- Mar 3, 2017
- 1,659
- 6,101
- 136
Not in the roadmap yet, too early for you sorryNo.
Neither exist so ughhh.
Yeap, AMD has been using chiplets for long time. But if they want to double core count, then they will have to integrate more cores by using more advanced process. Upcoming Turin Zen5 will employ 16 pcs 8-core CCD to get 128 cores. And here is even further down the road, Venice platform with 256-core, guess how many chiplets would AMD use?AMD moved to chiplets because nodes were expensive enough to make even a 2-chiplet SKU cheaper than the monolithic counterpart. Now that nodes are getting even more expensive, you expect them to go back to monolithic?
The future is disagg everywhere.This can easily change in the future, especially when the performance targets shift.
Is the 32C chiplet the "C" variant though? And the regular one will be just 16C, and that will be the one AMD is gonna use for desktop (as a step of from 2 8-cores ones, resolving the latency issue?) I see the last paragraph on that picture, but it kinda sounds too good to me - after x years lingering with 8 core chiplets, then jumping to 16C with Zen6 and 32C right after that with Zen7, and even using it for desktop/mobile? I mean, i would love for you to be right, but...Yeap, AMD has been using chiplets for long time. But if they want to double core count, then they will have to integrate more cores by using more advanced process. Upcoming Turin Zen5 will employ 16 pcs 8-core CCD to get 128 cores. And here is even further down the road, Venice platform with 256-core, guess how many chiplets would AMD use?
View attachment 92682
8 chiplets man. From 16pcs chiplets down to 8 chiplets with double core counts, ie. 256 cores. And one tCPU will be used for future AM6 platform to fight with upcoming Nova Lake...As stated by MLID, there is no mention about Zen7 in the roadmap like adroc said. Because there won't have seperate CCD for client, AMD just use the same 32-core chiplet for AM6 client platform. Therefore, Zen7 is actually a 32-core Zen6 chiplet build by N2 process combine with DDR6 IOD.
Zen 6 brings a different interconnect compared to previous AMD designs. Reportedly this also brings a split in the CCD design for desktop and server.Is the 32C chiplet the "C" variant though? And the regular one will be just 16C, and that will be the one AMD is gonna use for desktop (as a step of from 2 8-cores ones, resolving the latency issue?)
Evertything means everything.Low end mobile will probably stay monolithic
There's zero relation between server and client parts for Zen6.And the regular one will be just 16C, and that will be the one AMD is gonna use for desktop (as a step of from 2 8-cores ones, resolving the latency issue?)
Opinion discarded.According to MLID
Opinion discarded.
It's when it is.So you still think around april is realistic?
And STX1 is computex.I sure hope so.
No.but more because I think it was always fall 2024
You launch when it's ready and they got lucky.I mean April makes no sense whatsoever, apart from just wanting to launch a product in the most random way possible.
It's too good to show anything.AMD showed NOTHING so far, not even that ZEN5 is running
Comp pressure was real.Look at ZEN4: Teaser at CES, some information at Computex (with sandbagging), launch in fall
Because it's so good Zen4 sales would die overnight.Why wouldn't they show ANYTHING about ZEN5?
There is none.Boards definitely seem to be Fall 2024
No?AMD would completely destroy their own Strix Point 8x50 launch in June.
That would mean at least 50-60% single thread lead for me, because ZEN4 did upper 20s and didn't sell until X3D.Because it's so good Zen4 sales would die overnight.
Yeah, and would look old at launch event because of 8xxx branding.It's for premium AI PCs.
His delay claim isn’t about the motherboards - it was that Turin (and possibly Granite Ridge) needed a new stepping. I don't really buy that personally.According to MLID Zen 5 will arrive late because of Mainboard issues. He says that the announcement will probably be around computex with availability later...does not sound good to me. I thought april was kinda save...
yeah whatever that means for you.That would mean at least 50-60% single thread lead for me, because ZEN4 did upper 20s and didn't sell until X3D.
No, since it has the Windows AI premium badge.Yeah, and would look old at launch event because of 8xxx branding.
Just like Zen3 sales died overnight?Because it's so good Zen4 sales would die overnight.
That's a different platform with more costs and lower gen-on-gen perf bump.Just like Zen3 sales died overnight?
That's what I want to say...Just like Zen3 sales died overnight?
That's not the real number, the real number sounds like pure bs.Adroc says 32% in SpecINT
Alright, so you posted a made up BS number here in this forum months ago because the real number would sound like BS, got it.That's not the real number, the real number sounds like pure bs.
You implied 40 percent last week, so even that is too low?That's not the real number, the real number sounds like pure bs.
It just showcases the general direction of Z5 and future AMD cores (chunggus IPC bumps to make dense a very viable option across data center in general, not just cloud in particular).
Yeah, leaps like those just don't happen until they do.Alright, so you posted a made up BS number here in this forum months ago because the real number would sound like BS, got it.
how do I even say that without sounding schizo?You implied 40 percent last week, so even that is too low?