Question Zen 6 Speculation Thread

Page 297 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
L

LightningZ71

Platinum Member
Mar 10, 2017
2,601
3,288
136
Whatever LP is implemented with, it must be logically capable of supporting every instruction that the other cores can. They can go even further with densifying the cores at the expense of clocks, and with lower target speeds, spend fewer XTORS on critical paths. Depending on the instruction, they can even move to just supporting some things in microcode. It just must be feature complete.
 
  • Like
Reactions: Tlh97 and marees
Y

yuri69

Senior member
Jul 16, 2013
686
1,223
136
Gideon said:
Unfortunately, considering the crazy focus on AI, today's event might talk much less about CPUs to begin with.

My guess is we'll be lucky to get any info about Zen 7 (even as little as there was about Zen 5 in 2022), but there should be at least some hard metrics regarding Zen 6. I doubt we'll get exact IPC/Clock numbers, though. It's way too far out compared to the three months (for Zen 4) in 2022.
Click to expand...
* Just spades of MI400, Helios/Verano, and a repeated confirmation of the AI roadmap with MI500 and beyond!
* EPYC with a super cool number of the 256c Venice-D providing the best AI experience.
* A generic APU roadmap with many 2027 Medusas for the best AI PC.

AMD has been presenting roadmaps as "current year" -> "current year + 3" => 2025 -> 2028. I still doubt the MLID's A14 Zen 7 would hit 2028.
 
5

511

Diamond Member
Jul 12, 2024
4,855
4,415
106
yuri69 said:
* EPYC with a super cool number of the 256c Venice-D providing the best AI experience.
Click to expand...
EPYCs are good at many thing but not AI cause AMX exists and Pytorch is fully enabled to take advantage of it and all we are going to get is slide saying Zen 7 2028
 
G

Geddagod

Golden Member
Dec 28, 2021
1,572
1,642
106
yuri69 said:
* Just spades of MI400, Helios/Verano, and a repeated confirmation of the AI roadmap with MI500 and beyond!
* EPYC with a super cool number of the 256c Venice-D providing the best AI experience.
* A generic APU roadmap with many 2027 Medusas for the best AI PC.

AMD has been presenting roadmaps as "current year" -> "current year + 3" => 2025 -> 2028. I still doubt the MLID's A14 Zen 7 would hit 2028.
Click to expand...
The A14 part for 2028 is dubious. Unless it's very end of year, and AMD is a lead customer, and they manage to HVM it in early/mid year.

Besides since Zen 7 is also a tock core, they could realistically put it on a half node shrink onto A16 and still be fine. And by half a node shrink, I think A16 actually looks like that, and not a subnode improvement (which is way closer to the base node than it is to the next node) like what a half node improvement is usually called.

I do think Zen 7 is gonna be 2028 though. 2 year cadence.
 
  • Like
Reactions: Tlh97 and marees
5

511

Diamond Member
Jul 12, 2024
4,855
4,415
106
Geddagod said:
Besides since Zen 7 is also a tock core, they could realistically put it on a half node shrink onto A16 and still be fine. And by half a node shrink, I think A16 actually looks like that, and not a subnode improvement (which is way closer to the base node than it is to the next node) like what a half node improvement is usually called.
Click to expand...
A16 Has BSPDN the PPA improvement is from that alone and if Zen 7 is truly A14 it's H2 28 launch at best
 
T

Tangopiper

Junior Member
Nov 11, 2025
6
5
36
Geddagod said:
The A14 part for 2028 is dubious. Unless it's very end of year, and AMD is a lead customer, and they manage to HVM it in early/mid year.

Besides since Zen 7 is also a tock core, they could realistically put it on a half node shrink onto A16 and still be fine. And by half a node shrink, I think A16 actually looks like that, and not a subnode improvement (which is way closer to the base node than it is to the next node) like what a half node improvement is usually called.

I do think Zen 7 is gonna be 2028 though. 2 year cadence.
Click to expand...
N2P HVM is planned for 2H26, A14 HVM is planned for 2H28. Seems to fit quite well into AMD's cadence. Also bear in mind that A14 does not have BSPDN, that comes with A14P in 2029.

A0 parts sent out to CSPs etc. mid-year 2028, with 'official' introduction in either 4Q28 or 1Q29. Sure, low volumes are likely but margin is far more important anyway.
 
marees

marees

Golden Member
Apr 28, 2024
1,920
2,551
96
Geddagod said:
The A14 part for 2028 is dubious. Unless it's very end of year, and AMD is a lead customer, and they manage to HVM it in early/mid year.

Besides since Zen 7 is also a tock core, they could realistically put it on a half node shrink onto A16 and still be fine. And by half a node shrink, I think A16 actually looks like that, and not a subnode improvement (which is way closer to the base node than it is to the next node) like what a half node improvement is usually called.

I do think Zen 7 is gonna be 2028 though. 2 year cadence.
Click to expand...
I believe there could be a desktop zen 6+ refresh

But zen 7 only in 2029 for cient (desktop & laptop). The epyc, especially dense versions, could be rushed out in 2028 though.
 
T

Tangopiper

Junior Member
Nov 11, 2025
6
5
36
marees said:
I believe there could be a desktop zen 6+ refresh

But zen 7 only in 2029 for cient (desktop & laptop). The epyc, especially dense versions, could be rushed out in 2028 though.
Click to expand...
Far more likely they would backport a variation of Zen 7 to N2/A16 nodes akin to Zen 5 N3 to N4.

And only if A14 was not hitting targets.

uArch refreshes have been dead since Zen 3+. OEMs are content with rebrands and refreshes for products launching between new uArchs.
 
  • Like
Reactions: marees
L

LightningZ71

Platinum Member
Mar 10, 2017
2,601
3,288
136
The X3D chips and that tailing XT 5xxx series chip are effectively the "refreshes". The "renames" often offer zero improvement, and a few times regressed in clocks. As much as I would love to see Gorgon Point on N3E with no NPU and the MALL cache like it was originally intended, it's not happening. It would make for a great handheld product, but it's just a dream.
 
  • Like
Reactions: Tlh97 and Tangopiper
adroc_thurston

adroc_thurston

Diamond Member
Jul 2, 2023
7,513
10,289
106
CouncilorIrissa said:
No mention of changes to cache hierarchy even for Zen 7, interesting. It's somewhat expected for Zen 6, but Zen 7? ehhh
(although the target audience probably only cares about AI compute atm)
Click to expand...
They just added buzzwords.
It's not supposed to be informative (the average analyst is a human potato).
 
R

reaperrr3

Member
May 31, 2024
139
411
96
adroc_thurston said:
They just added buzzwords.
It's not supposed to be informative (the average analyst is a human potato).
Click to expand...
Yeah.
One only needs to focus on the 4 key buzzwords they used for Zen6 and Zen7:
"AI - AI - Matrix - AI"

Nobody but investors cares about that, while investors don't care about anything else atm.
 
D

Doug S

Diamond Member
Feb 8, 2020
3,686
6,519
136
Tangopiper said:
N2P HVM is planned for 2H26, A14 HVM is planned for 2H28. Seems to fit quite well into AMD's cadence. Also bear in mind that A14 does not have BSPDN, that comes with A14P in 2029.
Click to expand...

I have been assuming for a while now there will be another iteration of N2, call it "N2PP" for now, that they introduce in H2 2027 to fill the gap between N2P and A14, because I can't see them going two years without even a small improvement. If you want product to reach customer hands in 2028 that's what you're going to use, not A14.
 
  • Like
Reactions: marees
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom