Performance - Xeon E5 v3 vs v4 processor

Question

I was going to purchase an E5-2609v4 CPU for a new server - but then noticed that according to Passmark tests it has a significantly lower benchmark at 6973 then the V3 version at 9885 - but I can't understand why this is - can anyone advise ?

Here is a link to the Intel Ark comparison - which shows that the v3 is 1.5 years older, has fewer cores which are slightly faster (1.9 as opposed to 1.7). I would have thought that Intel would have made incremental improvements for the same model CPU with a different version number, but this seems not to be the case ? As someone who will never need more then 768GIGS of RAM in the server, is there any reason I should buy the newer CPU at the same price?

Answer 1

You're looking at the wrong benchmarks. Your first link is to a single v4 CPU, while your second link is for dual CPU, i.e. two v3 CPUs. Notice the [Dual CPU] in the name, and also the cpuCount=2 in the URL.

A single v3 is only ~6000 on PassMark.

And here's the correct PassMark comparison, where the v4 rates higher on overall performance but lower single-thread performance owing to the lower per-core clock speeds. Whether this is beneficial will depend on your use-case.

Answer 2

This may be very opinion based, but I guess the question comes back to you - what do YOU use it for?

Sure benchmark is a good set of baseline, but it is hard to compare as both CPU has different number of cores, and different speed.

Depends on what you use it for, the additional core may ends up more useful, or not (again - depends purely on how you use it). If you got application that is multi-core aware, then it will benefit more (as 8C @ 1.7 vs 6C @ 1.9). But if it a single-thread/core application, then the faster CPU will win the benchmark.

So in the end - it is up to YOU to check and find out what is the best for your need.

Lets go through some of the differences in the 2 CPUs:

• 6C @ 1.9GHz vs 8C @ 1.7GHz
• 15MB Cache vs 20MB Cache
• 51GB/s vs 59.7GB/s Memory Bandwidth
• NO TSX-NI vs TSX-NI

So I guess the question comes to:

1. Is your application Multi-core aware and able to use the multi-core, or just a single thread Application?
2. Is your application able to use and benefit from TSX-NI feature?
3. Do you have DDR4 1866 RAM? The v4 can benefit from it (and also from the bigger memory bandwidth) - if not, then this one may make no difference to you.

So yes, it goes back to you in the end to decide - based purely on your specific need.

Apologies if this is not an answer - but this is the closest thing to an answer to your question as why the benchmarking test may not be the best method for you to determine why a particular CPU is performing less than the other.