Seti@Home optimized science apps and information
Optimized Seti@Home apps => Windows => Topic started by: corsair on 17 Sep 2009, 02:44:41 pm
-
Hi group,
just as the subject, any available?? and better if would be for win x64.
Cheers all, Corsair ;)
-
There is something brewing backstage...
-
slurp :)
hope would be from lunatics, as usual ;)
I can't wait for it, if you need a beta tester please tell me.
-
slurp :)
hope would be from lunatics, as usual ;)
I can't wait for it, if you need a beta tester please tell me.
What GPU do you have?
-
ATI Raeon HD3850, but giving me some problems, in the next few day I'll upgrade to ASUS HD4770 :),
just awating from shop to be delivered :'(.
and by the way:
Phenom x4 9500
win xp pro x64 English (MUI Spaniard)
2 HD of 500Gb in RAID 1.
8 GB DDR2 RAM
-
Actually I more interesting in lesser card now than 4770.
There will be beta available probably, few day more maybe. Alpha almost finished for now (at least for current build).
-
Actually I more interesting in lesser card now than 4770.
There will be beta available probably, few day more maybe. Alpha almost finished for now (at least for current build).
will it work HD3850 or HD4770??
I looking forward to test it. :)
In the laptop I've Radeon X200M, but no thermal sensor in the GPU, I don't wan to burn my laptop by now,
altough I'm using in my computers TThrottle (marvellous for controlling CPU/GPU temp).
-
For now all beginning from HD2400 will go.
-
There is something brewing backstage...
I did hear there was an Astropulse app in the making. I would have thought it may have been easier to do MB seeing as it already has a cuda app.
Is it Astropulse or Multibeam?
It would be interesting to compare run times for ATI versus nVidia cards. By the way I have a HD4850 in one machine currently being built at the computer shop. Normally I go with nVidia but thought i'd try an ATI card now that apps are starting to come out.
-
I did hear there was an Astropulse app in the making. I would have thought it may have been easier to do MB seeing as it already has a cuda app.
Answered on SETI main forums already.
-
It would be interesting to compare run times for ATI versus nVidia cards.
CUDA MB has much better use of GPU resources for now.
ATI doesn'r offer same lavel SDK and library support as nVidia does so don't expect too much from ATI GPU for now. Raw computing power sometimes means much less than good framework...
-
It would be interesting to compare run times for ATI versus nVidia cards.
CUDA MB has much better use of GPU resources for now.
ATI doesn'r offer same lavel SDK and library support as nVidia does so don't expect too much from ATI GPU for now. Raw computing power sometimes means much less than good framework...
New ATI cards released today. The HD5850 has 1440 shaders. The HD5870 has 1600 shaders. The only thing holding them back is the lack of software support.
I would expect Nvidia will have to reply with something fairly soon otherwise they won't have much market share left.
-
Got my last i7 today. Its got a HD4850 with nothing to do :(
I will give it a whirl on MW in a bit (when it finishes off its cpu MW tasks). If you need a tester of the CAL/Brook+ app let me know.
Cheers,
MarkJ
-
Thanks all for offers. Let's await beta phase.
-
Any news on ATI Seti beta? :P
-
beta opened already:
http://lunatics.kwsn.net/18-astropulse-testing/astropulse-hybrid-cpu-ati-gpu-build-for-gpus-with-double-precision-support.msg22330.html;topicseen#msg22330
-
can't wait for single precision
-
can't wait for single precision
I wanna few success validations at least.
Situation complicated by no work from Berkeley servers for some alpha testers hosts.
It's impossible to test app online if there is no work for it ;)
-
i have a 2600 xt waiting on the sideline to run a cal app.
-
i have a 2600 xt waiting on the sideline to run a cal app.
Raistmer Meet zpm from Seti Beta.
-
tomorrow I think there will be app for 2600 too :)
@zpm
one more post and you will eligible for beta area ;)
-
Any idea what kind of performance we can expect from a 3850?
-
Any idea what kind of performance we can expect from a 3850?
Approx the same as for any other ATI GPU, capable to run brook.
~25% increase over last opt AP release [measured on Q9450/HD4870 combo. For slower CPU you will see bigger speedup and reverse]
Only part of algorithm ported to GPU (hence "hybrid" build) and ported part is heavely memory-bound so sometimes 2600 performs better than my HD4870.
Exact speedup should reveal beta testing.
-
boom!...
one more post, oh wait this is another post.. lol! in.
-
The topic or board you are looking for appears to be either missing or off limits to you.
I cannot enter.... :(
-
If you need, I have an Ati 4850 under linux for the beta testing... but I'm not able to enter beta section.
-
No linux port exist for now, maybe in future...
-
Version for single / multi-GPU, w/o double precision support posted in beta area
http://lunatics.kwsn.net/18-astropulse-testing/astropulse-hybrid-cpu-ati-gpu-build-for-gpus-w-o-double-precision-support.0.html
-
Only part of algorithm ported to GPU (hence "hybrid" build) and ported part is heavely memory-bound so sometimes 2600 performs better than my HD4870.
Exact speedup should reveal beta testing.
System memory bound or video memory bound?
-
GPU part GPU memory bound and CPU part - system memory (especially on P4 where L2 not enough to hold many of used arrays...)
AP likes big L2 cache, all builds including pure CPU one.
-
How large does it want? I assume that a large L3 cache will help a little also.
-
I just registered to help you testing with my HD 2600 mobility. It's running Collatz fine in about two hours if not interrupted. How can I get to the beta area to download the app?
-
I just registered to help you testing with my HD 2600 mobility. It's running Collatz fine in about two hours if not interrupted. How can I get to the beta area to download the app?
5 posts on Lunatics forums AFAIK :)
-
I just registered to help you testing with my HD 2600 mobility. It's running Collatz fine in about two hours if not interrupted. How can I get to the beta area to download the app?
5 posts on Lunatics forums AFAIK :)
Ah, so that's the magic ;D. Well, you can ignore the PM i just sent...
... now to spam the forum ... ;D
BR,
Vid
-
So, three to go now. Thnaks for the info.
-
Is there support for X2/Crossfire models?
-
Multi-GPU support included in r278 (untested though).
Crossfire just as SLI for nVidia cards should be disabled.
-
But in the my other SLI setup, I have SLI enabled (PhysX disabled) and crunch 1 DU per GPU just fine. Using 6.6.38 on latest official WHQL Nvidia drivers.
Oh, and Crossfire on X2 models is enabled by hardware (no option on drivers settings to do so)
-
But in the my other SLI setup, I have SLI enabled (PhysX disabled) and crunch 1 DU per GPU just fine. Using 6.6.38 on latest official WHQL Nvidia drivers.
Did you do speed comparison with SLI disabled /SLI enabled ?
Oh, and Crossfire on X2 models is enabled by hardware (no option on drivers settings to do so)
Well, you will have chance to check all assumptions about Crossfire by himself :)
-
I have dual HD4870X2 in quad crossfire mode, for gaming purposes at the moment. Would love to put it to some good use when I'm not gaming, don't mind if I have to disable crossfire, will test on and off. Beta, here I come...
-
I have dual HD4870X2 in quad crossfire mode, for gaming purposes at the moment. Would love to put it to some good use when I'm not gaming, don't mind if I have to disable crossfire, will test on and off. Beta, here I come...
Should be hugely interesting to see their throughput :-)
What does BOINC state for their FLOPS/S rating? 1.2 TFLOP SP each?
-
Sorry peepz - need 5 posts to try out the beta out ;-)
Will go back to 9.2 for a while.
-
If wanna try on multy-GPU host, please, download version marked as r278 even if it says for GPUs w/o double support.
-
Is there any way I could get a hold of this beta without a need to spam? Via mail maybe? :-\
PLS!!!
-
The semi-closed beta approach needed for selection of more or less responsible and motivated testers.
Currently there are more than 10 app downloads and NO FEEDBACK at all. Only questions how to install it and so on.
If download made for consumer purposes - it can wait a little until release time. Now I need feedback about usage experience.
-
Is there any way I could get a hold of this beta without a need to spam? Via mail maybe? :-\
PLS!!!
The semi-closed beta approach needed for selection of more or less responsible and motivated testers.
Currently there are more than 10 app downloads and NO FEEDBACK at all. Only questions how to install it and so on.
If download made for consumer purposes - it can wait a little until release time. Now I need feedback about usage experience.
Triumph, Thank you for your interest in helping. :)
Rasitmer's response is "spot on" and the Beta approach is necessary for the exact reason he states.
I appreciate your question, rather than spamming to get the requisite post count for access.
In all fairness, I've preferred members to act w/ integrity and make meaningful posts.
Unfortunately, there have been some less than substantive posts.
There are plenty of opportunities for meaningful participation within the community @ Lunatics.
Answering install questions and assisting new members is a good way to start.
Discussing S@H science and related is always welcome in the the Discussion Forum.
Going forward, posts made w/ no purpose other than to raise a members status will be immediately deleted.
-
AFAIK we're in the test phase of AP mixed CPU/GPU application, really stable and from my side as
I'm posting in the thread to the effect, it is going fast.
but inside this thread I've a question, there is developed and I've used (when I had installed a NVIDIA GPU)
really very happy with the performance of: CudaMB V12, I had to remove it because not crunching MW
(CUDA ver 1.1), but there is not same for ATI cards, will it be possible for the near future a version for
SH MB for ATI??
I could miss something because I haven't read all the posts in this forum, but can see anything about.
-
It seems it's better to wait until OpenCL port from CUDA for SETI MB.
-
It seems it's better to wait until OpenCL port from CUDA for SETI MB.
Is there a beta on-going testing for OpenCL on SETI MB?
-
Not yet as OpenCL is still not mature enough.
-
Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
ATI Radeon HD 4350
512 MB
110.0 MHz
Vist 64 bit
would this rig be usful for the testing?
-
yes :)
-
I just registered to help you testing with my HD 2600 mobility. It's running Collatz fine in about two hours if not interrupted. How can I get to the beta area to download the app?
Thank you Cristoph, This was the question i was looking for...
-
Currently there are more than 10 app downloads and NO FEEDBACK at all. Only questions how to install it and so on.
Any feedback so far? How fast is it compared to average CPU? How smooth is working on computer when ATI GPU client is running? Running Collatz my system gets kinda "glitchy"
-
cerrently beta suspended until databes on SETI beta site will be repaired.
Problem solved, beta continues.
-
Just in time for me to try it out finally!
I expect to be able to give some feedback on beta in a short while 8)
-
Not yet as OpenCL is still not mature enough.
Will the switch to OpenCL help with the 9.2/large array problem? What are the long-term plans regarding the large array problem?
-
Not yet as OpenCL is still not mature enough.
Will the switch to OpenCL help with the 9.2/large array problem? What are the long-term plans regarding the large array problem?
OpenCL probably will be free of this limitation. In any case it should be asked on ATI site, not here.
-
Another stupid quest, but given i'm not that familiar with CAL/Brook+ or cuda...
Isn't it possible to get an existing fft library (source) and compile that using the cal/brook+ compiler to get yourself a fft library that will work with ati. Assuming you could get it compiled of course. This would then allow MB and AP port to use ati card.
-
Another stupid quest, but given i'm not that familiar with CAL/Brook+ or cuda...
Isn't it possible to get an existing fft library (source) and compile that using the cal/brook+ compiler to get yourself a fft library that will work with ati. Assuming you could get it compiled of course. This would then allow MB and AP port to use ati card.
No, it's not possible.
Brook+ in not C++. It even not C. It's some small subset of C with GPU-specific extensions. Can you compile pascal or basic sources with C++ Intel compiler for example ? Same case here.
-
Yeah, the best case you could hope for is to develop an FFT algorithm making heavy use of the GPU-accelerated linear algebra routines present in the current Brook/CAL libraries.
But the benefits would be small.
-
Raistmer:
thanks for yor positive reply. I shall try to comply with the rest of the requirement,
conrad
-
Did anyone try running it on a 5870 yet? I'd like to know, could/should/will it run on 5870, win xp 64, catalyst 9.10 (driver ver. 8.66)?
BR,
Vid
-
It should run on this, why not.
But Catalyst 9.2 required, not 9.10. I have big doubts that ATI fixed address translation for 9.10 catalyst...
(worth to try though)
-
It should run on this, why not.
But Catalyst 9.2 required, not 9.10. I have big doubts that ATI fixed address translation for 9.10 catalyst...
(worth to try though)
I don't think 9.2 will recognize the 5870. However I will give it a try as it is.
Thanks.
BR,
Vid
-
It should run on this, why not.
But Catalyst 9.2 required, not 9.10. I have big doubts that ATI fixed address translation for 9.10 catalyst...
(worth to try though)
I don't think 9.2 will recognize the 5870. However I will give it a try as it is.
Thanks.
BR,
Vid
Let us know whether it ran successfully on the 9.10 CAL. I'm eager to try the ATI app but i really hate to downgrade drivers...
-
when run a hyperthreaded processor will the application use one core or one thread?
-
when run a hyperthreaded processor will the application use one core or one thread?
Only multithreaded app will use more than one thread, no matter is it HT CPU or multicore CPU. Currently AP has single worker thread.
It's possible to use Brook CPU backend to run it multithreaded in FFA part, but then there will be no GPU acceleration.
-
Interesting stuff guys. Have a couple of ati cards here. Just getting into the world of BOINC and SETI, I'm more familiar with folding. Anyways let me get my post count up and I'll get back to you on the beta :)
-
Is there any prefered driver? My ATI Radeon HD4350 is useing driver version 8.590.0.0 ssupplied by ATI.
-
Currently compatibility with all Catalyst 9.x line inverstigated. It produces valid work for 9.2, will see results of validation for 9.10 soon.
-
Currently compatibility with all Catalyst 9.x line inverstigated. It produces valid work for 9.2, will see results of validation for 9.10 soon.
Excellent stuff. Should hopefully get a chance to test it out this evening. ;D
-
Currently compatibility with all Catalyst 9.x line inverstigated. It produces valid work for 9.2, will see results of validation for 9.10 soon.
Hope results of this test will be positive! ;)
-
Hi,
any news about the beta client/tests?
-
Raistmer:
I have been running the 2.83 brook app some time now combined with Catalyst 9.10 drivers and the new brook .dll's.
Most of the results validate fine, some are inconclusive.
Only problem so far is that sometimes an error happens: ati2mtag "CPLIB :: OPM - Failed the HFS" ( from the event log )
The 3850 is not able to recover from this error and I have to reboot the computer.
One or two tasks makes no difference, card is running at standard freqs.
Arnulf
PS - should I upgrade to BOINC 6.10.17 or is there any problems with that version?
-
I ran one with no problem a couple of days ago.
-
Raistmer:
I have been running the 2.83 brook app some time now combined with Catalyst 9.10 drivers and the new brook .dll's.
Most of the results validate fine, some are inconclusive.
Only problem so far is that sometimes an error happens: ati2mtag "CPLIB :: OPM - Failed the HFS" ( from the event log )
The 3850 is not able to recover from this error and I have to reboot the computer.
One or two tasks makes no difference, card is running at standard freqs.
Arnulf
PS - should I upgrade to BOINC 6.10.17 or is there any problems with that version?
discussion in alpha thread.
-
the ATI Hybrid app is very interesting so far. I should have the new SSE version running on my system shortly.
-
Did i miss something is there now a SETI MB app for ATI?
-
Did i miss something is there now a SETI MB app for ATI?
No, its SETI@home AP app for ATI.
-
Damn any word on one coming out for MB
-
did you read the second post. Its still very alpha. Meaning it could be weeks-months or even never if they can't get it running correctly. I do have a great deal of confidence in the Lunatics Crew. they seem to live for the coding
-
did you read the second post. Its still very alpha. Meaning it could be weeks-months or even never if they can't get it running correctly. I do have a great deal of confidence in the Lunatics Crew. they seem to live for the coding
Second post was long time ago.
There is no MB ATI app, not in alpha beta or gamma, but there IS ATI AP in beta state, actually, in RC state IMO.
-
Ok thank you. :(
-
Now running an app, after rev.420, all WU's validated and took ~9hours, still better then 80 on CPU ;D.
Now running the ATI13ATI, 0.8CPU+ 1(ATI)GPU, low load, though.
Here are some results.
(http://setiweb.ssl.berkeley.edu/beta/results.php?userid=13581&offset=0&show_names=0&state=3)
-
change your link to the link to host.
Your link is personal and can't be loaded by anyone except you.
BTW, for what reason you stop using rev420?
It's recommended version for all GPUs that OpenCL capable.
hybrid AP now positioned only for older GPU and embedded chipsets.
-
This should get you Fred's Computer.
http://setiweb.ssl.berkeley.edu/beta/hosts_user.php?userid=13581
-
Well, he running "stock" hybrid AP instead of new OpenCL one.
-
Well, he's now running an app_info, and has some more new tasks, but he's not reported any completed tasks,
so what app is he running now?
Claggy