Get Ready for Folding Core 22 WUs

[Gorgon]

Stanford released the new OpenMM Core22 from Beta to Advanced last week with some interesting results.

 

Just as a recap the Folding Client just enables you to connect to Standford. When Stanford assigns you a Work Unit (WU) the client downloads the core (executable) as well as the raw WU. The WU is run on the appropriate Core.

 

Current GPU WUs, like the 117XX WUs that recently have stopped being issued and the newer 14XXX WUs we should all now be getting, use OpenMM Core version 21.

 

There is an OpenMM version 22 core that has been in Beta for a while and last week it was released to Advanced so those of you running with the Expert Option "client-type" set to "advanced" may have run some of these WUs. When they work they produce a significant bump in points.

 

However, they have been pulled from "Advanced" and relegated back to "Beta" due to some issues:

1. They tend to drive GPU utilization to 100% and on marginal rigs can cause the OS to become unresponsive
2. As the Utilization goes up the Power Consumed also goes up and if your near your Power Supply's capacity you may start getting UPS alarms or glitches due to the PS voltages drooping.
3. The utilization can be wildly variable and during low utilization on NVidia Cards the Boost algorithm may boost the card up into instability resulting a failed WU. This has happened even to users with no overclock applied but using Add-In Board partner (AIB) cards with factory overclocks.

So please be cognizant that these will be coming eventually and for those of us running overclocks we may need to remove the overclock and/or power-limit the cards back down to 100% so they run stably.

 

Note that if you are not running "client-type" "advanced" or "beta" you will not have seen this but presumably after some more investigation they will be re-releasing the new OpenMM Core22 back to advanced and eventually to "Production"

 

Note, there is currently no points advantage running "advanced" rather, as this shows, a disadvantage as you are far more likely to get WU failures.

[RollinLower]

9 hours ago, Gorgon said:

When they work they produce a significant bump in points.

 

i feel like we deserve a bit of a point boost after what the 14xxx WU's have been doing to my average PPD.

[Gorgon]

8 hours ago, RollinLower said:

 

i feel like we deserve a bit of a point boost after what the 14xxx WU's have been doing to my average PPD.

I’m runnng tests currently at lower Power Limits to see the change in efficiency of my rigs so though I’ve felt the overall PPD has gone lower that I’d have expected it to it is good to hear that others, presumably running at fixed values, have noticed a decrease in their daily production.

I have, however, also been collecting per WU statistics as a comparison to the weekly average values available from Extreme Overclocking and Stanford and have noticed that like the 117xx WUs some have better yields than others.

There was a 14xxx WU that was particularly brutal but strangely at the start of the week before last it appears to have had its base points adjusted to a value more in-line with its work done.

[Guest]

 

These WUs have been wild in terms of PPD. (Ignore the ones lower than 6K).

Substantially higher with the beta WUs.

 

 

[RollinLower]

On 3/19/2019 at 8:02 PM, King Poet said:

Substantially higher with the beta WUs.

would you recommend going beta for PPD? or is the potential instability too bad to have my only client doing beta WU's

[Guest]

4 minutes ago, RollinLower said:

would you recommend going beta for PPD? or is the potential instability too bad to have my only client doing beta WU's

Maybe if you're cold lol.

Noticed a bit more heat.

 

[RollinLower]

2 minutes ago, King Poet said:

Maybe if you're cold lol.

Noticed a bit more heat.

 

so basically no instability then? the heat wouldn't really be a problem i recon, just turn down the overclocks a bit.

[Guest]

6 minutes ago, RollinLower said:

so basically no instability then? the heat wouldn't really be a problem i recon, just turn down the overclocks a bit.

I haven't run OCs on the the beta units. It's such a boost in points that there's really no need. My dual 1080 Ti rig jumped 500k PPD and my 2080 Tis are both doing 3M regularly.

[RollinLower]

1 hour ago, King Poet said:

I haven't run OCs on the the beta units. It's such a boost in points that there's really no need. My dual 1080 Ti rig jumped 500k PPD and my 2080 Tis are both doing 3M regularly.

my Titan x just got a beta WU, and my PPD got a pretty substantial boost already! now i'm just waiting to see what the final difference would be once my 1070 gets a beta WU aswell.