Question Zen 6 Speculation Thread

[511]

that's a very long wait and still not 2nm? apple will definitely release 2nm in 2025

Nope 6C is N2 though also N2 is entering HVM in H2 25 so ain't no way we getting N2 Chips before 2026

 

[OneEng2]

Nope 6C is N2 though also N2 is entering HVM in H2 25 so ain't no way we getting N2 Chips before 2026

Agree. We could potentially see desktop Zen 6/6c on N3P in 2025, but I have been seeing late 2026 as the target (which makes sense compared to AMD's release cycle from past releases).

It may be that we see the server part before the desktop part for Zen 6.

It does give Intel a window of opportunity from late 2025 to late 2026 for both CWF and Panther Lake to have their moment in the sun.

 

[Win2012R2]

apple will definitely release 2nm in 2025

Apple pay in advance for the whole node to get exclusivity on initial volumes.

I got sneaky feeling that due to so many changes in N2 it will take at least one revision to get it good (A16?), so everyone else might as well wait a year.

 

[511]

I hope we don't get AI sticked on Zen6

 

[DAPUNISHER]

Deleted posts about Intel. Want to talk about Intel? Do it in an Intel thread.

Mod DAPUNISHER

 

[igor_kavinski]

I hope we don't get AI sticked on Zen6

Would be funny if there is an NPU at the center and it is surrounded by CPU cores. Largest silicon budget devoted to the NPU because local LLMs and image generation are supposedly the future.

 

[Chicken76]

Is there any early estimation on N2 yields? N7 was very good IIRC and N5 even better. N3 is worse than N7 but not by much, right?

 

[Win2012R2]

Is there any early estimation on N2 yields?

"According to a report by Liberty Times, industry sources indicate that TSMC’s 2nm trial production yield has exceeded 60%, surpassing expectations"

Source: https://www.trendforce.com/news/202...ile-ex-intel-ceo-gelsinger-questions-metrics/

 

[Doug S]

That conflicts with the news that they "reached risk production yield targets" in late May and officially entered risk production in early July. Based on their graphs showing yield when they've entered risk production for previous "good" nodes (i.e. ignoring N3B) that means ~ 80%.

But what they are using to measure that 80% and what is being reported as exceeding 60% may be different dies and/or different ways of measuring yield (i.e. not just "it works" but "it works and hits targeted power/performance")

 

[Win2012R2]

There are so many crazy cool new things (starting with the new transistor) in N2 that I find it difficult to believe they'll have less problems to solve than with N3(B). On the other hand wafer cost is reported to be 50% higher at $30k, which perhaps helps accepting lower than normal yields at this stage and still make it acceptable to produce, then shift everything to improved A16 in 2026.

 

[Bigos]

I have to agree with a recent Pat post, saying yield is x % without any context is really not saying anything. Yield has to be quantified with at least the die size, and better the actual structure that is being tested (i.e. SRAM often has redundant structures that improve yield at the cost of space).

 

[eek2121]

Agree. We could potentially see desktop Zen 6/6c on N3P in 2025, but I have been seeing late 2026 as the target (which makes sense compared to AMD's release cycle from past releases).

It may be that we see the server part before the desktop part for Zen 6.

It does give Intel a window of opportunity from late 2025 to late 2026 for both CWF and Panther Lake to have their moment in the sun.

I wouldn't get too caught up in this. I know the releases are slow, however they won't even be done with Zen 5 for a bit. Last I checked, there are even plans for some additional desktop SKUs that aren't in the "budget" or "APU" ranges, we'll see.

 

[OneEng2]

I wouldn't get too caught up in this. I know the releases are slow, however they won't even be done with Zen 5 for a bit. Last I checked, there are even plans for some additional desktop SKUs that aren't in the "budget" or "APU" ranges, we'll see.

If you look at it from a financial point of view, AMD has no reason at all to move to a more expensive node (N3P) from their existing node (N4P) as long as Zen 5 is competing favorably with Intel (which it is).

Should Intel release Panther Lake in late 2025 on 18A and it bests Zen 5, I think you will see Zen 6 sooner rather than later in 2026. Same for next gen EPYC. If Intel's Clearwater Forest on 18A does really well, then you will see Zen 6 Turin on N2 get released..... but not before.

This is the behavior of all big companies without competition. Milk your investment.

 

[DrMrLordX]

If you look at it from a financial point of view, AMD has no reason at all to move to a more expensive node (N3P) from their existing node (N4P) as long as Zen 5 is competing favorably with Intel (which it is).

Depends on how much institutional memory AMD brings to the table. One of Intel's continued failures is its inability to compete with itself. You could see it with Ivy Bridge (and to a lesser extent, Haswell) vs Sandy Bridge, and you can also see it in Arrow Lake vs Raptor Lake.

AMD will always need to improve itself, no matter what happens to Intel. The moment they forget that fact, they die.

 

[Hans Gruber]

If you look at it from a financial point of view, AMD has no reason at all to move to a more expensive node (N3P) from their existing node (N4P) as long as Zen 5 is competing favorably with Intel (which it is).

Should Intel release Panther Lake in late 2025 on 18A and it bests Zen 5, I think you will see Zen 6 sooner rather than later in 2026. Same for next gen EPYC. If Intel's Clearwater Forest on 18A does really well, then you will see Zen 6 Turin on N2 get released..... but not before.

This is the behavior of all big companies without competition. Milk your investment.

You do realize Zen 5 was a complete flop? It took the x3D chips to save it. On the server side, excellent. On the PC side, awful release. AMD should use N3P for Zen 6 and make the x3D chips standard. AMD should consider releasing a Zen 5+ on 3nm mid 2025. N3P for Zen 6 and Zen 7 would be the play for AMD.

 

[adroc_thurston]

AMD should consider releasing a Zen 5+ on 3nm mid 2025

No such thing exists.

N3P for Zen 6 and Zen 7 would be the play for AMD.

Zen7 is N2p and so on era.

 

[Win2012R2]

Zen7 is N2p and so on era.

No backside power delivery? Since N2 starts next year for Apple, doesn't that mean N2p ready 2027 max, where does that put Zen 6, early 2026?

 

[adroc_thurston]

No backside power delivery?

TSM canned BSPDN on N2(p) so nope.

doesn't that mean N2p ready 2027 max, where does that put Zen 6, early 2026?

Node being ready is a very relative thing, and Zen6 is H2'26.

 

[Win2012R2]

TSM canned BSPDN on N2(p) so nope.

I thought they had A16 for that, which is supposedly N2 half-node with same rules more or less, plus backside power delivery, a bummer AMD won't get it straight away, NVIDIA will probably grab it with both hands and Intel is supposed to have it on 18A.

 

[DrMrLordX]

You do realize Zen 5 was a complete flop?

Intel wishes it had a complete flop that profitable.

 

[RnR_au]

Should Intel release Panther Lake in late 2025 on 18A and it bests Zen 5, I think you will see Zen 6 sooner rather than later in 2026. Same for next gen EPYC. If Intel's Clearwater Forest on 18A does really well, then you will see Zen 6 Turin on N2 get released..... but not before.

I believe TSMC needs orders years in advance. Its not like AMD could release Zen 6 Turin on N2 6 months after a strong Intel release if AMD hadn't made plans to do so in advance.

 

[Thibsie]

Intel wishes it had a complete flop that profitable.

Sometimes I wonder... Are we all really in the same galaxy ? Dunno...

 

[DrMrLordX]

Sometimes I wonder... Are we all really in the same galaxy ? Dunno...

Some folks live in a bubble filled by some salty day 1 reviews of the 9950X. Should be interesting to see if the same thing happens with Zen6.

 

[yuri69]

No such thing exists.

This makes one wondering, how are they going to handle the client till a Zen 6-based APU arrive - either at Computex (June '26) or CES (Jan '27)?

Krackan and its refresh will stay relevant if priced correctly. Halo is a niche so it doesn't get really affected. But Strix? Strix is definitely going to be refreshed sometime during 2025. Alghough that's quite rough to make it to H2 26/Q1 27 with no major changes earning it a "plus", right?

 

[vanplayer]

Should Intel release Panther Lake in late 2025 on 18A and it bests Zen 5, I think you will see Zen 6 sooner rather than later in 2026. Same for next gen EPYC. If Intel's Clearwater Forest on 18A does really well, then you will see Zen 6 Turin on N2 get released..... but not before.

Last time I saw Panther Lake is still behind Strix Point especially in MT, maybe close to double digit at last. ST might be a tie.