Has anyone tried using AMD Thread rippers for optimisation and what would you recommend for a neuroshell machine? *I’ve read the support about cores/threads etc*
I see some people are using cloud/VPS, some are using XEON setups but I was looking at thread rippers and wanted to know what kind of setups are the ‘speediest’ for Neuroshell.1 year, 5 months ago baksa.peterParticipant
Hi neurospark511, I´m using thread ripper (Ryzen 2995) from beginning. Never saw performance disaster like this. Adv. is that more cores = more speed in case of optimisation, BUT when I compare these 64 cores of thread ripper with Intel i9 (9th gen) with 8 cores, I were expected 8 times (ok, lets be realist 5-6 times) more speed, but it was just 1.5 – 2.0 “speedier”.
Cannot explain.. NST is not performance optimized 🙁
Thanks for replying, this has been very helpful.
I was hoping the 64 cores would’ve worked better with a 3995WX…. Chaos Hunter might be able to utilise this alot better.
I’ve had some luck with over clocking the ram and pushing the CPU up to acceptable clock speeds which still had system stability for many hours.
Probably cannot push an intel system in this way but AMD is a whole box of fun when you got the right CPU and speedier ram.
Thanks for your feedback, was very helpful!1 year, 4 months ago roach2012Participant
i have the 3960x, in most simulations it’ll max out 16 of the available 24 cores.1 year, 4 months ago baksa.peterParticipant
I bought 3990X 64-core (128 Logical processors) but when I first run Prediction it uses only 64 cores. Also in NST7 settings is able to use 64 of 64 cores. Any ide how to enable all 128 Logical processors for NST ?
Peter have a read of this:
The Windows and Multithreading Problem (A Must Read)
Sometimes its actually the computer which is causing the problems, in this case it could be windows and not using the right ‘update’ – threadrippers have a tough time here because of some versions of windows and how they ‘see’ the cores. This is also down to how AMD created Zen2 and the architecture but you’ll have a better idea after reading the article.
I’ve been using Ryzen 9 5950x and have been able to use full core with sustained 4ghz+ on all cores running within temps of 45-55 degrees.
What I’ve also come to understand is…. some systems have the Greta Thunberg eco-mode activated and will be actively throttling for power saving purposes AND you may have to play around with overclocking to achieve something which will put in the work for you and run stable.
(You dont need to create magical black smoke to billow out, only for you to push the hardware alittle more… Ram is something which is funky and has an effect here but its not for the faint hearted and could affect your hardware warrenty so this is just for ‘educational’ investigation only… 😉
Time for optimisation is greatly affected if you choose Air cooling over AIO water cooling. The HEDT chips tend to run hot and then the throttling kicks in, I’ve found AIO cooling allows you to sustain higher clockspeeds over long periods of time with very acceptable temperature boundaries for such high end desktop chips.
you need Win 10 Enterprise or Win 10 Pro for workstations.
you could possibly get away with an update but i’m pretty sure this is whats holding that beastly 3990x back. Also I can say that Zen3 is freakish fast and worth it… get them TR Pro’s when Chagal line of processors come out 🙂
Roach – have a look at this – https://nstsupport.wardsystemsgroup.com/support/multiple-core-distributed-optimization-faq/
you might have some luck with optimising with more cores after making these changes1 year, 1 month ago michaelParticipant
i got the fastest ram i could get , i think that has made a good difference. 32 gigs of at 3600 .
Zen 3 , waiting for the new chip at the end of the year. that cache will be great .
You must be logged in to reply to this topic.