded app just not for your hardware/software still ( kick nVidia for new driver versions they should do that )
use notebook with 9700m gt but i can`t find nvidia driver 180.xx.
I find only 179.xx for Notebook...
I don't know anything about downloads ending in .rar. For a Windows PC what do I do with it to extract and run the optimized apps? Or, is it for Linux only?
Thanks
Art
I don't know anything about downloads ending in .rar. For a Windows PC what do I do with it to extract and run the optimized apps? Or, is it for Linux only?www.rarlab.com
Thanks
Art
It's offtopic to current thread but too curious not to ask :)
Is data set that possible to process with AP so smaller than MB one ?? We do MB many years already - do that huge data array not suitable for AP ?
3. how long did you see that? first 60 seconds (it's OK and no problem) or full task duration (it's bad and needs investigation)
I don't know anything about downloads ending in .rar. For a Windows PC what do I do with it to extract and run the optimized apps? Or, is it for Linux only?www.rarlab.com
Thanks
Art
To learn something new is always good ;)
And thanks for the replies on what to do with *.rar files......Everyone has to have that first experience with anything new. Now I know.Sure, no prob :)
Maybe I get it, as I get used to how things are done here. The file named *_V8a* is for AMD processors, the file named *_V8* is for Intel?No-no-no, for AMD will have_AMD in name. V8 and V8a differs in will it kill VLAR or will it process VLAR. Due current Berkeley's network troubles I would recommend V8.
V8 it is, thank you. Especially since I don't know what VLAR is. :-)
Hi Raistmer,
Would you mind updating your mod to switch between CPU/GPU for VLAR units (as you mentioned on the first page - if task is VLAR - stay with CPU, else - try pass task to GPU)? VLAR Autokill works perfectly, at last I have stable SETI setup using GPU and not requiring constant attention, but sometimes it is very pity to see that 20 WUs downloaded between constant "out of work" messages go to trash bin...
Thanks in advance :)
I understand your concern, but GPU-crunching of VLAR workunit requires at least the same time as CPU-crunching. So there will be almost no gain from allowing GPU-crunching of VLARs.Well, you recive one additional core instead 2 or even 3 - it worse of course but not bad too ;)
Also when I was using stock apps or V8 I was very worried by PC overload during VLAR crunching by GPU - PC starts working very-very slow and unstable, it is almost impossible to use it for any other purposes during that time...I see, but because experiment with tuning AK_v8 on different ARs failed (there was no speed increase if I did PGO only on VLAR dyn data set instead of complete set) adding AR recognition now on low priority really.
really I think that CPU efficace in such conditions also decrease so we loose any gain from GPU involvement.Not so sure. Don't forget, sluggish user interface - one thing and CPU app performance - completely another...
Of course, the best way is to postpone VLAR workunit if GPU is idle and go to next available workunit, but as I understand this is BOINC task so optimized apps can't achieve this without client support.Yes, interaction with BOINC API needed...
As you can see SSE2 build performs better than SSE3 one on SSE3-capable (early) AMD. Maybe on latest Phenom SSE3-situation was improved ?
Could someone of our pre-testers or regular users try to run KWSN bench for AK_v8b_SSE3 and AK_v8b_SSE2 on new Phenom CPU to shed light on current situation with AMD SSE3 support quality ?
Restarted at 100.00 percent.:o
The computer that ran on doesn't have a CUDA card in it. I don't know what happened to that wu, why it ended so soon.
Hm... interesting what will be with team app in that case...
Because bot tested apps is not AMD specific ones, better to re-test their analogues w/o team mod. That is, "standart" AK_v8b_SSE3/2 versions. Theys should give same speeds and much better support standalone runs than "teamed" ones.
Hm... interesting what will be with team app in that case...
Because bot tested apps is not AMD specific ones, better to re-test their analogues w/o team mod. That is, "standart" AK_v8b_SSE3/2 versions. Theys should give same speeds and much better support standalone runs than "teamed" ones.
I thought both of those apps were from your AMD packages.
The second test I posted is with the only other v8 apps i could find. I hadn't seen ak_v8b without gpu support. I don't have direct access to either of the phenom machines on my account, so i don't know when i'll be able to retest.
SSE2 one was in AMD-specific package indeed, now that version obsolete. Try to download new one (first page). There should be _AMD in app name.
AK_v8 available in "Download" section of this site.
Think you can just install new AMD-specific package. Included CPU-based app will slightly (~1-2% no more) faster than usual SSE2 AK_v8 on AMD chips, it tested already on Phenoms too (and on prev AMD SSE3-capable CPUs).SSE2 one was in AMD-specific package indeed, now that version obsolete. Try to download new one (first page). There should be _AMD in app name.
AK_v8 available in "Download" section of this site.
Ok, I hadn't seen the new version. What app should i run it against?
The results I labeled as nonCUDA were from the download section.
AMD SSE3 (X2 Athlon ) working on this build ?No, will do SSE3 soon.
BTW, does anyone need SSE2 and low ?
That is, does anybody have SSE2-only CPU with CUDA-enabled GPU ?
Unless you have Phenom-II no more testing required on this stage IMHO.
No, I will assemble working package and put on first page soon.AMD SSE3 (X2 Athlon ) working on this build ?No, will do SSE3 soon.
BTW, does anyone need SSE2 and low ?
That is, does anybody have SSE2-only CPU with CUDA-enabled GPU ?
Yep have old Opteron (SSE2) OC'd with a 9600 GTX under XP
So the AMD file on the first page will now work?
No, I will assemble working package and put on first page soon.AMD SSE3 (X2 Athlon ) working on this build ?No, will do SSE3 soon.
BTW, does anyone need SSE2 and low ?
That is, does anybody have SSE2-only CPU with CUDA-enabled GPU ?
Waiting with baited breath. ;)
Yep have old Opteron (SSE2) OC'd with a 9600 GTX under XP
So the AMD file on the first page will now work?
Currently only SSE3 and up supported. (provided AP is SSE3 one)
Just noticed one strange thing.
If BOINC (I use 6.4.5) downloads new workunits with shorter deadline it stops crunching previous WUs and switches to new ones. But GPU-crunching doesn't start - all workunits are processed only by CPU cores. As a result GPU stays idle till all short-deadline workunits are finished and computation of previous ones resumes.
I assume that it happens because when GPU crunching starts GPU-thread became fixed to parcitular workunit and nothing can be changed before this workunit is finished (completed or aborted). I don't know if such behavior is special for your 8a mod or it also affects stock app, but it will be really great of you defeat it :)
Pure example of CPU scheduling bug in Vista IMHO.
Pure example of CPU scheduling bug in Vista IMHO.
Nope, It's not a bug, it's on purpose. If it fills up the first processors first, then the other three can power down and save energy. You might want to double check that power saving is disabled in the Bios, and do exactly what you did, assign the single threaded program to the last core. You shouldn't need to place affinity controls on the ciuda feeder then, but it would be interesting to know if it then migrates to the last core.
XP Rules! ;)
Power saving disabled, moreover, you forgot that all CPUs still busy with another 4 CPU-based tasks (2 Ak_v8 2 einstein on moment of observation). So there is absolutely no power saving could be done there.LoL fair enough , you didn't say you were running other apps, and the scheduling algorithm will still attempt to oversubscribe the first core before moving on, So it is still the Windows strategy at work, despite all cores running.
And change affinity only for non-BOINC app is not enough. After I did that situation remains the same. And only when I exclude that core for CUDA app it begin work as usual.
I repeat this experiment few times so pretty sure in that.
Did you see the trac tickets #841 (http://boinc.berkeley.edu/trac/ticket/841) and #842 (http://boinc.berkeley.edu/trac/ticket/842) that Jord made us write the other day?
[Hmm... Might be nice if Boinc adjusted this automatically on the fly ....]
...In the future, we hope to see additional interfaces in operating systems to coordinate threaded applications including those built with TBB. We agree with those who have called for OSes to get out of the business of scheduling threads and focus instead on allocation of processors to applications. It’s an interesting topic to say the least.
yes....
But I thought in such "overloaded" condition where many procasses in ready state OS should just use round-robin shceduling (for example) on each priority level and give full quantum for each app. And it will do that except it can't realize that if one core busy with higher priority process on one core it could still execute low priority ones on another cores... It fails to do that.
Priorities were: non-BOINC thread - normal (16 ?), 2 AK_v8 and 2 einstein - 4 worker threads each with priority of 1. CUDA worker thread - priority of 3.
So, CUDA should preempt all other BOINC threads (and it do this usually) but can't fight with non-BOINC thread unless explicit affinity is setted...
My point is (why I call it "bug") that if I manually constrain resources available for CUDA app (exclude some cores) it works better (with the same load on system, I don't create free time slices that way, right?). What prevent OS to do the same on its own level?...
Will see what it does on dual core & 9600GSO OC's 20%, as I've run out of AstroPulse ::)
Well, I do production run with CUDA "team" combo for few days already (that's why so many updates last 2 days - I start to notice flaws and ways they could be eliminated ;D ) Will see how fast my RAC will climb back and higher.
LoL ... You need less cores, and to run AstroPulse ;D Let me see if I can remember my photobucket account.Well, think not. You could try to use this build as "standalone" CUDA app, w/o AK_v8 modified part. So you will do AP on CPU and MB on GPU...
[Now for this new app, I have my AstroPulse section, and Enhanced with this new exe with the ciuda * fftw DLL's right , Anything else needed ? ]
Oh I see, But there is no AP left, How I make it run AK & Cuda then?Download V8a combo (full package) from first page of this thread. Then update CUDA app to one of latest flavors :)
Will Do,
I think your quad has some way to rise yet. Here's the E8400 @ 3.6GHz + 9600GSO@20%OC running v7VLARKillMod on GPU & AP on CPU , 90% on main shown, 10% beta. and in 45 Degrees C weather ;D (Dip is where the AP validators were off those few days).
(http://i192.photobucket.com/albums/z94/jason_gee/E8400AP9600GSOMB.jpg) (http://s192.photobucket.com/albums/z94/jason_gee/?action=view¤t=E8400AP9600GSOMB-1.jpg)
I notice some "wrong" behaviour: if a task goes to urgent state and the client starts it "pushing out" the current CUDA task, the new task don't transit to CUDA but stays on the CPU, so there are 3 tasks on my 2-core CPU and the GPU is idle until one of the tasks is finished.
Agreed :)
Well, attached app little more "smart" in this sense and will leave worker thread on single (first) core. It could give some performance degradation indeed in general case (it can't fill idle window of another core) but it really helped in my situation. And achieved GPU temp is highest (the same was when no non-BOINC app runs).
I have this app running on Q6600+GPU. GUI performance on screen is very slow and jerky. It seems the app is not slowinng/stopping when I wish to use the gui. When I kill BOINC the user performance is restored.
I have this app running on Q6600+GPU. GUI performance on screen is very slow and jerky. It seems the app is not slowinng/stopping when I wish to use the gui. When I kill BOINC the user performance is restored.
What OS and gui theme are you using? Vista's Aero theme lagged with CUDA work for me, I had to drop back to Vista Basic.
Maybe you have cc_config in the wrong folder? It resides in BOINC, two levels up from the work folder.
I don't know then, I have a GeForce 8600GTS with 256MB ram and the only problems I have now are during task switches.
I think I'm seeing some of the symptoms you're describing Raistmer. I tried your moded app in your previous post, and I thought that the problem was solved as I don't think I made it past 24 hours without my GPU going idle. I applied the app last night and this morning my GPU temp was high telling me that it was being used. When I got home today I noticed that the temp was back to normal indicating that nothing was being used.for BOINC only you should not change affinity. Only if some non- BOINC CPU consuming task working....
So after reading your post about Vista (I'm using the 64bit version) which app do I change the affinty for? I'm not sure which one is used soley for the GPU. I have 4 processes running. (I only have a dual core)
AK_v8b_win_SSSE3x_GPU_CPU_team.exe is running at 3 instances 2 of them running around 22% and the 3rd one running at 00 CPU usage. The MB_6.08_mod_CPU_team_CUDA.exe file is running at 50% CPU. Which process would I change as all for of them have both CPU's checked off for the affinity.
Thanks
Vager
For core2 duo you should use SSSE3 version - it will faster. Moreover you probably forgot about cc_config.xml file. Ont core still idle in your configAgreed :)
Well, attached app little more "smart" in this sense and will leave worker thread on single (first) core. It could give some performance degradation indeed in general case (it can't fill idle window of another core) but it really helped in my situation. And achieved GPU temp is highest (the same was when no non-BOINC app runs).
OK, I have read almost all the thread, I have DL the SSE3 package (I have a Core2Duo) and this version of CUDA app. Mt task manager now shows one AK_V8 app at 50%, one AK_V8 at 00% and one MB_9.08 CUDA at 2%.
In Boinc manager I have 2 working units, one normal and another one that doesn't increase the CPU time field (but it is working).
Is this normal? I did make an config file with 3 CPU's and placed it on the work folder.
The CUDA app will run at 50% just for a while at the begining...
PC is very responsive.
Maybe you have cc_config in the wrong folder? It resides in BOINC, two levels up from the work folder.
That was it, silly me :) Thanks a lot, now my PC is slughish too ;D
Wold be nice that those tickets will be solved quickly, especially the one about suspending only the CUDA at will...
So you have responsive system w/o cc_config ? Then remove it. On good GPU it's better to have GPU working even if one core idle than to have all cores loaded but turning off GPU.
So you have responsive system w/o cc_config ? Then remove it. On good GPU it's better to have GPU working even if one core idle than to have all cores loaded but turning off GPU.
I am using the file with SSE3 in description for non-AMD - it's the right one I guess?Hm, as I wrote on your first question it's NOT correct.
I notice some "wrong" behaviour: if a task goes to urgent state and the client starts it "pushing out" the current CUDA task, the new task don't transit to CUDA but stays on the CPU, so there are 3 tasks on my 2-core CPU and the GPU is idle until one of the tasks is finished.
Look 1-2 pages before for update that cures this situation.
Agreed :)
Well, attached app little more "smart" in this sense and will leave worker thread on single (first) core. It could give some performance degradation indeed in general case (it can't fill idle window of another core) but it really helped in my situation. And achieved GPU temp is highest (the same was when no non-BOINC app runs).
It's can't be considered as upgrade cause it will help in some cases (when CUDA app appers freezing w/o reason) and can decrease performance (slightly if will do it) in other cases.
ADDON:
And some suggestion: If you experience some delays (PC behave sluggish) when running CUDA app and browsing Inet, playing game or watching video on multicore system, try exclude first CPU (By setting affinity for process in task manager) for non-BOINC app that experience delays (i.e. browser, game, media player). You could get better experience that way. Don't forget to upgrade to attached build of course in this case.
Raistmer thanks for all you have done i have use your app all the time from the start with gpu and all working fine. Thank's for your great support !!!
PS :i will post the results but i'm waiting the new watercooling kit for my system and push it to te limit now my Q6600 is OC to 3.4Ghz i think i will reach 3.8 with the watercooling will see :)
Best Regard's
D.Draganov
setiathome_CUDA: device 1 is emulation device and should not be used, supports 9999.9999
setiathome_CUDA: No SETI@home capabale CUDA GPU found...
SETI@home NOT using CUDA, falling back on host CPU processing
It was disscussed many times already. Look on elapsed times, not CPU ones....
in stderr of finished result.
On web site go to result info
in stderr of finished result.
On web site go to result info
Would that be this statement? If so I don't understand the stated time.
"Wall-clock time elapsed since last restart: 6902.6 seconds
called boinc_finish"
Also the above statement doesn't seem to exist in all completed WU that have been granted credit. Is this statement a part of your work and applies only to GPU-crunched WU?
Sorry for the bandwidth....just trying to understand what is happening.
The setting of <ncpus>0</ncpus> is NO LONGER the way to set it back to whatever the computer has (and I have this as a place holder on several machines). It has to be <ncpus>-1</ncpus> to work. Perhaps a warning about this should be put in the config documentation page is it is a rather nasty and hard to figure out problem.
That's right; I recently changed things so that <ncpus>0</ncpus> means use zero CPUs (this was in response to a request from someone who wanted to compute only with GPU).
To get the actual #CPUs, use -1 or omit the element.
If this is similar to the stuck tasks I 'used to' get, then there'll be no error messages, etc.. at all. It just sticks at zero progress. Restarting Boinc (service or boinc itself depending on installation type) should get them going normally.
If so, then restarting boinc is the only way I found to make them 'un-stick' and happens to me every couple of days or so. To stop from having to manually handle this, I installed the VBScript from the other thread here, then, as I run as a service, made the modification described later in that thread to restart the service instead of terminating the tasks. I haven't had to do any manual intervention since, but noticed it automatically restarts boinc every couple of days or so. I was not able to pin the 'sticky tasks' to any particular angle range or other factor that I could determine.
Jason
His picture shows that CUDA app still running ~4% CPU is normal ...yes but it is the AK v8 app that is stuck i haven't got any problem with the gpu it is from the boinc or the app i don't know you are the pro wow ::) ;D
1) How you could know GPU is working - you already used right method - looked at GPU temp.Hi,
If temp = idle temp - GPU idle, if temp increased - GPU working hard.
2) If you have no work from SETI then you probably doing another non-CUDA enabled projects. So your CPU will do 5 CPU tasks and no CUDA ones.
Try to get new tasks for SETI MultiBeam.
Link on your host, please ?
This result came from CUDA MB app.Hi,
http://setiathome.berkeley.edu/result.php?resultid=1159848143
BTW, update both CPU and GPU parts of package (links buried iun thread ~2-3 pages back) there is new apps available.
Maybe I do repack soon (don't want to touch it before opt AP new release)
Is the new op-ap ready yet?? :)Just after it passes internal validation tests. Consider it as "ready" ;)
I found the problem , when reinstalled Boinc 6.4.5 it has the option "Run in Protected Mode" and that is imcompatible with CUDA, just reinstall it again and uncheck the protected mode, now it´s crunching again with GPU
because of cuda use 4% ;D but can i do to work all the 5 :-\
just wanted to mention, Windows 7 is less than useful for this trick.
I ran this on Windows 7 64bit. Definitely stupid. The drivers are old, 179.23 (GPUZ ftw). From other threads, I assume 180 or something higher is required for this to function properly. (I figured that out after I realized something was wrong).
That said, I believe the teamwork algorithm was working superbly. Other than the fact that every CUDA WU errored out almost immediately, there were exactly the right number of tasks for my i7 system; 9 total, with 8 maxing the CPU, and one more only stealing part of 1 cpu long enough to do setup or teardown. No idea how much it would use while functioning, because they all died quickly. Also, 2 of the cpu tasks were AP's.
Here's hoping BOINC will fix the scheduling sooner rather than later, or Nvidia will release new drivers for me. Unfortunately, there's no way I'm going back to Vista. Win7 is "Vista, done right", imo.
if/when there are drivers that will let this work, is there any chance of a SSE4.1 build? On that note, I haven't been able to find what the speed difference for SSE4.1 vs SSSE3, SSE3, and so forth are, nor anything regarding if SSE4.2 provides anything useful to the project. If/when I'm able, I'll certainly experiment to see if 8x SSSE3 on i7 + 1x 9400 GT is faster than just 8x SSE4.1.
Great job and thank you!
I'm not having any troubles with vista x64, drivers are 181.21Had a crash on the cuda program x86 last night. Switched back to the V6 packages. One of the main problems with V8 and BOINC 6.45 is it requests less work by the day. It has a buffer of < 4 day with a setting of 10. (this was before there was no work) All task run way below 1 hour but the rest time counter always stay above 1 hour and mostly on 1:20 1:30. Seen a task complete at 20 Minutes and the rest time went ... up instead of down as it normally does.
Switching back to V6 is meaningless. They deprecated already. V8 CUDA app has all improvements from V6 + some more.I'm not having any troubles with vista x64, drivers are 181.21Had a crash on the cuda program x86 last night. Switched back to the V6 packages. One of the main problems with V8 and BOINC 6.45 is it requests less work by the day. It has a buffer of < 4 day with a setting of 10. (this was before there was no work) All task run way below 1 hour but the rest time counter always stay above 1 hour and mostly on 1:20 1:30. Seen a task complete at 20 Minutes and the rest time went ... up instead of down as it normally does.
Why? because I run some other projects as well and e.g. Wold Community Grid is greedy and downloads a lot of work and goes into High p mode. That stops the Cuda task altogether and worse SETI does not download any work, so the buffer gets very low. When you have a new AP ready I switch to that version. The completion time is now about 15 Minutes and on the V8 it goes to 1:30 and more even as the work is done much faster.Switching back to V6 is meaningless. They deprecated already. V8 CUDA app has all improvements from V6 + some more.I'm not having any troubles with vista x64, drivers are 181.21Had a crash on the cuda program x86 last night. Switched back to the V6 packages. One of the main problems with V8 and BOINC 6.45 is it requests less work by the day. It has a buffer of < 4 day with a setting of 10. (this was before there was no work) All task run way below 1 hour but the rest time counter always stay above 1 hour and mostly on 1:20 1:30. Seen a task complete at 20 Minutes and the rest time went ... up instead of down as it normally does.
What crash you had to decide such downgrade ? That is not the only reason, but the CUDA task just stopped.
Try V7, but not V6.Done, but the cuda exe is the same MB_6.08a_mod_VLAR_fix_CUDA.exe. And the AP is no longer working anyway.
Try V7, but not V6.Done, but the cuda exe is the same MB_6.08a_mod_VLAR_fix_CUDA.exe. And the AP is no longer working anyway.
i downloaded and installed raistmer's newest sse3 code and installed it into the seti@berkley folder and also installed boinc's newest 6.4.5 software and still only 4 wu's are going at once :/ gpu isnt doing anything..
need instructions on what to do in simplest terms possible i am by no means some computer genius and cant understand 95% the stuff you guys are talking about..
by the way i have a intel q6600 on windows xp with sp3
Playing games like WOW and doing cuda in the background doesn't go very well. Is there a way to disable/suspend only cuda while playing?No :)
something like a semaphore file?
3) This CPU-GPU team will not play nicely with other GPU-related projects like GPU-grid. It's because no BOINC GPU-management mechanism is used in this version.
This fact has positive side too - you do not need GPU-aware BOINC at all. You just need host with CUDA-supported GPU. It should work even with BOINC 5.xx
4) <ncpus>NUMBER_OF_CORES+1</ncpus> is REQUIRED for productive work. If you let BOINC manage CPU cores number you will end up with one idle core, trust me ;)
I didn't check that but it _should_work on BOINC 5 too.Still no fun on my 5.3.12 client. I've tried with cc_config on default (BOINC root) location and even copied it on project dir. but, as before, it end up with 3 occupied cores + cuda. There aren't any new lines on boinc messages tab
cc_config should be placed in BOINC directory.
I didn't check that but it _should_work on BOINC 5 too.
cc_config should be placed in BOINC directory.
Still no fun on my 5.3.12 client. I've tried with cc_config on default (BOINC root) location and even copied it on project dir. but, as before, it end up with 3 occupied cores + cuda. There aren't any new lines on boinc messages tab
during startup, so it looks like boinc doesn't detect the presence of cc_ file?
My box is std. XP_SP2 x86 config with latest nVidia 181.22@Asus 9600GT. Any thoughts on how to force the last core to run with V8a?
I didn't check that but it _should_work on BOINC 5 too.
cc_config should be placed in BOINC directory.Still no fun on my 5.3.12 client. I've tried with cc_config on default (BOINC root) location and even copied it on project dir. but, as before, it end up with 3 occupied cores + cuda. There aren't any new lines on boinc messages tab
during startup, so it looks like boinc doesn't detect the presence of cc_ file?
My box is std. XP_SP2 x86 config with latest nVidia 181.22@Asus 9600GT. Any thoughts on how to force the last core to run with V8a?
The cc_config.xml didn't exist until May 2006, the <ncpus> option was added for BOINC 5.5.4 on 21 June 2006. I did not check how it was used at that time.
Joe
Thanks Raistmer!!!
Having a small problem...I'm pretty sure I'm using v8b (the one called 'more clever autokill mod'?). Anyway, for the most part, everything runs great. However, when boinc downloads newer wu's with a more immediate date, boinc switches out one of the currently running wu's with the new one in high priority mode (in my case, this seems to almost always be the wu currently in GPU). When this happens, the new wu is not placed in the GPU and I end up having all 3 wu's processed by my 2 CPU cores. :o I can remedy the situation by shutting down boinc and restarting, but this isn't always possible (ie. the middle of the night). Any ideas on a possible fix?
Check what CPU based app you use, check what checkpoint interval you use (CPU app will check if GPU is idle only after checkpoint completion so big checkpoint nterval will delay idle GPU detection).
That mod sounds exactly what I need in this case. Unfortunately, my CPU only supports the SSE3 instruction set and not SSSE3 :( Oh well, guess I'll just have to live with the occasional idle GPU for the time being. Thanks for the prompt responses above, Raistmer ;DDon't worry, soon will be SSE3 support added too, just check this thread time to time.
It's almost as if the GPU is never flagged as free when the WU is swapped out and the team mod is sending any further work to the CPU cores.
Raistmer............Yes. It can't be repaired w/o direct interaction with BOINC. So as I said many times, kick BOINC devs for native CPU + GPU support for same app. They do this for srock apps already, now they should provide the way to do the same with app_info file installed.
Running this "team" version of software. Yesterday found one of my quad computers running 5 Astropulse work units on the cpu's. I paused one of the AstroPulse work units and resumed it later in order to start up a cuda work unit. Has this been reported before?
Playing games like WOW and doing cuda in the background doesn't go very well. Is there a way to disable/suspend only cuda while playing?No :)
something like a semaphore file?
Try to kick BOINC devs some more for it ;D
Sure you can suspend process via ProcessLasso, via ProcessExplorer or by any other means... but it's OS level solution, not BOINC level one.Playing games like WOW and doing cuda in the background doesn't go very well. Is there a way to disable/suspend only cuda while playing?No :)
something like a semaphore file?
Try to kick BOINC devs some more for it ;D
That must be you as i'm running apps from you for over 1 year now ;)
solution: c:\ut\pssuspend.exe MB_6.08_mod_CPU_team_CUDA.exe
Hi all,1) Probably your BOINC installation was unable to load files from berkeley's servers. Network connection overloaded now so corrupted downloads is very likely.
I'm running the AK V8 + CUDA MB team mod since last week, and everything was fine till the weekend. The WU's from my cache worked fine, but after finaly getting new work, I got quite a lot compute errors withs BOINC telling me "Outpute file xy_0 for task xy is absent". On one machine I get these errors for about 1 out of 20 WU's, but the other is much worse (so I switched it to get no new work).
Any suggestions for this problem?
BTW, when I started the mod the first time I observed the following:
The 6.03 WU's from my cache were processed fine, but all new work was 6.08 WU's (prcessed fine as well). As I understand it, the team mode should be able to process both types (as the are not realy different, just another name basically), then why is my BOINC just getting 6.08 WU's and not mixed work, depending on availability? This is not a problem, I'm just curious.
And Raistmer and all involved, keep up the good work. It realy is appreciated.
Andreas
1) Probably your BOINC installation was unable to load files from berkeley's servers. Network connection overloaded now so corrupted downloads is very likely.
2) BOINC will mark all new work with latest app version available to it.
Perhaps nothing :( It's network-related problems and if server drop connection client can't do anything with this...
1) Probably your BOINC installation was unable to load files from berkeley's servers. Network connection overloaded now so corrupted downloads is very likely.
2) BOINC will mark all new work with latest app version available to it.
Anything I can do to avoid download corruption other than wait for better times?
I got quite a lot compute errors withs BOINC telling me "Outpute file xy_0 for task xy is absent". On one machine I get these errors for about 1 out of 20 WU's, but the other is much worse (so I switched it to get no new work).
The 6.03 WU's from my cache were processed fine, but all new work was 6.08 WU's (prcessed fine as well).
Hi Andreas,
Output file is absent due to VLAR-killing, mod works ;) There is no need to NoNewTasks
This is replacement for CUDA part of V8 of my package. This build has VLAR autokill ability.
You need to decompress this file into SETI project directory (when BOINC switched off of course) if you don't want process VLAR on CUDA.
That is, if VLAR task goes to CPU app - it will be processed. If it goes to CUDA app - task will be aborted with error (BAD_HEADER).
EDIT: There was wrong file name (.exe twice). Archive replaced, please, remove secondary .exe by renaming manually or download new archive.
if I get this correct, VLAR-killing should produce a different error (bad-header and not missing output file), so I'll hope Raistmer is right, and the problem is corrupted downloads.
Leopoldo,
if I get this correct, VLAR-killing should produce a different error (bad-header and not missing output file), so I'll hope Raistmer is right, and the problem is corrupted downloads.
VLAR-killed result is indicated as missing output file in BOINC Manager (the idea was to not pollute science database with incorrect results, so result file is deleted by mod).
Truth about result was VLAR-killed or not You can see in your result at SETI@Home Tasks page ("Your results" button in BOINC Manager) - there is leftmost link at line with every WU, which leads to page with result. Raistmer's VLAR-killing mod add info about killng.
How can I recognice a VLAR WU in my BOINC manager? I want to assign them to cpu manually while gpu is bussy.
@echo off
findstr "<true_angle_range>0\.1[0-3] <true_angle_range>0\.0 " "C:\Program Files\BOINC\Data\projects\setiathome.berkeley.edu\*.*" > !VLAR!.txt
start !VLAR!.txt
exit
And what does the "name" of a WU tell me?
AP r103 is included as usually ! IT should work with AP too. But if you will have only AP tasks it will do NUMBER_OF_CORES+1 AP r103 and no CUDA MB. => GPU will be idle.Does this mean I can put AP app in beta folder? I have all files from the package in main folder only. :-\
(AP r103 provided "as is" w/o any modifications, it will use CPU just as will do any another app from another CPU-only project as Einstein for example. I expect some issues only with another GPU-enabled projects. And some inefficiencies because of BOINC inability to pair projects. This VERY needed feature still missing! We need option to pair one project with another....
I will post beta-site variant in pre-release area. This one will not work on beta.
Does this mean I can put AP app in beta folder? I have all files from the package in main folder only. :-\
Heres a silly question (or maybe not?)The engine core freq increase gave smallest speed increase so far. And very poor in OCing.
What part of the nVidia actually does the math. The cores or the Shaders?
I'm wondering if I need to overclock one the other or both. I am trying to get the max with out over heating my nVidia cards.
Any one know for sure?
Thanks.. I'll kick the shaders up a bit and see if that works.Heres a silly question (or maybe not?)The engine core freq increase gave smallest speed increase so far. And very poor in OCing.
What part of the nVidia actually does the math. The cores or the Shaders?
I'm wondering if I need to overclock one the other or both. I am trying to get the max with out over heating my nVidia cards.
Any one know for sure?
Memory give some speedup, need to be measured still (I still check upper memory freq OCing limit). It was reported on beta that shaders freq gives best speedup.
3/22/2009 2:42:47 PM|SETI@home|Message from server: Your app_info.xml file doesn't have a version of Astropulse v5.Upgrade to V10 or take AP v5 section from V10 app_info file.
Any ideas? I can't get the modified AP apps to work. I have four modified AK apps working, but four regular AP apps. Also running the CUDA app for 9 process total (using i7).