Folding Forum

I decided to see how extreme I can take my current folding rig in terms of performance/watt. I have spent some time to get the combined power usage of my CPU & GPU to about 160W (AKA less than 2 incandescent light bulbs). I lost about 19% performance on FAHBench with my GPU, while reducing my power draw by 50%. I have it folding now and will post results.

New HW entry: i7-8550U

New entries for:
i5-8400
i7-2600K
i7-8550U
R5 2600
RX 580
RX 590
RX Vega 64
Xeon E5649 @18t

Thanks to NoMoreQuarantine and AOD_N3URAL for the uploads.
I really appreciate the naming of the slots

Juggy wrote:

HaloJones wrote:I think to try to have all the possible variables accounted for is too much complication. OS, motherboard, cpu, gpu type, card maker, cooling spec, overclock settings, cooling, usage of machine while folding, Project, Run, Clone, Gen.

All this will do (which is still a lot) is give a range for a card that a potential buyer/user could expect.

Hmmm, for instance an OC card from XXX manufacturer may be $400 but from YYY may be $370.

There are lower priced manufacturers like Galax that still produce good kit.

Card manufacturer (Gigabyte)
Model Number (RTX 2080 Super)
Sub Model Number (Windfrorce OC)
Factory Clock Speed (1830MHz)

Surely that is not too much info?

If someone can give information in the CSV file like 'RTX 2080Ti @50% power limit' or sth. like this I will add this information in the new column Z.
When we have enough samples of such detailed information I will consider making an additional Table.
Currently there is not enough data/samples to do so.
I also think that manufacturer or Model number is too detailed. I think that clock speed would be enough information.

Clock speed is a variable anyway. Both cpus and gpus auto-boost these days so without having it fixed, the clock frequencies would be all over the place.

HaloJones wrote:Clock speed is a variable anyway. Both cpus and gpus auto-boost these days so without having it fixed, the clock frequencies would be all over the place.

That's true. Didn't think about that.

HaloJones wrote:Clock speed is a variable anyway. Both cpus and gpus auto-boost these days so without having it fixed, the clock frequencies would be all over the place.

That said, there will be an average clock speed when under full load, depending on ambient temperature and cooling solutions. A single datapoint just isn't enought to fully characterize a processor. That would require a sweep of frequencies at various temperatures to map the 3d spectrum.

Good luck trying to get an accurate report of the average clock speed over the course of a work unit! Mean, mode or median?

HaloJones wrote:Good luck trying to get an accurate report of the average clock speed over the course of a work unit! Mean, mode or median?

Good point, that would require an all new process to monitor the clock speed during each WU. I was being hypothetical, not practical

I uploaded a new dataset. Couple of older/slower cards mixed in this one, plus a bunch of the GTX970.

Kebast wrote:I uploaded a new dataset. Couple of older/slower cards mixed in this one, plus a bunch of the GTX970.

oops, I messed up a slot name. Anything that says Ubuntu_GTX970 should say Ubuntu_GTX970-1442Mhz

@Kebast:
no worries.

I updated the db with following entries:
i3-4130 @2t
i5-8265U
i7-6700HQ @7t
R3 1200 @3t
RX 460
RX 570
GTX 970
GTX 1050

Thanks to NormalDiffusion:
Data from ... uploaded:
Titan Xp
GTX 1050Ti
GTX 1650Super
GTX 750
GTX 960
i7-4770
i7-6700
i9-7900X
Xeon E5-1620v4
Xeon E5-1650v4
Xeon E5-2690

Thanks to NoMoreQuarantine and Juggy
New updates for:
RTX 2080 Super
RX 590
Intel i9-9900FK
Ryzen 5 2600

foldinghomealone wrote:Thanks to NoMoreQuarantine and Juggy
New updates for:
RTX 2080 Super
RX 590
Intel i9-9900FK
Ryzen 5 2600

My mistake, I labelled the slot incorrectly. It should be i9-9900KF @12t and replace the existing dataset you have

Apologies.