AMD/ATI 5470 failing on HPPdv7: UNSTABLE_MACHINE

[J_M_Ward]

Hi,

I have an HP Pavilion dv7 4141sa notebook PC with two graphics units: one is an AMD/ATI Mobility Radeon HD4200, which I understand does not support the F@H software, and the other is an AMD/ATI Mobility Radeon HD5470, which I believe should support it. I have had to update from the HP-supplied drivers to the latest AMD drivers that the machine will run, in order to have the necessary OpenCL files installed to be able to run the F@H software. The driver package is 13-1-legacy_vista_win7_win8_64_dd_ccc.exe.

The problem is that the GPU will only perform up to about 15% of its work unit before the process fails with the log report "Nonzero force sum on GPU... ...Folding@home Core Shutdown: UNSTABLE_MACHINE.

Is there anything I can do about this, or should I just remove the slot, leaving the CPU slot working? It seems counterproductive to acquire work units if they can't be finished.

An example logfile is shown below.

Code: Select all

*------------------------------*
Folding@Home GPU Core
Version 2.11 (Thu Dec 9 15:00:14 PST 2010)

Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 15.00.30729.01 for 80x86 
Build host: user-f6d030f24f
Board Type: AMD/OpenCL
Core      : x=16
 Window's signal control handler registered.
Preparing to commence simulation
- Looking at optimizations...
- Files status OK
sizeof(CORE_PACKET_HDR) = 512 file=<>
- Expanded 45182 -> 171163 (decompressed 378.8 percent)
Called DecompressByteArray: compressed_data_size=45182 data_size=171163, decompressed_data_size=171163 diff=0
- Digital signature verified

Project: 11293 (Run 20, Clone 39, Gen 42)

Assembly optimizations on if available.
Entering M.D.
Will resume from checkpoint file 00/wudata_01.ckp
Tpr hash 00/wudata_01.tpr:  1434198816 540116126 2021843840 3991596145 2502114219
Working on ALZHEIMER DISEASE AMYLOID
Client config unavailable.
Starting GUI Server
Resuming from checkpoint
fcCheckPointResume: retreived and current tpr file hash:
   0   1434198816   1434198816
   1    540116126    540116126
   2   2021843840   2021843840
   3   3991596145   3991596145
   4   2502114219   2502114219
fcCheckPointResume: file hashes same.
fcCheckPointResume: state restored.
fcCheckPointResume: name 00/wudata_01.log Verified 00/wudata_01.log
fcCheckPointResume: name 00/wudata_01.trr Verified 00/wudata_01.trr
fcCheckPointResume: name 00/wudata_01.xtc Verified 00/wudata_01.xtc
fcCheckPointResume: name 00/wudata_01.edr Verified 00/wudata_01.edr
fcCheckPointResume: state restored 2
Resumed from checkpoint
Setting checkpoint frequency: 500000
Completed   2000001 out of 50000000 steps (4%).
Completed   2500000 out of 50000000 steps (5%).
Completed   3000000 out of 50000000 steps (6%).
Completed   3500000 out of 50000000 steps (7%).
Completed   4000000 out of 50000000 steps (8%).
Completed   4500000 out of 50000000 steps (9%).
Completed   5000000 out of 50000000 steps (10%).
Completed   5500000 out of 50000000 steps (11%).
Completed   6000000 out of 50000000 steps (12%).
Completed   6500000 out of 50000000 steps (13%).
Completed   7000000 out of 50000000 steps (14%).
mdrun_gpu returned 54
Nonzero force sum on GPU

Folding@home Core Shutdown: UNSTABLE_MACHINE

[P5-133XL]

Personally, I would just delete the slot and not GPU fold. A 5470 is a very low-end GPU that will not produce much. The last known good ATI/AMD driver was 12.8 so you are actually only running a small % of the GPU's capability and it is inconvenient to revert to an older driver requiring a driver cleaner to make it work right. Currently for conventional GPU WU's the GPU core will also require a full CPU core decreasing the productivity of your CPU. Many notebooks tend to have temperature issues if you run both the CPU and GPU together.

i.e. lots of negatives and not a lot of gain. That being said, if you want, we will still try to diagnose the problem so you can GPU fold for as long as you can return the WU's on time then it is still good for the project.

The first thing I would do is start monitoring temp in the systray for both your CPU and your GPU top make sure nothing is overheating.

[J_M_Ward]

Thank you for this. If I revert to the last driver produced by HP, the GPU is unable to fold because OpenCL.dll is missing. Also, as you suggest, the CPU (but not the GPU) runs hot (92 degrees C) unless I turn down the folding power. I don't think that's really too much of a problem, since TjMax is 115 degrees, but, if effort to solve this problem is not worth the output of work units, I will take your advice and remove the slot, continuing on CPU only.

Thank you for your help.

Regards,

J M Ward