Top GPUs for Folding@Home

[Juggy]

On Windows fast GPUs need pcie 3.0 x4 to not bottleneck, pcie 3.0 x8 is nearly optimum.
On Linux fast GPUs can also be run on pcie 3.0 x1 risers with only small performance loss.

Core 22 and a 2080ti do saturate an x1 slot in Linux. Even an x4 slot is almost saturated (80%).
Windoze should run an x4 slot completely saturated.
At this point, the best hardware would be a dual core Celeron pushing 2 RTX GPUs, as no motherboard supports more than 2x pcie3.0 8x speed slots.

With pcie 4.0 on the horizon, one could run 3 or 4 GPUs on a $120 motherboard, though my experience with current hardware is, that it's very hard to find motherboards even supporting more than 2 Nvidia GPUs, (or 3 AMD GPUs) even if they have 3 full size slots and 3 or so x1 slots.
Something about manufacturers neglecting to program addresses for more than 2 or 3 GPUs in the bios.

There are quite a few motherboards that support more than 2x PCIe 8x. X399 and X299 supports 16x/8x/16x/8x in 4 slots for instance (differenmt combo's of this as well) simultaneously on PCIe 3.0.

[foldy]

This approach can do 6 GPUs with m.2 to pcie x4 adapters.
viewtopic.php?f=38&t=31991

See also these discution
viewtopic.php?f=38&t=32233

[scerbera]

Thanks foldy

Gigabyte RTX 2060 Super Gaming OC Windforce
PCIe 3.0 @ 16x
Windows 10 1909

Device - Geforce RTX 2060 Super
Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 86.0815
Scaled Score - 259.764
Atoms - 64614

Similar card as above ( MSI GeForce RTX 2060 Super Ventus GP OC) run at 67% power limit (just 125w) core oc 90
Score - 71.7368
Scaled Score - 216.477
Atoms - 64614

[bruce]

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

[NoMoreQuarantine]

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I haven't actually taken a look at them yet. I'm currently attempting to refine my process before I continue. Will post an update soon.

[foldinghomealone2]

Maybe the new FahBench doesn't work as expected:
viewtopic.php?f=38&t=24225&start=180#p325860

[tulanebarandgrill]

Looks from that like a pair of 5700s should really cook.

[JimboPalmer]

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I haven't actually taken a look at them yet. I'm currently attempting to refine my process before I continue. Will post an update soon.

I have a low profile no extra power GTX 1650, I get 320k PPD. A full height with better fans will be faster.

Previous to that, Low Profile GTX 1050 ti was 150k PPD, so the 1650 was a real step up.

[chabgood]

I am seeing 2.75M on my 2080 supers, running on ubuntu.

[Endgame124]

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I have a 1660 super - completely stock in a well ventilated case, it’s pulling 700k - 800k ppd in Windows. I highly suspect that I’ll be able to dramatically down clock the ram to save power and increase points per watt. I would really like to get a comparison with someone running a regular 1660 to see if they get points in the same range. If so, we can validate that the much lower memory bandwidth of the normal 1660 doesn’t make a difference in ppd

[ptchernegovski]

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I haven't actually taken a look at them yet. I'm currently attempting to refine my process before I continue. Will post an update soon.

I have a low profile no extra power GTX 1650, I get 320k PPD. A full height with better fans will be faster.

Previous to that, Low Profile GTX 1050 ti was 150k PPD, so the 1650 was a real step up.

Is that a normal 1650, or a super? I am deciding between 5500 XT and a 1650 currently

[BobWilliams757]

I have a low profile no extra power GTX 1650, I get 320k PPD. A full height with better fans will be faster.

That seems like a really solid return for a 75W card that can be had for that kind of price.

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I have a 1660 super - completely stock in a well ventilated case, it’s pulling 700k - 800k ppd in Windows. I highly suspect that I’ll be able to dramatically down clock the ram to save power and increase points per watt. I would really like to get a comparison with someone running a regular 1660 to see if they get points in the same range. If so, we can validate that the much lower memory bandwidth of the normal 1660 doesn’t make a difference in ppd

Same here, great bang for the buck. I'm curious to see where you can go with the power and points on this card as well.



Have either of you two used the new FOH Bench that Foldy posted? I'm just trying to get an idea how in line it is with PPD results.



Both the 1650 and 1660 Series cards look like a good value. It's interesting what they have done with the 1650 Super, it's more kind of a 1660 light in some respects. That is another card I'd be interested in seeing some real results on. Power wise it's at only 100W, so almost right between the two cards you two are giving numbers for.

[Endgame124]

I have a low profile no extra power GTX 1650, I get 320k PPD. A full height with better fans will be faster.

That seems like a really solid return for a 75W card that can be had for that kind of price.

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I have a 1660 super - completely stock in a well ventilated case, it’s pulling 700k - 800k ppd in Windows. I highly suspect that I’ll be able to dramatically down clock the ram to save power and increase points per watt. I would really like to get a comparison with someone running a regular 1660 to see if they get points in the same range. If so, we can validate that the much lower memory bandwidth of the normal 1660 doesn’t make a difference in ppd

Same here, great bang for the buck. I'm curious to see where you can go with the power and points on this card as well.



Have either of you two used the new FOH Bench that Foldy posted? I'm just trying to get an idea how in line it is with PPD results.



Both the 1650 and 1660 Series cards look like a good value. It's interesting what they have done with the 1650 Super, it's more kind of a 1660 light in some respects. That is another card I'd be interested in seeing some real results on. Power wise it's at only 100W, so almost right between the two cards you two are giving numbers for.

I’ve been wanting to try the new bench the next time I see the machine without a WU. However, that machine has been running every time I look at it, so I’ve been letting it go.

[iceman1992]

And what about the GTX 1600 series. FAH gains nothing from the extra features in the RTX series ... so unless your into AI or ray-tracing....

I have a 1660 super - completely stock in a well ventilated case, it’s pulling 700k - 800k ppd in Windows. I highly suspect that I’ll be able to dramatically down clock the ram to save power and increase points per watt. I would really like to get a comparison with someone running a regular 1660 to see if they get points in the same range. If so, we can validate that the much lower memory bandwidth of the normal 1660 doesn’t make a difference in ppd

I have a regular GTX 1660. Getting 500-630k PPD (More frequently hangs around 540k). I'm not sure if that's actually low or expected for this card. Anyone else with a 1660?

[NoMoreQuarantine]

I decided to take a look at the actual specs of the GPUs. Shown below are the NVIDIA Turing GPUs ranked by price. I'm going to do the same with Navi GPUs in a separate table. Since Navi GPUs seem to not perform as well as Turing GPUs for FAH, my hope is that by keeping the comparison apples-to-apples I can give some more accurate guidance on what kind of performance a user can expect.