In my opinion, from a 3950x, a reasonable upgrade would be a 24 or 32 core CPU. 5950x is great, but does not provide that much computing power compared to a 3950x in multithreaded workloads. 10-15% uplift does not justify the extra cost. If anyone has a 8 or 12 core CPU on AM4 platform, then yes, 5950x is the ultimate way to go.
I guess that is why I ended up on the Epyc platform. I was thinking in terms of the significantly lower price to a new 5950 as well as getting a new vs used CPU.
That said whenever I fire up my 3950x it is very robust still.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Does anyone else see the potential security problems I can see?
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Depending the CPU speed and load it can take between 3 and 8 hours per task. The number of volunteers running makes a difference too.
Many volunteers run 8-16 thread machines at most.
We saw a couple of systems make it in the Top 50 running 256 threads. But they were later model servers.
So the estimate is an estimate. I have no clue how it is calculated. But CPU tasks don't pay as much as GPU tasks so fewer machines run them.
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
I also find that the Gamma Ray task does not scale with threads, but with cores, at least on my machine. Half of my CPU threads sit idle because using all of the threads does not make an advantage.
I also find that the Gamma Ray task does not scale with threads, but with cores, at least on my machine. Half of my CPU threads sit idle because using all of the threads does not make an advantage.
I do know that running with two to four idle threads seems to be more productive than trying 100 percent.
I agree that general rules of thumb are always trumped by experimental results.
CPU crunching is the lowest common denominator. Even cellphones can do it!
All hale CPU crunching!
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Hey all, just curious, probably mentioned before but does anyone recall what a system like the Supermicro X10QB with 4P, 4 x Xeon E7-8894 v4 in one chassis does? I usually BOINC in winter and have two dual cpu x99 mobos. But love the idea of seeing 96C/192T in Task Manager and letting them rip in BOINC. Debating to grab one for the fun of it but there's so few posts about it on the internet from the DIY perspective.
Hey all, just curious, probably mentioned before but does anyone recall what a system like the Supermicro X10QB with 4P, 4 x Xeon E7-8894 v4 in one chassis does? I usually BOINC in winter and have two dual cpu x99 mobos. But love the idea of seeing 96C/192T in Task Manager and letting them rip in BOINC. Debating to grab one for the fun of it but there's so few posts about it on the internet from the DIY perspective.
Might be very successful on Rosseti at home. If your running windows install with Oracle VM. Which would mean you could also run LHC at Home (cern). If your running Linux you would run Rosseti without issues.
I haven't succeeded in getting Oracle VM to run reliably under Linux.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
stsfred wrote: In my
)
I guess that is why I ended up on the Epyc platform. I was thinking in terms of the significantly lower price to a new 5950 as well as getting a new vs used CPU.
That said whenever I fire up my 3950x it is very robust still.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
https://www.techradar.com/pro
)
https://www.techradar.com/pro/worlds-first-wireless-cpu-prototype-breaks-cover-and-could-pave-the-way-for-a-world-of-seamless-modular-computing-but-it-will-still-require-a-physical-power-connection-for-now
Does anyone else see the potential security problems I can see?
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
FGRP5 search
)
FGRP5 search progress
1507.8 days (estimated)
1500 days to go? What?
Filipe wrote: FGRP5 search
)
Depending the CPU speed and load it can take between 3 and 8 hours per task. The number of volunteers running makes a difference too.
Many volunteers run 8-16 thread machines at most.
We saw a couple of systems make it in the Top 50 running 256 threads. But they were later model servers.
So the estimate is an estimate. I have no clue how it is calculated. But CPU tasks don't pay as much as GPU tasks so fewer machines run them.
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
I also find that the Gamma
)
I also find that the Gamma Ray task does not scale with threads, but with cores, at least on my machine. Half of my CPU threads sit idle because using all of the threads does not make an advantage.
taketwicedailey wrote: I
)
I do know that running with two to four idle threads seems to be more productive than trying 100 percent.
I agree that general rules of thumb are always trumped by experimental results.
CPU crunching is the lowest common denominator. Even cellphones can do it!
All hale CPU crunching!
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
In other News if you like the
)
In other News if you like the DeepCool cpu coolers you need to pick them up soon:
https://www.techspot.com/news/103400-deepcool-may-halt-us-business-after-sanctioned-over.html
Hey all, just curious,
)
Hey all, just curious, probably mentioned before but does anyone recall what a system like the Supermicro X10QB with 4P, 4 x Xeon E7-8894 v4 in one chassis does? I usually BOINC in winter and have two dual cpu x99 mobos. But love the idea of seeing 96C/192T in Task Manager and letting them rip in BOINC. Debating to grab one for the fun of it but there's so few posts about it on the internet from the DIY perspective.
Won’t be worth it for
)
Won’t be worth it for Einstein.
but a system like that can still be useful for some PrimeGrid projects.
_________________________________________________________________________
TjwLv8ke7qQdB1yqAQzQ7De1ZPH
)
Might be very successful on Rosseti at home. If your running windows install with Oracle VM. Which would mean you could also run LHC at Home (cern). If your running Linux you would run Rosseti without issues.
I haven't succeeded in getting Oracle VM to run reliably under Linux.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!