Today, I installed the BOINC 7.0.28 beta, which can install normally under Win8, and use GPU acceleration. However, it is only using one of my graphics cards. I have plenty of tasks waiting, but BOINC does not seem te realise it can run two threads at a time. Is there a known solution to this problem or is it a new issue with Windows 8?

Post your Boinc 7.0.28 startup messages from the Event log please.

Claggy

02/06/2012 01:14:33 | | Starting BOINC client version 7.0.28 for windows_x86_64
02/06/2012 01:14:33 | | log flags: file_xfer, sched_ops, task
02/06/2012 01:14:33 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
02/06/2012 01:14:33 | | Data directory: C:\ProgramData\BOINC
02/06/2012 01:14:33 | | Running under account Yannick
02/06/2012 01:14:33 | | Processor: 4 AuthenticAMD AMD Phenom(tm) II X4 965 Processor [Family 16 Model 4 Stepping 3]
02/06/2012 01:14:33 | | Processor: 512.00 KB cache
02/06/2012 01:14:33 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 htt pni cx16 syscall nx lm svm sse4a osvw ibs skinit wdt page1gb rdtscp 3dnowext 3dnow
02/06/2012 01:14:33 | | OS: Microsoft Windows 8: Developer Preview x64 Edition, (06.02.8250.00)
02/06/2012 01:14:33 | | Memory: 7.12 GB physical, 7.12 GB virtual
02/06/2012 01:14:33 | | Disk: 111.45 GB total, 16.60 GB free
02/06/2012 01:14:33 | | Local time is UTC +2 hours
02/06/2012 01:14:33 | | NVIDIA GPU 0: GeForce GTX 460 (driver version 296.17, CUDA version 4.20, compute capability 2.1, 46080MB, 8377165MB available, 1048 GFLOPS peak)
02/06/2012 01:14:33 | | NVIDIA GPU 1: GeForce GTX 460 (driver version 296.17, CUDA version 4.20, compute capability 2.1, 46080MB, 845MB available, 1048 GFLOPS peak)
02/06/2012 01:14:33 | | OpenCL: NVIDIA GPU 0: GeForce GTX 460 (driver version 296.17, device version OpenCL 1.1 CUDA, 1024MB, 8377165MB available)
02/06/2012 01:14:33 | | OpenCL: NVIDIA GPU 1: GeForce GTX 460 (driver version 296.17, device version OpenCL 1.1 CUDA, 1024MB, 845MB available)
02/06/2012 01:14:33 | | Config: use all coprocessors
02/06/2012 01:14:33 | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 6687462; resource share 100
02/06/2012 01:14:33 | SETI@home | General prefs: from SETI@home (last modified 23-Mar-2012 22:14:49)
02/06/2012 01:14:33 | SETI@home | Host location: none
02/06/2012 01:14:33 | SETI@home | General prefs: using your defaults
02/06/2012 01:14:33 | | Reading preferences override file
02/06/2012 01:14:33 | | Preferences:
02/06/2012 01:14:33 | | max memory usage when active: 3647.59MB
02/06/2012 01:14:33 | | max memory usage when idle: 6565.65MB
02/06/2012 01:14:33 | | max disk usage: 16.65GB
02/06/2012 01:14:33 | | suspend work if non-BOINC CPU load exceeds 25 %
02/06/2012 01:14:33 | | (to change preferences, visit the web site of an attached project, or select Preferences in the Manager)
02/06/2012 01:14:33 | | Not using a proxy
02/06/2012 01:14:33 | SETI@home | Restarting task 31dc09ag.5879.14791.13.10.40_0 using setiathome_enhanced version 610 (cuda_fermi) in slot 0
02/06/2012 01:14:33 | SETI@home | Starting task 31jl10ad.31893.10296.9.10.61_0 using setiathome_enhanced version 610 (cuda_fermi) in slot 1


I did some more research and fixed the issue by saving a cc_config.xml file in C:/ProgramData/BOINC

Now both GPU's are being utilized and I can hear the coil whine again :D

Still, seems like an issue that this isn't enabled automatically.

Still, seems like an issue that this isn't enabled automatically.

BOINC will automatically use the best GPU, or more than one if there are the same GPUs in the system. In your case they may look the same, however, they aren't, as the GPU 0 is rated best due to that available memory fluke you have: 8377165MB of memory versus only 845MB available.

02/06/2012 01:14:33 | | NVIDIA GPU 0: GeForce GTX 460 (driver version 296.17, CUDA version 4.20, compute capability 2.1, 46080MB, 8377165MB available, 1048 GFLOPS peak)
02/06/2012 01:14:33 | | NVIDIA GPU 1: GeForce GTX 460 (driver version 296.17, CUDA version 4.20, compute capability 2.1, 46080MB, 845MB available, 1048 GFLOPS peak)

So some detection/driver adjusting in Windows 8 still needs work, it seems.

---

That memory fluke would be due to enabling SLI without restarting. This bug has afaik been present in Nvidia drivers for at least a year now. I had turned it off and back on again to see if that would fix the issue. The memory fluke happened chronologically after this bug occurred, and therefore cannot be the cause.
Usually both cards are reported as having about 800-900 out of the total 1024MB memory available, and I don't think BOINC should not utilise a card because it has 3MB less memory available.

By the way, would disabling Aero in Windows, which does free up some memory, speed up computation? Right now usage is at 754MB on both cards, so memory doesn't seem to be a limitation.

02/06/2012 01:14:33 | | NVIDIA GPU 0: GeForce GTX 460 (driver version 296.17, CUDA version 4.20, compute capability 2.1, 46080MB, 8377165MB available, 1048 GFLOPS peak)
02/06/2012 01:14:33 | | NVIDIA GPU 1: GeForce GTX 460 (driver version 296.17, CUDA version 4.20, compute capability 2.1, 46080MB, 845MB available, 1048 GFLOPS peak)
02/06/2012 01:14:33 | | OpenCL: NVIDIA GPU 0: GeForce GTX 460 (driver version 296.17, device version OpenCL 1.1 CUDA, 1024MB, 8377165MB available)
02/06/2012 01:14:33 | | OpenCL: NVIDIA GPU 1: GeForce GTX 460 (driver version 296.17, device version OpenCL 1.1 CUDA, 1024MB, 845MB available)

Please read the Sticky thread in Number Crunching at Seti, I didn't get the thread put there for everyone to ignore it:

<core_client_version>7.0.28</core_client_version>
<![CDATA[
<stderr_txt>
setiathome_CUDA: No CUDA devices found
setiathome_CUDA: Found 0 CUDA device(s):
setiathome_CUDA: CUDA Device 2 specified, checking...
Device cannot be used
SETI@home NOT using CUDA, falling back on host CPU processing
setiathome_enhanced 6.09 Visual Studio/Microsoft C++
libboinc: 6.3.22

Work Unit Info:
...............
WU true angle range is : 0.444742
Optimal function choices:
-----------------------------------------------------
name
-----------------------------------------------------
v_BaseLineSmooth (no other)
v_GetPowerSpectrum 0.00016 0.00000
v_ChirpData 0.01078 0.00000
v_Transpose4 0.00540 0.00000
FPU opt folding 0.00434 0.00000
setiathome_CUDA: Found 2 CUDA device(s):
Device 1 : GeForce GTX 460
totalGlobalMem = 1073741824
sharedMemPerBlock = 49152
regsPerBlock = 32768
warpSize = 32
memPitch = 2147483647
maxThreadsPerBlock = 1024
clockRate = 1800000
totalConstMem = 65536
major = 2
minor = 1
textureAlignment = 512
deviceOverlap = 1
multiProcessorCount = 7
Device 2 : GeForce GTX 460
totalGlobalMem = 1073741824
sharedMemPerBlock = 49152
regsPerBlock = 32768
warpSize = 32
memPitch = 2147483647
maxThreadsPerBlock = 1024
clockRate = 1800000
totalConstMem = 65536
major = 2
minor = 1
textureAlignment = 512
deviceOverlap = 1
multiProcessorCount = 7
setiathome_CUDA: CUDA Device 1 specified, checking...
Device 1: GeForce GTX 460 is okay
SETI@home using CUDA accelerated device GeForce GTX 460
Restarted at 4.82 percent.

Flopcounter: 41963995546487.562000

Spike count: 3
Pulse count: 1
Triplet count: 0
Gaussian count: 0
called boinc_finish

</stderr_txt>

Claggy

Yea I'm running into that issue too...guess I'll just disable sleep for now.