Forum > GPU crunching
Optimized CUDAs on Windows Seven
MarkJ:
--- Quote from: Josef W. Segur on 27 Sep 2009, 01:21:33 pm ---Your "Total GPU memory 335544320 free GPU memory 268029952" says that Win7 is using 64 MiB even without Aero, which seems high to me but I'm a minimalist type :P I suppose it indicates you're running at 1600 x 1200 resolution or similar.
Joe
--- End quote ---
We see this on dual GPU setups (GTX295's and dual card setups). It appears that the 1st card allocates 64Mb for something, I assume for communication. It does this even under XP. It seems to be a consistent 64Mb and the 2nd card doesn't appear to lose 64Mb.
We had those hassles in BOINC 6.6.20 (ish) days. Dr A decided they must be different because they report different amounts of memory. BOINC has since been changed to allow +/- 30% variance.
liath555:
I'm also having problems with the vlarkill files for win7. I have aero and whatnot turned off, and I think I still have above the minimum graphics memory, but I've gotten constant errors. here's the relevant part of my error text.
<core_client_version>6.10.18</core_client_version>
<![CDATA[
<message>
- exit code -1073741819 (0xc0000005)
</message>
<stderr_txt>
setiathome_CUDA: Found 1 CUDA device(s):
Device 1 : GeForce 8500 GT
totalGlobalMem = 268435456
sharedMemPerBlock = 16384
regsPerBlock = 8192
warpSize = 32
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 918000
totalConstMem = 65536
major = 1
minor = 1
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 2
setiathome_CUDA: CUDA Device 1 specified, checking...
Device 1: GeForce 8500 GT is okay
SETI@home using CUDA accelerated device GeForce 8500 GT
V12 modification by Raistmer
Priority of worker thread rised successfully
Priority of process adjusted successfully
Total GPU memory 268435456 free GPU memory 234024960
setiathome_enhanced 6.02 Visual Studio/Microsoft C++
Build features: Non-graphics CUDA VLAR autokill enabled FFTW USE_SSE x86
CPUID: Intel(R) Core(TM)2 Quad CPU @ 2.40GHz
Cache: L1=64K L2=4096K
CPU features: FPU TSC PAE CMPXCHG8B APIC SYSENTER MTRR CMOV/CCMP MMX FXSAVE/FXRSTOR SSE SSE2 HT SSE3
libboinc: 6.3.22
Work Unit Info:
...............
WU true angle range is : 1.486312
After app init: total GPU memory 268435456 free GPU memory 123924480
Cuda error 'cudaMalloc((void**) &dev_PowerSpectrumSumMax' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcceleration.cu' in line 383 : out of memory.
setiathome_CUDA: CUDA runtime ERROR in device memory allocation (Step 1 of 3). Falling back to HOST CPU processing...
I am running windows 7 ultimate x64. does that version of vlarkill require more video ram than the x86 version? because on xp-32 which I was on a few days ago everything worked just peachy
Raistmer:
Not CUDA MB, but Windows 7 requires much more GPU memory than XP.
Look at this:
total GPU memory 268435456 free GPU memory 123924480
Your GPU has less than 50% memory free only...
Josef W. Segur:
--- Quote from: Raistmer on 10 Dec 2009, 02:01:06 pm ---Not CUDA MB, but Windows 7 requires much more GPU memory than XP.
Look at this:
total GPU memory 268435456 free GPU memory 123924480
Your GPU has less than 50% memory free only...
--- End quote ---
Hmm, Total GPU memory 268435456 free GPU memory 234024960
After app init: total GPU memory 268435456 free GPU memory 123924480
The difference is 105 MiB and used to be 89 MiB, are the new drivers loading more? Perhaps the new larger cufft.dll actually puts that much more kernel stuff on the GPU, or something similar.
From earlier observations it seemed that less than 128 MiB (134217728 bytes) free after initialization was the problem area.
Joe
Raistmer:
What new DLL you refer to? 2.3 or 3.0 ?
From my host:
Total GPU memory 536870912 free GPU memory 492482560
After app init: total GPU memory 536870912 free GPU memory 492482560
:o :o :o No difference at all.
http://setiathome.berkeley.edu/result.php?resultid=1438996810
CUDA 2.3 + Vista + 190.38 driver
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version