Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 14
|
![]() |
Author |
|
RawLu
Cruncher Canada Joined: Oct 19, 2006 Post Count: 17 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
what matters most when running WCG CPU projects?
![]() Single-threaded performance or Multi-threaded performance??? For example: Would the top of the line AMD Threadripper be better than the top of the line Ryzen CPU? I would assume the top of the line Ryzen would maybe blow away Threadrippers Single-thread performance? But would maybe be able to do more WU's at a time? etc. Pls note, I'm not up at all on the latest Threadripper or Ryzen models etc., just examples. TY ![]() |
||
|
Acibant
Advanced Cruncher USA Joined: Apr 15, 2020 Post Count: 126 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() |
I would lean toward multi-threaded performance. For Africa Rainfall Project and Mapping Cancer Markers you have to wait for another person to validate your results, so completing a work unit as fast as possible often won't get you points any sooner.
----------------------------------------Further, the goal should be to complete the most science possible. You could turn off hyperthreading or even limit the number of work units running simultaneously (such that you also get higher boost clocks on fewer cores versus lower boosts across all cores) and that would get really good single-threaded performance. But the overall efficiency of your use of the processor would go down. And if you're interested in the fun "badges" then that also means less overall runtime as well. ![]() |
||
|
Bryn Mawr
Senior Cruncher Joined: Dec 26, 2018 Post Count: 345 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Most work done regardless of whether it’s one thread doing 2x work or two threads doing 1x work makes no difference
|
||
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 609 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() |
According to the list I use: https://www.cpubenchmark.net/cpu_list.php the top of the line AMD Threadripper is the Ryzen Threadripper PRO 3995WX 64/128 and the top of the line Ryzen is the Ryzen 9 5950X 16/32. So it is kind of unfair to compare a 64/128 to a 16/32 and is safe to assume the Threadripper PRO 3995WX would kick the butt of a Ryzen 9 5950X.
Now the link I gave is not the gospel but just the same rating methodology for all CPUs and I'm sure there are other things to factor in. So this makes me agree to most work done. |
||
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1673 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Threadripper is nice but not affordable for me.
----------------------------------------My main criteria is the size of the Level 3 cache, since for some projects - i.e. ARP1 or MIP1 (now "completed") - the needed memory has a significant impact on performance. Imho, the key aspect is less the CPU time than the done work. In this case, the performance for me is at first the number of computed WUs per hours resp. per day. If you consider that the time is a more important criteria for you, you have only to consider the number of threads. Cheers, Yves |
||
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 609 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() |
Threadripper is nice but not affordable for me No kidding! From amazon dot ca:AMD Ryzen Threadripper PRO 3995WX 64-core, 128-thread Desktop Processor Without Cooler 6999.99 CAD/5452.99 USD 1 A batteries required but not included Is there anyone on WCG using this monster [Edit 1 times, last edit by BobbyB at Aug 20, 2021 2:48:47 PM] |
||
|
Acibant
Advanced Cruncher USA Joined: Apr 15, 2020 Post Count: 126 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() |
Keep in mind that rumors are that new Zen 3-based Threadrippers might be announced by the end of this month and available by next. (Or by November from another rumor source.) So you see third-parties scalping product that is perhaps getting slim in the retail channel. I know it's not saying much since the 64-core non-pro version still has an MSRP of USD $4000, not counting the board that will be around $500 and $500+ for the RAM. (And case, PSU, SSD...) Still better than Intel's pricing, though.
----------------------------------------![]() |
||
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Any 16 or 32 systems priced reasonably?
----------------------------------------![]() |
||
|
bikeaddict
Cruncher Joined: Apr 11, 2020 Post Count: 32 Status: Offline Project Badges: ![]() ![]() |
The best value I've found so far in used eBay hardware is a barebones HP Z440 workstation with no CPU/GPU, paired with a 14-core Xeon E5-2690 V4 CPU. The Z440 was selling for $240 a while ago but has increased in the last month for some reason. The Xeons seem to be flooding the used market from some server farm that is being upgraded, and sometimes sell for $210-225 and frequently for $250-270. People are buying them up 10, 20, 30, 40 even 100 at a time.
|
||
|
RawLu
Cruncher Canada Joined: Oct 19, 2006 Post Count: 17 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
For Africa Rainfall Project and Mapping Cancer Markers you have to wait for another person to validate your results, so completing a work unit as fast as possible often won't get you points any sooner. hmm I did not know that ![]() |
||
|
|
![]() |