Upcoming Windows 11 24H2 unlocks further Ryzen performance

SpyderTracks

We love you Ukraine
So early findings on gaming performance on the 9700x and 7700x in the soon to be released windows 11 update unlocks between 7 and 20% fps in games, due to optimisation in the windows scheduler.


Good news for a few of us!
 

TonyCarter

VALUED CONTRIBUTOR
Does installing this also mean I have to hand over all my personal, confidential information to Microsoft's AI?
 

TonyCarter

VALUED CONTRIBUTOR
Absolutely. All your clothing has to have Microsoft branding, your back garden MUST replicate the XP desktop background, and everytime you answer the phone or post on social media you have to first say “AI is great” before doing anything else
I suppose if I have to, I have to...anyway my I ❤️ Steve Jobs gear is looking a bit threadbare now, so I need something shiny & new!
 

Martinr36

MOST VALUED CONTRIBUTOR
Absolutely. All your clothing has to have Microsoft branding, your back garden MUST replicate the XP desktop background, and everytime you answer the phone or post on social media you have to first say “AI is great” before doing anything else
And that's all while doing the oki kokie with the likes of Alexa and friends
 

SpyderTracks

We love you Ukraine
windows 11 update unlocks between 7 and 20% fps in games

So it's like the real life equivalent of "download more ram for improved performance".
No, downloading more RAM is impossible.

Optimising the scheduler is a real thing, it dictates which processes hit which cores at what load, there are usually only 2 peak cores on a CPU, the rest are lower binned and can't reach the same frequencies. The scheduler intelligently moves threads from one core to another if it's no longer requiring that same peak boost, for instance apps or services that are open but not currently being used. They last did it for the release of Ryzen 3000 and it significantly boosted performance.
 

TonyCarter

VALUED CONTRIBUTOR
No, downloading more RAM is impossible.
Well, I downloaded and installed Symantec/Norton RAM Doubler back in the day.

Maybe it was Connectix…
 

SpyderTracks

We love you Ukraine
Well, I downloaded and installed Symantec/Norton RAM Doubler back in the day.

Maybe it was Connectix…
That’s a very different thing though, that was providing cutting edge OS tech that Apple themselves hadn’t yet implemented. Pretty damn savvy

My father worked for a Bank back in the 60’s and 70’s and managed them putting in the IBM Mainframes. After initial testing of the first one, the bank decided they needed more RAM. In those days, youre talking the mainframe took up an entire room and you had to call in an IBM techy at mega thousands per hour. And this was in South America so the guy probably had to be flown in from the states or somewhere, I’m betting first class too

The guy rocked up, said sure, took my Dad along and he flicked a switch somewhere which enabled another bank of RAM which was installed by default. Mental. That’s proper RAM downloading!

That bank chain never paid IBM for any further RAM upgrades
 

TonyCarter

VALUED CONTRIBUTOR
I remember when we got the first lot of PCs on our COBOL course (after running on AS/400 for the first 3 months), we were told not to push the 'turbo' button on the front of them as it used too much power and was unstable.

Of course we all did, will no ill effects, and it would be the first thing we'd do in the mornings. But just before we graduated, one of the senior lecturers pointed out that it simply turned an LED from off to on and wasn't connected to anything else.
 

TonyCarter

VALUED CONTRIBUTOR
If you don't want to sign up for the Insider programme to get this, or don't want to wait, OR don't want to be forced to install the possible-data-stealing-for-ai-training 24H2 upgrade, then this standalone update is for you!!!


 

SpyderTracks

We love you Ukraine
If you don't want to sign up for the Insider programme to get this, or don't want to wait, OR don't want to be forced to install the possible-data-stealing-for-ai-training 24H2 upgrade, then this standalone update is for you!!!


Oh, now that's very cool, thanks, definitely going to install that
 

TonyCarter

VALUED CONTRIBUTOR
Just running some quick benchmarks (3-run average of a couple of games and synthetics) to see if there's any difference...please wait...

Not a lot of, if any, difference there at all (maybe doesn't add much for X3D CPUs)...and the only difference I did notice was my GPU utilisation went up from c.90-93% to c.95-98%. But I'm running all these tests under my normal user account, not a superuser-admin-root account!

2024-08-28_17-23-07.jpg


Forza Horizon 5 (Before / After):
(Ultra settings, no DLSS, no DLAA, no FG)
  • 3440x1440 (RT Ultra) = 153 FPS / 154 FPS (0.7% - no statistical difference)
  • 3440x1440 (RT Off) = 163 FPS / ??? FPS (0.0% - no difference)
  • 1080p (RT Ultra) = 178 FPS / 181 FPS (1.7% - biggest gain, and all 'after' runs were 1% or more faster than the 'before' runs)
  • 1080p (RT Off) = 192 FPS / 193 FPS (0.5% - no statistical difference)

Assassin's Creed Mirage (Before / After):
(Ultra settings, DLSS Native, no DLAA, no FG, No RT)
  • 3440x1440 (fullscreen) = 159 FPS / 159 FPS (0.0% - no difference)
  • 1080p (borderless) = 202 FPS / 205 FPS (1.5% - 2nd biggest gain, but run-to-run variances were bigger 'after')

3DMark Steel Nomad (Before / After):
  • DX12 9,140 / 9,104 (-0.4% - no statistical difference)
  • Vulkan 10,230 / 10,223 (-0.1% - no statistical difference)

Unreal EZBench (Before / After):
(DLSS off)
  • Particles
    • 3440x1440 = 11,810 / 11,826 (0.1% - no statistical difference)
    • 1080p = 16,308 / 16,340 (0.2% - no statistical difference)
  • Photogrammetry
    • 3440x1440 = 16,826 / 16,888 (0.4% - no statistical difference)
    • 1080p = 24,628 / 24,574 (-0.2% - no statistical difference)
 
Last edited:

SpyderTracks

We love you Ukraine
Quite an in depth test of various games with a spread of Ryzen 9000, 7000 and the 5800X3D

He does note off the bat: A first insight was, or is: our workstation or synthetic benchmarks such as Cinebench, Blender, V-Ray, Corona, Handbrake or 7-Zip and the Geekbench do not benefit from the optimizations.

Some interesting improvements, in Cyberpunk the 7800X3D saw an 11.5% improvement although at 720p??? Seems an odd resolution to bench at, maybe there's something I'm not aware of on that games performance scaling. But a number of the tests are done at 720p which I do find strange.

It's in German, you should be able to set translation in the browser

 
Last edited:

SpyderTracks

We love you Ukraine
not a superuser-admin-root account!
Please if anyone has seen the performance gains in this admin account, don't mistake this for some standard admin user, it's the actual root admin account that should never be logged into in that way as it's a HUGE security risk since they hid it back in NT4 days. Please don't unhide it and log in, it really isn't worth it!
 

TonyCarter

VALUED CONTRIBUTOR
From the tests I'd already seen I had a feeling it might only benefit the games which were CPU bound, so threw in some 1080p runs to see if that was the case....and that report you linked to seems to confirm that...as you can see greater gains at 720p than at 1080p, so if they'd done the same at 1440p/UW1440p/4K I guess the gains would be smaller at each step up.

If I'd been a professional tester, then I would have turned everything to low and done some 360p testing to see how many thousand FPS I could have got...but that's why I'm a 'documentation analyst'!

I would have done more, but they were taking an age, and I wasn't sure which games had built in benchmarks for me to run...and then I got a notification that I needed to restart my machine to install the latest update (which had triggered automatically).

I'm just happy that it hasn't actually broken anything ;)
 

Paul1964

Silver Level Poster
Brett on the YouTube channel UFD Tech has said that you will only see significant gains if you have not already done optimisations yourself. So if you're an enthusiast who has tinkered around you are less likely to see much improvement but if you're running "out of the box" the the gains could be quite significant.
 
Last edited:
Top