Unstable Machine Errors with F@H

[Drsignguy]

Lately, I have been doing some switching hardware around to make things run more efficient with my folding experiences. I looked into all my hardware and thought it out to what Rig could use each piece of hardware to its full advantage. In my sig, It is now pretty much up-to-date of what each Rig has.

As I have been reinstalling and updating the video drivers, installing the F@h Gpu clients, I found out something that kind of put things into more perspective. For example; My MSI GTX 260 has Stock settings of 620/1296/1080. When trying to run F@H, it kept giving me that "unstable machine" error and I wondered why as everything was on par as per windows (xp x64 in this case). I benched the card and it was stable as per 3DMark, FurMark and another vid test program. But when I tried to run F@H, it gave me the error after only 1% completion every time. After everything I could do with the drivers and program installations, I just decided to drop the default clock speeds (620/1296/1080) to 576/1296/1000 and guess what, it worked.

After a complete successful run, I started inching my way back up and I am now presently at 612/1512/1000. I did this with all my cards and so far I haven't received an error.

Yep, I am happy now.

I have now vented!

 

[petrusbroder]

Hmmm, very interesting: so it seems that F@H is also more sensitive to errors than other applications.
I guess, that in graphics a miscalculation would not have much of an impact (a pixel blinks at the wrong time in the wrong color) because the very fast framerate (you do not see pixel errors which occur faster the once every 25th frame) ... but in calculations it matters.
It also proves that the F@H client has some error detection and possibly error correction built in.
Thanks for the info!

 

[Duvie]

I agree...nvidia optimization said 687 core but I had one error so I dropped it down to 675 with shader clock linked and no issues so far...

Nvidia optimization is likely for gaming where it may be more forgicing to have a limited amunt of bad data now and then

 

[Drsignguy]

Originally posted by: Duvie
I agree...nvidia optimization said 687 core but I had one error so I dropped it down to 675 with shader clock linked and no issues so far...

Nvidia optimization is likely for gaming where it may be more forgicing to have a limited amunt of bad data now and then

Good point Duvie. Seems that gaming is more forgiving than that of folding proteins.






Also. your welcome Peter.

 

[Markfw]

It could also be power supply. I burned up an Antec 550 on my 260 card and a quad@3.5. Now I am running a Fortron 700.

My 260 says (by nvidia system tools) it can run at 704, but it got the unstable machine, and I put it down to 689, and now its fine.

 

[Drsignguy]

True. But I did fail to mention that I did test the power supply. It was the first thing I tested and all was good. It is a Corsair TX 750w so it is very capable to run it. I got it so I can run a second 260.....soon.

Good call to mention that Mark....:thumbsup:

 

[Denithor]

You guys see a difference in F@H results when adjusting the core+shaders versus just shaders?

 

[Drsignguy]

Originally posted by: Denithor
You guys see a difference in F@H results when adjusting the core+shaders versus just shaders?

Slight. I have tweaked them a bit to see how much and to tell you the truth it's only minimal. The temps are a bit lower and that lets me lower my fan speed a touch.

What I have been doing now is making sure the defaults are dropped a bit before I start a client and let it run complete, then I start inching my back up. Left the core and memory clocks at default and then started to crank the shaders. I know how high I can go with them but am not going to push it so I back them down a few notches and let 'er rip. PPD For the 260/216 reaches over 8000 on projects 5770 (353 points) down to 5500 on project 5753 ( 511 Point). My 260/192 is slightly lower on each.

 

[Shebnay]

I'm using a 8800 GTS at stock speeds 678/1728/972 and I seem to get errors every other WU. Going to try 656/1728/954, hopefully that will work.

There's a huge error reporting thread on the f@h forums. I guess it may be WU specific?

 

[Drsignguy]

Originally posted by: Shebnay
I'm using a 8800 GTS at stock speeds 678/1728/972 and I seem to get errors every other WU. Going to try 656/1728/954, hopefully that will work.

There's a huge error reporting thread on the f@h forums. I guess it may be WU specific?

Sometimes it is specific, but it can mainly be your vid card. When you start seeing "unstable machine", thats when you know its your card or hardware related. EUE's are most common by the occational bad client. By backing down your clocks at or below defaults, can give you a good indication on how well your card will perform and then you can start stepping up the clock speeds once you know your card is stable at the lowest speed settings.

 

[Shebnay]

Originally posted by: Drsignguy

Originally posted by: Shebnay
I'm using a 8800 GTS at stock speeds 678/1728/972 and I seem to get errors every other WU. Going to try 656/1728/954, hopefully that will work.

There's a huge error reporting thread on the f@h forums. I guess it may be WU specific?

Sometimes it is specific, but it can mainly be your vid card. When you start seeing "unstable machine", thats when you know its your card or hardware related. EUE's are most common by the occational bad client. By backing down your clocks at or below defaults, can give you a good indication on how well your card will perform and then you can start stepping up the clock speeds once you know your card is stable at the lowest speed settings.

It looks like I'll have to keep backing down a bit more. Another "unstable machine" last night, below stock speeds :\

 

[Markfw]

Either I am real lucky, or doing this just right. With 15 GPU clients, I have not had an unstable machine or GPU EUE in weeks to months. I had one card a month or two ago overclocked too much, thats all in months.

 

[Drsignguy]

Originally posted by: Shebnay

Originally posted by: Drsignguy

Originally posted by: Shebnay
I'm using a 8800 GTS at stock speeds 678/1728/972 and I seem to get errors every other WU. Going to try 656/1728/954, hopefully that will work.

There's a huge error reporting thread on the f@h forums. I guess it may be WU specific?

Sometimes it is specific, but it can mainly be your vid card. When you start seeing "unstable machine", thats when you know its your card or hardware related. EUE's are most common by the occational bad client. By backing down your clocks at or below defaults, can give you a good indication on how well your card will perform and then you can start stepping up the clock speeds once you know your card is stable at the lowest speed settings.

It looks like I'll have to keep backing down a bit more. Another "unstable machine" last night, below stock speeds :\

Remember, it still can be other hardware related as well. If you overclock your Cpu and or Ram, this can also cause the error. You have to have a stable machine or F@H will pick up on that.

Originally posted by: Markfw900
Either I am real lucky, or doing this just right. With 15 GPU clients, I have not had an unstable machine or GPU EUE in weeks to months. I had one card a month or two ago overclocked too much, thats all in months.

I think you're doing it right. :thumbsup: