Intel Core i9 9900KS Linux Performance Benchmarks

[Michael]

What's the likelihood of being able to compare the R0 stepping 9900k's context switching (I assume you mean P0 stepping is what your current 9900k is)

Unlikely unless someone donates the funds, Intel never offered any other 9900K (or even informed me of the stepping with mitigations) and no funds to buy one.

[loganj]

i think amd still has some unsolved issues with power consumption on linux. there is an article here since 2 months ago

[mackal]

Unlikely unless someone donates the funds, Intel never offered any other 9900K (or even informed me of the stepping with mitigations) and no funds to buy one.

I just saw a few comments using your results as proof the 9900ks isn't just a special binned 9900k, but we can't really say anything given that it's more likely a binned R0 stepping and your results were with P0 stepping :P

[tildearrow]

Typos:

while the Core i9 9900KS now jsut solidifies that lead for Linux gamers.

But these Intel Cofffee Lake CPUs did tend to deliver

than the particularly Ryzen 9 3900X comparison system.

(did you miss an adjective?)

Demo: Magnetic Reconnection - Renderer: SciVis

(is this supposed to be a graph?)

the AMD Ryzen 9 3900X performance with its geomeric mean being just under 10% faster.

[Michael]

What happened to your wife, Michael?! Typos:

(did you miss an adjective?)

(is this supposed to be a graph?)

Thanks, fixed. Complications from pregnancy, she's due in 2 weeks. She's fine for now.

[audi.rs4]

An interesting test. I fully expected the 9900KS to be competitive performance wise but I am surprised the 9900KS system drew so much less power than the 3900X.

I know there are other components contributing to that (PCIe 4.0 chipset for one?) but it's still a bit of a surprise.

Along with the vulnerability mitigations, I would consider this a pretty good showing overall for Intel.

I think I'd still get the Ryzen myself but, as always, competition is good....

Couple things to keep in mind about the power consumption comparison.

1. As you mentioned, yes the new AMD systems do consume more due to the PCIe 4 chips. It'll be interesting to see if Intel's boards draw as much once they add PCIe 4 support.
2. The 9900KS is on a very refined (and old...) node process. So, they don't have any many chips with "leaks". All Ryzen CPUs, there chips are actually pretty much the bottom of the barrel. The "good" chips, as in there most efficient chips that don't have power "leaks", are put in Epyc's. To make them as power efficient as possible, because the data center cares heavily about this. Ryzen gets the leftovers.
3. The 3900X has 50% more cores than the 9900KS. Subtract the higher usage of the PCIe 4 system, and my "guess" is that it is less than a 50% increase in power usage, making the 3900X more efficient overall, which it should be as it is on the 7nm node.
4. As someone else also mentioned, the new Ryzen platform also appears to use more in Linux. This could be further inefficiencies with PCIe 4 in Linux with the drivers than in Windows, but I don't think it is exactly known as to why yet.

Point is, all things considered, the Ryzen platform power draw, isn't too bad, all things considered. And could have huge improvements, if PCIe 4 high draw in Linux (and even Windows) is due to improvements that could be made in the drivers. We'll just have to wait and see.

[tildearrow]

Are you sure that this is the proper place and proper way of how to ask such questions?

Sorry. Removed.

I was just worried...

[_Alex_]

Judging from the power consumption which is very similar on average to the 9900k I think the 9900ks was not boosted for long. It seems some mobo with some mobo settings can keep it at 5ghz for a long time:

https://youtu.be/aVLuKqfyVyw?t=199

Might be interesting to make a re-run with such a mode (?).

[smitty3268]

Judging from the power consumption which is very similar on average to the 9900k I think the 9900ks was not boosted for long. It seems some mobo with some mobo settings can keep it at 5ghz for a long time:

https://youtu.be/aVLuKqfyVyw?t=199

Might be interesting to make a re-run with such a mode (?).

It may have been thermally throttling itself? I think I've read others who assumed the 9900ks would need water cooling to keep the all-core 5ghz speed going for any length of time.

Edit: Scratch that. You're right, it appears these results are right around what you'd expect if the motherboard didn't enable permanent boosting. From some windows tests elsewhere, they were getting approximately:

9900k - 90w
9900ks throttled after boosting - 125w
3900x - 145w
9900ks boosted 5ghz - 185w

[torsionbar28]

Wow intel is really falling behind, and it shows. This is clearly a "catch up" sku in an attempt to retain some marketshare from Ryzen's technical superiority. Even so, I'm not sure why any sane person would choose a 14nm Pcie3 chip when a 7nm Pcie4 chip costs the same or less.