teamanandtech.org

Hey, I was lurking over at AT and found out about the race, will there be any individual stats, or team stats only?

crashtech wrote: ↑Thu Mar 19, 2020 9:16 pm Hey, I was lurking over at AT and found out about the race, will there be any individual stats, or team stats only?

So far there are only Ken's team stats, http://tastats.atspace.cc/ta-th-20/.
If anybody wanted to do user stats, that would be an Herculean effort this time. :-)

Yeah, there's several hundred active users from both teams!!
Tony's weekly stats will have a very long list of FAH contributors.

I'm going back to SETI. I don't think the TeAm needs my help beating Tom's, and F@H is getting plenty of compute power, maybe more than they need.

crashtech wrote: ↑Fri Mar 20, 2020 11:17 am I'm going back to SETI. I don't think the TeAm needs my help beating Tom's, and F@H is getting plenty of compute power, maybe more than they need.

F@H definitely has more contributor power than they can properly use right now. But so does SETI@home, where half the world is fighting for the last tasks while the database has been left in a bad shape since earlier this year.

StefanR5R wrote: ↑Sat Mar 21, 2020 6:48 pm

crashtech wrote: ↑Fri Mar 20, 2020 11:17 am I'm going back to SETI. I don't think the TeAm needs my help beating Tom's, and F@H is getting plenty of compute power, maybe more than they need.

F@H definitely has more contributor power than they can properly use right now. But so does SETI@home, where half the world is fighting for the last tasks while the database has been left in a bad shape since earlier this year.

True. My decision is purely selfish. I neglected SETI for years; it's rather frustrating that the year I decided to catch up in it is the year they decide to go dark...

Well, you are keeping company with the project and its fans during the weeks towards the project's conclusion. That's a good thing to do; enjoy it. :-)

I'm all in with Folding at the moment, with everything that makes any sense to throw at it, CPUs and GPUs alike.

Unfortunately, it seems like the CPU WUs for COVID-19, like 14550, yield extremely poor credit. So, I'm am going to leave them to those who only care about the science.

14550 is the only problematic WU that I have seen, but it's a full time job to eliminate them. Stopping the receipt of all beta work would seem to be the only solution at the moment.

Edit 1: It sure would be nice to be able to exclude 14550 from running on my clients. I don't see a way to do that without excluding all beta WUs.

Edit 2: It seems they may have stopped distributing 14550, as of 6 April 1900 UTC I am no longer having to dump them. What a relief!

Earlier this week I stashed some Rosetta tasks away, to be worked on until the FB sprint starts and after the sprint. With the sprint cancelled, I took part in the PrimeGrid challenge instead, but went on to finish the Rosetta cache afterwards.

The cache did not suffice until the new Rosetta work appeared on the server, hence I did the same as you and switched the CPUs to Folding@home. Unlike earlier, I had to fight a bit with CPU slot thread count on the Intel computers which have core counts which are a multiple of 7, or worse, of 11. I knew from our past Winter races that this can be a problem at times. (WUs error out right away because domain decomposition fails. The CPU slots need to be reconfigured to a thread count then which is a multiple of 2 and 3 at most.)

I got many of these 14550 units too, which with their low TPF and absolute ridiculous credit became a huge nuisance. Therefore I was glad to find Rosetta back in business today in the morning and switched the CPUs back there before I went to work.

According to responses at the Folding@home forum, credit including QR does reflect the value which the work has for the scientists. Hence, that 14550 project must have had practically no value to them. It's a mystery why they blocked precious resources with it anyway.

Re the domain decomposition failures, I've seen the client automatically reconfigure thread count downward. My one attempt at over-subscribing (to hopefully land closer to full utilization) on a GPU-less machine was blocked by the client. Maybe that behavior could be tweaked in a config file, but it's not clear that it would be worth the effort. Regardless, since F@H work flow seems to be fairly constant, some of the CPU units essential, and the problematic WUs being eliminated, I will continue full bore on F@H for a while longer.

I'm ramping down participation in F@H, if I see an Nvidia GPU idle, it's getting switched to GPUGrid.