Folding Forum

Maybe a future GTX 1080 TI will be the GPU to buy for folding at better price then this Titan X2.

Adam A. Wanderer wrote:Just a thought, but 20,000 1080s in one rack dedicated to folding would run about 20,000,000 dollars. Just a thought.

It would also run 3.6MW of power for just the gpu's and 14Bn PPD

Adam A. Wanderer wrote:How sad they don't make GPUs exclusively for folding. But, the market would probably be too limited.

An ideal card exclusively for folding would be able to eliminate all the hardware associated with rasterizing and displaying the video ... i.e.- only the hardware needed to process the 3D Computations required for CUDA and/or OpenCL support ... and thereby making it unsuitable for gaming. You'd need a separate (perhaps integrated-) GPU to connect your monitor.

... I don't think it'll ever happen ... but perhaps some percentage of RMA GPUs with defective video circuitry could be re-sold at a reduced rate (after removing the video connectors) -- perhaps supported ONLY for PhysX.

Thought I'd add my data to this post. I bought a GTX 1070 and wanted to post some real world Folding output.

The Build:
i7-4770 on a Gigabyte Z87X-UD4H-CF
8GB DDR3 & Samsung 830 SSD
Rosewill Capstone 'Gold' PSU
CM Elite chassis w/ 4 fans
Windows 7 Pro 64-bit

The Card:
EVGA GTX 1070 SC
Nvidia 368.81 drivers

I wrote down a handful of the work units and the average output that the Advanced Control screen was showing.
(this is the output with Just the video card Folding)

P9160 - 535,900 PPD
P13133 - 556,500 PPD
P9151 - 691,300 PPD
P9152 - 667,500 PPD
P13110 - 643,800 PPD
P9159 - 554,000 PPD
P11414 - 645,700 PPD


Power draw from the wall has ranged from about 186W to 210W.

Adam A. Wanderer wrote:How sad they don't make GPUs exclusively for folding.

It is very fortunate that they don't. Our costs for Folding card development are now being paid almost exclusively by the gamers. If we had to pay them, you wouldn't want to buy the card.

Nathan_P wrote:

Sven wrote:Even more FoldingPower at an extraordinary price:

http://www.geforce.com/hardware/10series/titan-x-pascal

We need that gpu with half the memory to pull the costs down. Going out on a limb here but I'd say about 1.2m PPD from the card

I wasn't far off, PPD for a Titan X (Pascal) has been reported at 1.12m PPD

https://hardforum.com/threads/titan-x-p ... d.1907506/

http://www.anandtech.com/show/10540/the ... -review/14

The GeForce GTX 1060 Founders Edition & ASUS Strix GTX 1060 Review
OpenCL test, utilizing the OpenCL path for FAHCore 21.

Nathan_P wrote:I wasn't far off, PPD for a Titan X (Pascal) has been reported at 1.12m PPD

https://hardforum.com/threads/titan-x-p ... d.1907506/

Did they mention if the measure was taken with Windows or Linux ?

toTOW wrote:

Nathan_P wrote:I wasn't far off, PPD for a Titan X (Pascal) has been reported at 1.12m PPD

https://hardforum.com/threads/titan-x-p ... d.1907506/

Did they mention if the measure was taken with Windows or Linux ?

No they did not. And not that it matters much. 1 Million PPD is ludicrous.

toTOW wrote:

Nathan_P wrote:I wasn't far off, PPD for a Titan X (Pascal) has been reported at 1.12m PPD

https://hardforum.com/threads/titan-x-p ... d.1907506/

Did they mention if the measure was taken with Windows or Linux ?

They didn't. i'll ask.

Some other Titan X pascal reports

http://forums.bit-tech.net/showthread.php?t=308023

http://forums.evga.com/Anyone-have-a-Ti ... 27251.aspx

The tests were run on windows 10 64bit, driver was released 2nd August

1x Titan X Pascal for $1200 make 1.1m PPD at 250W
2x gtx 1080 for $1300 make 1.6m PPD at 360W
3x gtx 1070 for $1300 make 1.8m PPD at 450W

Looks like GTX 1080s are currently the sweet spot.

So, finally, how much ppd is a 1060 doing?

MARSTG wrote:So, finally, how much ppd is a 1060 doing?

from what I understood reading on other forums it is a bit above a 970; around/above 300kPPD