Best CPU for R studio server

[uberweiss]

Hi!

My company is going to upgrade our R Studio Server machine and I have been given the following options in terms of CPU and memory:

Option #1

• Xeon Gold 5222
• 192 GB RAM 3.2 GHz Registered ECC

Option #2

• Xeon E-2274G
• 128 GB RAM 2.67 GHz Unbuffered ECC

They mentioned that if R benefits from larger cache i should go for option 1 and if R benefits from higher clock speed than option 2 is better.

Which one should i go for? Tried searching for information about which is more important, cache or clock speed, but couldnt find any good sources.

We are 5 data scientists in the team using the server. Doing descriptive/predictive analysis. Working with maybe up to 50 mill rows each in our environments.

[alexkgold]

#1 is probably better -- because R loads all data into memory and is single-threaded, more RAM and a faster processor is usually a better call.

[uberweiss]

@alexkgold thanks! I just realized that option 1 only has 4 cores and that my team will probably benifit from more cores. Because even if R is single threaded, that only goes for each individual session? So if we're 6 analysts we will probably benifit from having 8 cores instead of 4, right?

[alexkgold]

Hi @uberweiss,

Really up to you. For the most part it depends on how heavy-duty your analytics jobs are. If you're doing lots of long-running simulations or model building, you might start running into resource contention. If you're not, you may well find that 2/3 of a core per user is just fine.

Best,
Alex