• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[H]ardOCP vs TeAm Anandtech Folding@Home race: December 1st - 8th

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
[H]Coleslaw said:
Not many of the FAH crew left unfortunately. And the few from those days left are still PO'd at Pande Group. So, this is probably all she's gonna get unless someone just crawls out of the woodwork and surprises us.
Click to expand...

This is a good looking graph though! Our production is close to double what I had expected. Science wins in the end. 🙂

1575456801301.png
 
Out of curiosity, I switched on my dual-processor computers with CPU folding slots. They basically work like I remember from previous occasions:
The 56-thread slots make 650...730 750 k PPD.
The 88-thread slots are a mixed bag: They might go up to 1.0...1.3 1.4 M PPD, or they might error out. — During our race with Brony@home, I was briefly able to run them with 88 threads, then no longer got work this size and dialed back to two 44-thread slots.
 
Last edited:
TennesseeTony said:
Shall we turn this into a 3-way race as suggested elsewhere? TAAT team one vs TAAT team two vs [H]?

I say yes. 🙂
Click to expand...
It might be a headache for biodoc what with random members dropping in and out. On the surface it would make things more interesting, and give [H] a fighting chance.
 
TennesseeTony said:
Shall we turn this into a 3-way race as suggested elsewhere? TAAT team one vs TAAT team two vs [H]?

I say yes. 🙂
Click to expand...
Pokey said:
It might be a headache for biodoc what with random members dropping in and out. On the surface it would make things more interesting, and give [H] a fighting chance.
Click to expand...
The 3-way race would last only until Sunday morning. Hence, members may still pop in, bot probably not yet out.

StefanR5R said:
Out of curiosity, I switched on my dual-processor computers with CPU folding slots. They basically work like I remember from previous occasions:
The 56-thread slots make 650...730 750 k PPD.
The 88-thread slots are a mixed bag: They might go up to 1.0...1.3 1.4 M PPD, or they might error out. — During our race with Brony@home, I was briefly able to run them with 88 threads, then no longer got work this size and dialed back to two 44-thread slots.
Click to expand...
Over night, each 88-thread box became stuck in loop with a WU saying "There is no domain decomposition for 72 ranks that is compatible with the given box and a minimum cell size of ... nm". I switched them both to 2x 44 thread slots and hope they keep going. The 56 thread slots are still fine, as expected.
 
Well er umm you see, it's like this, coleslaw did me a solid and you know how I like running with the underdogs, and then there is the good cause factor and my contribution aint worth a hill of beans in this crazy world and now I'm running out of things to say so good luck fellow dcer's.:beercheers:
 
Folding on CPUs is a bit of a PITA.

Last night, one of the 2x44-thread boxes' client went down after a series of segfaults. When I found it like this in the morning, I deleted the work folder and restarted the client.

Today, one 44-thread slot got stuck with a WU which said "There is no domain decomposition for ... ranks that is compatible with the given box and a minimum cell size of ... nm". I switched the slot to 64, 32, and finally 16 threads, each to no avail. I paused the slot, deleted the work/xx folder, and restarted the slot.
 
PS,
deleting the work folder and restarting the client or the slot was the (effective) remedy in these situation. Perhaps a call to FAHClient --dump ... would have been the canonical approach to such a situation.

Regarding the strange WU which didn't want to work with power-of-2 thread counts: It didn't occur to me to try a multiple-of-3 thread count at the time, although the "ranks" given in the error messages were 3 * 2^n. --- It's a rather unacceptable failure mode though, as it causes downtime while unattended, and is certainly a big problem to solve for less technically minded contributors.
 
Last edited:
I had a 15 thread instance with the same error, had to also delete the slot and then add it. I am surprised by the ppd of my CPU fleet, another 2.2M.

[H] has some sleepy-heads who just noticed the challenge is going on: "a 2080ti, 2x titan Xp and AMD VII " being added to their team from one member, and a laptop 1070Ti from another. 🙁

Splitting up into triple race has certainly made this more fun! 😀
 
yodap said:
Well er umm you see, it's like this, coleslaw did me a solid and you know how I like running with the underdogs, and then there is the good cause factor and my contribution aint worth a hill of beans in this crazy world and now I'm running out of things to say so good luck fellow dcer's.:beercheers:
Click to expand...

Much cheaper than firing up GCC instances. Glad to help you out regardless. And thanks for the assist.
 
Any possibility you guys would want to move this challenge to being a yearly recurrence but in January or February since there is typically only a PrimeGrid challenge in each month? Formula-BOINC doesn't normally start until March. I doubt we do much with FB next year, but that would certainly conflict less with other events and holidays. Not to mention everyone could use their new Christmas toys for the event.
 
[H]Coleslaw said:
Any possibility you guys would want to move this challenge to being a yearly recurrence but in January or February since there is typically only a PrimeGrid challenge in each month? Formula-BOINC doesn't normally start until March. I doubt we do much with FB next year, but that would certainly conflict less with other events and holidays. Not to mention everyone could use their new Christmas toys for the event.
Click to expand...
I do like the yearly event suggestion!!. How about the day after Christmas to the day after new Years? Would be 8 days like this is now.
 
The only problem with a month long race is people starting hammering their credit cards for new GPU purchases on about the 15th day, lol.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top