Question Vega 64 and 5930k DC performance

[Shmee]

Hey, so I have a 2nd desktop with a Vega 64 Nitro+ and an i7 5930k @ 4.4GHz, and 16GB of RAM if that matters. I am wondering, would it be worth folding with this? How is the PPD on a Vega 64? Any specific driver set or clocks / settings I should use? Thanks. I have used the Vega 64 for mining ETH before so I am not a total noob to DC on it. Also WAAAY back I did BOINC on 5870s.

 

[borandi]

Folding@home GPUs PPD Price Watt - Google Drive

docs.google.com

Vega64 ~ 800k PPD.

 

[Shmee]

Ok cool, that's pretty good right?

 

[Assimilator1]

It's a decent amount of points, certainly worthwhile , although the high end Nvidia cards go way over that.
My little RX 580 does ~400-450k ppd for comparison .

As for the CPU, the ppd you'd get from that for F@H will be very low, if you want to make good use of it I would get them to run Rosetta (via BOINC) instead (which is also doing Covid 19 research).

Btw, note that both projects are struggling to keep up with the WU demand, but you will get some sooner or later!

 

[StefanR5R]

Btw, note that both projects are struggling to keep up with the WU demand, but you will get some sooner or later!

To add to that: Since the recent media coverage and quick increase of contributing individuals and organizations, the Folding@home servers currently have less storage bandwidth than they could use in order to send work and receive results fast enough to/from everybody. Rosetta@home on the other hand enjoyed some attention and increased participation too, but not as much by far. Their servers are still keeping up with the traffic (last time I checked, which was a few hours ago). Rosetta@home sometimes has periods without work after some work batches were finished and until new batches are prepared and put on the server, e.g. during the past couple of days.

 

[Assimilator1]

Forgot to mention earlier, for Rosetta we can increase the default task time (8hrs?) to 24hrs or more, I don't know if going to the upper end occurs any kind of penality (not that I know of), I've set mine to 24hrs & will be monitoring ppd (bear in mind atm I'm also running a little LHC@H as I d/led some WUs by mistake ).
This reduces the amount of WUs you need to d/l of course .

Further to what Stefan said, F@H had about a x20 (yes twenty!) increase in new participants! Whilst with Rosetta it was x2-3.

 

[Shmee]

Well I am getting about 900k estimated PPD on the Vega, so seems to be doing ok so far. I even undervolted it to 1050mV on Pstate7, and saved about 60W!

 

[Assimilator1]

Nice!
And I was thinking that my RX 580s drop to 1050mV, saving 20w was good! 60w is an amazing saving! What is its default voltage?

 

[Shmee]

1200mV according to wattman.

 

[blckgrffn]

Well I am getting about 900k estimated PPD on the Vega, so seems to be doing ok so far. I even undervolted it to 1050mV on Pstate7, and saved about 60W!

How many watts did it decrease from? My 5700xt is between 1M and 1.1M PPD and running about 160 since I dropped a -10% power limit on it.

Automated undervolting showed that I don't likely have a very good sample for cutting voltage.

 

[Shmee]

Its running about 160W now, before it was around 210-220W when folding at stock settings.

This is according to the wattman software for the card only. I have not tested it in games with this undervolt, but so far it seems stable in folding.

 

[Shmee]

I just checked the Vega power usage again, now with a different WU it is using less! about 110-130W. Likely this WU is not as demanding.

 

[Leeea]

My Vega 56 generates about 550-750 points per 19 hours. ( I have tiered rate electricity, so I do not run it in the expensive time )
I think that would be 800ish ppd. It tends to be all over the place.

It sits right around 140-150 watts, but I have the fan speed turned down so it does get very hot and thermal throttle. Rest of settings are default.


Folding @ home has specific covid-19 cpu projects that cannot be run on GPU, so I have been running those once in a while also.


I also have client-type: advanced and next-unit-percentage: 90 which helps a bit to.