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

[DisEnchantment]

Speculate at will

 

[FlameTail]

Strix was always Q3 anyway, X Elite remains too late to change the landscape, sorry.

I disagree.

So Computex for the desktop "reveal" show. Then launch some time in Aug. This makes sense unlike the April rush.

Strix and Turin later => Sep-Nov. Real availability of Zen 5 laptops is still 2025.

As Yuri point out, even if Strix gets announced sometime in 2024H2, It will take a while for widespread availability of laptops with it, as is usual for AMD. So 2025 it is.

X Elite is coming in mid-2024, with widespread availability of devices. Ian Cutress said he heard from some industry people that 17 laptops with X Elite will be revealed at Computex 2024. At that time, X Elite's competition is going to be Hawk Point and Meteor Lake, and from the performance previews Qualcomm has shown, it's better than either of them.

So X Elite has a healthy time window of 3-6 months to make an impression, before Lunar Lake, Arrow Lake Mobile and Strix all arrive.

 

[branch_suggestion]

I disagree.

As Yuri point out, even if Strix gets announced sometime in 2024H2, It will take a while for widespread availability of laptops with it, as is usual for AMD. So 2025 it is.

X Elite is coming in mid-2024, with widespread availability of devices. Ian Cutress said he heard from some industry people that 17 laptops with X Elite will be revealed at Computex 2024. At that time, X Elite's competition is going to be Hawk Point and Meteor Lake, and from the performance previews Qualcomm has shown, it's better than either of them.

So X Elite has a healthy time window of 3-6 months to make an impression, before Lunar Lake, Arrow Lake Mobile and Strix all arrive.

That is nice and all, but the issue is that AMD will reveal the performance of all of their H2 products at Computex. Strix should be ready for some rudimentary benchmarking, even if final Si isn't ready.
This would be AMD's biggest keynote since Computex 2019, full spectrum Z5 reveal+RDNA4. Sure the Advancing AI event is more important, but I think this will stick out to more people.
So before Qualcomm gains any meaningful traction, people will be waiting for Z5 instead. Better luck next time.

 

[coercitiv]

I'm old enough to remember AMD saying that they are adhering to a 12 - 18 month schedule for their CPUs. first slipped by Zen 4 due to CXL and Covid. Now it's again nearly 22+ months.

Quote from Mark Papermaster for confirmation:

Dr. Ian Cutress: So far AMD’s rate of new products is on track to produce a new core almost every year. The roadmaps quite proudly showcase Zen 3 as almost ready, Zen 4 in development, and Zen 5 further out. Is this cadence sustainable?
Mark Papermaster: We’re on a 12-18 month cadence, and we believe that is sustainable. It’s what the industry demands from us.

 

[adroc_thurston]

has a healthy time window of 3-6 months

No?
It's September.
You really should forget about QC doing anything PC, they don't get the market.

 

[MarkizSchnitzel]

Is it expected for Zen5 to be competitive on efficiency with Elite X?
I will be in a market for mid range laptop and literally just need Chrome/Chromium. Battery life will be deciding factor.

 

[Thunder 57]

https://twitter.com/x/status/1752472072339472806

Paul Alcorn from Tomshardware is saying Consumer Zen 5 will be launched in H2 2024. This is directly from AMD.

Paul Alcorn is no fan of AMD. Just read some of his "Best CPU" articles. He may be right, but I can't see Zen 5 slipping to H2 which would generally translate to Q4.

 

[branch_suggestion]

Is it expected for Zen5 to be competitive on efficiency with Elite X?
I will be in a market for mid range laptop and literally just need Chrome/Chromium. Battery life will be deciding factor.

Both are N4P, so it will be a straight battle of who did better. So many variables with mobile SoC's.

 

[FlameTail]

Is it expected for Zen5 to be competitive on efficiency with Elite X?

Depends on what your expectations are and how you define competitiveness.

Zen 5 will certainly be more powerful than Oryon Phoenix (first gen Oryon core). However what's not certain is Zen 5's efficiency.

Oryon Phoenix is essentially having Apple Silicon P-core levels of efficiency.

We all know how formidably efficient Apple Silicon is. So then the question becomes whether Zen 5 will be reaching the efficiency level of Oryon/Apple Silicon.

I don't think so.

I will be in a market for mid range laptop and literally just need Chrome/Chromium. Battery life will be deciding factor.

X Elite you mean?

 

[Kolifloro]

They literally decided the launch date during the conf call

They didn't but all three relevant parts were H1 last November.

Really ?????????

Does anybody has any kind of cue about why to postpone ?

What were the 'reasons' about the change ?

Was it something regarding 'marketing' , 'over-stock' , 'Intel' , 'supply chain' , 'TSMC available slots' , 'NPU' ... maybe 'technical' reasons ?

 

[Kepler_L2]

Really ?????????

Does anybody has any kind of cue about why to postpone ?

What were the 'reasons' about the change ?

Was it something regarding 'marketing' , 'over-stock' , 'Intel' , 'supply chain' , 'TSMC available slots' , 'NPU' ... maybe 'technical' reasons ?

For Turin it seems another IOD stepping was needed to reach target memory frequency (DDR5-6400). Not sure why GNR was delayed also.

 

[MarkizSchnitzel]

X Elite you mean?

I do.
Thanks for the reply. I have a feeling I will need a laptop sooner then Zen5 will be actually on the shelves in mid range laptops. But then again, I have no idea will there be 800-1000€ laptops with X Elite even.

The point of my question was that a lot of people will take silent efficient laptop over x86 software compatibility. If Snapdragon can deliver there in various price points, they will do well. I hope it will not all be in the prosumer or 1000€+ bracket..

 

[misuspita]

I'm not sure people buy medium expensive laptops that are softwareless... yet. In time, sure, the software will ketchup, but now?

I personally am interested in what AMD will do to Strix Halo, because just relying on system memory will be a giant bottleneck. My current 5700G still delivers, 8xxx series hasn't made upgrade a itch for me yet.

 

[FlameTail]

I do.
Thanks for the reply. I have a feeling I will need a laptop sooner then Zen5 will be actually on the shelves in mid range laptops. But then again, I have no idea will there be 800-1000€ laptops with X Elite even.

The point of my question was that a lot of people will take silent efficient laptop over x86 software compatibility. If Snapdragon can deliver there in various price points, they will do well. I hope it will not all be in the prosumer or 1000€+ bracket..

X Elite is not the only SKU. There was leak about a lower end X Plus as well. Follow the Qualcomm thread for the news and discussion!

Regarding software compatibility, the majority of home/casual software already has ARM native versions. Web browsers, Microsoft Office, etc...

 

[Shivansps]

Software is not that much of a problem, it mostly run behind translation, drivers are the main problem. Arm64 drivers are very very rare.

 

[CouncilorIrissa]

inb4 it’s another RDNA3-like dumpster fire with missed targets 3 months before release

 

[Glo.]

When there is silence about a product in the ether - its good sign.
When there rumors are being spread like madness - its a bad sign.

The best example is Turing vs RDNA2 generation. Bad news about Nvidia spread immediately, when AMD's best GPU generation in years was hidden under tight lips of industry sources.

Now the same thing as RDNA2 is happening with Zen 5 products. Well, if we exclude what Adroc is doing, of course.



Same thing can be said about Intel. If there is silence - its good news. If the leaks are happening - its bad news for the product.

 

[eek2121]

I disagree.

As Yuri point out, even if Strix gets announced sometime in 2024H2, It will take a while for widespread availability of laptops with it, as is usual for AMD. So 2025 it is.

X Elite is coming in mid-2024, with widespread availability of devices. Ian Cutress said he heard from some industry people that 17 laptops with X Elite will be revealed at Computex 2024. At that time, X Elite's competition is going to be Hawk Point and Meteor Lake, and from the performance previews Qualcomm has shown, it's better than either of them.

So X Elite has a healthy time window of 3-6 months to make an impression, before Lunar Lake, Arrow Lake Mobile and Strix all arrive.

Not really.

No?
It's September.
You really should forget about QC doing anything PC, they don't get the market.

This.

Is it expected for Zen5 to be competitive on efficiency with Elite X?
I will be in a market for mid range laptop and literally just need Chrome/Chromium. Battery life will be deciding factor.

Doesn’t matter for most folks.

Remember that even if the qualcomm part hits performance targets with no finagling by qualcomm, you have to multiply any benchmark numbers by around 0.65 to account for WOA (or <insert emulator of choice> overhead unless there is a native port.

Emulation is also not battery life friendly.

That is why many are pessimistic about ARM. The x elite would essentially need to be twice as fast as Zen 4 to be able to run non-native software at Zen 4 speeds.

Zen 5 moves that goal post further away.

Before you start blabbing about software being ported over: As a developer, why would I bother with a native port for a niche platform when I can simply let them run it under an emulator or not at all? We are talking about a platform that has smaller marketshare than Linux.

Zen 5 is going to significantly outsell the X elite. I would rather spend my time on optimizing for AMD instead.

 

[FlameTail]

Before you start blabbing about software being ported over: As a developer, why would I bother with a native port for a niche platform when I can simply let them run it under an emulator or not at all? We are talking about a platform that has smaller marketshare than Linux.

Zen 5 is going to significantly outsell the X elite. I would rather spend my time on optimizing for AMD instead.

Developers certainly seem to be a big roadblock for Qualcomm/Microsoft's aspirations for Windows On ARM.

I came across such an app developer on reddit, who had to say some 'not so nice' words regarding the matter.

We can continue the conversation in the Qualcomm thread.

 

[adroc_thurston]

Developers certainly seem to be a big roadblock for Qualcomm/Microsoft's aspirations for Windows On ARM.

The big roadblock is Qualcomm itself, they have no idea how to approach PC as a market.
It's been 6 years since WoA got started and it has less market presence than ever.

 

[FlameTail]

The big roadblock is Qualcomm itself, they have no idea how to approach PC as a market.
It's been 6 years since WoA got started and it has less market presence than ever.

Well, you should know that Qualcomm hasn't been taking WoA seriously before. Their previous SoCs were essentially phone chips repurposed for laptops.

The wind began to sail in the opposite direction when Qualcomm acquired Nuvia in 2021.

Now, we are finally seeing the fruit of that acquisition. WoA finally is getting good hardware.

And Qualcomm will not remain as the sole WoA silicon provider for long. Once the exclusivity agreement runs out this year, the doors are open for others. Especially, Nvidia will be making an ARM SoC in 2025.

 

[adroc_thurston]

Well, you should know that Qualcomm hasn't been taking WoA seriously before

Yea they did, the issue is them treating PCs as a margin padding for their premium handset biz.

Their previous SoCs were essentially phone chips repurposed for laptops.

No they weren't.

The wind began to sail in the opposite direction when Qualcomm acquired Nuvia in 2021.

Ugh. No?
Hamoa solves roughly no problems with QC approach to PC.

WoA finally is getting good hardware.

Need to put like 30 asterisks here.
They start somewhere around "unusable GPU".

 

[soresu]

Well, you should know that Qualcomm hasn't been taking WoA seriously before. Their previous SoCs were essentially phone chips repurposed for laptops.

Not sure how you got that idea.

When Snapdragon 8cx and Microsoft SQ1 launched (basically the same overall hardware with minor differences) they had something like twice the GPU ALU's as the smartphone SD flagship of that year.

Beyond that there really wasn't much they could do about the CPU problem, given that the smartphone flagship SoC's were already trying to get as much juice out of the big Axx core designs as possible

 

[FlameTail]

Not sure how you got that idea.

When Snapdragon 8cx and Microsoft SQ1 launched (basically the same overall hardware with minor differences) they had something like twice the GPU ALU's as the smartphone SD flagship of that year.

Beyond that there really wasn't much they could do about the CPU problem, given that the smartphone flagship SoC's were already trying to get as much juice out of the big Axx core designs as possible

Yeah I was mainly talking about the CPU, not GPU.

 

[soresu]

Yeah I was mainly talking about the CPU, not GPU.

Exactly.

They tried and failed to make a viable CPU core in the v8-A generation with Kryo.

Having fallen flat on their face with that and been forced to eat humble pie once already to go back to ARM Ltd for A73+ they weren't willing to make that same mistake again.

(until they did so for ARM servers instead with Falkor)

 

[soresu]

Under those constraints there really wasn't much of an option for QC in this.

Apple snuck up on everyone with A7.

Not only far beating A57 and Kryo in raw performance, but also beating the first A57 silicon implementation to market - becoming the first ARM64 SoC in the process.

IMHO QC's modem market dominance went to their heads, and they got sloppy in planning for the future.