Nvidia Pascal GPU cards???

[artoar_11]

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

nVidia:

Titan X Pascal - 1400k PPD, ~250w, 1300 dollar.
Titan X - 700-850k PPD, ~230w, 1200 dollar (old price, now 1100-1200?)
GTX 1080 - 700-900k PPD, ~160w, 700 dollar
GTX 1070 - 600-750k PPD, ~120w, 430-470 dollar
GTX 1060 - 330-420k PPD*, ~90w, 290 dollar (around 47% of 1080, based on clocks and chip)

420k seems quite high(?)

viewtopic.php?f=38&t=28870&start=45#p287749

[rwh202]

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

nVidia:

Titan X Pascal - 1400k PPD, ~250w, 1300 dollar.
Titan X - 700-850k PPD, ~230w, 1200 dollar (old price, now 1100-1200?)
GTX 1080 - 700-900k PPD, ~160w, 700 dollar
GTX 1070 - 600-750k PPD, ~120w, 430-470 dollar
GTX 1060 - 330-420k PPD*, ~90w, 290 dollar (around 47% of 1080, based on clocks and chip)

420k seems quite high(?)

viewtopic.php?f=38&t=28870&start=45#p287749

Yeah, scaling for clocks and cores doesn't account for QRB - I'd expect nearer the 300k end, but hoping to get one later this week for my work pc and will report back numbers

[rwh202]

Installed a 1060 on friday, and with 13 WU completed it's showing 320k PPD min, 390k PPD max and 352k as an average across WU in HFM.

That's on win 10 with a slightly factory overclocked 6 GB card (1556/1771 but actually boosting to nearer 1850). Not quite 400k, but higher than I was expecting. The newly announced 3 GB cards might be the price/performance sweetspot though with 10% fewer shaders and 20% less expensive.

[foldy]

The little slower 3GB version of gtx 1060 is now available for 200 dollar. May do 300-350k PPD.
http://www.newegg.com/Product/Product.a ... -_-Product

[Foxbat]

I'm sure there are plenty of Folders who know about this site, but I just came across it a few weeks ago. http://www.nowinstock.net/computers/videocards/nvidia/ lists all the NVIDIA GPUs and tracks their availability and prices (if available). Click on the flag at the top of the site to check for other countries, too.

[7im]

Please be careful. This is a little close to the line of this forum's standard for being hardware agnostic, maybe even a little over the line.

[foldy]

Use that link and you can choose between AMD and Nvidia
http://www.nowinstock.net/computers/videocards/

[Foxbat]

Please be careful. This is a little close to the line of this forum's standard for being hardware agnostic, maybe even a little over the line.

Sorry, 7im, you're right. Since the thread was about NVIDIA's Pascal cards, I wasn't thinking about equal opportunity for AMD's latest silicon. Thank you, Adam, for balancing it out.

[kiore]

I have a GTX 1080 hybrid on the way, recommended driver version that will still allow it to play well with a GTX 980ti on win7?

[7im]

Current drivers work all the way back to Fermi cards, so install the latest drivers with no issues.

[kiore]

Current drivers work all the way back to Fermi cards, so install the latest drivers with no issues.

Thanks, on version 373.70 GTX980ti and GTX 1080 working well together, not seeing an enormous lift in PPD for the Pascal over a 980ti yet, but definitely faster, need to see how they go on the same work unit. Changed as I wanted to run both slots on hybrid during the hot weather will add my temporarily resting aircooled GTX 980ti back in a 3rd slot once cooler weather arrives.

[kiore]

I currently don't run CPU slots due to heat and reduced GPU output, but when cooler may try to see what the entire system can contribute.

[foldy]

For nvidia GPUs you need one CPU core per GPU to feed them. So if you have 3 nvidia GPUs like kiore and a 4 core CPU then only one CPU core is left for CPU folding.

If he now would setup a CPU slot with 4 cores then it would be in conflict with 3 GPUs which means it slows them down.

[rwh202]

For nvidia GPUs you need one CPU core per GPU to feed them. So if you have 3 nvidia GPUs like kiore and a 4 core CPU then only one CPU core is left for CPU folding.

If he now would setup a CPU slot with 4 cores then it would be in conflict with 3 GPUs which means it slows them down.

I'm not sure exactly why, but the effect is worse than that though. If I run a CPU:4 slot on an 8 thread i7 with 2 GPU, I get an overall reduction in points. There must still be some contention and possibly increased latency to CPU and memory resources. Also, with more cores active, CPU boost speed goes down, and an extra 50 W or so dissipated in the chassis will increase the thermal throttling of the GPUs. Less than a 1% reduction in GPU performance (2x 980) will wipe out the PPD from a 4 thread CPU slot.That's all before you start considering the efficiency of things where the CPU comes out pretty poorly at the moment.

I'm sure there is still important science being run by CPUs, but in my opinion, that is probably best left to dedicated CPU machine, such as majority running with just integrated graphics or the multi-cpu servers and workstations where their CPU resources can really make a difference. If you're running GPUs, I'd concentrate on the getting the most out of them, rather than complicating things with a CPU slot.

[7im]

I will agree that there is a slight contention, and even a very small impact on a high end GPU reduces the bonus more than the CPU can add. But due to the ludicrously high QRB points added with very short frame times, a few seconds a frame ends up being distorted in to thousands of points.

Go this route if points are your only concern. Fold on both the GPU and CPU when science means more to you than points.

And if you need proof, compare the change in only the base points, with or without the CPU folding.