Folding Forum

Also note that with the exponential nature of the bonus points curve, the PPD will vary quite a bit, especially with higher end cards that are further along on the points curve.

Take a look at the variations in the chart on this FAH News Blog article: http://folding.stanford.edu/home/a-peek-into-core-17-benchmarking-2/

For reference, the fahbench comparisons... 7970 vs. R9 290 is about a 22% difference, at theoretical best. Unfortunately, driver issues play in to actual folding, and the AMD drivers lag a little behind NV drivers.

I have two 7970s, one per system. They are using the same drivers, Fahcore, clocks (core and memory), and similar systems (3770k and 4770k). The only difference is one is Windows 8.1 and one is 7. Also, one is from the first batch, and the other brand new. Result? The new one is 5-10k PPD better than the old one. Strange things can cause PPD variation even between the same card.

$ilent wrote:...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...

It's my understanding that you initially had AMD R9 290X and you later changed it to an AMD R9 290, is it correct? If so, then your AMD R9 290 will be getting less points than AMD R9 290X.

Regarding the drop in points, are you, by chance running any GPU intensive applications while folding? Watching HD Videos with hardware acceleration or other stuff? If so, then it may account for (some?) drop of points since the TPF has increased.

no matter what I do I cant get over 130,000ppd even on the better 7810 and 7811 units.

Here is part of your confusion.

8900 did much better on AMD cards...on my 7970 realistic final ppd of 89-92k
7810 and 7811 might do better on NV, but on my AMD realistic final ppd of 68-70k

Hence the combination of reduced shaders and the variability in WU, and your results make sense.

compare the 290x to the 290 on the 8900 WU for a more accurate comparison.

also watch temps and fan settings as the cards will clock down aggressively...to reach full potential you will have to make sure they stay cool.

Edit: on closer reading I see others have mentioned and you tried 100% fan.
I am holding off for non-reference or water cooling myself...so thanks for confirming

mdk777 wrote:

no matter what I do I cant get over 130,000ppd even on the better 7810 and 7811 units.

Here is part of your confusion.

8900 did much better on AMD cards...on my 7970 realistic final ppd of 89-92k
7810 and 7811 might do better on NV, but on my AMD realistic final ppd of 68-70k

Hence the combination of reduced shaders and the variability in WU, and your results make sense.

compare the 290x to the 290 on the 8900 WU for a more accurate comparison.

You must have a super slow 7970 if your only getting 70k ppd on a 7810/7811. I know for a fact 7970s can get up to 120k ppd when overclocked.

PantherX wrote:

$ilent wrote:...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...

It's my understanding that you initially had AMD R9 290X and you later changed it to an AMD R9 290, is it correct? If so, then your AMD R9 290 will be getting less points than AMD R9 290X.

Regarding the drop in points, are you, by chance running any GPU intensive applications while folding? Watching HD Videos with hardware acceleration or other stuff? If so, then it may account for (some?) drop of points since the TPF has increased.

I dont mean to be rude, but there is no way that the difference from a 290x to 290 is like 30% folding@home performance, the cards when benchmarked are like 5% difference.

Unless you're referring to the fahbench benchmark any other benchmark means very little to PPD. Also note that fahbench uses a very small data set when doing the benchmark so in real world situations, large work units can vary quite a bit from small work units, in final PPD. The same applies when comparing one GPU to another GPU and the number of shaders is different.

I dont mean to be rude, but there is no way that the difference from a 290x to 290 is like 30% folding@home performance, the cards when benchmarked are like 5% difference.

Shader count, which is important to folding is 10%

But like I said, why don't you just compare like WU?

PS
Tom's reports worse TIM on 290 VS. 290X
http://www.tomshardware.com/reviews/rad ... ,3678.html

As I mentioned, with continuous load of FOLDING, performance is going to be very THERMAL dependent.

Open rig? ambient?
air flow. ?

guys please stop saying ppd varies by work units...im not a noob. I know for a fact that ppd does NOT vary by 30% across the same damm work units?

@mdk777 I know performance changes with thermal limits, but ive said this numerous times now temperature is not the issue here. I am well below my thermal limit before the gpu starts throttling. Open rig, ambient less than 20c.

im not a noob. I know for a fact that ppd does NOT vary by 30% across the same damm work units?

sorry, not here long enough then.
yes, given the multiplying affect of quick return bonus, differences in the compute nature of individual wu(#of atoms, forces) and the way different cards handle memory(while shaders is a large part, other architecture features do come into play)....

long story short, yes ppd have in the past and most likely will in the future vary by that much.

EDIT:

ps not saying that it ideal or intentional, just that it does happen.

I am well below my thermal limit before the gpu starts throttling

you have monitored the gpu activity and it stays at 100%...never cycling down?
possibility...but not likely.