300+ Benchmarks With AMD Threadripper 3960X vs. Intel Core i9 10980XE

[Danielsan]

It would be also interesting a comparison at three with the Power9...

[Pariston]

Nice test. I would also like to see a comparison between CPUs with close MSRPs. Something like a 10940X vs a 3950X would be good to see.

[Michael]

Nice test. I would also like to see a comparison between CPUs with close MSRPs. Something like a 10940X vs a 3950X would be good to see.

Unfortunately and as said in the article, I don't have a 3950X and when checking with AMD on Monday it didn't sound certain whether I would get one... I guess 3950X supplies really are tight.

[betam4x]

This is good and all, but why compare CPUs from different price brackets? I believe intel CPU is competitor vs Highest end Ryzen processors Ryzen9 [whatever], and not TR.

Intel's chip is HEDT, the lowest HEDT AMD CPU is the 3960X.

[wizard69]

Very nice reporting Micheal. All the whining about processor comparisons don’t mean much to me anymore. The simple fact is for a long time it wasn’t even rational to consider AMD based on performance needs. Now it is foolish not to consider AMD for just about any work load. That is a most refreshing change in the situation.

[nuetzel]

And again.

Compare with 'decrippled' Intel MKL

https://www.phoronix.com/forums/forum/linux-graphics-x-org-drivers/intel-linux/1141322-intel-core-i9-10980xe-linux-performance-benchmarks?p=1141355#post1141355

[sabriah]

Very impressive! No, I didn't necessarily mean the 3960X, I meant the test and the article! Nice!

[cybertraveler]

Interesting.

I'd be curious to see some gaming benchmarks comparing popular, high-end AMD and Intel CPUs. The tests would all use the same graphics card. Only the CPU would change. Though... it might be worth using two graphics cards: one AMD and one NVIDIA. The different drivers might yield different CPU usage characteristics.

[reavertm]

This does not change the fact that the 3960X is somewhat overpriced, because a single computer with 3960X is more expensive than 2 computers with 3900X, even if for many applications with high parallelism and little communication, e.g. for program compilation, the double 3900X will be as fast or faster. The same is true when comparing a single computer with 3970X versus 2 computers with 3950X.

This analogy is flawed imho. Threadripper comes not just with double the cores but double the memory channels and more than double PCI-E lanes.
Twisting your argument even further, it's overpriced compared to 3900x because it's not twice faster in single threaded appliances. They are expensive agreed, but it's not home users market anyway. Threadrippers like EPYCs are meant to pay for themselves. 3900x and 3950x are monster-enough home CPUs imho.

[Linuxxx]

The defaults were used. Though for the most part in these workloads, generally I see minimal difference with P-State powersave vs. performance, usually it's only with gaming tests where sometimes is such a difference.

No, I really meant a comparison between AMD & Intel with BOTH using ACPI-cpufreq 'performance'!
It would be an eye-opener for many, since Intel's P-state CPU driver really is bad, especially in the latency department; so gaming benchmarks with a focus on minimum framerates & frame-times would be quite interesting.

Also, do You include Google's "Web Latency Benchmark" in the PTS?

https://google.github.io/latency-benchmark/