the hunt for instability

[velusip]

Running into a lot of errors since upgrading from a low watt AM2 to a Phenom core. Also changed memory from Mushkin to G.Skill. Lots of testing in store to find the problem, but here it goes:

[17:07:11] Completed 60000 out of 500000 steps  (12 percent)
[17:09:43] Quit 101 - NaN detected: (ener[20])
[17:09:43]
[17:09:43] Simulation instability has been encountered. The run has entered a
[17:09:43]   state from which no further progress can be made.
[17:09:43] This may be the correct result of the simulation, however if you
[17:09:43]   often see other project units terminating early like this
[17:09:43]   too, you may wish to check the stability of your computer (issues
[17:09:43]   such as high temperature, overclocking, etc.).
[17:09:43] Going to send back what have done.
[17:09:43] logfile size: 8783
[17:09:43] - Writing 9333 bytes of core data to disk...
[17:09:43]   ... Done.
[17:09:43] - Failed to delete work/wudata_01.arc
[17:09:43] No C.P. to delete.
[17:09:43] - Failed to delete work/wudata_01.sas
[17:09:43] - Failed to delete work/wudata_01.pdo
[17:09:43] Warning:  check for stray files
[17:11:43]
[17:11:43] Folding@home Core Shutdown: EARLY_UNIT_END
[17:11:43]
[17:11:43] Folding@home Core Shutdown: EARLY_UNIT_END
[17:11:46] CoreStatus = 7B (123)
[17:11:46] Client-core communications error: ERROR 0x7b
[17:11:46] This is a sign of more serious problems, shutting down.

Dialog:
"Folding@home has run into a serious error running the core. and will shutdown."

I just came home to this after my first trial of the non-gpu smp client. So many things have changed recently it will be tricky to track down. It's really discouraging because up until my last computer upgrade folding at home ran flawlessly on multiple computers for the past several years non stop.

 

[tkam]

The windows smp client is pretty notorious for being unstable. You could try installing vmware and setup a ubuntu instance to run the linux smp client on. The linux smp client is very stable and will get you a higher ppd to boot.

 

[velusip]

Just came back from a memtest run and just stopped the test at 18% as over 1100 errors found all over the place. Set timings under SPD settings and came back... popped some nachos in the oven and looking at memory prices. Going back to Mushkin for sure.

 

[Xilikon]

Here you go SMP client is very sensitive to overclocking and hardware instabilities so it works as expected in your case. I'm sure that replacing the memory will fix the problem for you.

 

[nomad8u]

Interesting. My last 6 builds have all been with GSkill and I've had nothing but good luck with them. I have heard that some boards don't like them though. Or maybe just a bad stick.

Either way, you can't go wrong with Mushkin. Excellent memory.

 

[velusip]

My previous build was entirely passive cooling which worked exceptionally well with a low wattage AM2 and huge syncs on the gpu and cpu. However with today's high watt everything I dropped two 120mm fans in the side. The gates on the side of the case are creating noise so I will cut those off shortly.
(you can see them clearly here:

Anyway, at full load all around with 4 cpu folding clients running and the gpu client the cpu temp is 52 degrees and the gpu core is bouncing between 64 and 65 degrees. I can't run the smp client anymore due to some non-sensical IO errors that I really can't be bothered to investigate they are so cryptic. I'm guessing it has to do with permissions on named pipes.