Epyc Rome or Threadripper Build

Hi everyone,

So I’m torn and not sure what the pro’s and/or Con’s are (Besides Costs) of choosing to build a Threadripper build versus building a dual or single Epyc Rome build. My considerations I guess would be…

1.) I do a lot of financial modeling and not sure whether the higher clock rate should make me go Threadripper.

2.) Other things I would like is a quiet build in a 4U case, rather than a screamer in a 2U.

3.) What the extra memory channels can do for me in Epyc rather than threadripper 2970x.

4.) Would be running windows 10 Pro for workstations. Not sure whether the issues with

5.) Do both the new architectures of Rome and Threadripper make the old storage interface of LSI/Broadcom sas/sata raid cards to an expander backplane to 24 bays create a bottleneck? What are the new better ways under these new hardware realities be?

I have an old Supermicro 846 4U box that can take a E-ATX board with dual 1400 watt platinum SQ power supplies. Hoping to use this as a foundation unless you think otherwise.

I’m a bit of a novice, but would appreciate your thoughts and build suggestions.

Thanks for your help and thoughts.

Best,
Sean

If you’re using Windows you want 3rd gen for sure. As far as epyc vs threadripper, if you’re just using it as a personal workstation, I’m not sure if you need the reliability and features of Epyc. But I’m not a SA so take my words with a grain of salt

If you are going to use the system in a “competitive” way as in to compute faster than your competitors (in a stock market scenario for instance), get the largest Threadripper and water cool it.

If you just need to scale to multiple cores with minimal fiddling, go for Epyc, especially if GHz speed is not particularly critical.

EDIT/DISCLAIMER: I am not a sysadmin, I just recalled an LTT video where he had shown a watercooled/overclocked Xeon and he said the use case was competitive computations for stock market purposes where time is critical so that you can bid earlier than your competitors because you can compute faster.

There was this recent GN Video featuring Buildzoid analyzing Wendell’s Epyc Server and he said, Epyc is not overclockable (or at least what Wendell bought - a Gigabyte Epyc Server).

So it’s not really trying to be first in a live situation but conducting research. I’m just thinking expand-ability (Epyc) versus plenty of cores and a higher frequency, albeit a single CPU, Rusen 3 Threadripper.

Did Microsoft fix the kernel issues in Windows that Wendel spoke about which was holding back AMD performance? Also was that kernel issus specific to Threadripper, Epyc or both? It would seem that Windows 10 Pro for workstations was suppose to be specifically for supporting higher end hardware etc?

-S

Most places say the EPYC has a SoC (system-on-a-chip), but I haven’t seen any claims that the Threadripper has it… I think it is a great design, but I don’t understand how Threadripper & EPYC are so similar - while the Threadripper is missing the SoC.

You could take a leap of faith and look into the Lenovo Threadripper Pro workstations. They seem to provide the best of TR and Epyc. I know, I know, buying a “prebuilt” damages the ego of a hardware enthusiast but if you really want the best compromise, look into it. Apart from being the best of both worlds, the TR Pro platform is “qualified” for being rock solid, which is probably very important for a task like financial modelling.

If you’re adamant about building a system yourself and need a LOT of memory, I’d say the frequency optimized Epyc 7F72 is your best bet, otherwise the TR 3970X is probably the way to go.

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.