F@H Project 6811 is brutal on GPUs

[AdamK47]

Good for system stress testing I suppose. Heats up my new GTX 580s like no other game/app.

Gets the fans close to 4,000 RPM.

 

[brownstone]

Wow, three 580's! That is just sick! :awe:

 

[Electrode]

Heh, my poor little GT430 can barely handle it

 

[Soccerman06]

92c seems a tad high, even futuremark and uniengine barely pushes 2 580s to 85c here. I think its watercooling time for you!

 

[AdamK47]

92c seems a tad high, even futuremark and uniengine barely pushes 2 580s to 85c here. I think its watercooling time for you!

Try folding one of these.

 

[Ken g6]

Try folding one of these.

Is there a way to control what I get? I keep getting 6800s. Not that working on Alzheimer's is bad; I'm just curious.

 

[Electrode]

Is there a way to control what I get? I keep getting 6800s. Not that working on Alzheimer's is bad; I'm just curious.

I'd assume that if you use -advmethods you'll get some. Might also depend on what card you are using...

 

[AdamK47]

Got three of them going at the same time now and damn it's hot in here. I may not need to turn on the furnace this winter. All I need to do is fold these.

 

[ZipSpeed]

My MSI 460 Cyclone on previous WUs would peak around 49C but these new ones easily 60+.

 

[Peter Trend]

I just got two of these and it made my system lag to the point of being totally unusable! Even pressing Num Lock (which I use a lot to check if my system has actually crashed or is just lagging) it takes about 25 seconds for the Num Lock LED to come on.

I found this. 6811 is certainly a very big WU. http://www.evga.com/forums/tm.aspx?m=675735&mpage=1

Does anyone have any ideas to help me reduce my GUI lag?
Currently at 840MHz on both GTX 460s, going to restart and turn GPU3 priority down to idle from low to see if that helps. But I've been trying to shut down the clients for about 45 minutes now! TOO MUCH LAG!!!

 

[biodoc]

I turned off advmethods on the gpu3beta client and have been getting a nice stream of project 6800 WUs which are fast and well behaved. Life is too short IMHO for the some of the "junk" I was getting with advmethods "on".

 

[Peter Trend]

Yeah, I don't blame you! Fortunately, although I had to do a forced shutdown, turning priority down from low to idle for both clients has stopped the lagging so far. I did try changing priority only on the GPU that my monitor is connected to, but the other client did it again. So two forced-shutdowns needed but crunching away again now

Just hope the forced shutdowns didn't damage my registry or corrupt any Windows files. Only time will tell...

Temps are only 58C and 44C with fans at 100%. Didn't reduce my overclock, either. MSI Frozr is a very impressive HSF

Edit: I turned fans from 100% to auto and it immediately started lagging again. Obviously not a good idea even with a good HSF! Very, very tempted to delete -advmethods after these WUs end!

Edit: Deleted -advmethods and had to ditch those WUs. It happened again with room temp at 10C and fans at 100%. I'm sure Stanford will have got a LOT of canned/corrupt 6811s!

Edit 200th post!

 

[Ken g6]

GUI lag hasn't been a problem for me since I switched my monitor to my Intel graphics chip. But I'm on Linux; I'm not sure how to do that for Windows.

 

[Peter Trend]

Yeah, I could try enabling my onboard GPU and use that for display. May actually help PPD as the GTX 460s would be free for f@h to use. As long as there isnt too much of a northbridge bandwidth/CPU time overhead for doing it.

When I did this with my 4870, my BIOS simply set display output to whatever device had a monitor plugged into it, and Windows followed suit without any problems. However, this time I will have the added fun of running NVIDIA and ATI drivers concurrently... Maybe will try it in January, don't want to kill my registry before the race even starts :|

Edit Oh wow! Now I'm a "Senior Member" hehe!
Just realised last post was no. 200

 

[Peter Trend]

Hmm...now it's happening with 6800WUs, think I have a deeper problem, going to downclock and try 258.96. Last time I ran Driver Cleaner it bricked my PC, but maybe that's what I need to try :|

Anyway, don't let me put the team off using -advmethods, I think my problem is rooted elsewhere.

 

[biodoc]

Hmm...now it's happening with 6800WUs, think I have a deeper problem, going to downclock and try 258.96. Last time I ran Driver Cleaner it bricked my PC, but maybe that's what I need to try :|

Anyway, don't let me put the team off using -advmethods, I think my problem is rooted elsewhere.

I've got the same problem on win 7. My daughter got the latest Sims3 expansion pack and wanted to play it on my quad with the GTX295. After she was done, I tried the GPU3 client with advmethods turned off (260.99 driver) and the machine locked up 3 times. I finally finished the WU but I had to leave the machine alone...no browsing the internet, etc.

 

[Peter Trend]

I've got the same problem on win 7. My daughter got the latest Sims3 expansion pack and wanted to play it on my quad with the GTX295. After she was done, I tried the GPU3 client with advmethods turned off (260.99 driver) and the machine locked up 3 times. I finally finished the WU but I had to leave the machine alone...no browsing the internet, etc.

Are you running the systray or console GPU3 client? Mines the console beta which I possibly shouldn't be using(?).

I spent the last 3 hours wiping the old drivers and trying to get 198.56 on. Turns out if you're going to use Driver Cleaner you need to tick NVIDIA Chipset if your running a mobo with an ATI chipset.

198.56 didn't solve the problem.

It only seems to happen when I start the second gpu client (which isn't even my display card!)

Hope I haven't damaged one of my cards. I doubt it though as that is the one that gets the least hot, as it's sat below the one used for display. So it's 10~20C cooler. Will try swapping them over when I get in anyway.

Also going to try setting CPU Usage to 99% on the GPU3 clients and set WU size to small. Just to rule a couple more things out.

It all worked fine for about 6 hours before I got my first 6811 unit. Could be a coincidence but I'll be very annoyed if this killed my card!

Might be worth letting Stanford know if this is affecting previously stable systems (i.e. it's not just me!)

 

[biodoc]

Are you running the systray or console GPU3 client? Mines the console beta which I possibly shouldn't be using(?).

I was running the console client too. I just fired up the GPU2 app (console) to see if that's stable. I've kept the 260.99 drivers for now.

There's a link over at the F@H forums for a GPU mem test app. I'll see if I can find it.

 

[biodoc]

Peter Trend, you have 2 GTX460's in that machine and are running 2 console clients? Are you using the -gpu 0 switch for one and -gpu 1 switch for the other? It's also a good idea to use the -local switch for both. Also I believe you need 256.xx or later drivers for the fermi cards?

Here's the link for the GPU mem tester app. I've haven't used it yet.

 

[Peter Trend]

Peter Trend, you have 2 GTX460's in that machine and are running 2 console clients? Are you using the -gpu 0 switch for one and -gpu 1 switch for the other? It's also a good idea to use the -local switch for both. Also I believe you need 256.xx or later drivers for the fermi cards?

Here's the link for the GPU mem tester app. I've haven't used it yet.

Yes I'm running two console clients on two GTX 460s.

I was using 260.99 when this first started occuring. I was using the -gpu 0 and -gpu 0 flags, as well as the -forcegpu nvidia_fermi flag on the second gpu client. I wasn't using -local but will try that tonight. Will also try swapping to the systray client and run that gpu mem tester. I hope it doesn't take too long.

Thanks biodoc, and good luck with your box!

 

[VirtualLarry]

What does -local do? I'm not using that.

I've noticed some problems with P6800 WUs. Both GTX460 1GB are doing one, yet the first GPU is only getting 5000 PPD, and the second GPU is getting 11000 PPD, according to HFM. I have no idea why the first GPU is running at half the speed of the second.

Shouldn't a GTX460 get more than 5000 PPD? I don't think that the cards have been overheating. They are Gigabyte cards with the dual fans. Temps according to Precision are around 65C.

Edit: Do GTX460s throttle when they lack enough power? I wonder. I tried overclocking them to 820 to fix the PPD, and the second card went from under 10000 PPD to over 11000 PPD, but the first card went DOWN from 6000 PPD to 5000 PPD.

I'm running the GPU3 systray client on XP, 260.99 drivers.

Edit: SOLVED. The reason that one of my GPUs (1) was running at half speed, is because - it was running two clients! I had installed GPU3, and then manually copied the work dir and created two shortcuts. Little did I realize that it automagically sets up a shortcut in the "Start" menu, so when I rebooted, a client fired up and prompted me to configure it. I clicked the "Close" button on the dialog, expecting it to cancel. But apparently, it fired up a client and kept running.

 

[VirtualLarry]

I turned off advmethods on the gpu3beta client and have been getting a nice stream of project 6800 WUs which are fast and well behaved. Life is too short IMHO for the some of the "junk" I was getting with advmethods "on".

Fast and well behaved, eh? Hmm.

 

[biodoc]

Fast and well behaved, eh? Hmm.

On linux, yes. I've got a 1090T w/ 2 GTX460s that's running smoothly without the -advmethods flag.

On my Q9550 w/ a GTX295 running 64 bit Win 7 pro, I've had problems. I reinstalled the 260.99 driver using the custom option and clicked on the "clean install" box. I'm running the GPU3 beta client now with no problems so far. Hopefully this continues.

 

[Peter Trend]

What does -local do? I'm not using that.

Edit: Do GTX460s throttle when they lack enough power? I wonder. I tried overclocking them to 820 to fix the PPD, and the second card went from under 10000 PPD to over 11000 PPD, but the first card went DOWN from 6000 PPD to 5000 PPD.

I'm running the GPU3 systray client on XP, 260.99 drivers.

Could be. I'm running a Zalman ZM-850, in case that helps.
Interestingly enough when I said I was having lock-ups on the second card (which isn't my display card), I noticed HFM claimed ~10900 for GPU 0 and ~5500 for GPU 1.

My only other power supply is "busy" but I suppose I could take it out, short pin 14 to ground, and use it to power my second card. That would at least rule out lack of power.

Could somebody with a folding forums account make a thread there and link to this thread please?

Also, does anybody know what Pande Group's bug logging system is?

Thanks

Edit: Thought this through and shorting it to ground isn't quite what I meant, don't do this if anybody is considering it! I meant link pin 14 and ground to my power switch so they both turn on at the same time. It's not as simple with GPUs as it is to run a couple of fans off a PSU, is it?
Anyway sorry, OT, just had to correct myself in case anyone fried a GPU on my advice :|

 

[Peter Trend]

I'm running the GPU3 beta client now with no problems so far. Hopefully this continues.

Did you keep the same WUs?

Fingers crossed for you.