Implications of FahCore_17 on Current GPU WUs

[7im]

Or don't be good, and risk getting your points zero'd out and your user name permanently blocked. Wouldn't be the first time!

[Xavier Zepherious]

BTW, this is what proteener has to say about the current GPU Projects (excluding FahCore_17):

If donors dump FahCore_15/FahCore_16 WUs, it just means that the same WU would have to be done multiple times which will slow down the completion of the Projects and possibly hurt the researcher(s). Something that must be avoided by us donors. If we all folded WUs from previous FahCores without dumping, then it is quite possible that those Projects can be finished quickly which means that those FahCores can be retired early

Just how many people are gonnna bay their system 24/7 to get prime WU's on rigs with multiple GPU's

with bigadv they take days so you can sit and wait on them...one could do this with little effort

with Wu's that take a few hrs - and some would be possibly completing GPU's WU's every hour...Whose gonna stick around to drop and pick up a new one- and possibly drop again until you get one you like

I sure am not - not worth the time or the loss of sleep
(I wouldn't anyways - science first)


as others have noted you could also make the completion success % more stringent - like 90% - so you would want to avoid as many errors as you can
and include GPU in the success rate

this would discourage dropping Wu's and would probably get those that OC (and fail a few WU's too many)... to reduce clocks to get more stable systems

[bruce]

As others have noted you could also make the completion success % more stringent - like 90% - so you would want to avoid as many errors as you can and include GPU in the success rate.

this would discourage dropping Wu's and would probably get those that OC that drop a few WU's too many... to reduce clocks to get more stable systems

Maybe the QRB needs to be variable, based on the success rate. Rather than cutting it off ad 80%, they could award QRB/X where X starts off at 1.0 with 100% and increases according to some schedule that progressively reduces QRB -- say to half (X=2.0) at 90% and still keep the present QRB=Zero at 80%.

As long as failures are something the Donor can control, that could be made to work. Unfortunately there's still the occasional BadWU and no matter how perfect the PG wants their part of the system to be, "stuff" happens. Stanford has been known to correct missing baseline points when it's clear that it was beyond the control of the Donor, but the've never made any adjustments to bonus points and I don't expect they ever will. I suspect that the information required to recompute the QRB manually isn't retained anywhere so they can't, even if they wanted to.

[Bleeder]

Would it not be possible to recompile the Projects to work with Core_17? I have no idea how Projects are created but I would guess that there is a database holding the actual work and that work is compiled into WU's in some automated way. Maybe it would be fairly simple to recreate the Core 16 and 15 Projects into Core 17 Projects?

[7im]

It might be possible, but unlikely to happen for scientific consistency. Wouldn't want to introduce any potential for data integrity to be compromised.

[Jesse_V]

Bleeder, proteneer basically answered that question on the Reddit AMA.

[7im]

Bleeder, proteneer basically answered that question on the Reddit AMA.

Great. What did he say?

[bollix47]

http://www.reddit.com/r/Folding/comment ... eloper_at/

Unfortunately the old core is tied to the GROMACS way of doing things. Although it's possible, it'd involved a bunch of work done by the researchers in the PG group.