Question Zen 6 Speculation Thread

Page 220 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Win2012R2

Golden Member
Dec 5, 2024
1,208
1,246
96
There is no other choice.
Intel is in "managed decline" state which will become unmanageable once AMD grabs another 15%+ of market share (by money, not units), plus AMD will have juicy "AI" income, until the bubble bursts at least, which is probably a little further than a galaxy far far away ...
 
  • Like
Reactions: Tlh97

StefanR5R

Elite Member
Dec 10, 2016
6,672
10,561
136
I'm assuming that they will attack the problem from a fresh perspective.
I for one am assuming that they will attack the problem from the perspective of folks who know their trade. Why am I saying that? Just recently, they took their processors and swapped semi-wide cores (Zen 4) out for quite wider cores (Zen 5) but kept the IFOP and IOD. (As far as desktop is concerned.) That's the part which everybody noticed. But they did more. Fatter cores --> more data traffic --> what happens if data traffic increases through a given fabric? Right: Latency tanks... if you do nothing about it. However, they apparently actually did something about it, although almost nobody took note. :-)

Also, I read a story somewhere that they received complaints about lackluster latency figures in a popular memory benchmark. So they implemented a BIOS option as a concession to these complainers which alters prefetcher policy or something, such that the benchmark figures look nicer... while realworld workloads tend to regress when this option is enabled. So, maybe what we want is good latency, not least latency just for least latency's sake?

Of course, my assumption goes out the window if their engineers simply give up and accept high RAM latencies as a fact of life.
AMD make central processors, graphics processors, FPGAs, recently they dabble in networking I think — I don't see them becoming a DRAM/MRAM/xyzRAM developer in addition. :-) Rather, I think that they will stick a little longer with the old recipe of adding a little SRAM into the memory hierarchy as a means to deal with DRAM latency.
 

adroc_thurston

Diamond Member
Jul 2, 2023
7,094
9,850
106
like i said above they botched the interconnect on ARL
that's not the problem.
2.5D slabs just don't do anything by themselves.
Where there is a will, there is a way. If Intel is serious about clawing back lost marketshare, they will do it. There is no other choice.
that's not how any of that works, kiddo.
this ain't battle shounen, you're not gonna just will your way thru fabric topologies.
 

LightningZ71

Platinum Member
Mar 10, 2017
2,508
3,190
136
Intel has been doing advanced packaging and multi chip modules since Lakefield in 2019 and more primitively in Kaby Lake G. There's no excuse for Intel to be whiffing this badly by this point
 

511

Diamond Member
Jul 12, 2024
4,523
4,144
106
They did Chiplets before anyone did tbh they just love to fumble the ball so hard for the past 15-20 years.
 
  • Haha
Reactions: Thibsie

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,183
136
no? Foveros is far superior to the cheap IFOP it's thier clocks and the bad fabric design that is causing this issue on Intel
Is it really superior?

For what AMD is using it for, short distance link between dies, InFO is fine.

But Foveros is far more costly. Could be up to 7x the cost.
 

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,183
136
I am "speculating" in a thread specifically opened up for "speculation".

There is no point on going on a tangent that is almost certainly not going to happen.

I would actually think that if AMD was going to use N2 on anything, it would be mobile where they could make some headway in market share in mobile.

Mobile is where the greatest need is, for AMD, to gain competitive advantage. I don't see anything in the leaks that would accomplish that.

Whole mobile CPU on N2 is a non-starter due to cost.

What one of the leaks showed was an N3p monolithic chip with some LP cores, few full, few dense cores that would then attach to the big N2 12 core chiplet. Which would be the Medusa Point mobile CPU.

Which would leave the monolithic chip alone for some lower end segments. Which is a good re-use of chiplet, but maybe not exactly the silver bullet AMD needs in mobile.

I have a feeling that even a N3P Zen 6 on the desktop would be sufficient to eclipse the best Intel will have within that same time frame. I could be wrong though.

We have another AMD quarter behind us (since we addressed this subject), and the 2 AMD leading ladies on the call were again all giddy about the sky high ASPs and market share gains in desktop.

Zero reason to reverse the strategy, that lead them to insurmountable lead against Intel. If anything, I could see them doubling down on the strategy.

Which I would foresee as:
9600x3d, 9800x3d low to mid range
Zen 6 N2 12 cores V-Cache premium
non-V-Cache entry level

Which would leave Intel only with big / expensive bLLC as a competitive chip and everything else would be uncompetitive. Including all of the Intel fabbed deskto CPUs.
 

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,183
136
Yes there was a comparison
The IO density/power characteristics are superior sadly don't have actual bandwidth figure

View attachment 128955

Sorry, I got the acronyms crossed.
You were talking about IFOP - Infinity Fabric on Package
while I thought you were comparing Foveros with InFO - Integrated FanOut

Which is what my reply addressed.
 
  • Like
Reactions: Tlh97 and 511
Jul 27, 2020
28,002
19,125
146
That would be some glorious trolling AMD could do.
Nah. They could've done so much more to absolutely crush Intel. They have a philosophy of co-existence and I think it's because they are not confident that they can be the sole torchbearer for x86. They need to get to 80% marketshare before they can justify the R&D expense that responsibility would entail. So they are just letting Intel suffer just enough to increase their marketshare gradually and incrementally. They don't suddenly want Intel's volume because they are not equipped to handle it.
 

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,183
136
Intel doesn't have the budget for anything and the second layer won't do anything.
You can't force a race with ~nothing.

That's the trolling aspect of it. Intel does not have money to redesign the die, it would be terribly expensive.

And when Intel thought they caught up with AMD in L3 Lisa - I mean Lucy goes:

giphy.gif
 
Last edited:

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,183
136
Nah. They could've done so much more to absolutely crush Intel. They have a philosophy of co-existence and I think it's because they are not confident that they can be the sole torchbearer for x86.

If you think AMD is holding back to keep Intel alive - no. AMD actually lost share in mobile last quarter to Intel (in units)

They need to get to 80% marketshare before they can justify the R&D expense that responsibility would entail. So they are just letting Intel suffer just enough to increase their marketshare gradually and incrementally. They don't suddenly want Intel's volume because they are not equipped to handle it.

AMD is already spending om client R&D at the level to cover all of the market. But AMD has not penetrated the Intel-OEM lock. And really, has not offered a product in mobile that would convince OEM that it is so good they can leave Intel behind and switch to AMD.

So why not try trolling?
 
Jul 27, 2020
28,002
19,125
146
So why not try trolling?
They are going for profit over market penetration. It's like they don't think they can continue to make great products so may as well make as much money as possible. Prime example is not changing the IOD for Zen 5 for DDR5-8000 1:1 RAM speeds. Also, no Strix Halo with quad channel DDR5-8000 for desktop. They can certainly do a lot better but they are more occupied with making money and playing it safe rather than making a serious effort to drive Intel out of the market with products that Intel has no hope of beating.
 
Jul 27, 2020
28,002
19,125
146
what's the thread title.... that's right.... zen 6
We don't have any indication that Zen 6 will be anything more than 15% ST and maybe 40% MT. Meanwhile, Intel is bringing core spam to the party with Nova Lake. Even if Intel messes up the ST performance, they still end up being competitive in the MT department with real threads instead of virtual ones. That's not what the market wants. What the market really wants is

1) Intel being forced to bring out an expensive to manufacture CPU that beats Zen 6 decisively, at least in MT and approaching HEDT performance of Threadripper.

2) Intel being forced to sell their lower performing CPUs at a hefty discount compared to Zen 6.

No.1 compels AMD to strike back with more power in whatever way they deem necessary.

No.2 forces AMD to respond in kind and get into a serious price war.

We, the consumers, watch with glee and eat popcorn and enjoy better, cheaper CPUs.

The upcoming face-off between Zen 6 and Nova Lake seems to be more of the same. Boring status quo. Nothing exciting.

Competition needs to heat up.
 

DrMrLordX

Lifer
Apr 27, 2000
22,901
12,970
136
like i said above they botched the interconnect on ARL
And Meteor Lake (and Arrow Lake-U, presumably). And Lakefield. There seems to be a recurring problem here . . . in any case, it's hard to see how they can compete with Zen6 when they keep pushing inadequate chiplet designs.
 
  • Like
Reactions: Tlh97 and Joe NYC