8th Annual Holiday Season Folding@Home Race! The Race is on!!

[Pokey]

....................... Maybe the temp is hitting the limit and then the GPU takes a break?

Everything I have read says these cards are hard to break. They either crash or self modulate down. You may have pushed the limit on something. From the meter it looks like you have exceeded temp limit. (Your temp target is 82 but your temp is 83). Might play with fan curve........... or something else.............. also you might want to push power target up to 115%.

Just my two cents, YMMV.....

 

[Ken g6]

The GPU temp and fan speed slowly creeps up to 82 C and 60%, runs for a while and then the GPU load drops to nothing for a few seconds (GPU temp and fan speed go down during those few seconds and then GPU load kicks in and the process starts again. Weird, huh? Maybe the temp is hitting the limit and then the GPU takes a break?

Well, my GTX460 does the same thing. It seems to stop at every checkpoint for a little while. I suspect this is the way the F@H client was designed, to stop, copy current results off the GPU, and maybe do something with them on the CPU before continuing.

 

[brownstone]

Well, my GTX460 does the same thing. It seems to stop at every checkpoint for a little while. I suspect this is the way the F@H client was designed, to stop, copy current results off the GPU, and maybe do something with them on the CPU before continuing.

This is what I've seen as well on my gpu. I'm thinking it's just some kind of save point or checkpoint.

 

[ZipSpeed]

I have a 7950 and 7870XT that drops off to 0% for a brief moment as well. With my 2 x GTX 460 and the other 7950, I have yet to see it behave like that. Not sure why some cards are doing it and others aren't.

 

[Assimilator1]

The radeon 4870 is on the "white list" of supported GPUs

0x1002:0x9440:::RV770 [Radeon HD 4870]

https://fah-web.stanford.edu/file-releases/public/GPUs.txt

Something's wrong with that list, it includes X1900s & much older GPU's which I know aren't supported. I'll have to dig around , thanks anyway

[edit]I think the above list is just a list of recognised GPUs, also I found this.

◾The end of support for ATI/AMD 4xxx series and below was announced in March 2011, and scheduled to end in September 2011. The last of the Fahcore_11 (GPU2) work units actually lasted until June 2012, more than a year after the original announcement.

Damn, that rings a bell now I've read it.

Ugh, what a day. Monitor literally breaks off of it's stand (and it wasn't the stand that broke, but the monitor), and now that I'm on a single monitor, FAH isn't coming up anywhere... :|

This is why I hate F@H

You hate it because it broke your monitor??
What do you mean it isn't coming up anywhere? I found it on google
Seriously, what did you mean?

 

[Ken g6]

You hate it because it broke your monitor??
What do you mean it isn't coming up anywhere? I found it on google
Seriously, what did you mean?

You mean when you bring up fahcontrol, it thinks it should be in the place it was when you had two monitors, and so it's off-screen? With its tray icon selected, try [Alt], [space], "m", and click your mouse in the middle of the screen. Does it appear?

 

[petrusbroder]

Here are the stats as of December 5, 2013, approx. 21:00 UTC:

Sorry for the inferior quality of the image - but I am using an old laptop today ...
Edit: posted the correct image.

 

[Ken g6]

Er, Peter, where are the names? In column A?

May I suggest two different screenshots if you can't fit them all in one?

 

[biodoc]

I have a 7950 and 7870XT that drops off to 0% for a brief moment as well. With my 2 x GTX 460 and the other 7950, I have yet to see it behave like that. Not sure why some cards are doing it and others aren't.

Looks like the 780Ti is taking a 3/4 second break after each frame/step. I think the default checkpoint is every 15 minutes and I'm not sure how to monitor that event. My TPF is under 3 minutes so the checkpoint would be approximately 1 per 5 frames or steps.

 

[Neurodog]

Hi everyone!

I'm a little late to the race...any room to fit another?

 

[GLeeM]

Thanks for the stats, Peter

I think the default checkpoint is every 15 minutes and I'm not sure how to monitor that event.

Years ago there used to be a core that would checkpoint even if the time per frame was less than the checkpoint time.
I think that a checkpoint happens with each frame (percent).

Hi everyone! I'm a little late to the race...any room to fit another?

Peter will add you - start crunching :thumbsup:

 

[Fullmetal Chocobo]

You hate it because it broke your monitor??
What do you mean it isn't coming up anywhere? I found it on google
Seriously, what did you mean?

FAH broke my monitor... And here I was blaming the cat.

It isn't running anymore after I went to a single monitor. Makes no sense. I could understand if it was stuck on the other monitor, but my processor usage is nil. :| I found some more command line options that I might be able to use.

Is anyone else getting failed results on GPU tasks? It keeps happening on one machine. I will have to mess with it later.

 

[GLeeM]

It isn't running anymore after I went to a single monitor.

Which OS is it that has to have the GPU connected to a monitor (or have a dummy plug installed) before the GPU will crunch?
What OS do you have?

 

[Fullmetal Chocobo]

Which OS is it that has to have the GPU connected to a monitor (or have a dummy plug installed) before the GPU will crunch?
What OS do you have?

Running Windows 7 64-bit.

There is one monitor connected. I have had issues with some programs going to the non-existing monitor, but they are at least running. FAH just quit running, literally. Makes absolutely no sense.

 

[Markfw]

Running Windows 7 64-bit.

There is one monitor connected. I have had issues with some programs going to the non-existing monitor, but they are at least running. FAH just quit running, literally. Makes absolutely no sense.

What Client ? I have win xp 64, and one monitor and 2 cards, and it works on both cards always.

 

[Markfw]

Alright, finally have some good WU's !!!! 550k ppd.

The other 3 boxes are off, since they only generate 30k ppd each, but use a lot of power.

 

[petrusbroder]

Er, Peter, where are the names? In column A?

May I suggest two different screenshots if you can't fit them all in one?

Sorry; :$ I used my 10-inch laptop and missed that I did not get the whole spreadsheet on the screenshot. Mistake corrected in that post!

 

[petrusbroder]

Here are the December 6, 2013 stats, as of 21:10 UTC

Neurodog and Assimilator1 have been added to the racing teams.

Here are some production stats:

 

[Assimilator1]

Just realised, although I said it in a PM to Petrus, I hadn't said it here, but I'm in with my old Quad .

Atm it's just crunching on 2 cores whilst it burns off the remaining Asteroid WUs, actually that could take a while as it had a 4 day cache of them, for 4 cores!
Wondering if I'd be better off going all on A@H & then switching when the WUs are depleted, don't suppose it'd make much difference though.....
Client's showing ~4k ppd with the 2 cores.

Btw, thx for the stats Petrus

 

[petrusbroder]

Hi everyone!

I'm a little late to the race...any room to fit another?

Welcome to the race, Neurodog! You are in - as seen in the stats above! ()

 

[petrusbroder]

Just realised, although I said it in a PM to Petrus, I hadn't said it here, but I'm in with my old Quad .

Atm it's just crunching on 2 cores whilst it burns off the remaining Asteroid WUs, actually that could take a while as it had a 4 day cache of them, for 4 cores!
Wondering if I'd be better off going all on A@H & then switching when the WUs are depleted, don't suppose it'd make much difference though.....
Client's showing ~4k ppd with the 2 cores.

Btw, thx for the stats Petrus

As you can see from the stats, Mark, you are in too!

 

[Assimilator1]

:thumbsup:

 

[biodoc]

Thanks for the stats, Peter!

Welcome to the race Neurodog and Assimilator1! :thumbsup:

 

[Rattledagger]

Just realised, although I said it in a PM to Petrus, I hadn't said it here, but I'm in with my old Quad .

Atm it's just crunching on 2 cores whilst it burns off the remaining Asteroid WUs, actually that could take a while as it had a 4 day cache of them, for 4 cores!
Wondering if I'd be better off going all on A@H & then switching when the WUs are depleted, don't suppose it'd make much difference though.....
Client's showing ~4k ppd with the 2 cores.

FAH uses Quick Return Bonus (QRB) on most work, meaning running 4 cores for 24 hours will likely give significantly more than running 2 cores for 48 hours.

Using two of my own results as basis, and just assuming linear decrease due to #cores (not really the case for HT) together with the bonus-calculator located http://www.linuxforge.net/bonuscalc2.php

Project-9004:
1 core: 519 PPD -> 4151 PPD if could multiply by 8 to use 8-core
2 cores: 1468 PPD -> 5870 PPD if could multiply by 4 to use 8-core ==> 41 % higher than 1-core
4 cores: 4137 PPD -> 8273 PPD if could multiply by 2 to use 8-core ==> 99 % higher than 1-core.
8 cores: 11700 PPD ==> 182 % higher than 1-core.

Project-8571:
1 core: 501 PPD -> 4011
2 cores:1418 PPD -> 5671 ==> 41 %
4 cores: 4012 PPD -> 8023 ==> 100 %
8 cores: 11336 PPD ==> 183 %

So, based on this, running 4 cores for 24 hours would get 41 % more points than running 2 cores for 48 hours.

 

[GLeeM]

Project-9004:
Project-8571:

It's been so long I don't remember for sure but I think there is a "Kfactor" that figures into how much the Quick Return Bonus gives. If you look at the last column on this page you can find Kfactor for each project: http://fah-web.stanford.edu/psummary.html

9004 has Kfactor of 0.75
8571 has Kfactor of 2.99
some have Kfactor of 0.00