Top GPUs for Folding@Home

[Juggy]

In the beginning upload once per day. But as soon as we have several samples a longer period would be better.
But it's up to you.
As long as you don't send me every WU separately

Perfect, I will submit another sample in the morning (7 hours from now) as I have already done another 3 with the 2080.

The crappy thing is I have a 2060 Super as well but cannot run it due to there not being enough space in my mATX build. The 2080 Super is 10mm thicker than the 2060 Super so I cannot run them side by side on my MSI Z390M

[foldinghomealone]

I uploaded my hfm.net results to the database.

Did you? I can't find the data...

I just uploaded a new one. Hopefully, it is their this time.

Found it and updated the db

[foldinghomealone]

/quote]

The crappy thing is I have a 2060 Super as well but cannot run it due to there not being enough space in my mATX build. The 2080 Super is 10mm thicker than the 2060 Super so I cannot run them side by side on my MSI Z390M

Did you try to use a PCIe extension cable (aka riser cable) to put the 2060 Super somewhere else?

[Juggy]

Did you try to use a PCIe extension cable (aka riser cable) to put the 2060 Super somewhere else?

My case is mATX as well so no space.

NZXT H400i

[Juggy]

I uploaded my hfm.net results to the database. It only has 8 samples. I've had to baby sit my computer all day just to get those 8. I'm still struggling to get WUs.
The results are for my EVGA RTX 2080 XC Black. The card is overclocked and running a hybrid liquid cooler.
__________________________________________________________________________________________________________________________________________

GPU - 2025 to 2050 MHz depending on GPU boost 4.0
Memory - 8000 MHz
Voltage - 1050 mV
Power Target set at - 136%
GPU Temp - 52 C

Is this a standard 2080 or a Super?

If it's a standard 2080 you're getting pretty good results.

[foldinghomealone]

If it's a standard 2080 you're getting pretty good results.

I wondered as well and therefore I named it 'RTX 2080 OC' (for OverClocked)

[Spongebob25]

If it's a standard 2080 you're getting pretty good results.

I wondered as well and therefore I named it 'RTX 2080 OC' (for OverClocked)

The card is Over Clocked but is a standard non super.

https://www.techpowerup.com/gpu-specs/e ... lack.b6241

[bruce]

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.

If you want a machine without a WU, use the FINISH command in FAHControl or the STOP FOLDING option of WebControl. Both options are supposed to provide the option to continue folding but to not download a new WU.

The other choice is to PAUSE the active WU(s) and run FAHBench, then unPAUSE.

[Endgame124]

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.

If you want a machine without a WU, use the FINISH command in FAHControl or the STOP FOLDING option of WebControl. Both options are supposed to provide the option to continue folding but to not download a new WU.

The other choice is to PAUSE the active WU(s) and run FAHBench, then unPAUSE.

Thanks for the tip. I’ve been using finish here and there, but the big motivation for not stopping the in progress host was the Anandtech vs Tom’s Hardware points race. Now that it’s over, I’ve been much more liberal with pausing / finishing.

[foldinghomealone]

We're still looking for HFM.net exports to increase sample size for the PPD database:
viewtopic.php?f=38&t=34510

Would appreciate if you could upload your HFM.net export.
If you have questions, ask

[gordonbb]

Are you basing your list on theoretical performance rather than measured results?

PassMark DirectCompute results.

GTX 1660Ti 700kPPD
RX5700XT 963kPPD
RX5700XT 1.01MPPD
RTX 2060 1.1MPPD
RTX 2060S 1.2MPPD
RTX 2070A 1.3MPPD
RTX 2070S 1.85MPPD
RTX 2070S 1.88MPPD

Interesting, you are measuring major differences even between the two 5700 XTs. When you say they are running stock, are they running at the recommended NVIDIA/AMD clock speeds? A lot of cards come factory OCed by the videocard makers and have to be told to operate at standard clock rates. Do all cards have max RAM operating at recommended frequencies?

These cards are operating at the stock settings provided by the manufacturer and so have higher boost clocks than the reference models.

These are actual F@H results from FAHClient logged once a minute to a Zabbix server and averaged over several days with whatever current WUs are in the Beta work queue.

Normally a cards single precision TFLOPs value as calculated based on the manufacturers specs should give you a good idea of the relative performance and do so in the case of NVidia but AMD Navi are performing in the real world much lower than their specifications suggest they should.

IMO the overclock.net PPD database is almost useless these days since the maintainer removed viewing of the individual data points and so you are limited to just an aggregate value so you don't know:

What Core is being used. Many of the results are based on Core21 and Core22 has dramatically changed performance for Windows and a lesser extent for Linux
What OS was used
What the graphics clock was running at (i.e. was this a result from a under-clocked or over-clocked system)
What project was this on. There is a wide variation in PPD by project. IMO the best PPD result should be an average taken over at least 3 days of continuous WUs.

[foldinghomealone]

@gordonbb:
If you want to have a short look at our new database
viewtopic.php?f=38&t=34510

It shows most of your points, however clock speed and OS is not incorporated
- clock speed can vary from WU to WU and even during the same WU, so I don't know how anyone should know which is the correct clock speed
- there is not enough feedack on which OS was used. IMO, as a fast GPU model of the same type under Windows makes more PPD than a slow GPU under Linux it is not really important. Otherwise PCIe connection would be equally important.
To keep it simple I would calculate 5 to 10% more PPD for linux.

We are in contact with Ryan from HFM.net to discuss more HW and OS information for the HFM export.

[Endgame124]

@gordonbb:
If you want to have a short look at our new database
viewtopic.php?f=38&t=34510

It shows most of your points, however clock speed and OS is not incorporated
- clock speed can vary from WU to WU and even during the same WU, so I don't know how anyone should know which is the correct clock speed
- there is not enough feedack on which OS was used. IMO, as a fast GPU model of the same type under Windows makes more PPD than a slow GPU under Linux it is not really important. Otherwise PCIe connection would be equally important.
To keep it simple I would calculate 5 to 10% more PPD for linux.

We are in contact with Ryan from HFM.net to discuss more HW and OS information for the HFM export.

HW and OS information would likely have some decent value, as I know my A10-5800K is starving my 1660 Super even with 1 dedicated GPU and someone else with a different CPU may get better results. There is likely a certain point where a single CPU feeding a GPU is enough, but I don't really know where that is. I would also really love something like a splunk agent, where I set it up on my client PC and it just uploads the logs to the central repository as soon as they are generated - no user interaction required after initial install.

[MeeLee]

The crappy thing is I have a 2060 Super as well but cannot run it due to there not being enough space in my mATX build. The 2080 Super is 10mm thicker than the 2060 Super so I cannot run them side by side on my MSI Z390M

You can always use a ribbon cable to run the 2060 Super outside your case.
use a 6 pin extension cable from your PSU outside the case, and place the RTX 2060 on top of your PC, or besides it (where ever it is safe).

Even on a PCIE x1 slot, you'll still run about 75-80% of the stock performance on a x16 slot. An x4 slot or greater is preferred though.

[NoMoreQuarantine]

I would also really love something like a splunk agent, where I set it up on my client PC and it just uploads the logs to the central repository as soon as they are generated - no user interaction required after initial install.

We all would. The issue is that we are a couple of non-programmer volunteers trying to get something done with limited time & resources. That said, I am persistent so if anyone with experience has advice on how to implement such a solution I am all ears.