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

[DisEnchantment]

Speculate at will

 

[Timorous]

You claim that Intel has been stagnant for 10 years, and yet AMD has not managed to significantly overtake uarch. Zen1 barely caught up with Broadwell (Haswell with minor tweaks). If it were that easy, they would have crushed Zen1 by now.

Since 2017, AMD and Intel have been going together like a couple in a dance (Intel sometimes loses its steps a bit).

Are you saying this will suddenly change? I believe that Zen4 is so good that a 40% IPC jump from generation to generation is impossible, considering the project implementation time and the possibilities of the transistor packing process per mm2.

Intel had a massive headstart in architecture, node and OEM support. It should have been impossible for AMD to catch up let alone surpass Intel on what looks to be all 3 fronts. They have done so through Intel utterly screwing up on multiple sides and by AMD taking a leaf out of NVs book and becoming an execution machine.

Zen 5 is the 1st ground up design where AMD have had decent R & D money. They could squander it of course but given their execution record over the last 7 years I don't think they will.

Compared to other designs I don't think Zen 4 is that great, it is simply the least bad x86 design.

 

[Kepler_L2]

I wanted to emphasize that in the past, even if there was a very large IPC jump, it was only against the background of the poor solution of the previous generation. Find another example if there was an exception.

Apple, they did many 20-30% IPC increases in a row despite already having the best core.

 

[APU_Fusion]

At this rate I speculate zen5 is slower than zen4 and = to zen3. 🙄

 

[PJVol]

Zen 5 is the 1st ground up design where AMD have had decent R & D money. They could squander it of course but given their execution record over the last 7 years I don't think they will.

To me this “ground up design” assessment (given pretty much to each Zen CPU gen) seems exagerrated. I've always wondered what exactly this "ground up" entails, even compared to the 1st Zen?

 

[moinmoin]

To me this “ground up design” assessment (given pretty much to each Zen CPU gen) seems exagerrated. I've always wondered what exactly this "ground up" entails, even compared to the 1st Zen?

The only Zen gens with ground up designs are 1, 3 and the upcoming 5. Ground up means that the floor plan of the core is being significantly changed. For Zen 2 and 4 it was just expanded upon.

 

[Markfw]

Compared to other designs I don't think Zen 4 is that great, it is simply the least bad x86 design.

Zen 4 includes server. Also, its the 1st generation to include basically a full avx-512. Also, with speed as well as generalization its far from Zen 3 in both departments. And if you add efficiency in server or in desktop using ECO mode, its light years past Zen 3. Speaking from someone who had massive Zen 3 boxes and upgraded everything to Zen 4, including server hardware.

 

[PJVol]

The only Zen gens with ground up designs are 1, 3 and the upcoming 5. Ground up means that the floor plan of the core is being significantly changed. For Zen 2 and 4 it was just expanded upon

Based on your definition, each Zen CPU generation is "ground up". But even then, the most radical change in the design was with Zen 2 (not counting Zen 1, of course)

 

[Joe NYC]

The clock regression was supposed to happen because Zen 5 was originally supposed to be on 3nm (N3). TSMC (N3) silicon is more energy efficient but sucks at clock speed compared to N5/N4. A bulkier wider core and more bandwidth is not going to affect the max CPU clock speed. AMD will not be on 3nm until a Zen 5 refresh.

For efficiency sake, it would have been nice to see Zen 5 on 3nm.

It's unlikely there will be a Zen 5 refresh.

 

[adroc_thurston]

AMD got definitely not a stellar track record of making "brand new things".

What is zen3

 

[PJVol]

What is zen3

Hella lot of post-layout optimisations, Vth swapping and bias tuning, plus new L3 and its ring bus?
ed: forgot TSVs

 

[Fjodor2001]

The only Zen gens with ground up designs are 1, 3 and the upcoming 5.

Is Zen3 really as big of a redesign as Zen1 was?

Also, when people have said that Zen5 will be a complete redesign, I was kinda hoping it would be like with Zen1. But should Zen5 more be compared to Zen3 in this regard? Should kinda lower the performance improvement expectations for Zen5 in that case.

 

[inf64]

Hella lot of post-layout optimisations, Vth swapping and bias tuning, plus new L3 and its ring bus?

Wrong about Zen 3.
Also, what is Zen1?

 

[PJVol]

Wrong about Zen 3.
Also, what is Zen1?

In what exactly?
Are we talking about the CCX design only or the CPU as a whole?
And don't quite understand about Zen 1

 

[adroc_thurston]

Is Zen3 really as big of a redesign as Zen1 was?

yea.
yuo can see in on the floorplan.

 

[Joe NYC]

I don't know if there is anything new here, or just summary of the "consensus"

https://twitter.com/x/status/1788507350707617897

 

[Goop_reformed]

I don't know if there is anything new here, or just summary of the "consensus"

https://twitter.com/x/status/1788507350707617897

Recycled infos. IPC/ST uplift, which is the selling point for zen 5, is still missing. Sad

 

[Fjodor2001]

I don't know if there is anything new here, or just summary of the "consensus"

https://twitter.com/x/status/1788507350707617897

For Granite Ridge: 2x RDNA3 CU

Previously it was believed to be RDNA2, same as for Zen4. If actually RDNA3, it will mean new IOD, so not same as for Zen4.

But I suspect it could be a typo, since it also says ”Raphael IOD (N6)”, i.e. same IOD as for Zen4.

 

[Joe NYC]

For Granite Ridge: 2x RDNA3 CU

Previously it was believed to be RDNA2, same as for Zen4. If actually RDNA3, it will mean new IOD, so not same as for Zen4.

But I suspect it could be a typo, since it also says ”Raphael IOD (N6)”, i.e. same IOD as for Zen4.

Good catch. Probably a mistake / typo.

 

[del42sa]

I don't know if there is anything new here, or just summary of the "consensus"

https://twitter.com/x/status/1788507350707617897

having RDNA 3,5 clocked at 2,6 - 2,7Ghz means they managed to achieve worse GPU frequency than Phoenix RDNA3 GPU....

 

[inf64]

In what exactly?
Are we talking about the CCX design only or the CPU as a whole?
And don't quite understand about Zen 1

In a way that people who designed it disagree with you, and I tend to believe them more:

“Zen 3”: AMD 2nd Generation 7nm x86-64 Microprocessor Core

“Zen 3”: AMD 2nd Generation 7nm x86-64 Microprocessor Core - Download as a PDF or view online for free

www.slideshare.net

For Zen 1 it's a huge uplift versus previous core and a major redesign - they used bits and pieces from BD and Bobcat cores and expanded on them; they also implemented brand new stuff into Zen 1 core. Similar will be the case with Zen 5.

 

[Panino Manino]

So true, Zen5% has been downgraded to Zen0%.

People are laughing now, but soon I'm the one who will laugh seeing the tears of disappointment.

Zen 5 will be good, yes, but not good enough.

 

[gdansk]

I'm guessing the 10%ers are as wrong as the 32%ers
I don't think either side will be crying but cherrypicking.

 

[jpiniero]

People are laughing now, but soon I'm the one who will laugh seeing the tears of disappointment.

Zen 5 will be good, yes, but not good enough.

That's ok, because the only thing that matters currently is the AI performance

 

[adroc_thurston]

I'm guessing the 10%ers are as wrong as the 32%ers
I don't think either side will be crying but cherrypicking.

Nope.

 

[Timorous]

I'm guessing the 10%ers are as wrong as the 32%ers
I don't think either side will be crying but cherrypicking.

What is to cherry pick?

The claim is very specific in that it refers to Specint 2017 1t performance.

Either it hits the claimed figure or it does not, there is no wiggle room on it.