Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

[DisEnchantment]

Speculate at will

 

[majord]

How would this be expected to compare with win 10?

 

[branch_suggestion]

Looks like all CPU testing will need to be redone for 24H2.
Will be interesting where things end up, the 7800X3D may pull a decent ways ahead of anything Intel has, will be very interesting if first party ARL comparisons are done on 24H2 or an older Windows version.
While AMD certainly carry some blame here, Microsoft are once again proven to be incompetent.
Thing already works fine on Linux, which is the first platform AMD builds for as they can upstream earlier in development and it matters more for EPYC.

 

[coercitiv]

How would this be expected to compare with win 10?

 

[leoneazzurro]

I still don't know why they continue to test at 1080p with high or even ultra settings. In many cases (and it shows) they get into GPU limited situations. It is also clear that in some cases the platform has some room for improvement because when they get in GPU limited situation sometimes Zen5 is slightly slower than Zen4 and that is also something that could change in future updates.

 

[KompuKare]

For CPU benching you never want anything GPU limited.
Now if certain games and engines do actually put more stress on the CPU with higher settings, then for CPU benchmarks I would rather they run at 480P with high settings than 1080P at low but only if the CPU does something extra at the high settings.

 

[Tup3x]

So with the new Windows 11 build, you are getting a 10% boost on avg for Zen 4 & 5. Bad news is that it doesn't make Zen 5 any better than Zen 4 in gaming, because their relative position doesn't change.

Pretty sick stuff. Doesn't really change anything vs. Zen 4. Now I'd like to know how much this improves Zen 3.

 

[MS_AT]

Looks like all CPU testing will need to be redone for 24H2.
Will be interesting where things end up, the 7800X3D may pull a decent ways ahead of anything Intel has, will be very interesting if first party ARL comparisons are done on 24H2 or an older Windows version.
While AMD certainly carry some blame here, Microsoft are once again proven to be incompetent.
Thing already works fine on Linux, which is the first platform AMD builds for as they can upstream earlier in development and it matters more for EPYC.

While M$ can be blamed for lots of stuff, it is in AMD best interest that their platform is shown in best possible light. They should have dedicated teams that work on both Linux and Windows to keep them as optimized as possible. If we are to believe the rumors and

Looks like this 'bug' is that under true admin account the spectre/meltdown patches are either working in a different way, or not working at all. After I've disabled it, the difference between the benchmark results of FC6 and CP2077 disappeared. Could be something else, but I've rerun the tests several times and did several reboots to be sure.

this post gives strong support to the rumors that the issue was related to some security mitigations for older Zen CPUs. It's rather the case of AMD not telling M$ to add exclusion where applicable and ensuring its there. But yes AMD and Microsoft relationship looks really weird with this first to Pluton no Surface, AI chips no features, oh well

Pretty sick stuff. Doesn't really change anything vs. Zen 4. Now I'd like to know how much this improves Zen 3. Also does that change anything for Intel?

It shouldn't if the regression is caused by Zen specific security mitigations, but I guess somebody will test soon enough

 

[branch_suggestion]

But yes AMD and Microsoft relationship looks really weird with this first to Pluton no Surface, AI chips no features, oh well

AMD has been in the friendzone for time immemorial. Even good old Wintel is being usurped by upper exec obsession over ARM.

 

[IEC]

11% average gains just from a windows update? 6+ games with 20-35% improvement? WTH?

I wonder if the 7800X3D will also gain 10-11%...

 

[exquisitechar]

AMD sitting on 10% free gaming performance until almost 2 years after Zen4 launched is such an AMD thing to do. I’m curious about X3D gaming performance too.

 

[fastandfurious6]

wow these gainz are YUGE

gonna make big difference for phoenix / zen4 mobile

also overall system responsiveness beyond games I guess?? these are big numbers

even less reason to upgrade lol!!!!

 

[IEC]

Wondering if it extends to Zen 3 as well. That would keep my 6800H/6800S laptop relevant for even longer.

 

[JustViewing]

So with the new Windows 11 build, you are getting a 10% boost on avg for Zen 4 & 5. Bad news is that it doesn't make Zen 5 any better than Zen 4 in gaming, because their relative position doesn't change.

This is insane. I am confused on who to blame for this debacle. AMD/MS/Intel. Was this also present in windows 10? Maybe Intel optimization/bugfix did this? Would this also improve application performance?

 

[Josh128]

So with the new Windows 11 build, you are getting a 10% boost on avg for Zen 4 & 5. Bad news is that it doesn't make Zen 5 any better than Zen 4 in gaming, because their relative position doesn't change.

This is quite significant , unexpected, and perhaps mostly-- embarassing. So the next question is, how does it affect non-gaming applications? Judging by the gaming gains, it appears that it would actually be prudent to re-review the processors entirely, and Zen 4 for that matter as well, lol.

This is a blunder of massive proportions for AMD. Zen 4 gaming performance has apparently been suppressed for its entire life, apparently. Also very interesting how it doesnt affect Intel, outside of a few potential one offs. The question now is, were the mechanics of this performance suppression due to Spectre/Meltdown mitigations? If so, is Zen 4 and/or Zen 5 now vulnerable to these exploits with the new Windows update?

I mean, this is good objectively, NOW-- but what about the entire Zen 4 generation?? Is this a panic maneuver by AMD due to claw back mindshare for their new processors after what was almost universally recognized as a disastrous launch and disastrous sales numbers, that throws caution to the wind for the sake of "muh performance"?

They are trying to put out the dumpster fire, but at what cost?

 

[JustViewing]

Intel users, after one year of usage loose 10% performance
AMD users, after one year of usage gain 10% performance .

 

[lightmanek]

According to video Wendel posted at L1T, Microsoft finally changed branch prediction unit flushes not to occur so often on AMD. It was flushing good data way before AMD massive branch target buffer was full, a remnant from the past it seems.

 

[MS_AT]

This is quite significant , unexpected, and perhaps mostly-- embarassing. So the next question is, how does it affect non-gaming applications? Judging by the gaming gains, it appears that it would actually be prudent to re-review the processors entirely, and Zen 4 for that matter as well, lol.

This is a blunder of massive proportions for AMD. Zen 4 gaming performance has apparently been suppressed for its entire life, apparently. Also very interesting how it doesnt affect Intel, outside of a few potential one offs. The question now is, were the mechanics of this performance suppression due to Spectre/Meltdown mitigations? If so, is Zen 4 and/or Zen 5 now vulnerable to these exploits with the new Windows update?

I mean, this is good objectively, NOW-- but what about the entire Zen 4 generation?? Is this a panic maneuver by AMD due to claw back mindshare for their new processors after what is almost universally recognized as a disastrous launch and disastrous sales numbers, that throws caution to the wind for the sake of "muh performance"?

They are trying to put out the dumpster fire, but at what cost?

If we are to believe the rumors mitigations were necessary only for Zen1 and Zen2. It should therefore have nothing to do with Intel [I know about GoW results, but this insider build is doing more than just patching this problem] and is indeed embarrassing it was not fixed for Zen3,Zen4,Zen5 as soon as possible. But is also likely they thought the patch was applied and simply did not notice if their own internal testing infra had the patch applied...

 

[Hitman928]

This is quite significant , unexpected, and perhaps mostly-- embarassing. So the next question is, how does it affect non-gaming applications? Judging by the gaming gains, it appears that it would actually be prudent to re-review the processors entirely, and Zen 4 for that matter as well, lol.

This is a blunder of massive proportions for AMD. Zen 4 gaming performance has apparently been suppressed for its entire life, apparently. Also very interesting how it doesnt affect Intel, outside of a few potential one offs. The question now is, were the mechanics of this performance suppression due to Spectre/Meltdown mitigations? If so, is Zen 4 and/or Zen 5 now vulnerable to these exploits with the new Windows update?

I mean, this is good objectively, NOW-- but what about the entire Zen 4 generation?? Is this a panic maneuver by AMD due to claw back mindshare for their new processors after what was almost universally recognized as a disastrous launch and disastrous sales numbers, that throws caution to the wind for the sake of "muh performance"?

They are trying to put out the dumpster fire, but at what cost?

Later Zen generations had hardware mitigations for spectre (meltdown didn’t affect Zen) and similar vulnerabilities so the software mitigations were never necessary on the later generations. Same thing applies on the Intel side with Alderlake. Microsoft, it seems, didn’t have a problem in not applying the software mitigations for Intel in updated hardware, but continued to apply the software mitigations for AMD unnecessarily. This is MS’s screw up and probably explains a good deal of why AMD always performs better in Linux.

AMD probably should have caught this earlier from their end but without knowing the details, it’s hard to say how much they knew when or why their testing setups didn’t catch it earlier. MS, outside of Xbox, doesn’t seem to have much interest in really working with AMD very closely.

 

[PJVol]

This is unfair on MS (?) part to all Zen4 owners

 

[CouncilorIrissa]

This is unfair on MS (?) part to all Zen4 owners

It's partly on AMD as well to ensure MS does not do stupid stuff on their CPUs, and MS being stupid is pretty much guaranteed these days.

 

[StinkyPinky]

Steve himself does say he wants other people to check his data as he thinks some of these gains are too good to be true. These gains are even higher than AMD themselves are saying.

 

[CouncilorIrissa]

Steve himself does say he wants other people to check his data as he thinks some of these gains are too good to be true. These gains are even higher than AMD themselves are saying.

AMD might want to be conservative with their estimations given all the negative press they've received recently.

 

[coercitiv]

Come on Microsoft, give something for Zen 3 too, won't you?. Id doesn't have to be much, just 3% for Zen 3.

 

[Glo.]

All of this stuff gives more fuel for...

Console gaming, over PC gaming.

This is horrendous.

 

[Saylick]

Dang, I did not expect AMD Fine Wine to apply to their CPUs…