AMD Processors Lose 15% Gaming Performance with Windows 11, L3 Cache Latency Tripled

[windwhirl]

good luck for all the beta testers out there

Eh, it's already fixed in Dev, so MS will probably just backport it to RTM

 

[zlobby]

Eh, it's already fixed in Dev, so MS will probably just backport it to RTM

It begs the question how did this slip during internal testing? Oh, wait!

 

[Arkz]

Pretty sureit will be fixed soon. Dont know why many jumps into conclusion microsoft prefers Intel. Their xbox system in practically all AMD since their existence.

Since Xbox one. So only 2 console generations. 360 was an IBM PowerPC CPU and OG Box was a Pentium 3/Celeron hybrid.

Either way it will be fixed, they'll probably update the Xbox Series kernel to W11 at some point too so obviously they'd need it fixed in that.

 

[billEST]

Ah, I posted that in another thread. Here you go:

Benchmarked: Do Windows 11’s Security Features Really Hobble Gaming Performance?

How much do they hurt, and how to opt out.

www.tomshardware.com

Tom's Hardware's tested VBS on and off on Intel's 10700K and 11700K and AMD's 3800X and 5800X. Mind you, this are merely tests on VBS impact alone, it's got nothing on whatever is going on with the L3 cache issue.

The 15% impact is an AMD's estimated measure on select games
View attachment 219743

there is a fail in the test they use Z490 for both 10 and 11 g .. or there is real difference in perf between Z490 and Z590 ...

 

[Crackong]

Wintel at its finest

 

[windwhirl]

there is a fail in the test they use Z490 for both 10 and 11 g .. or there is real difference in perf between Z490 and Z590 ...

No, Z490 supports both 10 and 11th gen Intel CPUs

It begs the question how did this slip during internal testing? Oh, wait!

Yeah, it's strange as hell.

 

[R-T-B]

All right, now I wonder where is this 15% performance loss? so far benchmarks I have done, have no regression x windows 10, all within a margin of error.

Try AIDA64 cache benchmark. It shows there bad in the L3 cache section.

I have not noticed it either but I doubt AMD would claim this were it not true.

Likely to do with VMR (acronym might be VmS or something, no VBS read the thread) ie fully virtualized OS security.

It's not even on in fresh installs, so no.

There were reports it'll affect ALL CPUs as much as 25%.

Not to my knowledge. No benchable figures like AMD.

 

[AlwaysHope]

You are a brave soul I never touch a new OS for at least 6 months.

+1 here too!

 

[Minus Infinity]

+1 here too!

Make it 12 months or more in this case.

 

[Lionheart]

I’m really looking forward to Adler Lake reviews…on both Windows 10 and Windows 11. We’re going to see just how much had to be changed in software to hit performance claims, and it’s not the first time Intel’s “help” with software results in performance degradation when genuineIntel is not detected. It could quite possibly be a short-term marketing game. Early reviews will show Adler Lake looking significantly better than Ryzen on the latest and greatest Windows 11. A post-review patch from MS will come about and fix some of the glaring deficits, and only the enthusiast will know any better. Just needs to look good enough for investors to be happy and a line of commercials to air. MS gets the marketing might of Intel selling PCs with Windows 11. I can even see it being something like “only Intel 12 series CPUs are designed for Windows 11.” It all sounds like conspiracy until you look at history.

I like how you can see the domino effect & past the obvious BS.

Pretty sureit will be fixed soon. Dont know why many jumps into conclusion microsoft prefers Intel. Their xbox system in practically all AMD since their existence.

It's not about preferring Intel, it's just obvious business practices, annoying dodgy ones at that, Intel has more mindshare advertisement muscle, gotta sell that "new" Windows 11 & most PC's out there are Intel.

I mean AMD knows about it... MS knows about it...

Reality is MS loves AMD (Xbox is amd based), and is familiar with AMD architectures... doubtful this will be around for very long.

I'd say couple months at least, gotta make those Windows 11 sales with Intel based Machines, it's scummy & brilliant at the same time. lol

 

[AlwaysHope]

Make it 12 months or more in this case.

That is most certainly not out of the question either.

Having said that though, I am looking at a platform upgrade in the next couple months, hardware wise at least.

 

[Mysteoa]

I'd say couple months at least, gotta make those Windows 11 sales with Intel based Machines, it's scummy & brilliant at the same time. lol

It's already fix in Dev build.

 

[R-T-B]

It's already fix in Dev build.

That is what I just migrated to temporarily.

Is indeed fixed.

 

[DemonicRyzen666]

Not a surprise to me. Similar thing happened going from Core2duo to Nahalem i7 on XP/Vista 64 OS. to windows 7 OS. 15% up lift for Intel on Nehalem i7 chips.
All down to scheduler fixes.

 

[Patriot]

That is what I just migrated to temporarily.

Is indeed fixed.

Now if its not pushed to public before alders lake benchmarks is where this returns to a conspiracy.

 

[eidairaman1]

question: If I do update to w11, I should also resetup all the apps/configs?

Yes

 

[TheDeeGee]

Judging by the comments noone read the article all the way to the end.

Maybe change the headline to: "Windows 11 fix underway for AMD 15% performance loss bug."

 

[phanbuey]

exactly - fix will be out before Alderlake even hits the market.

 

[mama]

All right, now I wonder where is this 15% performance loss? so far benchmarks I have done, have no regression x windows 10, all within a margin of error.

What are your specs?

 

[Bomby569]

It begs the question how did this slip during internal testing? Oh, wait!

or what more has to be discovered and never been tested

 

[Melvis]

ok cool so ANY leaked benchmarks of up and coming hardware is now complete and utter rubbish/BS, awesome! thanks for clearing all that up for us!

I can see it now......terrible, bad, inaccurate reviews will pop up all over the net in amonths time where A is running on 11 but B is not etc etc.

 

[zlobby]

or what more has to be discovered and never been tested

Yes. It was meant as a sarcasm. It was implying that Microsoft don't do any internal testing at all, even for the most critical parts.

Even if they did, someone should really lose their job for missing stuff like this.

ok cool so ANY leaked benchmarks of up and coming hardware is now complete and utter rubbish/BS, awesome! thanks for clearing all that up for us!

I can see it now......terrible, bad, inaccurate reviews will pop up all over the net in amonths time where A is running on 11 but B is not etc etc.

But all the verdics will be 'intel takes back the crown', 'intel rullz, amd sucks', etc. Many people will cream their underwear and rush to the stores to get the latest and greatest from intel.

 

[Wirko]

I'm amazed to hear that cache latency can be controlled by software in any way.

 

[zlobby]

I'm amazed to hear that cache latency can be controlled by software in any way.

Actually, many, many things happen before a processor 'executes' an instruction. And even then there are sets of internal (to the processor) things that take place.
More so, with all the branching, speculative executions and other kinky stuff, the entire system becomes really volatile from users' perspective.

 

[Vayra86]

This summarizes the upgrade quite well I think. Its really something you don't need and might never make use of, is it? But marketing says its better. Not sure if its louder...