Page 1 of 1

Fermi vs Kepler

Posted: Sun May 12, 2013 1:41 am
by Hellmel
Just wondering if my collection of fermi's are almost obsolete? Is the new core going to be optimized for them as well as they have for the Keplers, or is it time for me to start upgrading them. I don't have a large contribution, but I like to keep it current. Unless the core will be optimized for Fermi's as well. I have 8 GPU's that would need to be replaced. Any ideas would be appreciated. Currently running four GTX 570SC's three GTS 450's and one GTX 460.

Re: Fermi vs Kepler

Posted: Sun May 12, 2013 1:54 am
by PantherX
Welcome to the F@H Forum Hellmel,

Please note that the FahCore_17 is currently being Beta Tested so any suggestions are subjected to change and can be speculative. I would advise you to wait and see how the QRB (Quick Return Bonus) pans out with the new Projects running FahCore_17.

AFAIK, the 400 Series do use significant amount of power so if you want to be more efficient and save some money on the electric bill, you could start by upgrading those GPUs to the more power efficient ones.

Re: Fermi vs Kepler

Posted: Sun May 12, 2013 7:34 am
by tlg
From my understanding, the AMD cards have blown past the nVidia cards just because of their acceptance of OpenMM. This may change, but the 7970 is gtetting 100k+ a day with QRB. Again this is just what I've read here.

Re: Fermi vs Kepler

Posted: Sun May 12, 2013 8:42 am
by AndyE
tlg wrote:From my understanding, the AMD cards have blown past the nVidia cards just because of their acceptance of OpenMM. This may change, but the 7970 is gtetting 100k+ a day with QRB. Again this is just what I've read here.
(As PantherX explained, things are subject to changes until the new core is final.)

For now, I can support your observation (running currently 4 GTX Titans and 4 AMD 7970 GE)

With stock frequencies,
the Titan average 120k with 160 watt/card
the 7970 average 95k with 180 watt/card

"Energy efficiency"
Titan: 750 PPD per watt (+41% vs. 7970)
7970: 530 PPD per watt

"purchase cost efficiency"
Titan: 1000$, 120 PPD / $
7970: 400$, 237 PPD / $ (+98% vs. Titan)

"3 yrs dedicated folding efficiency" (assumptions: purchase deprecated for 3 years, energy cost 12 cents /Kwh)
Titan: 80 PPD / $
7970: 98 PPD / $ (+23% vs Titan)

Please consider:
1) This is the best case for the Titan with these energy prices, as it is assuming dedicated folding
2) If the average % of folding time goes down, the case for the 7970 gets stronger
3) These numbers are based on the current beta work units. With other WU, these will probably change (in absolute terms).
4) The ratio between the cards might also change (due to potential different optimizations in the final version of the core)
5) Don't have numbers for the 450$ GTX 680

Things are in flux, treat it just as a current snapshot

Re: Fermi vs Kepler

Posted: Sun May 12, 2013 10:05 am
by artoar_11
I am a member of the beta team. I would advise you to wait. When do the final release of Core_17, we'll see which cards are better.
New software, optimization (drivers) sometimes change the behavior of the hardware. Nobody can predict it now.
Right move now to wait (waiting game).

(How long will it take: -- Soon ----- Soon-ish --) :)

And welcome to the F@H Forum Hellmel :D