GPUv2 Client with HD4870

A

aznofazns

Limp Gawd
Joined
Sep 14, 2008
Messages
273
I'm trying to run a combination of GPUv2, SMP, and x86 console clients 24/7, but the GPUv2 client keeps giving me trouble. I usually don't get an error message regarding "NANS detected on GPU" or whatever. Instead, Windows simply freezes up for a few seconds, my screen turns black, then everything reappears, but my HD4870's manual fan control turns off and the GPUv2 client stops. It doesn't exit or even display an error message... the progress just stops. The SMP and x86 clients keep running just fine.

Even with my HD4870 at stock clocks and voltage, the client always fails. Sometimes it takes 2 hours, sometimes 8. I've tried Catalyst 9.8, 9.9, 9.10, and 9.11 drivers. They all do the same thing. I've also switched from ATI Tray Tools to Rivatuner to see if that would make a difference.

CPU is Prime95 stable, but I'm not 100% sure if memory is stable (overclocked to 850mhz at 1.92v, stock 800mhz/1.80v). The SMP and x86 clients don't give me problems so I think it's just the graphics card...

Does anyone else have this problem or know what's up?

System specs in sig.
 
I'm sorry this will not be of much help, but have you tried this in both XP and 7?
 
No I have not yet tried 7. I only have the expired beta anyways so I would much rather run it in my primary OS, XP.
 
Turn up your fan to 100% if it's not already, try make sure it's not getting too hot, and/or delete the .../folding@home-gpu/work folder and start again.

I would be hard pressed to find an explanation for this that included bad RAM.

As for the screen going black, that should be the VPU recover kicking in. I'm surprised that you don't get an error message, something about not responding to driver commands or something. But otherwise, yeah, that's the GPU automatically rebooting because it's locked up.
 
if your worried about the ram set it to 2v.. thats what it should be at for anything over 800mhz on the gskill..

but like Mr. P said.. check the temps.. use the system monitor in rivatuner and check all the temps as its folding and see if anything looks odd.. if the GPU temps getting way up there try changing the fan speeds.. see what works for you best..
 
Thanks for the replies.

I'm pretty certain temps are not the problem. Even with my fan locked at 35% speed, my GPU temp doesn't exceed 66C. The HD4870 doesn't throttle until it hits a much higher temp.

I just managed to run GPUv2 for about 8 hours nonstop so we'll see if the problem recurs. If it does, I'll try upping my RAM voltage a bit and/or install Catalyst 9.6.
 
You might want to take a look at this thread on the Folding Forum and see if it sounds familiar. Looks like you might not be alone.
 
Yeah, the core is most likely overheating on your card, so check your fan speed. The video shut down IS a core crash recovery process, but as stated before, im amazed you were not getting an error message. Update your catyl drivers too, that may be a part, and get the newest CCC for the fan speeds and monitoring on the card.
 
@FrEaKy: As I stated earlier, my GPU core is not overheating. It's far from it. Also, if you read my original post, you'll notice that I've tried multiple Catalyst drivers... to no avail.

@tjmagneto: Thank you! This is precisely the problem I'm having. It looks like I'll be spending some time on this tonight...
 
So I tried downgrading to Catalyst 9.4 and used CCC in place of Rivatuner, but ended up getting the VPU Recover thing after a little while. I noticed that in the thread tjmagneto linked, the people who got it working used "experimental DLL's" and "environmental variables," but I'm not sure where they got the DLL's and what those variables are... does anyone know?
 
Not muchhelp as I run a purely nvidia farm, I just thought I was the only person on here at 0300 hours lol
Posted via [H] Mobile Device
 
Mr. Pedantic,
Thanks for the links... I think I got it working now. I'll test it overnight (what's left of my night, anyway) and see if the crashing recurs.
 
Awesome, I've been running GPUv2 + SMP + x86 for 7 hours now and no problems whatsoever. The experimental DLL's and environmental variables seem to have fixed the problem.
 
grats, nice to know you got it up and running, we can use this information on the GPU information thread too.
 
For future reference:

If your ATI video card randomly crashes during GPUv2 folding, do the following:

1) Install the GPUv2 folding client (I used the console version).
2) Uninstall whatever version of Catalyst driver you have and reboot.
3) Install Catalyst 9.4 with Catalyst Control Center and reboot.
4) Download the Experimental DLL files and extract them to desktop.
5) Go into C:\Windows\system32 and replace the three ati DLL's with the three experimental DLL's.
6) Go into your folding@home folder. Delete the TWO DLL's that correspond to the experimental DLL's. They begin with "amd" instead of "ati," so you need to delete those two and rename the experimental DLL's to begin with "amd." Make sure you don't delete the third DLL, which doesn't correspond to the experimental ones.
7) Right click My Computer > Advanced > Environment Variables. Under System Variables, create four new variables:

name ...............................value
BROOK_YIELD ................2
CAL_NO_FLUSH ..............1
CAL_PRE_FLUSH ............1
FLUSH_INTERVAL ...........128 (You can experimental with this value if you wish. Setting a lower value should help eliminate the VPU Recover error).

8) Apply the changes and relog into Windows.
9) Start folding@home and pray that it no longer crashes.

If you want to know exactly how this method eliminates the crashing, click the links in Mr. Pedantic's post above me.
 
Just wanted to add that my overall PPD has boosted from ~4900 to ~5400 after I made the modifications.
 
You must log in or register to reply here.
Back
Top