Thursday, August 22nd 2024

AMD Works with Microsoft to Improve Zen 3 thru Zen 5 Performance in Windows 11 24H2

AMD Ryzen 9000 "Granite Ridge" desktop processors have been out for a couple of weeks now, and the "Zen 5" based processors have fallen short of gaming performance expectations, set mainly by some of the numbers AMD put out in its Computex 2024 reveal for the processors. The consensus among the tech press is that these processors are roughly 3-5% faster than Ryzen 7000 series "Raphael," but with noteworthy improvements in energy efficiency. AMD set out to study why there are such vast deltas in performance between its numbers and those of reviews, and arrived a few possible explanations. The company also stated that it is working with Microsoft to fix this in the next major update to Windows 11.

One of them is that AMD's testing was done on Windows 11 23H2 with Admin mode (i.e. a local system administrator account was used), while some reviewers tested with a regular user account that has some admin privileges. Apparently this affects the way the branch prediction units of "Zen 5" processors work. "Our automated test methodology was run in "Admin" mode which produced results that reflect branch prediction code optimizations not present in the version of Windows reviewers used to test Ryzen 9000 Series," AMD said in a statement.
AMD also pointed out that in its first-party tests that compare Ryzen 9000 series processor models to Intel's, the company ran the 14th Gen Core chips with the same DDR5-6000 memory speed with the same timings, and used Intel's baseline power profile that uses stock power limits for these processors (125 W base- and 252 W maximum turbo power in case of the Core i9-14900K and the i7-14700K, for example).

AMD said that it's working with Microsoft to bring the "correct" branch prediction behavior seen in admin mode to regular Windows 11 user accounts. These updates will be incorporated in the retail release of Windows 24H2, although you won't have to wait until then. Microsoft will release this as an "optional update" sooner than that, so it could be implemented on Windows 11 23H2. Here's the best part—it turns out that the admin mode discrepancy even affects "Zen 4" and "Zen 3" processors, which means even Ryzen 5000 thru Ryzen 7000 series processors should get a performance uplift in regular Windows 11 user accounts.
Source: AMD
Add your own comment

83 Comments on AMD Works with Microsoft to Improve Zen 3 thru Zen 5 Performance in Windows 11 24H2

#1
natr0n
modern windows mostly 11 is a nightmare of hobbled together bit and pieces.

instead of improving and adding on it's breaking and repairing what was fine and working
Posted on Reply
#2
Chaitanya
natr0nmodern windows mostly 11 is a nightmare of hobbled together bit and pieces.

instead of improving and adding on it's breaking and repairing what was fine and working
Its a spyware that cannot even work properly.

Watched this video in morning and FC6 and CP2077 are outliers it seems.
Posted on Reply
#3
LabRat 891
When/If there's an LTSC IoT Ent. version of 24H2+, maybe I'd put Win11 back on my rig. MAYBE.

Then again, I wonder if WIn10LTSCEntIoT even has such issues that need addressing?
Posted on Reply
#4
natr0n
LabRat 891When/If there's an LTSC IoT Ent. version of 24H2+, maybe I'd put Win11 back on my rig. MAYBE.

Then again, I wonder if WIn10LTSCEntIoT even has such issues that need addressing?
I'm using that now actually. everything works. I disabled just about everything that updates.

average person can use this without any issues.

Posted on Reply
#5
eidairaman1
The Exiled Airman
natr0nmodern windows mostly 11 is a nightmare of hobbled together bit and pieces.

instead of improving and adding on it's breaking and repairing what was fine and working
Its a continuation of 10, which still is a hobbled mess
Posted on Reply
#6
SL2
LabRat 891Then again, I wonder if WIn10LTSCEntIoT even has such issues that need addressing?
I don't see how it could include updated "branch prediction code optimizations" that AMD posted info about eight hours ago?
Posted on Reply
#7
LabRat 891
SL2I don't see how it could include an updated branch predictor that AMD posted info about eight hours ago?
I don't know if Win11's branch prediction regressed, or not. I assume so, but I'm not certain.
-if older was already better, there wouldn't be any need for a new fix/update/component. (which, wouldn't be the first time such has happened.)
Posted on Reply
#8
SL2
Zen3 is four years old, I guess it should have been done a long time ago.
Posted on Reply
#9
JWNoctis
LabRat 891I don't know if Win11's branch prediction regressed, or not. I assume so, but I'm not certain.
-if older was already better, there wouldn't be any need for a new fix/update/component. (which, wouldn't be the first time such has happened.)
Another useful question might be, exactly how does the OS impact something as basic as branch prediction the way it did?

Though I guess the true answer is probably behind at minimum a couple NDAs and a graduate-level course in modern processor architecture.
Posted on Reply
#10
Pumper
Wonder how Zen5 is performing vs Zen4 in Win10, but I guess no one is testing outside of Win11 at this point.
Posted on Reply
#11
Falck
Why don't they test in the "normal" user/"admin" enviroment, like most users use Windows.
Posted on Reply
#12
azrael
LabRat 891When/If there's an LTSC IoT Ent. version of 24H2+, maybe I'd put Win11 back on my rig. MAYBE.

Then again, I wonder if WIn10LTSCEntIoT even has such issues that need addressing?
Probably yes.
natr0nI'm using that now actually. everything works. I disabled just about everything that updates.

average person can use this without any issues.

I guess the question isn't as much if you can use Win10 IoT LTSC, but if the bug exists in Win10 (in general) as well.
SL2I don't see how it could include updated "branch prediction code optimizations" that AMD posted info about eight hours ago?
That's just it. Win10 probably suffers from the same bug.
JWNoctisAnother useful question might be, exactly how does the OS impact something as basic as branch prediction the way it did?
Though I guess the true answer is probably behind at minimum a couple NDAs and a graduate-level course in modern processor architecture.
My guess is that it may have to do with context switching of sorts (user mode to/from kernel mode).
PumperWonder how Zen5 is performing vs Zen4 in Win10, but I guess no one is testing outside of Win11 at this point.
It's probably the same difference in performance. Problem is, no one bothers with Win 10 any longer, now that "shiny" Win11 needs to be pushed. Certainly no one at Microsoft.
Posted on Reply
#13
LabRat 891
azraelI guess the question isn't as much if you can use Win10 IoT LTSC, but if the bug exists in Win10 (in general) as well.
Does it really though, for "most users"?
At least by the L1T/Wendell vid, I got the jist that the issue revolves around 'virtualized security'. Is this even an issue at all for those not running VBS, Linux in Win10, HyperV, etc.?
Posted on Reply
#14
tabascosauz
LabRat 891When/If there's an LTSC IoT Ent. version of 24H2+, maybe I'd put Win11 back on my rig. MAYBE.

Then again, I wonder if WIn10LTSCEntIoT even has such issues that need addressing?
I think 24H2 IoT is technically just a preview at the moment. Seems to update and work just fine. Moved to it a few weeks ago because my usual 11 Pro just keeps on racking up integrity errors and breaking itself for no reason (other than Microsoft being Microsoft). Nothing in BIOS has been changed and this install is not breaking itself.

Knowing the optimizations in the past few years specifically for MPO and flip model, there's no way I would go back to anything earlier than 23H2, let alone Win10 for that matter. Even if it was LTSC.

But then again, Win10 never really had performance problems with bigger Zen 3 CPUs whereas Win11 has. So maybe for Zen 3 I'd just stick with Win10.
Posted on Reply
#15
azrael
LabRat 891Does it really though, for "most users"?
At least by the L1T/Wendell vid, I got the jist that the issue revolves around 'virtualized security'. Is this even an issue at all for those not running VBS, Linux in Win10, HyperV, etc.?
Haven't seen the vid, so all I can go by are my own assumptions. VBS certainly sounds like a valid reason as well.
Posted on Reply
#16
john_
why there are such vast deltas in performance between its numbers and those of reviews
This is really BS. Either the team running those benchmarks internally at AMD faked the numbers on their own, or the marketing department decided to fake the numbers, or Su herself asked from it's employees to create a presentation with fake data. If they believe that their numbers can be confirmed and that tech press haven't running Zen 5 systems at their optimal configuration, they could set a meeting with Gamers Nexus for example, call them at AMD's labs and show them how they got those numbers.
Posted on Reply
#17
Bwaze
""Zen 5" based processors have fallen short of gaming performance expectations, set mainly by some of the numbers AMD put out in its Computex 2024 reveal for the processors."

I find it baffling that large thing like this (performance degradation with common Windows setup) is missed in testing the product before release.

They do test the product before release, right?
Posted on Reply
#18
izy
Are they faster in WIN10 then?
Maybe if running the games/apps with "Run as administrator" will work the same as running them with the Administrator user?

I see that lately we have to update the BIOS each month and do a lot of tweaks in BIOS to get "stock" performance and stability (on Intel and AMD now) + OS tweaks! :)
Posted on Reply
#19
forman313
When I have time, I´m gonna install Win 11 with VT disabled so that Windows runs on bare metal, then do some benching. Its been a while.
Win 11 Bare metal >= Future patch
Probably
Posted on Reply
#20
wolf
Better Than Native
Interesting, AMD can be quite communicative about issues when they want to be, last years simple question was met with "We have no comment at this time" for months, which is interesting given the relative complexity/simplicity of both situations.

Not holding my breath for large gains to be had from admin/scheduling fixes, but it's welcome that they're committed to trying to get people the touted performance.
Posted on Reply
#21
svan71
Oh, wow a reason to update to Windows 24H2 Microsoft fixes their bug and big brother can watch EVERY SINGLE THING YOU DO as he jams more advertisements down your throat.
Posted on Reply
#22
izy
I was on BETA channel and i was still on 23H2 for some reason , i just switched to Windows Release Preview Channel and im getting the 24H2 , i thought that BETA is receiving updates faster than WRPC but it looks like it is behind.
Posted on Reply
#23
Wirko
eidairaman1Its a continuation of 10, which still is a hobbled mess
At least 10 stopped receiving "feature" updates, so it can be considered stable now.
Posted on Reply
#24
Broken Processor
PumperWonder how Zen5 is performing vs Zen4 in Win10, but I guess no one is testing outside of Win11 at this point.
It's been there ever since UAC I remember reading about it back in the day and everyone Benchmarked in Admin account I still benchmark in Admin account to this day I don't know how an entire community forgot this.
Posted on Reply
#25
Wirko
One TPU forum member devised a power plan for Windows 10 on Alder Lake P+E long ago, when it became clear that Wintel wouldn't do that. Described all the details here. I can't find it now, also because forum search does not want "10" as a search keyword. Maybe someone can make something similar for Windows 10 and Zen 5 as well.
Posted on Reply
Add your own comment
Nov 1st, 2024 23:53 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts