Hi, I'm running this rig It has 2 GTX295s that are oc'd to 665/1100 (from i believe 575/1000).

The thing is the reported clockrate value on card 2 and 4 are reduced over time. After a reboot they report correctly again. GPU-z, Riva tuner and Nv Control panel always report the correct (665) value.

The theory I'm going with now is that this is some sort of thermal throttle kicking in, though I have never seen a card go above 85C.

Example from a task stderr:

setiathome_CUDA: Found 4 CUDA device(s):
Device 1 : GeForce GTX 295
totalGlobalMem = 939261952
sharedMemPerBlock = 16384
regsPerBlock = 16384
warpSize = 32
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 1440000
totalConstMem = 65536
major = 1
minor = 3
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 30
Device 2 : GeForce GTX 295
totalGlobalMem = 939196416
sharedMemPerBlock = 16384
regsPerBlock = 16384
warpSize = 32
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 601714
totalConstMem = 65536
major = 1
minor = 3
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 30
Device 3 : GeForce GTX 295
totalGlobalMem = 939261952
sharedMemPerBlock = 16384
regsPerBlock = 16384
warpSize = 32
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 1440000
totalConstMem = 65536
major = 1
minor = 3
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 30
Device 4 : GeForce GTX 295
totalGlobalMem = 939261952
sharedMemPerBlock = 16384
regsPerBlock = 16384
warpSize = 32
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 799200
totalConstMem = 65536
major = 1
minor = 3
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 30
setiathome_CUDA: CUDA Device 4 specified, checking...
Device 4: GeForce GTX 295 is okay

They run Raistmers V12 app.

Should I trust the Boinc report or GPUz/Riva tuner/Nvida Control Panel? Is this a problem at all?

/Anton

---

This is really strange..

Maybe post this at the opt. crew site: [http://lunatics.kwsn.net]

IIRC, after two weeks or something no reboot and my 4 OCed GTX260-216 show all the time the correct shader speed in the stderr_txt output.

If you look to Vypers 4 GTX295 cruncher, IIRC the output show all the time all GPUs at same speed.

An example: [http://setiathome.berkeley.edu/result.php?resultid=1344064209]
O.K., I don't know how often he reboot this PC..

---

update,
It seems to be the thermal throttle after all. after reviewing the logs from gpuz I can see that the cards do throttle every once in a while, only in short spurs lasting perhaps 10-15 seconds.

So my guess is boinc picks up on that and reports it, eventhough the downclock isn't permanent.

/Anton