Forum > GPU crunching
V10/11 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use
Raistmer:
--- Quote from: MAOJC on 04 Mar 2009, 06:54:25 pm ---Did you ever build on for older SSE2 AMD CPUs + a GPU?
--- End quote ---
What config do you have? For what CPU you need SSE2 app ?
Raistmer:
--- Quote from: yodi on 04 Mar 2009, 08:35:25 pm ---
--- Quote from: yodi on 03 Mar 2009, 07:03:54 pm ---If you're interested I can test the stock apps if it's working in general or not. ;) Already installed the stock apps on one system today to see if it's running without a remote desktop connection and would test it with remote desktop tomorrow.
--- End quote ---
jFYI, the stock apps are working fine logged in via remote desktop. :-\
http://setiathome.berkeley.edu/result.php?resultid=1177433675
--- End quote ---
The single difference between my build and stock one that can influence of CUDA app startup is CUDA version used at compile time. I use CUDA 2.1, stock version - probably 2.0.
So, different driver versions are required to run. Check if you have 180.20 and up installed. There should be no difference in remote desktop related behavior.
Try to describe your test case in details so I could reproduce it on my hosts.
Raistmer:
--- Quote from: elec999 on 04 Mar 2009, 10:07:59 pm ---My GTX 260 is been sitting there for three days now. I complety removed Boinc and reinstalled it.
Thank you
--- End quote ---
1) Install BOINC 6.4.5 as proven to work. I will not even try to explore situations with higher BOINC versions - they are KNOWN to have bugs (maybe 6.5.0 is OK, I don't know, I personally know that 6.4.5 is OK for V10 configs).
2) Install V10a pack. Check that number_of_gpus and cc_config.xml are edited to reflect your number of CUDA-enabled GPUs and number of cores + number of GPUs respectively.
3) Start BOINC.
If it will not work, then probably screenshots will be required.
HSchmirPo:
My nvidia GTX 295 seems to run only on 1 core. Sometimes even this core stops. :(
System: Q9450 @ 3,4 (water), geforce GTX 295.
HSchmirPo
yodi:
--- Quote from: Raistmer on 05 Mar 2009, 12:43:08 am ---
--- Quote from: yodi on 04 Mar 2009, 08:35:25 pm ---
--- Quote from: yodi on 03 Mar 2009, 07:03:54 pm ---If you're interested I can test the stock apps if it's working in general or not. ;) Already installed the stock apps on one system today to see if it's running without a remote desktop connection and would test it with remote desktop tomorrow.
--- End quote ---
jFYI, the stock apps are working fine logged in via remote desktop. :-\
http://setiathome.berkeley.edu/result.php?resultid=1177433675
--- End quote ---
The single difference between my build and stock one that can influence of CUDA app startup is CUDA version used at compile time. I use CUDA 2.1, stock version - probably 2.0.
So, different driver versions are required to run. Check if you have 180.20 and up installed. There should be no difference in remote desktop related behavior.
Try to describe your test case in details so I could reproduce it on my hosts.
--- End quote ---
Hello Raistmer,
Please excuse the wrong information posted from my side. Guess I've checked some wrong workunits yesterday. They were ok, but must be processed without a remote desktop session. I used the PC today via remote desktop only, checked again some workunits and they were listed with the outcome "Success". Having a closer look on those tasks, all were showing the same error like optimized apps - "initialization error". The differnce is that stock apps are falling back from GPU to CPU processing. So they will be computed with the CPU and a result send to the server. As it's does not make sense/is not possible for optimized apps to fall back from GPU to CPU (already used) the workunit will not be continued and reported with a "Client error"...
Stock app result:
---
setiathome_CUDA: Found 1 CUDA device(s):
Device 1 : Device Emulation (CPU)
totalGlobalMem = -1
sharedMemPerBlock = 16384
regsPerBlock = 8192
warpSize = 1
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 1350000
totalConstMem = 65536
major = 9999
minor = 9999
textureAlignment = 256
deviceOverlap = 0
multiProcessorCount = 16
setiathome_CUDA: device 1 is emulation device and should not be used, supports 9999.9999
setiathome_CUDA: CUDA Device 1 specified, checking...
Device 1: Device Emulation (CPU) is okay
SETI@home using CUDA accelerated device Device Emulation (CPU)
Cuda error 'cufftPlan1d(&fft_analysis_plans[FftNum], FftLen, CUFFT_C2C, NumDataPoints / FftLen)' in file 'c:/sw/gpgpu/seti/seti_boinc/client/cuda/cudaAcc_fft.cu' in line 49 : initialization error.
Cuda error 'cufftPlan1d(&fft_analysis_plans[FftNum], FftLen, CUFFT_C2C, NumDataPoints / FftLen)' in file 'c:/sw/gpgpu/seti/seti_boinc/client/cuda/cudaAcc_fft.cu' in line 49 : initialization error.
setiathome_CUDA: CUDA runtime ERROR in plan FFT. Falling back to HOST CPU processing...
setiathome_enhanced 6.03 Visual Studio/Microsoft C++
libboinc: 6.3.22
---
So nothing to waste additional time on from your side. :D
I am sorry for any circumstances and thanks a lot for your great work! :)
Cheers,
yodi - with already installed TightVNC - that might work fine
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version