Page 1 of 2

Low ppd on r9 290

Posted: Sun Nov 17, 2013 6:12 pm
by $ilent
Hi all

I'm having issues with a r9 290 when it comes to folding performance. For instance on a 7810 work unit I'm only getting like 120,000 ppd whereas a 7970 will get similer or more points. I have seen the ppd on this unit jump up to 190,000 once, but it's just gone back down.

I have tried amd beta 13.11 9.2 driver, amd r9 290whql drivers in conjunction with amd Ccc and nothing works, the ppd is still really low, when it should be as good or better than a Gtx 780.

Anyone know why my ppd is so low?

Re: Low ppd on r9 290

Posted: Sun Nov 17, 2013 6:29 pm
by EXT64
Hi, Are you using V7 Control or Web to get your PPD numbers? If so that is the problem. There is a glitch in V7 that causes the PPD to fluctuate a lot on 781x WU. If you use a program like HFM ( http://code.google.com/p/hfm-net/ ) it can average over the entire WU and get the correct PPD. Or you can look in the log and read the time per frame manually and average and then input into a PPD calcuator ( http://www.linuxforge.net/bonuscalc2.php ) to get the true PPD.

Also, my 7970 at 1.05GHz (so minor OC) only gets 95-100k PPD on a 7810 WU, so to get 120k would take quite an OC.

And finally, how are your temps? If the R9 290 gets too hot it can underclock a lot and lose a ton of performance.

Re: Low ppd on r9 290

Posted: Sun Nov 17, 2013 6:41 pm
by $ilent
Hi

I am using V7 client but im monitoring ppd with HFM. TPF stinks on the 290 at the minute, its like something is wrong with it. I used to have a r9 290x and it got 170,000ppd on an 8900 with a mild overclock, with this new r9 290 ive got no matter what I do I cant get over 130,000ppd even on the better 7810 and 7811 units. I have tried different drivers, different setup (using ati CCC and using msi afterburner, then neither etc) and ppd still stays the same.

Temperatures are perfectly fine, never any core clock throttling. Something is definitely wrong with it, I dont see how in the space of 3 weeks my ppd has dropped by 30% even though im trying same setup in terms of driver. Also my gtx 670 gets close to 110k ppd on a 780, so for this to only get an extra 15k ppd I know is wrong. I see 7970s getting over 120,000ppd, so I am almost certain something is wrong with this.

Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:03 pm
by $ilent
Hi

around 3 weeks ago on my 290x I was able to attain 170,000ppd on a 8900 on my amd R9 290X, with a TPF around 2min 52sec. Now my ppd has dropped massively and its taking ~3min 27sec to do the same work unit at the same clocks. Have Stanford increased the difficulty of the core17 work units or something recently?

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:05 pm
by P5-133XL
No

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:08 pm
by $ilent
Any ideas then why myself and others used to be able to attain higher ppd a few weeks ago, but since core17 seems to have gone public ppd has dropped?

I have tried literally every combination before you ask of new drivers, older drivers, overclocking, not overclocking and the ppd has remained the same. Something must have changed either with the units getting less bonus' points or something is making them hard to complete.

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:10 pm
by jimerickson
there is considerable variability amongst work units.

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:10 pm
by $ilent
30,000ppd worth of variance?

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:22 pm
by P5-133XL
A likely possibility is that you are using FAHControl to monitor PPD. It tends to have large swings in PPD/TPF estimates with the actual PPD somewhere in the middle. If you compare your high and a low you are well within those swing levels. HFM.net tens to have much more accurate and stable estimated PPD/TPF's. This possibility is irrelevant if you are getting the TPF's from the logs rather than FAHControl (It is FAHControl's estimates that are poor not the actual log times).

Pure guess: R9 290's are particularly sensitive to temp. They dynamically drop their clock rates (and perhaps your OC monitoring utility isn't fast enough to detect the drops) to keep their temperature in check.

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:23 pm
by $ilent
Ive used folding@home with my fan speed at 100% (and lost 50% of my hearing in the process), which has kept max gpu temp to @65C, which would eliminate thermal throttling. I also use HFM to monitor ppd.

Re: Have Stanford changed core17 difficulty?

Posted: Sun Nov 17, 2013 9:25 pm
by P5-133XL
Then we are stuck with basic WU variability as jimerickson suggested unless you are pausing the slot occasionally.

Re: Low ppd on r9 290

Posted: Sun Nov 17, 2013 9:31 pm
by $ilent
hmm im not pausing the slot either, ah well no worries. Thanks for your help guys, ill just live with it lol

Re: Low ppd on r9 290

Posted: Sun Nov 17, 2013 9:35 pm
by 7im
There is a big difference between 290s and 290xs. Where as the 290 is more similar to the 7970. You need a serious overclock to get 120k on a 7970. 100 to 110 k is more typical. I don't see the big difference here.

Re: Low ppd on r9 290

Posted: Sun Nov 17, 2013 9:36 pm
by 7im
Have you used driver cleaner to pull out the NV drivers, and previous AMD drivers? Conflicting drivers can slow down the GPU.

Also, are you folding on the CPU as well? Did you reserve 1 core for GPU folding?

Re: Low ppd on r9 290

Posted: Sun Nov 17, 2013 9:41 pm
by $ilent
Ive tried removing both sets of drivers, done clean windows install, only folding on the 1x290, folding set to full priority.