GTX 1080

[Bryan]

I've been playing around with the new card and comparing it to my 7970s. With the exception of MW (DP) the GTX 1080 outperforms the 7970. In some cases by wide margins and in others not so much. This will at least give you some numbers for the GTX 1080 so you can compare it to whatever card you have and help you decide if a upgrade is worthwhile. The 1080 definitely performed better than my dead GTX 780 so that upgrade was worthwhile.

GTX 1080 - clock rate 1911-1965 MHz selected by card and it changed with project
7970 - clock rate 1100 MHz - selected by me

Both cards in identical I7-3930K machines with 12 threads available to the GPU.


Credits / day

COLLATZ x2 WU
1080 5.2M
7970 3.0M

PG x2 WU
1080 1.2M
7970 405k

MW x3 WU
1080 133k
7970 433k

POEM
1080 1.2M (x4 WU)
7970 600k (x3 WU)

MOO
1080 580k (x1)
1080 623k (x2)
7970 507k (x1)

GPUGrid - both cards tied at zero credits :D They have to change the program to match the architecture of the new NVidia cards and they can't do that till they get one.

The number of simultaneous WU were chosen to get the cards to 100% loading.

NOTE: it is possible the GTX 1080 would perform better on a PCI 3.0 mb but you can't prove it by me :)

Both cards were "optimized" on Collatz.

I didn't run Asteroids or Einstein and I damn sure didn't run Seti ... I only run Seti when I need 100 hours for WuProp stars!!

[nanoprobe]

Did you run the sieve app at collatz on the 1080 and is there an opti app for it?

[Bryan]

Yes, that was the sieve - to my knowledge that and the "micro" are the only ones available. If there is an optimized app then I'm not aware of it.

[nanoprobe]

Yes, that was the sieve - to my knowledge that and the "micro" are the only ones available. If there is an optimized app then I'm not aware of it.

Thanks. I wouldn't normally run collatz but being the badge ho that I am, well, you know........ =))

[Bryan]

Thanks. I wouldn't normally run collatz but being the badge ho that I am, well, you know........ =))

Yeh, I can relate to that =))

BTW, there is a thread on the forum concerning OPTIMIZING your card. Check that out because it does make a BIG difference versus running in the default mode.

[STMahlberg]

I've been playing around with the new card and comparing it to my 7970s. With the exception of MW (DP) the GTX 1080 outperforms the 7970. In some cases by wide margins and in others not so much. This will at least give you some numbers for the GTX 1080 so you can compare it to whatever card you have and help you decide if a upgrade is worthwhile. The 1080 definitely performed better than my dead GTX 780 so that upgrade was worthwhile.

GTX 1080 - clock rate 1911-1965 MHz selected by card and it changed with project
7970 - clock rate 1100 MHz - selected by me

Both cards in identical I7-3930K machines with 12 threads available to the GPU.


Credits / day

COLLATZ x2 WU
1080 5.2M
7970 3.0M

PG x2 WU
1080 1.2M
7970 405k

MW x3 WU
1080 133k
7970 433k

POEM
1080 1.2M (x4 WU)
7970 600k (x3 WU)

MOO
1080 580k (x1)
1080 623k (x2)
7970 507k (x1)

GPUGrid - both cards tied at zero credits :D They have to change the program to match the architecture of the new NVidia cards and they can't do that till they get one.

The number of simultaneous WU were chosen to get the cards to 100% loading.

NOTE: it is possible the GTX 1080 would perform better on a PCI 3.0 mb but you can't prove it by me :)

Both cards were "optimized" on Collatz.

I didn't run Asteroids or Einstein and I damn sure didn't run Seti ... I only run Seti when I need 100 hours for WuProp stars!!

Dang... is this correct? A 69% decrease? :-/

MW x3 WU
1080 133k
7970 433k

[Al]

Dang... is this correct? A 69% decrease? :-/

MW x3 WU
1080 133k
7970 433k
Bryan is the numbers guy...and yes, MW loves fast ATI cards and their enormous advantage on DP projects.

[nanoprobe]

Yeh, I can relate to that =))

BTW, there is a thread on the forum concerning OPTIMIZING your card. Check that out because it does make a BIG difference versus running in the default mode.

Played with the optimization. Does 2 minutes per task sound about right on a GTX 970? There were no posts specifically about the 970 settings.

[scole of TSBT]

Played with the optimization. Does 2 minutes per task sound about right on a GTX 970? There were no posts specifically about the 970 settings.
Yep.

[nanoprobe]

Yep.
Thanks

[DrPop]

@Bryan - WHAT? No SETI comparison?! :)) :D I guess I'm out of the running for one then. ;) haha

[FourOh]

Thanks for the data, Bryan!

For Einstein, you can expect 160-165k on 1070 and 180k-190k on 1080. May improve as drivers catch up.