18454 BioFilm PPD issues

[cbigfoot]

im checking to see Is there a reason the PPD Gets Decimated when we get assigned Biofilm Reaserch? my 5700xt and 5600m seem to get their PPD cut down from 1.5-1.9mPPD and 6-700K AVG to ~900K and 300k
5700xt
18454, 19, 21,110
18454, 7, 39, 55
18454, 7, 3, 96

5600M
18454, 11, 11, 61 Due to complete ~11:48 PM CST 11/6/2022

[BobWilliams757]

It's a small atom count project compared to most. Often the more powerful GPU's with more cores/shaders can't be utilized to their full potential due to this, and suffer a PPD hit. The bandwidth of the memory and clock speeds help, but they simply can't push a higher end GPU. On the inverse side, often GPUs that are older or less powerful overall see a boost in PPD with small atom counts. They can be fully saturated and used, and as such their efficiency goes up.

If you use HFM to monitor your progress you can often see trends if you look closely at the data it keeps. You can also use sources like LAR systems, and see the averages for various hardware.

https://folding.lar.systems/projects/

You can click on any project number and see averages for any specific project. Keep in mind they deal with averages of input, so it may include folders that don't fold 24/7, power limit hardware, as well as those overclocking their hardware to the limits. But the numbers should give you an idea of how other hardware is or isn't impacted.

[cbigfoot]

Thx very aware of Lars. HFM is how i spotted the trend on my side just thought that it was strange that it would take so long and produce such little Points. mostly worried that people would be rejecting them and trying other units to keep their PPD up. sadly, the way they group the GPUS with AMD i tend to ignore the data as my 5600M AND MY 5700xt are in the same group and 2 different beasts Nvidia however I take it as the bible...

[BobWilliams757]

There is currently work underway to lessen how broad the AMD species assignments work. Though it likely won't be perfect, it might bridge the gap on some of these projects that are assigned to powerful cards often. But really with the broad range of hardware out there, it appears that variances of 200-300% are not really uncommon. They still have to get the work done, and sometimes we get less an optimal projects, but it's better than not getting projects. And I do understand the pain, I folded for a couple years on a iGPU Vega 11, and due to species assignment it was often folding projects at 50-70k PPD and barely meeting the 1.5-2 day deadlines. But that little thing would thrive on these small atom count projects, getting up in the 250+K range at times.

As for the people cherry picking, it's sad that it happens. Unless it's something beyond my control I just fold what I get and accept the roll of the dice. I hope that most others do as well. Those so engrossed in the points contest that they slow the science down should try to look more at the big picture. Hopefully those types are few and far between.

[muziqaz]

p18454 was never meant to be sent out to anything above 5700x. It was supposed to be limited to slower cards. We are still waiting for researcher to notice our requests. Limitation probably has never been applied correctly :/

P.S. 900k on 5700x is only 20% lower than standard average for that card. But hopefully researcher will apply our suggestions soon