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
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 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.
83 Comments on AMD Works with Microsoft to Improve Zen 3 thru Zen 5 Performance in Windows 11 24H2
In the past from AMD Ryzen 1000 to 3000, AMD release the AMD Ryzen power plan in the chipset driver.
This power plan helps to optimize the processor performance and efficiency.
When 5000 series come out, power plan is drop but I does question is there a need?
As it's obvious Windows doesn't work/perform well with AMD unlike Intel.
Since Intel CPU architecture remains the same for most of it's generation.
Whereas AMD Ryzen wise CPU architecture is completely different one thus concurrent Windows aren't
optimized to run properly? Which is why Win 11 24H2 needed since it's yet to be release.
Microsoft and AMD can work out to optimize Windows for Zen 5?
Point here is as Win 11 24H2 is yet to be release, what can be done to improve improvement.
Bring back AMD Ryzen power plan?
And how even past the initial wave of updates the better half was still not scheduled right?
I agree though. But this seems to happen everywhere, because if it happens too early in the process, I think there are competitive / fair market related issues.Yeah its hard to say for me as I went for 11 along with a new CPU but honestly Win 11 is definitely snappy. Not seeing much of any choppiness either. There are more smooth transitions in desktop use which kinda gives the impression it might be slower but its really not, those aren't there to give the illusion of smoothness while the OS is hard at work like it is on Apple devices (yeah...), they're just smooth transitions.
Now that I had the time to read the thread more thoroughly and watch HUB's video in its entirety, by trying to shift blame onto reviewers and consumers, coming up with wild stories like "admin account changes branch prediction engine behavior", I'm gonna argue that this AMD's very own "You're holding it wrong" moment
when did that get released.
learn.microsoft.com/en-us/windows/release-health/windows11-release-information
Also the bios auto enabled/disabled problem.
But the video did feel a lot of stuff could have been cut. The bios part I was waiting for was right at the end.I am going to double down on my reviewer guide comments, get rid of the reviewers guide. If there is an NDA, remove it so if the guide has to stay, reviewers can fully publish it.
The concept of doing reviews becomes flawed if a reviewer is expected to match the vendor's results. You have to take into account different software and games tested, different way of testing games, game settings, different windows configuration, different drivers, different bios version, different bios settings, impact of storage used, the list goes on.
If different reviewers get different results, thats fine, its a review, its not a validation process.
I am glad HUB made this video though, it offers us plebs some insight. AMD should have waited a month or so for the update to get rolled out so then this issue of incorrect prediction behaviour would not have came up (at least for Windows 11, which all the reviewers use now I think).
Wendell from leve1ltechs in his video interestingly also mentioned all the AMD fan boy tribal behaviour in their channel comments, it seems both GN and level1techs have started to get sick of it now as well.
And I also don't see how "use admin account" makes their story any better, even if it is exactly what they say it is. Both Intel and AMD offloading more and more of the firmware's work onto Windows (even if it is inevitable/justified to some degree due to any design reasons) has always resulted in no shortage of performance problems and isn't something to be proud of.
It only really enables most - not all! - privileges, including those who grant the right to debug, profile and modify the system performance and scheduling priorities, by executing as NT Authority
I've never heard anything regarding executing code in privileged security levels changing the hardware's behavior (such as changing how the branch predictor units work), to anyone who has an even moderate understanding of the Windows NT system architecture this seems incredibly far-fetched.
I wonder if PR was told "quick, think of some niche thing that sounds geeky enough, the suckers will buy the excuse wholesale!" and they came up with that!
I had an Intel x25-M 80GB before I moved from XP to W7, and remember testing W7 on an original pre-Sandforce OCZ Agility 128GB SATA2 SSD before committing to W7. Mainstream for me means "available to purchase by regular folk in regular stores", rather than "enterprise kit that consumer resellers don't stock". These weren't special purchases, I just handed over about £400 for a drive to my regular online retailer each time.
A 4080Super is eye-wateringly expensive, but it's mainstream because anyone can walk into a store and pick it off the shelf, or log on to one of dozens of (inter)national retailers and buy five of them without any hassle. An Nvidia H100 or L40 accelerator is not mainstream because it's not something you can buy. You have to be a server manufacturer to procure them as an Nvidia-vetted OEM partner for resale in your servers offered for sale to corporate customers.
If by mainstream to mean "popular" then it's going to be way behind the curve of a typical purchase today - the 5-year-old GTX1650 has only just lost it's #1 spot on the Steam hardware survey this year, replaced by a 4-year-old, discontinued GPU, and I think and the 8-year-old 1060 is still in the top-10 most popular GPUs today.... ;)
There is the SYSTEM level, I will paste the output, the easiest way to elevate to that is using a tool called PowerRun.Now the effect of disabling UAC?
You already posted the permissions that are "after" UAC elevation.
It is like this without UAC elevation. Which if I remember right is not possible on the default hidden admin account, everything is auto elevated, or am I wrong on this?
Running without UAC elevation will disable most permissions except those required for restricted user accounts to work, this is how it achieves its intended sandbox effect.
Check this for reference ~
answers.microsoft.com/en-us/windows/forum/all/windows-10-home-getting-trustedinstaller/6664a5f1-5c71-45fc-94be-805075db682a
and I’m pretty certain the only reason amd is doing that is because it’s easier to clone a windows installation and get it up to date with all drivers without having to click yes dozens of times When it’s deployed over dozens of systems
UAC can be left enabled but without the prompts on the normal account, meaning everything is not auto elevated, things that dont ask for elevation will stay not elevated, whilst things that want elevation will be silently approved.
What they did is a big no no. Disabling the entire UAC system which the admin account defaults to.
Under my account name it says "account is disabled" and it's unchecked, I either did it when I first installed it or that's how LTSC is I do not remember :D To check just go here:
"Press the Windows Key + R and type in lusrmgr.msc. Click on the Users folder. Double-click on Administrator. Uncheck Account is disabled and click on OK." If that's the case then us LTSC owners have nothing to worry about in terms of performance, at least that's what I think, I can be wrong LOL. :p
While I like AMD Ryzen, and did a deep dive into this new Ryzen 9000 performance issue, the problem boils down to one BAFFLING thing.
AMD didn't validate CPU performance on a normal Windows setup?
(they used "Admin" mode which is something a normal Windows Administrator (i.e. home user) doesn't actually have enabled)
How? How does that slip through the cracks? It's sort of like "validating" the shocks you designed on a flat race track and then somebody reports a problem when you hit bumps IN THE REAL WORLD.