CPUFreq Governor Tuning For Better AMD Ryzen Linux Performance

[MrCooper]

schedutil vs intel_pstate ?

What are you asking? No Intel CPUs here, so I can't say how intel_pstate compares.

[sarfarazahmad]

What are you asking? No Intel CPUs here, so I can't say how intel_pstate compares.

ah my bad nevermind then.

[indepe]

Check the Adored video that Geopirate linked above, starting around 13:05. Looks like the actual testing was done by Joker Productions and referenced in the Adored video.

Ah, got it now. Yes I already wrote in some other posts that in those realtime HUD displays, of each core's CPU-usage during gaming execution, one can see that the Ryzen has a lot of unused capacity (usually *all* cores below 70%, or at least below 80%, with only few exceptions.)

[dungeon]

Yeah, there are two conflicting views on this - one is the "performance at low res is the best predictor of performance you will get in the future, particularly if you upgrade your graphics card" and the other is "sure, assuming nothing else changes like games getting better at using multiple cores and your 2- or 4-core high clock CPU maxing out all its cores and becoming the bottleneck".

Gamers Nexus is arguing in favour of the first view, Adored is arguing for the second (and there are a lot of people in between); both viewpoints are worth a watch. The supporting evidence for the second view is benchmarks where the cores on a 7700K are pretty much (but not quite) maxed out, and the argument is that this is already starting to become visible in the form of higher minimum frame rates than you get with Ryzen.

Mainly, i don't see any other arguments between these two other than just Present vs Future^tm or Arguments vs Guesses, isn't it

Gamers Nexus just represent more of a Present arguments, where Adored represent more of the Future guesses

Nothing wrong if someone name gaming CPU review as "Intel (still) better for Today, AMD (maybe) for Tomorrow"

[dungeon]

This is clearly a software issue because if it was a hardware issue it would be exposed in Cinebench. The performance is there and this isn't bulldozer.

It is not clear software issue, it is somewhat combination of both and of course this isn't Bulldozer

AMD claim they are on singlethread equal to Broadwell-E , but ~7% slower than Kabylake plus clock advantage of Kabylake which seems to be 12% let say. So taken both into account that should be average expected difference and if you see more diff than these could make in some cases, then that is a bug - or software issue Just watch that Gamers Nexus video, someone from AMD said that there

Basically, it is slower than Intel's recent KabyLake even without software issues. How much? So lets just believe AMD - ideally 7% plus clock diff.

Sinking the ship at non ideal cases of present reality where issues are clear and you got sometimes even these results - better than Bulldozer, slower than KabyLake...




Now imagine how scalability works proper among asics, put a slowest GPU there instead, cut hundreads of zeros and you get large diff still when software issue is present, so that ends up as litteral 20 vs 30 vs 50 fps

So cleary it isn't Bulldozer, but clearly it is not KabyLake and because of let guess software issues sometimes might be even much bellow Broadwell-E with which it should on par on paper... i think even worse reviewers were more than correct on average, regardless of hype

[liam]

AFAIK Intel splits L3 cache up even more than we do - each core has a chunk of LLC, and a ring bus allows any core to access any chunk.

They do, but arbitrary l3 fetches aren't then pushed onto a relatively high latency fabric.

Probably worth waiting for an updated AIDA64 before drawing any conclusions:

https://forums.aida64.com/topic/3768...en-processors/



The obvious follow-on is "average with what workload ?". During normal operation discarded L2 entries from cores on each CCX will end up in the L3 partition associated with that CCX (ie the closest one), in the same way they do with Intel parts. What I don't know is the extent to which the current code was designed around the way Intel LLC operates and hence how much difference in results will be seen once Ryzen is factored into the code.


**SNIP**

Sure, i wouldn't be surprised to see some movement. That l3 had the same access times as main memory was surprising.
Btw, are you aware if ryzen is going to support the new resctrl fs interface?

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=60cf5e101fd4441ab112a81e88726efb6fd7542c

[dfyt]

Also, reality is that 95% people currently steam uses 1080p monitors or less... yes or less, also there are much more people using less than 1080p than anything beyond that

On the surface this comment sounds fair and objective but it's NOT. There should be a comma.

"Also, reality is that 95% people currently steam uses 1080p monitors or less," "and of that 95% 99% have 60hz monitors" - which means it really is a propaganda comment. I'm in PC sales it's VERY easy to sway public opinion with half truths / statements.

[dfyt]

As someone who sells 99% Intel and has owned a good mix of AMD and NVidia as well as a few AMD but mostly Intel, The more I think about it a test with a R7 1700 vs my 5 i7 rigs I really am struggling to see a single advantage of buying an i7. I have them connected all via 4k DP's and even the loading times on the AMD is visibly faster in games, benches etc.

I have also tested the linear performance of striping SSD's. From the old X4 BE 970 - 8370 990FX the perf on Gnome-Disks is almost 90% linear ie 500MiB/s 1x ssd, 3xssd = 1.4GiB/s. I haven't tested 7th gen i7 but have from 1st to 5th and this on performance Z boards was slack to say the least. 2xSSD=750MiB/s, 3xSSD was around 980MiB/s. I can't explain it other than to say AMD boards seem like good overall packages. All SATA3 ports were used and the EXACT same SSD's. Anyways just something interesting I thought I'd mention as I always stripe. Live images were used.

I haven't had luck with Ryzen and Linux though - just won't get into DE. Kernel Panic and thats with Latest Tumbleweed. MSI Tomahawk B350. ;(

[bridgman]

I haven't had luck with Ryzen and Linux though - just won't get into DE. Kernel Panic and thats with Latest Tumbleweed. MSI Tomahawk B350. ;(

Another poster observed that including acpi=off as a boot parameter helped...

[indepe]

On the surface this comment sounds fair and objective but it's NOT. There should be a comma.

"Also, reality is that 95% people currently steam uses 1080p monitors or less," "and of that 95% 99% have 60hz monitors" - which means it really is a propaganda comment. I'm in PC sales it's VERY easy to sway public opinion with half truths / statements.

It is more complex than that. We are discussing those who consider paying more than $300 for the CPU alone, especially those who are considering either a i7-7700K or a Ryzen 7. That's a much more specific group of customers. Then, we are discussing those who, at the point of purchase, put a high priority on gaming performance. A 1920*1080 gaming monitor with 144 Hz can be had for less than $200. Many gamers prefer the smoothness of >60Hz to a higher resolution. You can buy them even at Best Buy, which usually sells at the lower end of the spectrum. Furthermore it is true that new graphics cards improve in performance rapidly. Some even use two in parallel. Including, I guess, those who make such reviews.