F@H GPU problem

[Thump553]

I've been running F@H on a modest desktop for a year or two, and run the GPU version as well when the machine is inactive. Starting last night the GPU verision would download and install a new core five times, install it successfully then say it was outdated. the videocard is a stock GEForce GT240. Here's the log:

*------------------------------*
Folding@Home GPU Core
Version 2.22 (Wed Feb 16 09:48:20 PST 2011)
Build host amoeba
Board Type NVIDIA/CUDA
Core 15

Window's signal control handler registered.
Preparing to commence simulation
- Looking at optimizations...
DeleteFrameFiles: successfully deleted file=work/wudata_06.ckp
- Created dyn
- Files status OK
sizeof(CORE_PACKET_HDR) = 512 file=<>
Need version 225
Error: Work unit read from disk is invalid

Folding@home Core Shutdown: CORE_OUTDATED


----------------

Any ideas? Desktop is a Win7 64 system, Compaq Presario CQ5320F, Athlon II x2 240 Dual Core, nothing overclocked, with an upgraded PSU and the GEForce GT 240 added.

 

[Rattledagger]

Any ideas? Desktop is a Win7 64 system, Compaq Presario CQ5320F, Athlon II x2 240 Dual Core, nothing overclocked, with an upgraded PSU and the GEForce GT 240 added.

Where's nothing you can do, except pause the slot for now until FAH has fixed their server-problems.

A possible fix was made on Friday and after some internal tests started using it today, but the fix didn't work afterall so they'll need to try another code-change.

 

[petrusbroder]

Yeah, that is a problem - 4 of my GPUs have nothing to do ...
Patience is the only remedy ...

 

[Thump553]

Thanks for the info. The CORE OUTDATED message made me inclined to think somehow my setup got corrupted.

 

[DooKey]

My understanding is Stanford is having server probs with the gpu wu server. If you have another machine running the 2.25 core you can copy it over to the machine having problems and then you'll get more wu.

 

[Rattledagger]

My understanding is Stanford is having server probs with the gpu wu server. If you have another machine running the 2.25 core you can copy it over to the machine having problems and then you'll get more wu.

This will maybe work in some instances, but FAH is having many problems this weekend with GPU's:
1: Kepler-cards getting Fermi-wu's, this was the oldest reported problem.
2: Nvidia-cards getting stuck in "outdated core"-loop.
3: Amd-cards getting NVIDIA-wu's.

#2 is AFAIK now fixed, but, this popped-up next error:

4: Some cards now gets GPU_MEMTEST_ERROR instead.

It's not expected #3 will be fixed before Monday, and #4 will probably demand a new core-version so chances are will take longer to fix.

Oh, and even after #3 is fixed, FAH is total crap with the current Amd-drivers & beta-drivers, performing much worse than even the mediocre performance Amd-cards always have had under FAH.

 

[Thump553]

My (nvidia) GPU began working again last night, but it seems to be processing work units a whole lot slower than before.