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

[DisEnchantment]

Speculate at will

 

[Hitman928]

Let me get this straight, does the core parking driver thing turn off the other CCD when gaming, or does it just force the game to run on only the first CCD? Because I typically run lots of stuff, even when gaming, and I want 16 cores for a reason, the first of those alternatives are a big no for me.

It puts the other CCD to sleep, essentially, but it doesn't completely shut it off. So, if you can use them, all cores will be available to you.

 

[yottabit]

I highly doubt core parking driver is causing those high latencies. My guess is they shipped the core parking driver BECAUSE of the high inter-CCD latency. Why is it so high? I have no idea. I'd love to be wrong and we see the Inter-CCD latency disappear with an AGESA update, but I'm not holding my breath

EDIT: Like, there's no way the core complex would wake up from sleep in 120 nanoseconds, right? That's the delta in inter-CCD latency from Zen4 to Zen5.

No one has tested inter-CCD latency in Linux yet have they?

 

[Det0x]

Det0x save us with your tuned 8200 MT/s 🙏

Have made a owners thread on a other forum and posted some results and findings there

[Official] Zen 5 Owner's Club (9600X / 9700X /...

edit 11.09.24 So more links to posts made in this thread 8400MT/ + 2200mhz FCLK done stable with hynix m-die Pictures from delid 9950X ES Core cpu binning results for 3* 9950X retail vs ES Blameless 4x 9700X binning results 8200MT/s + 2200mhz FCLK done stable with hynix a-die...

www.overclock.net

 

[BoomerD]

From the little I've read, this reminds me of the 2011 Bulldozer release. (Faildozer) BIG hype, little-to-no gain in performance.

I expect AMD will fix the problems...if they can be fixed short of recalling all CPUs sold.

 

[Tuna-Fish]

I'm having a hard time understanding why the performance is next level in Linux while being underwhelming in windows...

Linux has been used a lot in large NUMA systems and has had a lot of work done to support it, and because of this has the best NUMA scheduling in the industry.

 

[Hitman928]

From the little I've read, this reminds me of the 2011 Bulldozer release. (Faildozer) BIG hype, little-to-no gain in performance.

I expect AMD will fix the problems...if they can be fixed short of recalling all CPUs sold.

Seems a bit of an exaggerated take on the situation.

 

[Hitman928]

I highly doubt core parking driver is causing those high latencies. My guess is they shipped the core parking driver BECAUSE of the high inter-CCD latency. Why is it so high? I have no idea. I'd love to be wrong and we see the Inter-CCD latency disappear with an AGESA update, but I'm not holding my breath

EDIT: Like, there's no way the core complex would wake up from sleep in 120 nanoseconds, right? That's the delta in inter-CCD latency from Zen4 to Zen5.

No one has tested inter-CCD latency in Linux yet have they?

I don't think the whole CCD is turned off, the cores are just put to sleep. I'm not sure what the time to wake on modern CPUs is, but it wouldn't shock me if it can be done in around that time if they are just clock gated.

 

[gdansk]

From the little I've read, this reminds me of the 2011 Bulldozer release. (Faildozer) BIG hype, little-to-no gain in performance.

I expect AMD will fix the problems...if they can be fixed short of recalling all CPUs sold.

They can maybe fix it to be as good as Linux in some workloads. But it's never gonna become good at gaming. Maybe even with X3D.

As for Bulldozer, Zen 5 is still showing 11% INT and 23% FP improvements at the same clock in industry standard benchmarks. Bulldozer....was not.

But yes it is more vulnerable to Arrow Lake than I had previously imagined.

 

[biostud]

Yeah, to me it really feels like RDNA 3 redux. AMD thinking that they can unlock performance in the drivers and then being unable to.

Hopefully this time they will be manage to get it done with an AGESA update.

We all know if anything needs to be “fixed“ in software you might as well wait for Zen6.

 

[gdansk]

We all know if anything needs to be “fixed“ in software you might as well wait for Zen6.

Disagree on that. At first thread director didn't work on my 12600K because the antique BSD didn't support it. But the next version did and performance improved.

It can sometimes happen. Probably not for Windows though, Microsoft doesn't care about all 60 Zen 5 DIY users.

 

[biostud]

They can maybe fix it to be as good as Linux in some workloads. But it's never gonna become good at gaming. Maybe even with X3D.

As for Bulldozer, Zen 5 is still showing 11% INT and 23% FP improvements at the same clock in industry standard benchmarks. Bulldozer....was not.

But yes it is more vulnerable to Arrow Lake than I had previously imagined.

Basically Zen5 should have launched at current zen4 prices, if they wanted to sell some units. With current pricing there's no reason for most users to choose it over zen4.

 

[biostud]

It can sometimes happen. Probably not for Windows though, Microsoft doesn't care about all 60 Zen 5 DIY users.

This.

What happens in Linux world I know very little about, but remember how some magical driver was going to fix RDNA 3?

 

[yuri69]

From the little I've read, this reminds me of the 2011 Bulldozer release. (Faildozer) BIG hype, little-to-no gain in performance.

I expect AMD will fix the problems...if they can be fixed short of recalling all CPUs sold.

This launch shares many parallels with Bulldozer, although the absolute performance and power characteristics of the product are nothing like Bulldozer.

Anyways, both got:
* Horrible hype and projections
* Slow and painful hype train derailing process
* Incompetent marketing fluff
* Unorthodox x86 architecture solutions
* Last minute software "adjustments"
* Linux performance gains more impactful than the Windows ones
* etc.

But probably the worst part having striking resemblance to Bulldozer are the expectations for future products - "fixing the shortcomings" and "unlocking the true potential". This sounds exactly like all those people talking about the great foundation for the future. It remains to be seen.

 

[gdansk]

This.

What happens in Linux world I know very little about, but remember how some magical driver was going to fix RDNA 3?

Yep, don't buy things for what they could be but what they are. And Zen 5 on Windows is a nothing burger at a higher price. It's an easy skip if you use Windows.

But I have seen software take better advantage of certain hardware with time.

 

[gdansk]

But probably the worst part having striking resemblance to Bulldozer are the expectations for future products - "fixing the shortcomings" and "unlocking the true potential". This sounds exactly like all those people talking about the great foundation for the future. It remains to be seen.

Actually it is worse.
AMD basically stopped shipping iterations of Bulldozer for DIY desktop. But they really did polish that turd almost every year through 2015! Zen 5 won't be getting that yearly treatment.

 

[Glo.]

I'm having a hard time understanding why the performance is next level in Linux while being underwhelming in windows...

Wendel claims that even gaming has cases in which Zen 5 CPUs are faster under Linux in the same game than on Windows which puts 9950X on top of Linux gaming benchmarks, but not on Windows in those specific cases.

 

[positivedoppler]

Wendel claims that even gaming has cases in which Zen 5 CPUs are faster under Linux in the same game than on Windows which puts 9950X on top of Linux gaming benchmarks, but not on Windows in those specific cases.

The plot thickens, many of the Youtube comments are blaming the Windows kernel being worse than Linux

 

[Rheingold]

No new cores counts until IOD resign so Zen 6 this has been known forever.

They could use the same hardware and still have added cores, just by renaming SKUs: 8 for R5, 12 for R7, 16 for R9, and drop the higher R9.

 

[Rheingold]

Let me get this straight, does the core parking driver thing turn off the other CCD when gaming, or does it just force the game to run on only the first CCD?

Watch the JayzTwoCents video. Apparently the problem isn't new, and his fix was for Zen 4. Moves the game to VCache CCD, keeps background stuff on the frequency CCD.

 

[Glo.]

The plot thickens, many of the Youtube comments are blaming the Windows kernel being worse than Linux

This whole situation makes absolutely zero logical sense.

 

[yottabit]

I don't think the whole CCD is turned off, the cores are just put to sleep. I'm not sure what the time to wake on modern CPUs is, but it wouldn't shock me if it can be done in around that time if they are just clock gated.

I really doubt that. I mean I’m not an expert, but it it’s been shown in other reviews it takes on the order of ~ 1 ms to hit boost frequency

~ 100 ns is on the order of RAM access latency, I can’t imagine it coming out of any kind of sleep anywhere near that fast. The voltage regulation, everything else will respond on timescales much longer than that. Just based on my intuition and respect for orders of magnitude I don’t think there can be a software (Windows driver) reason to explain the extra 120 ns. Microcode, maybe.

 

[Rheingold]

iterations of Bulldozer

Now that I think of it... that series also had an "over 9000" moment... the FX-9590 with boxed water cooler.

 

[StefanR5R]

View attachment 105293

Screenshot from Bilibili in 360p, sorry for the bad clarity.
It seems like performance uplift is there, so why is the avg application performance showing single digit uplift according to Techpowerup?

Is there a detailed description anywhere of what we are seeing in this graph and how it was produced?

I hope the regression versus Zen 4 < 125 W is an artifact of the particular desktop setup. (Because Turin vs. Genoa eventually needs to have a very different outcome than this one...)

[It's not a rehash of the older 9950X *ES* measurements, is it?]

 

[Fjodor2001]

So are we still expecting X3D in September? Mid or end of? Based on what?

Any chance it will magically be able to save Zen5?

 

[MS_AT]

The Core Parking seems to me more like a band aid to the problem [we know if we let threads across CCD communicate it will tank performance] rather than reason for the problem. I mean they did not bother to update 7900x or 7950x even though they could, but probably there it wasn't worth it when they were releasing x3d. That would not hurt x3d sales but could have been used as nice marketing trick. Anyway this is all speculation on my part.

But yes, after decoder mystery got solved, we got new one. Zen 5 the gift that keeps on giving