AMD Ryzen 5 5600G Linux Performance

[phoronixk]

I seem to have random crashes with opensuse tumbleweed kernel 5.13 on my new system with ryzen 5600g. Could this be a too old kernel?

[gregzeng]

I seem to have random crashes with opensuse tumbleweed kernel 5.13 on my new system with ryzen 5600g. Could this be a too old kernel?

Read the other comments here. OpenSuse is very different from Ubuntu. Tumbleweed is not famous for stability, by deliberate design. The exact model & version of them motherboard is very important. You need test the next version of the Linux kernel, 5.16, in probably the alpha or beta versions, to get better test results from the AMD range of CPU.

[gregzeng]

This year report is far too modest. It compares not only the one small version of AMD Ryzen 5, but the full range of the latest Intel & AMD main CPU, on the standard mainstream operating system. Power usage and performance costs are also compared. Omitted are the new motherboards required, of the users choose the latest Intel CPU.
Phoronix is almost ready to make executive summaries of its test results now. AMD seems better than Intel CPU production on every basis now. On performance per price however, Intel might be preferred for the impoverished computer users. The only barrier for Intel users are the necessary brand new motherboards needed for each new version of the Intel CPU. Unlike Intel, AMD does not force a brand new version of motherboard for each upgrade to the CPU.

[pal666]

11600K should be comparable to 5600X, not trail it far behind:

i guess this chart was generated on intels with all possible vulnerabilities enabled

[nranger]

[QUOTE=jochendemuth;n1276774]

Phoronix: AMD Ryzen 5 5600G Linux Performance

Although only tangentially related to this post, I have the following question: I noticed that, the Ryzen 5900 is outpacing the Ryzen 5950x in CPU Power consumption. In my own installation I found the Ryzen 5900 challenging to cool (even with AIOs). How were the Ryzen 59xx CPUs cooled in the testing? and Do you think that the Ryzen 5950 was thermo throttled in some of the tests?

Launch reviews found the 5950X taking advantage of better bins to achieve the same or better power efficiency than the 5900X. As a result it isn't any harder to cool. Stock package power lands in the neighborhood of 140 watts, and stays pretty consistent. So at 12 core loads, the 5950X should run at the same or higher clocks. Then for loads up to 16 cores, the 5950X will run at either lower all-core clocks (it does have a lower base clock than the 5900X) but also was able to run all cores at lower voltages to keep within the package power limit.

https://www.guru3d.com/articles_pages/amd_ryzen_9_5900x_and_5950x_review,7.html

https://www.anandtech.com/show/16214/amd-zen-3-ryzen-deep-dive-review-5950x-5900x-5800x-and-5700x-tested/8

[avem]

i guess this chart was generated on intels with all possible vulnerabilities enabled

Rocket Lake just like Tiger Lake has roughly the same vulnerabilities as AMD CPUs and a 52% performance loss cannot be caused by enabling or disabling mitigations, besides Michael didn't disable anything to run benchmarks for this or earlier test. Your guess is squarely invalid.

Here's an article on Phoronix which proves your wrong:

https://www.phoronix.com/scan.php?pa...re-rocket-lake

Out of 96 different tests ran (see all the data in full over on OpenBenchmarking.org) in the default and then mitigations=off state on the Core i9 11900K, there was 2% difference overall for the geometric mean.

[trueblue]

Would you test for ECC working on the 5600g? AMD response to my query was that ECC was enabled on all 5000 processors, but I think they did not really understand my question. Specifically, I asked if the Pro series were required for ECC unbuffered memory support.

[smitty3268]

If if it's a different motherboard / microcode revision the 11900K should have taken a hit as well. Its results are the same as before. I've trusted Michael's benchmarks previously, not any more, sorry. Not a single reply from him to address in certain cases a 52% performance loss.

I see nothing in the article stating that the 11900k and11600k are using the same motherboard.

[avem]

I see nothing in the article stating that the 11900k and11600k are using the same motherboard.

What other excuses do you have to justify the results which run contrary to all other results on the net and to Michael's own previous results?

[avem]

Would you test for ECC working on the 5600g? AMD response to my query was that ECC was enabled on all 5000 processors, but I think they did not really understand my question. Specifically, I asked if the Pro series were required for ECC unbuffered memory support.

The normal 5000 APUs do not support ECC, you'll need the Pro versions for that, e.g. Ryzen 5 PRO 5650G.