FAH "Unstable Machine" EUE limit etc...

[chrisk]

I have an issue with the machine in my sig, on Windows XP. This has been happening the last week or so:
[04:25:13] Completed 25%
[04:25:13] mdrun_gpu returned
[04:25:13] NANs detected on GPU
[04:25:13]
[04:25:13] Folding@home Core Shutdown: UNSTABLE_MACHINE
[04:25:18] CoreStatus = 7A (122)
[04:25:18] Sending work to server
[04:25:18] Project: 4744 (Run 0, Clone 591, Gen 2)
[04:25:18] - Error: Could not get length of results file work/wuresults_00.dat
[04:25:18] - Error: Could not read unit 00 file. Removing from queue.
[04:25:18] EUE limit exceeded. Pausing 24 hours.
[05:25:39] + Working...

I have been getting a similar error every time, and by reading around the various forums, I decided to try the following:
- delete the folding folder in app data to force a download of the latest core
- install new Catalyst drivers (use driverclean to completely get rid of old ones first), currently 8.10s
- uninstall FAH and reinstall.
- reduce cpu OC (run at stock)

I have tried the above steps one at a time, and have also done a combination of these steps (ie. uninstall drivers and fah, deleted the app data folder, run at stock, then reinstall each). I have heard of issues with the new FAH cores and ATI cards in my travels, but the above attempts usually solved this for others.

This setup worked perfectly fine for a few weeks, but now I don't think it has successfully completed any work units at all. The only other issue with this setup I had was the adapter getting reset under Vista, and the "atikmdag" issue, which I am not sure are related. Otherwise since I moved back to XP, the computer is completely stable, and folded fine for at least a couple of weeks.

Any ideas?

[LCB001]

I am not to familiar with ATI stuf as I run nvidia cards, from a quick look at the Folding forums all I can see is:
run the console client, not the systray
don't let it install in the program files [x86] or user data directory's
run as administrator
increase the fan speed on the GPU

Also are you running Boinc as well, if so try stopping Boinc for a bit to see if it helps. You might need to run it at less than 100% cpu since the GPU needs an entire core on XP

[frontier204]

I can't shake my UNSTABLE_MACHINE errors either... I seem to be able to complete 4732 units but nothing else. Computer's in my sig, and the errors I get are posted below. Maybe it's my PSU or motherboard but I don't have the time or money to go swapping components in and out.

I solved the "NANs detected on GPU" problem in Vista by using Administrator mode, but I never saw the error in Windows XP. Ususally it wouldn't let me complete even 1% before I get a NANs error. The Nonzero force error I get seems to be random, sometimes waiting until I'm 89% or so complete before destroying my run.

I tried:
- Catalyst versions 8.7 thru 8.10 on Windows XP 32 and Vista 64
- Disable antivirus and firewall
- Disable Sidebar
- GPU (stock cooler) fan speed 100%
- Underclock to as low as the CCC (500/750) would allow running 100% fan speed
- Closing CCC after applying fan mod
- Using ATI display driver without CCC
- Uninstall FAH and reinstall (tried both console and systray variations, stored in E:\fah)
- Granting my user account "full control" over the folder.
- CPU-starving the client by allowing it to only use 50% of one core

My system is stable for anything other than GPU folding, and when I get UNSTABLE_MACHINE, no other program is affected. For stability, I verified CPU with Prime95, and then 6 runs of IntelBurnTest. I also verified my memory with memtest86, and my GPU by running FurMark (renamed the EXE) overnight on stock fan speeds.

Running either Windows XP MCE 32-bit or Windows Vista HP 64 bit I get:
[13:00:27] Completed 1%
[13:13:17] Completed 2%
[13:28:19] Completed 3%
[13:28:19] mdrun_gpu returned
[13:28:19] Nonzero force sum on GPU
[13:28:19]
[13:28:19] Folding@home Core Shutdown: UNSTABLE_MACHINE
[13:28:27] CoreStatus = 7A (122)

[sswilson]

Mine started getting a lot of these as well. I'm thinking it's probably a bug of some form, but there wasn't a lot of info on the FAH forums when I checked earlier this evening.

[SugarJ]

Originally Posted by frontier204

I can't shake my UNSTABLE_MACHINE errors either... I seem to be able to complete 4732 units but nothing else.

I had the same issue with my 4870, drove me nuts so I stopped folding again. Boinc FTW!

[chrisk]

I have read in a few forums of people having issues with the new core and ATI cards. Pretty annoying. I won't have much time until this weekend to troubleshoot. I am thinking that I might turn off Boinc and see if I can complete a few work units to see if thats an issue (just got my RAC up in Boinc so i don't want to shut down).

I am actually thinking of going nvidia as I have had problems with my gpu since I went from my 8800GT to 4870...just don't have the time or cash this instant, and the 4870 is supposed to be a great card and I don't want to give up yet.

[frontier204]

Update: Last night, I wanted to check if the problem really was my GPU, so I set an overclock that I KNEW WAS UNSTABLE (690/1170). With this unstable OC, I would fail FurMark in about 2 seconds, but I was actually able to COMPLETE a 4742 work unit. I had the failing FurMark open alongside my GPU console client, which was still happily churning away. I EUE'd shortly after on a series of 4744 work units. I also had the CCC window open to the Overdrive menu the whole time, and my CPU folding / BOINC was off at the time.

From that experiment, it looks like the bug lies somewhere else, or I should've immediately (or at least somewhere along the line) EUE'd with the unstable overclock.

Next experiment:
- Virtualization tech off

[sswilson]

I've stopped running the GPU client on my ATI card until they get this figured out.

[chrisk]

I just walked in the door with a brand-spankin new EVGA GTX 260 Core 216 card...gonna install it now. I'll be selling my 4870! Shame...I wanted to keep this card for a long time but I have had issues with drivers and folding since I got it. Never ever had such issues with nVidia cards.

[Prof. Dr. Silver]

Guys, this is not just an ATI issue. I've gotten tons of these EUE's lately(2/3/4 per day). They believe it has something to do with the 58XX WU(for nVidia). Has nothing to do with my system since everything is running stock. I do hear that nVidia does a little better when it comes to the EUE issue. So I just downloaded the 178.24 drivers and core 1.18 for the Systray application and the results are looking up. I might get them once every two days now.

Oh and for the latest update.....I just saw that there is a new core available: 1.19. It apparently increase your PPD a bit and generates less EUE's YAY!