SG BOINC Pentathlon 2021

[velociraptor]

Should I throw my RTX 2070 at primegrid or Einstein?
Greetings vel

[dannyridel]

That's why I advise to run PrimeGrid as GPU project, so you can use your hyperthreading for the other projects in the Pentathlon. And while Wieferich & Wall-Sun-Sun may favor Turing GPUs it does quite well on the 1600 series t00.

1600 is Turing too.

[dannyridel]

PrimeGrid certainly! At least in my view. @velociraptor

[velociraptor]

Will do so @dannyridel . Awesome, more bars stops at the city run than the field run.

[vaughan]

Yes I recommend PPS_Sieve for fast points from your GPU.

[Dirk Broer]

Should I throw my RTX 2070 at primegrid or Einstein?
Greetings vel

When PrimeGrid is finished you have some days left to do your Einstein thing, so....

[dannyridel]

I have a weird idea: at the end of the throw we could bunker tasks for the next throw and bunker GPU tasks for the city run, without hampering any progress. Will that likely work? Dumping SiDock bunkers should also be possible, I guess.

[SuicideCabbage]

The issue I see with that, is only the middle of the three is counted for the end result, so if we do 1.5m 1.5m 1m 750k 750k or 1.5m 1.5m 1m 0 0, in either case our scored throw is 1m. In my opinion, today sets the bar, and the next two throws we only need to get just as much as we got today, any extra is "wasted", also crunching more than three days is also a total day "wasted" unless we feel we can vastly change placement by giving it a 4th or 5th attempt. So it is my view that resources above what is required to equal today's attempt are better spent on other disciplines. I'll let others weigh in on what they feel we should do though, and support any majority decision.

[Dirk Broer]

1600 is Turing too.

Big thing of the Turing architecture is the Tensor cores hype (3-dimensional matrix calculations), and the 1600 series is without those -so more like the previous 2000 series.

BTW: The Volta architecture includes Tensor cores too and, quite surprisingly, can be found within the nVidia Jetson Xavier series -where it BOINC-wise can't be used....

[dannyridel]

I think we should do the first three throws all our best, since they don't interfere with PG as a CPU subproject. Then we should straight abandon the last two.