Folding Forum

Hi,
I've noticed my GPU folder slot has been doing solid 8018's for the last 2 days. The GPU shows it is back to running FahCore 0x15 after running 17 in advanced mode for two months without any problems. The configuration still has the client-type advanced flag. It seems to be a problem with the project rather than the client. Is the 8018 project limited to running only on the 15 core? Can it run in 17? I've noticed my PPD for that slot fell from 100-120K PPD down to 27K PPD. Looking at my team I noticed many people's PPD seem to have been cut in half lately so it looks like this is effecting lots of people. If the project can only run on 15 then no worries - I read the description and it is interesting science - glad to contribute!
Thanks very much,
Telecentricity

client type advanced is only a preference setting that says you are willing to fold late stage beta work units and accept that additional risk of WU instability. This is not a guarantee that the client will download a specific project. You could get an advanced WU with core 15 just as easily as an advanced WU with core 17. Ultimately, Pande Group decides which projects get folded in what order.

Each project # runs on a specific core and cannot be changed.

client-type = advanced does not guaranty a Core_17 project nor does It even guaranty an advanced project. PG sets a probability (i.e. priority) for every project and then all the advanced projects compete with each other for assignment. So you can get an advanced Core_15/16 project easily enough. If they run out of advanced projects then you will be shifted to a general release server to get your WU i.e. no Core_17 projects.

All that client-type = advanced is is an acceptance of risk that a specific client/core/project will fail. Projects are internally tested and if the failure rate is within acceptable levels then they go to beta testing where the public can comment. There may be modifications or travel backwards all the way to internal but If they still do not have too many problems then they go to advanced and finally general release. Core_17 has been travelling down this path towards general release.

Core_11/15/16 have all traveled down the same path and have been running under general release for a long time. They are quite stable. For high-end GPU's there will be a large PPD hit running any WU's requiring them because they do not contain QRB (Quick return bonus points) which give a large amount of points for quickly returning a completed WU.

Every project has a specific core (and minimum version of a core) that it will run under and those cores are not interchangeable so no a p8018 can not use Core_17.

One thing of note for Only AMD GPU's: Core_17 needs a very recent video driver that contains OpenCL however earlier cores only work well with AMD drivers v12.8 or earlier. That may be a problem for all you AMD video card folders.

In any case, P8018 is a poor producer, at least on a low end Fermi like my GT430 - only 3500 PPD. As for other Core_15 projects, 7624-7626, 8054 and 8074 have produced about 6750 PPD. P7660 has produced a little less, about 5500 PPD. P8018 seems to be a new low as far as core_15 projects are concerned.

7im and P5 thanks very much for your quick reply. I didn't know specific projects were tied to specific cores. I've found the FAH forums to be very friendly and helpful, and I've always learned something when I've asked questions.
Thanks again,
Telecentricity

I think it was Joe_H who mentioned it yesterday, but is there an issue with the server assigning Core_17 work units? Something about being in "accept" mode and not assigning work?

Either way I'm not overly concerned as long as there is work to do, but it sure feels better when my GPU slots crank out 5x the scientific value per watt used on Core_17 versus Core_15

NookieBandit wrote:...snip
...but it sure feels better when my GPU slots crank out 5x the scientific value per watt used on Core_17 versus Core_15

Only the points are different. I very much doubt the amount of science your GPU can do has changed in any way.

NookieBandit wrote:I think it was Joe_H who mentioned it yesterday, but is there an issue with the server assigning Core_17 work units? Something about being in "accept" mode and not assigning work?

Apparently that was a temporary situation. It's not true now.

Thanks 7im, like I said I'm indifferent as long as there are work units needing processing. At the same time, if there is a problem with a server assigning Core_17 work units, it would be nice to get an ETA on the fix.

I certainly agree with you on the amount of science my GPUs are doing has likely remained unchanged, but the value of the science hasn't. As I understand it -- and I could be wrong -- the QRB is intended to provide an incentive to process work units promptly, which infers some notion of the old adage, "time equals money". If true, then the shorter time period to return work units equals increased scientific value. My (minor) angst is centered on not being able to fully leverage the scientific value of 16 high-end GPUs because they're otherwise occupied on non-QRB work units, yet pulling just as much power as if they were, hence my comment "scientific value per watt used". Again, not a complaint, simply an observation

I understand what you are trying to say, but it is a misconception. Only points per watt has changed. Scientific value has not changed. WUs are not getting done slower, just differently.

bruce wrote:

NookieBandit wrote:I think it was Joe_H who mentioned it yesterday, but is there an issue with the server assigning Core_17 work units? Something about being in "accept" mode and not assigning work?

Apparently that was a temporary situation. It's not true now.

Yes, I also suggested in my post that people check back later in the day yesterday, as it was early in the morning at Stanford - 7:30-8 AM.

7im wrote:I understand what you are trying to say, but it is a misconception. Only points per watt has changed. Scientific value has not changed. WUs are not getting done slower, just differently.

Remember when WUs earned you usually one (1) point, and occasionally even half (.5) of a point? 3 or 4 PPD was cranking 'em out. We're mighty spoiled these days, but some of us just don't know it yet. No offense to the younger folk, just trying to help you see things a little differently.