AMD Ryzen 5 8600G Linux Performance

[phoronix]

Phoronix: AMD Ryzen 5 8600G Linux Performance

Last week the new AMD Ryzen 8000G series processors with Zen 4 CPU cores and RDNA3 graphics became available in retail channels. The initial launch-day Linux testing was focused on the Ryzen 7 8700G processor, which showed great uplift for the integrated graphics over the Ryzen 5000G series with Vega/GFX9 graphics and the Ryzen 7000 series with their cut-down RDNA2 integrated graphics. In today's article the AMD Ryzen 5 8600G is the focus of the Linux testing along with an expanded set of comparison processors re-tested.

https://www.phoronix.com/review/amd-ryzen5-8600g

[artivision]

The correct way to test APUs is with Ram speed synchronized with the SoC Bus at max speed, this is 3Ghz Bus combined with 6Ghz Ram. When you run the this first test only then you run another test with the fastest budget Ram (today 7.2Ghz) to see what the Bus will do and what the performance will be versus the first test. If the performance is not on par or better then you lock the Bus at 3Ghz with the Ram at 7.2Ghz and test again. If you cannot lock then you buy another motherboard. Then you repeat the tests with the fastest Ram on the market for future reference. Those test here show nothing, non interested in the deference between Apu generations. Everyone want to know if they can match low eng Gpus with some perfect configuration.

[pinguinpc]

Michael Good tests however if you can test 8500G iGPU will be good compared 5600G iGPU

thanks

[grigi]

Interesting that the GPU on the 8700G isn't ~50% faster as the specs indicate it should. It seems to be between 20% and 35% faster only.

Is the 8700G *that* bandwidth starved?

[schmidtbag]

Everyone want to know if they can match low eng Gpus with some perfect configuration.

So long as people keep insisting on cranking up AA and texture details, iGPUs will never never keep up. As you have alluded to, memory bandwidth is the Achilles heel of iGPUs and those 2 settings are very demanding on VRAM (AA might not be as taxing but relative to the visual fidelity it offers, it's quite taxing on VRAM). Considering iGPUs seem to scale linearly with the fastest RAM you can throw at them, this really shows that until iGPUs get access to a large dedicated cache, people are going to have to change their expectations of what they can do. The Ryzen 8000 series is proving to be impressive but these GPUs are still starved.

Despite everything I said, I'm curious how much less starved for bandwidth the 8600G is from the 8700G.

[JEBjames]

Michael

Typo /wording page 1:
"Of course, the newer Mesa nad kernel you are on is typically better for yielding the best performance and features."

Maybe
"Of course, the newer Mesa and kernel versions typically yield better performance and features."

Was surprised how well the 8600g did in these tests.

Maybe add clearly linked amazon affiliate links for all processors reviewed here at the end of the article.

[gukin]

I'm a big fan of APUs but am having a little trouble justifying the purchase of one of these. I first got on the A10-7850K Kaveri and it was really quite good compared to everything else at the time, but going from the 7850K to a Raven Ridge 2400g was huge, both a giant leap in GPU as well as CPU performance; the only downside was that DDR4 was horribly expensive in 2018. Going from a 2400g to a 5700g was also big although the GPU increment was only marginal the CPU bump was dramatic and the power consumption of the 5700g was also impressive.

The GPU leap now is . . . 50% to 80% and the CPU is 10% to 20%, 15% - 30% for the 8700g but the hardware is almost twice the price of AM4 stuff (comparing B450, B550 to B650, DDR4 to DDR5 and AM4 to AM5 processors). Heck throwing in a wimpy RX 6400 beats the 8700g by 50% and probably doesn't use more power. Still there is a definite uptick on both CPU and to a greater extent the GPU, it's also on AM5 which may endure for a long time and I've played on my 5700g enough to know that a 50% - 80% bump in the GPU would make a huge difference.

On the other hand, AMD skipped releasing the 4000 APUs to retail because they couldn't beat the Vega 11 on Raven.

I'll have time to stew since it appears that the motherboard I was looking at, the Asrock b650i ITX Lightning, is currently unavailable.

[zexelon]

I think every CPU should have an iGPU in it... including server grade. I would be curious to know what the power consumption and such of an iGPU versus say an external Rage128 knock off on your rack MB is. I mean you are still running that ancient gpu for console no matter what, would not an iGPU in your EPYC not be better?

These iGPUs make awesome DOM0 consoles Really wish the ThreadRipper had one.

I am guessing though that wiring such a thing up on an already fairly obscene die would be deeply painful to someone somewhere...

[blackshard]

I'm a big fan of APUs but am having a little trouble justifying the purchase of one of these. I first got on the A10-7850K Kaveri and it was really quite

...

I'll have time to stew since it appears that the motherboard I was looking at, the Asrock b650i ITX Lightning, is currently unavailable.

I dIdn't go through the same APU journey, but I share the same thoughts.

I expected a much more performance per power and performance per cost ratio from these new 8x00 APU series, but the price is not competitive (both the APU and the whole platform) and power consumption neither (perf/watt in GPU benchmarks show a regress, CPU benchmarks did not show any dramatic improvement).

I would expect important changes in benchmarks when the STAPM issue will be fixed, but I guess a fix will probably not make perf/watt better.

edit: benchmarks from other sites also show different and sometimes weird numbers (see anandtech for example, where 8600g is plenty faster than 8700g in some cases!)

[avis]

There's a known STAPM issue with certain BIOSes and desktop Zen 4 APUs, I wonder if Michael has been affected.

https://www.youtube.com/watch?v=MFzegmwHxPM