Folding Forum

foldy wrote:Which one is faster? What vendors and models?

Yes pray tell. Also happy to see Pascal and Fermi folding on same board.

It seems to be about the same amount of PPD compared to 980 ti but under half the Watt. I need to spend some more time. Guess the 1070 is faster than the 980 ti.

Please take the size of the image down a little. It's blowing out the size of the screen in the forum, and your sucking up all my data to download it on my phone. Please be considerate of others. Thanks.

Nice PPD number, BTW.

@Sn1ken: Can you do a FahBench on your gtx 1070?

http://i.imgur.com/1diiQEb.jpg

Mod edit: Changed overly large image link to an URL

I don't know if you care about max power usage but maybe try to reduce power limit in msi afterburner e.g. to 75% to see if TPF increases much?

Adam A. Wanderer wrote:

Adam A. Wanderer wrote:

Adam A. Wanderer wrote:I upgraded to GTX-1070 today. I'll keep y'all posted on how well it works out.

The GPU folding slot failed. I'll let you know how the troubleshooting goes. I'm starting with installing a new driver.

Ah, much better now. It's taking eight point five (8.5) seconds per frame, plus or minus a second or two. Knock on wood.
I'll keep y'all posted.

Glad you got one ... And running. FPS is less important for us; but TPF, PPD, temps, TDP and power draw are of interesst.

TPF works as long as we have the project number - we can use the bonus points calculator to work out ppd

The problem with pascal is that its in short supply and some people are unwilling to pay for an FE card or put up with a blower cooler. I'm waiting for better availability and cards with a different cooling solution. Even then I'm going 1070 not 1080 and replacing a 970 - can't justify those costs for a 1080 yet.

As a comparison 11400 on my 980ti gets about 490k PPD

GTX 1080 on some work units gets 1 Million PPD but on other only 650k so average will be 800k.

GTX 1070 average will be 600k PPD.

Gtx 1080 has 2560 shader while gtx 1070 has 1920 shader both have same clock.
So gtx 1080 is 2560/1920 = 33% faster in theory which matches 800k / 600k = 33%.

Adam A. Wanderer wrote: TPF is one minute and 58 seconds. The Project number is: 11400. The PPD is 638,691.

@ foldy, calculate always with TPF. PPD is not a linear function.

GTX 1070; p11400; TPF: 01:58; according to Bonus Point Calculator: PPD: 628k

If GTX 1080 33% faster than GTX 1070; (TPF: 01:58=118 seconds); 118/1.33= 89 seconds (TPF: 01:29). According to Bonus Point Calculator: for TPF: 01:29, PPD= 958k.

958k/628k= 52%

@artoar_11: I see so gtx 1080 is 18% faster than gtx 1070 when usign TPF compare and 33% faster when using PPD compare.

foldy wrote:@artoar_11: I see so gtx 1080 is 18% faster than gtx 1070 when usign TPF compare and 33% faster when using PPD compare.

Demonstrates the QRB points curve. Not all "performance" increases are the same.

Adam A. Wanderer wrote:snip...
Do we have any way of telling how many Pascal units have come on line so far?

No one in this forum can. Would need to be asked of PG on reddit.

Please keep in mind that as shown on the OS Stats page - http://fah-web.stanford.edu/cgi-bin/mai ... e=osstats2, currently there are over 20,000 Fermi or later nVidia GPU's represented in the total folding stats. It will probably take at least 1 or 2 thousand Pascal GPU's being added to the mix before that contribution makes a notable change in the summary stats.