AMD Ryzen 5 2600X + Ryzen 7 2700X Linux Benchmarks

[phoronix]

Phoronix: AMD Ryzen 5 2600X + Ryzen 7 2700X Linux Benchmarks

The embargo on the Ryzen 5 2600X and Ryzen 7 2700X processors has expired now that these Ryzen+ CPUs are beginning to ship today. We can now talk about the Linux support and the initial performance figures for these upgraded Zen desktop CPUs.

http://www.phoronix.com/vr.php?view=26226

[shmerl]

What about freeze on idle? Is it fixed in new Ryzens?

[Michael]

What about freeze on idle? Is it fixed in new Ryzens?

I haven't had any freezing issues at all.

[Brisse]

I haven't had any freezing issues at all.

I could say the same thing about my first gen Ryzen 1700X which I got at launch day. Doesn't pass the kill-ryzen test though.

[gukin]

Thank you Michael for including the 2400G in your testing, I gather that the 2400G was also stable? Maybe I missed something but which motherboard chipset did you use?

[ddriver]

Weird results.

Testing on windows shows that the 2700x is about 10% slower in single/low threaded scenarios, but fairly consistent 30% faster in tasks that put all threads to good use, such as 3d rendering.

[Brisse]

Weird results.

Testing on windows shows that the 2700x is about 10% slower in single/low threaded scenarios, but fairly consistent 30% faster in tasks that put all threads to good use, such as 3d rendering.

That makes no sense and does not check out with the Windows benchmarks I've seen thus far (Gamers Nexus, Sweclockers). Are you sure you are looking at a trustworthy review, and are you looking at just one or several?

[raonlinux]

Typo at page 7:
"The Ryzen 7 2600X does deliver better performance-per-dollar than the Ryzen 7 1800X"
must be:
The Ryzen 5 2600X does deliver better performance-per-dollar than the Ryzen 7 1800X"

[dungeon]

That makes no sense and does not check out with the Windows benchmarks I've seen thus far (Gamers Nexus, Sweclockers). Are you sure you are looking at a trustworthy review, and are you looking at just one or several?

Nothing without watching at least 10 reviews, then removing spikes (the worse and the best review) and at a very end calculating average

Likely AMD have that data from much more reviewers, which are probably more than 300 in the world... and i guess who get consistently worse results, won't get review samples next time I think all that could be even simulated to get average, but one never know what would happen in the wild

That is how you get on average what to expect - never believe to just one review, but also the worse nor even to the best one

[L_A_G]

Would be nice to see if those cache latency improvements were put to the test as they, together with overclocking past 4 GHz, were the original series' Achilles heel. Don't think there's anything for it testing cache latency in the current set of benchmarks however and most people testing it use SiSoft Sandra or AIDA64, neither of which have a Linux version (which in the case of AIDA64 is kind of confusing considering it's got versions for Linux-derived platforms like Android, ChromeOS, Tizen and SailfishOS).

As for the overclocking, no need to test that as it's outside of Phononix' scope and done by loads of other outlets.