Dual Xeon 2696 v3 upgrade for chess

[yorkman]

I'm thinking of upgrading my dual E5-2696 v3 cpu's to something that will be significantly faster when used with chess engines such as Stockfish.

Could someone with experience suggest the most significant and efficient upgrade for chess?

Things to consider:

-will be used in a house so electricity usage should be lower but can be higher if speed increase is significant
-cpu cost

On a slightly other note. Can someone explain this to me?

Servethehome (https://www.servethehome.com/amd-ep...imply-peerless/amd-epyc-7742-chess-benchmark/)

When you look at the chart the dual 8180 is almost as fast but when comparing these same two dual cpu's on cpubenchmark.net the difference is huge (about 18,000 points)

PassMark - [Dual CPU] AMD EPYC 7742 - Price performance comparison
PassMark - [Dual CPU] Intel Xeon Platinum 8180 @ 2.50GHz - Price performance comparison

If I was to guess selling my dual e5-2696 v3 and buying a dual xeon 8180 makes more sense than buying a dual epyc 7742 considering the huge cost difference.

 

[Evan]

Benchmarks on test usually a specific function with software compiled in a certain way. Some CPU’s are better at different things.
For example if anybody needs AVX512 then it’s intel you need not AMD.

 

[alex_stief]

The benchmark you referred to on STH compares ONE Epyc 7742 against TWO Xeon 8180.
That should make it pretty easy to work out which solution gives you better performance/dollar. Chess simulation is one of those things that is almost trivial to parallelize. So you can make your purchase decision based on number of cores multiplied by all-core boost frequency.

 

[yorkman]

The benchmark you referred to on STH compares ONE Epyc 7742 against TWO Xeon 8180.
That should make it pretty easy to work out which solution gives you better performance/dollar. Chess simulation is one of those things that is almost trivial to parallelize. So you can make your purchase decision based on number of cores multiplied by all-core boost frequency.

Ahh yes, you're absolutely right. I don't know how I didn't realize that. Thanks.

 

[LukeP]

actually cores multiplied by all-core boost frequency multiplied by operations per tick

 

[alex_stief]

actually how is someone supposed to obtain values for this "operations per tick" metric?

 

[LukeP]

you would look at the program you are running and what AVX modes it utilises. each cpu can do a certain number of AVX per tick.
if the program uses AVX-512 the Xeon scalables are twice as fast as 2nd gen EPYC per tick. if its just AVX, they will be the same per tick. older 1st gen EPYC are half as fast at AVX per tick.

given the above it pays to at least check.