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

[DisEnchantment]

Speculate at will

 

[JM Popaleetus]

Did AMD not address every complaint people had with Zen 4's launch? The MSRP is cheaper, they're more efficient, and they're faster. Stock vs stock, on average the 9700X is 5% faster using 35% less power than the 7700X in applications. As a cherry pick, der8auer's video has the 9700X using 44% less power than the 7700X in HWBot X265.

The only people that find this release disappointing are likely gamers, who should be waiting for the X3D chips anyway (or buying a discounted 7800X3D). Even then, from that video, the 9700X is still faster and more efficient:

• Assassin's Creed: 2.19% faster using 9.67% less power.
• Cyberpunk: 1.25% faster using 19.83% less power.
• Counter Strike: 2.43% faster using 4% less power.
• Valorant: 6.02% faster using 5.2% less power.

SFF community will love these chips, the 9950X should be a hit, and I eagerly await the 9800X3D (especially if the rumor it's completely unlocked is true).

 

[Hans Gruber]

Wrong. First, it provides either 22% OR 11%. Not at the same time. If you ever saw a single node comparison chart, you'd know that.
View attachment 104728

Wrong again, only desktop parts used N5. Phoenix and Hawk Point both use N4.


We don't have desktop parts at the moment with the same power limit running at the same clocks (we will with the release of 9950X, but even still, you have to account for the IOD power, making obtaining the data for the comparison very difficult). Hence, the comparison between 8940HS in Razer Blade 14 and HX 370 in the Zenbook -- neither hits a power limit under 1T workload, and both parts are monolithic with no savings coming from the node.

Your original claim was that all of power savings were explained by the node changes and not by architecture.
It is disproven by the graphs from the Computerbase's Strix Point review.

You are the only person here comparing laptop processors to desktop processors. The performance uplift and power efficiency gains can be both and not one or the other. The only way you would know what you are getting is if they ran a Zen 5 9700x with N4P vs a Zen 4 7700x with N4P. Same clock speeds and the same TDP. Then you would see the power and performance uplift. I am certain AMD has a lab somewhere they evaluates silicon performance. It would be even more helpful to run a 9700x vs 7700x with both on N5 silicon. That would clearly show the differences in architecture performance.

It's up to AMD how they want to use the efficiency and performance gains in N4P that may not show up in consumer processors. When reviews talk of the better power efficiency of Zen 5 vs. Zen 4. I attribute that to the N4P silicon. The performance numbers TSMC has released are accurate. That is how they sell silicon. AMD had a huge uplift in efficiency and performance when they switched from GloFlo to TSMC 7nm with Zen 2.

 

[Kepler_L2]

Sorry, but I find it extremely hard to believe you have a source with that level of information, who simultaneously didn't have enough information to steer you away from your 40% IPC insanity.

Counter point: I was also wrong about RDNA3 performance but right about Navi4c's cancellation.

 

[CakeMonster]

Did AMD not address every complaint people had with Zen 4's launch? The MSRP is cheaper, they run cooler, and they're faster.

I don't disagree with that, I just think the ~22 month cycle is too long for what is after all very decent iterative changes. If this had arrived in May/June, or we knew Z6 would be here in Dec 2026, I wouldn't have worried.

 

[HurleyBird]

Counter point: I was also wrong about RDNA3 performance but right about Navi4c's cancellation.

Counter point to that: Everyone was wrong about RDNA3, and that was downstream of bad information coming out of AMD. Being wrong (for the right reasons) about RDNA3 while being right about Navi4c's cancellation is not incongruous.

Being 180 degrees wrong about Zen 5 IPC, while also somehow being right about Zen 6 intentionally being targeted so late would be incongruous. Because if you have a robust source and are not engaging in inappropriate extrapolation, you shouldn't be getting only one of those right (again, for the right reasons. You could easily end up right on Zen 6 for the wrong reasons).

And if you know anything about the way management tends to set goals, or how engineers tend to make predications, and it's odd to think that you wouldn't, then 2027 Zen 6 stuff feels so out there that even if you do have a robust source this time around, even if it's coming out of AMD itself, rather than say, reading the tea leaves on some very specific server part based from some far downstream third party source, then, if I were you I'd still be extremely cautious that AMD isn't intentionally putting out bad information as part of an infosec strategy.

Like, the only way it seems it could possibly be right is if AMD has some stuff coming down the pipe in the meantime it feels is competitive, while Zen 6 is an extreme outlier in terms of performance increase. Otherwise, it's like, just about everything seems like it's theoretically doable in a two year span. So planning for the failure with a 'tick' that only comes out three years after an already significantly delayed 'tock', with Zen5c already acting as a 3nm pipecleaner... it simply strains credulity.

You can say "it aligns with DDR6," and that sounds not totally insane until you think about when Zen 5 would have been planned to come out. At that point we're be talking about an intentional four year gap. Was AMD assuming that Intel would implode in the meantime? Insanity.

 

[H433x0n]

In case anybody is interested in perf/watt comparisons gen on gen:

Here’s HUB getting +7% performance at isopower.

Below is TPU getting +9% performance at isopower (both 9700X and 7700 are pulling 80W package power)

 

[LightningZ71]

If bald eagle point dies actually exist, and if it has 16MB of SLC, it MIGHT be a decent uplift over Strix Point. The L3 cache on the C ccx is still goingvto be an issue that pollutes the IMC in heavy MT situations, but in gaming, it should make a notable improvement in iGPU usage situations and somewhat of an improvement in dGPU situations. Should be a great APU for tiny desktops.

 

[carancho]

Council was talking about 1T. The Z1 Extreme is a 8C/16T part and your comparing MT to a 4P+4E/8T M2. Of course Z1 is going to score higher in MT, not to mention the disadvantage of using R23 for ARM based products.

—-

Anyway better in the meantime AMD, starts work on custom ARM silicon. Get a seperate team and kill Qualcomm push before Nvidia makes a move.

This is important. If Intel tanks hard(er), will the supply chain (laptop and server makers) be happy to rely solely on AMD? Or will they push strongly to get multiple competitive ARM providers (Qualcomm, Nvidia, Mediatek), now that the switching cost is so low and that entry into that market is easy?

 

[adroc_thurston]

If Intel tanks hard(er)

They won't, mobile-wise LNL-onwards is neat.

be happy to rely solely on AMD

They're ok, only MS has that weird obsession with WoA.

 

[gdansk]

They're ok, only MS has that weird obsession with WoA.

If they saw the Kepler_L2 Zen 6 roadmap it'd all make sense.

 

[FlameTail]

Anyway better in the meantime AMD, starts work on custom ARM silicon. Get a seperate team and kill Qualcomm push before Nvidia makes a move.

Custom ARM CPU cores?

That will need some significant R&D.

 

[poke01]

That change nothing, from the 20W Z1EX MT score we can deduct that it use 2W per core for 1020 pts per core excluding SMT.

To score the same as the M2 1585 pts ST score a Z1 core would use about 5W, wich according to the power measurements is the power used by a single M2 core to score those 1585 pts.

So despite the lower frequency a M2 performance core is not more efficient than a Zen 4 core at same ST score, and it s undoubtly quite less efficient than a Zen 5 core in the same conditions.

I see what you mean now.

Edit: I like to add that was R23. An M2/M3 core is significantly better in CB 2024.

 

[Goop_reformed]

If they saw the Kepler_L2 Zen 6 roadmap it'd all make sense.

Link?

 

[poke01]

Custom ARM CPU cores?

That will need some significant R&D.

Not really, AMD is rich now.

 

[Jan Olšan]

Tom's single thread benchmarks look pretty nice. My colleague reviewing 9600X also specifically praised the single thread performance (while he got those unimpressive gaming results).

9600X and 9700X rule Windows Geekbench ranking too and the gains were not exactly "5%" (2465 points in Geekbench 5 single-thread score - disclaimer: selfpromo).

 

[gdansk]

Link?

Kepler_L2 said Zen 6 is planned for 2027.
It must be wrong like other things he said. But if it is true and Microsoft heard of the roadmaps it is rational, not obsessive, to redouble their ARM efforts.

 

[carancho]

They won't, mobile-wise LNL-onwards is neat.

They're ok, only MS has that weird obsession with WoA.

Intel Product survives the current situation unscathed? All of the downsizing falls on the legacy node foundries?

 

[yottabit]

View attachment 104719
View attachment 104720
View attachment 104721

View attachment 104722

Did AMD not address every complaint people had with Zen 4's launch? The MSRP is cheaper, they're more efficient, and they're faster. Stock vs stock, on average the 9700X is 5% faster using 35% less power than the 7700X in applications. As a cherry pick, der8auer's video has the 9700X using 44% less power than the 7700X in HWBot X265.

The only people that find this release disappointing are likely gamers, who should be waiting for the X3D chips anyway (or buying a discounted 7800X3D). Even then, from that video, the 9700X is still faster and more efficient:

• Assassin's Creed: 2.19% faster using 9.67% less power.
• Cyberpunk: 1.25% faster using 19.83% less power.
• Counter Strike: 2.43% faster using 4% less power.
• Valorant: 6.02% faster using 5.2% less power.

SFF community will love these chips, the 9950X should be a hit, and I eagerly await the 9800X3D (especially if the rumor it's completely unlocked is true).

Sure… but… I mean we generally expect a grounds-up new architecture to be at least a little faster too, at least confidently outside the margin of error. The things you’re describing sound like what I’d associate with a Zen4+ or some minor refresh. Plus you factor in the ancient Mike Clark hype (and Intel’s flawed execution) and I think it’s clear to see why folks are a little disappointed.

I agree they will be great for SFF, productivity, homelab stuff

 

[adroc_thurston]

Intel Product survives the current situation unscathed

They have a roadmap and they're gonna stick to it.

All of the downsizing falls on the legacy node foundries?

Intel is bloated, layoffs may not impact their operations that much if at all.

 

[FlameTail]

They won't, mobile-wise LNL-onwards is neat.

Doubt

They're ok, only MS has that weird obsession with WoA.

It seems MS grand strategy with WoA is to artificially inject competition into the PC hardware space by making WoA viable. MS want to compete with Apple Silicon Macs. At the end of the day, I dont think they care if it's x86 or ARM. If Intel/AMD can deliver the silicon they want, so be it. Otherwise, they use ARM.

 

[adroc_thurston]

Doubt

LNL is good, and nothing about PTLs seems sus.

WoA is to artificially inject competition into the PC hardware space by making WoA viable

They don't have any besides vague "maybe Wintel alternative magically emerges from the ether".
Also it's Windows on Snapdragon for the most part.

 

[Thunder 57]

9700X doesn't beat 7800X3D? Insert surprised Pikachu face!

 

[gdansk]

9700X doesn't beat 7800X3D? Insert surprised Pikachu face!

Someone at AMD even tried to warn us.
But that it'd hardly beat the 7700 is unexpected.

 

[poke01]

They have a roadmap and they're gonna stick to it.

Umm didn’t Intels roadmap change. Like royal is now split. Nothing is set in stone.

 

[adroc_thurston]

Umm didn’t Intels roadmap change.

Everyone's do.

Like royal is now split

That's super duper far away.
PTL and NVL are near horizon.