Question Zen 6 Speculation Thread

[Joe NYC]

And since the mobile Zen 6 parts apparently aren't launching until 2027Q1 it might be two "generations" of losing ground to Intel.

If Zen 6 is to be one of the leading N2 products (which a TSMC leak had Zen 6 as part of N2 qualification, it's more likely that Zen 6 will be released by mid 2026. Maybe not all models and versions at once, but the trailing designs will more likely be a monolithic low end die in 2027. The group of modular designs. sharing the common CCD will likely come first in 2026.

 

[gdansk]

Irrelevant to most PC users.

I don’t even do software development on a Mac anymore.

If AMD doesn't aim at Apple they still lose by proxy. Stagnation is a company killer. Both AMD and INTC stock prices are down more than 50% since last March. AI PC didn't pan out? Client sales for AMD in particular are bleak (Intel makes more in a quarter than AMD in a year) and that's AMD with what I'm told is a generally superior product for years straight. Intel seems able to keep up with the Zen cadence (even if it is with two generations instead of one) and so prevent AMD from gaining much market or revenue share in client.

And I don't think Zen 6 is an Intel-killer based on available information so far. It all points to a server-focus yet again (~10% IPC, small CCD, 12 cores, huge L3).

 

[basix]

"10%+" IPC can be anything. And we do not know frequency and power efficiency characteristics yet. So in essence: We know nothing.

The increase L3$ will at least help in gaming, especially for the non-X3D SKUs. 12C CCD will also be good regarding MT Workloads, where the current 8C CPUs fall a little behind compared to the competition.
And N2 ist likely to be a big enhancement regarding performance characteristics. But we do not really know, how far that can take Zen 6.

 

[adroc_thurston]

Intel has a yearly cadence

ehhhhh.

SPECint 2017 1T unless AMD improves by 35%+ for laptop parts

well yeah it's gonna pump that one subtest with 3x the accessible L3.
It's also gonna have moar freq gain since unlike DT they're not gonna have to cope with decreasing Vmax.

Zen 6 is designed to be a smaller core so they can fix 12 in 70-80mm

No?
it's just a double-shrink thing.

It all points to a server-focus yet again (~10% IPC, small CCD, 12 cores, huge L3).

what.
The CCD isn't small, and 12 cores are just a consequence of 2 shrinks.

 

[Gideon]

Zen 3 and M1 were on par. 5950X was 9.15 and M1 Pro was 9.2 in SPECint 2017 1T.
Zen 5 and M4 are not on par. 9950X is 12.6 and M4 Pro is 13.6. Firestorm derivatives are pulling away.

In moblie the difference is even more pronounced. For desktops the ST clock speed rose from 4.9 GHz (5950x) to 5.7 Ghz ( 9950x) While in mobile it was around 5 Ghz already for the 6000 series (zen3+). 2 Generations later and they are still stuck at 5.0 - 5.1 even with Strix Halo.

Apple in the same period (2022 - 2024) went from 3.5 Ghz (M2) to 4.4 Ghz (M4)

 

[adroc_thurston]

In moblie the difference is even more pronounced. For desktops the ST clock speed rose from 4.9 GHz (5950x) to 5.7 Ghz ( 9950x) While in mobile it was around 5 Ghz already for the 6000 series (zen3+). 2 Generations later and they are still stuck at 5.0 - 5.1 even with Strix Halo.

Apple in the same period (2022 - 2024) went from 3.5 Ghz (M2) to 4.4 Ghz (M4)

yes that's the problem.
Fortunately, fmax will see a big bump with moar shrink.

 

[Gideon]

And from the patent trail from last quarter you'd imagine AMD is a packaging specialist than CPU designer. Not good. Typically patent trail is good indicator of active research in novel ideas but not since last two quarters unfortunately. At least not much applications in core design.

I have a (perhaps silly) hope that AMD is working on a radically new core and doesn't want the details to be public yet (thus also no patents).

I refuse to believe that AMD's long-term plan is to keep iterating Zen forever at the pace of "5-10% IPC improvements every 2 years," especially considering what is already public knowledge. The insiders must be much more aware of industry trends.

There are two main reasons supporting my belief:

1. Let's look at what the competition is doing:

• Qualcomm has a rather agressive roadmap that will definitely outpace "5-10% IPC in 2 years" designs for several generations to come (as they have plenty of low-hanging fruit to pick)
• Apple's M1 is already 5 years old, M4 shows that wide designs can achive good clock speeds. In low-power environments it has an almost silly fficeny edge when running flat out (4.4 vs 5.1 GHz)
• While cancelled, Intel's Royal Core was a significat redesign in the pipeline for Intel (rumored 2X IPC in some cases though with terrible area efficency). Now the same architects are developing something something likely equally radical at Ahead Computing
• While it may end up not delivering, Tenstorrent is planning to release a massive 16-wide clustered RISC-V design in 2-3 years

Some of these parts will not pan out, but most represent more aggressive approaches than slow iterations on existing cores every 20 months More importantly, none of these developments are new. Royal Core was a known entity for years before cancellation. It would be a terrible strategy for AMD to simply assume "Intel will screw up" and allow themselves to stagnate. Only the paranoid survive, and AMD needs to have a ground-up redesign in their pipeline as a contingency.

2. . Zen Architecture's Maturity

Despite considerable changes in Zen 5, the Zen microarchitecture is just getting long in the tooth. Here's how often AMD has started from scratch (first level indicates new designs, second level shows iterations):

• K5 - 1996
• K6 - 1997
• K7 (Athlon) - 1999
• K8 (Athlon 64) - 2003
  • K10 - 2007
• Bulldozer - 2011
  • Piledriver - 2012
  • Steamroller - 2014
  • Excavator - 2015
• Zen - 2017
  • Zen 2 - 2019
  • Zen 3 - 2020
  • Zen 4 - 2022
  • Zen 5 - 2024
  • Zen 6 - 2026

Granted, there are major low-level changes in Zen 3 and Zen 5 finally made considerable high-level improvements, but these are nowhere near the comprehensive redesigns seen with Bulldozer or Zen. Jim Keller often refers to these as "Clean Breaks" that managers typically heavily resist (as real engineering means real tradeoffs and new parts might regress in places). Nevertheless these breaks are essential for significant future improvements.

Of course AMD can continue iterating on Zen, but betting their entire CPU division's future solely on this approach would be strategically unsound.

Speaking of that ... where are the Zen 7 rumors?

We had information about Zen 4 and Zen 5 circulating 3-4 years before their launch. There is nothing substantial out there about Zen 7. Perhaps this absence is also telling?

 

[yuri69]

Apple launched Firestorm 5 years ago. Yet, no ARM/x86/RISC-V vendor took the inspiration from it. People have been working on Apple Silicon for years, yet... We have Zen 5% in client workloads.

But yea, we still have that "Zen 5 client 1T IPC gain = server SMTed 96c gain / 96" guy preaching how L3 will save Zen 6.

---

Zen 6's "10+%IPC" can be anything between 10 and 15% since that's how the slide was designed.

---

Intel throws $ and pushes products fast. It doesn't matter that there is a lot of tape to hold it together. But AMD doesn't seem to have *anything new* in mobile - besides rebrands - for the rest of 2025-2026. That's quite sad.

---

As for Zen 7, that one has been referred by that MLID clown like year ago. He pushed the narrative that Zen 7 was re-evaluated since AMD did not want to risk yet another "major redesign" core after the Zen 5 experience. Reportedly Zen 7 takes the idea of serving different markets with different core variants even further.

 

[Kepler_L2]

I have a (perhaps silly) hope that AMD is working on a radically new core and doesn't want the details to be public yet (thus also no patents).

I refuse to believe that AMD's long-term plan is to keep iterating Zen forever at the pace of "5-10% IPC improvements every 2 years," especially considering what is already public knowledge. The insiders must be much more aware of industry trends.

There are two main reasons supporting my belief:

1. Let's look at what the competition is doing:

• Qualcomm has a rather agressive roadmap that will definitely outpace "5-10% IPC in 2 years" designs for several generations to come (as they have plenty of low-hanging fruit to pick)
• Apple's M1 is already 5 years old, M4 shows that wide designs can achive good clock speeds. In low-power environments it has an almost silly fficeny edge when running flat out (4.4 vs 5.1 GHz)
• While cancelled, Intel's Royal Core was a significat redesign in the pipeline for Intel (rumored 2X IPC in some cases though with terrible area efficency). Now the same architects are developing something something likely equally radical at Ahead Computing
• While it may end up not delivering, Tenstorrent is planning to release a massive 16-wide clustered RISC-V design in 2-3 years

Some of these parts will not pan out, but most represent more aggressive approaches than slow iterations on existing cores every 20 months More importantly, none of these developments are new. Royal Core was a known entity for years before cancellation. It would be a terrible strategy for AMD to simply assume "Intel will screw up" and allow themselves to stagnate. Only the paranoid survive, and AMD needs to have a ground-up redesign in their pipeline as a contingency.

2. . Zen Architecture's Maturity

Despite considerable changes in Zen 5, the Zen microarchitecture is just getting long in the tooth. Here's how often AMD has started from scratch (first level indicates new designs, second level shows iterations):

• K5 - 1996
• K6 - 1997
• K7 (Athlon) - 1999
• K8 (Athlon 64) - 2003
  • K10 - 2007
• Bulldozer - 2011
  • Piledriver - 2012
  • Steamroller - 2014
  • Excavator - 2015
• Zen - 2017
  • Zen 2 - 2019
  • Zen 3 - 2020
  • Zen 4 - 2022
  • Zen 5 - 2024
  • Zen 6 - 2026

Granted, there are major low-level changes in Zen 3 and Zen 5 finally made considerable high-level improvements, but these are nowhere near the comprehensive redesigns seen with Bulldozer or Zen. Jim Keller often refers to these as "Clean Breaks" that managers typically heavily resist (as real engineering means real tradeoffs and new parts might regress in places). Nevertheless these breaks are essential for significant future improvements.

Of course AMD can continue iterating on Zen, but betting their entire CPU division's future solely on this approach would be strategically unsound.

Speaking of that ... where are the Zen 7 rumors?

We had information about Zen 4 and Zen 5 circulating 3-4 years before their launch. There is nothing substantial out there about Zen 7. Perhaps this absence is also telling?

Zen3/5/7 are new cores, Zen4/6/8 are iterations.

 

[Gideon]

As for Zen 7, that one has been referred by that MLID clown like year ago. He pushed the narrative that Zen 7 was re-evaluated since AMD did not want to risk yet another "major redesign" core after the Zen 5 experience. Reportedly Zen 7 takes the idea of serving different markets with different core variants even further.

Yeah, I could have thought all that out myself too.

Besides I think it's quite likely AMD will at some point do a "Arm X" core version of their big core as well. E.g. Zen6c, Zen6 and Zen 6X. Something that bloats up area, a bit has more cache and clocks higher. It would be quite useless in servers but very helpful for client CPUs.

IF I were MLID, i'd just make a 20 minute episode on a "leak" of a Zen 7 APU having 2x Zen7X cores, 4x Zen 7 cores and 8 Zen7c cores or whatnot

 

[adroc_thurston]

Something that bloats up area, a bit has more cache and clocks higher. It would be quite useless in servers but very helpful for client CPUs.

Cortex-X is no longer derived from mainline Cortex-A IP.
And it shows, X5 is a bloatmongus all in the name of keeping 1t CAGR alive.

 

[yuri69]

Yeah, I could have thought all that out myself too.

Besides I think it's quite likely AMD will at some point do a "Arm X" core version of their big core as well. E.g. Zen6c, Zen6 and Zen 6X. Something that bloats up area, a bit has more cache and clocks higher. It would be quite useless in servers but very helpful for client CPUs.

IF I were MLID, i'd just make a 20 minute episode on a "leak" of a Zen 7 APU having 2x Zen7X cores, 4x Zen 7 cores and 8 Zen7c cores or whatnot

Well, this is MLID but we are talking about a core IP used in 2028-2029 (or later?) products. I wouldn't expect much details at this point.

 

[jpiniero]

Cortex-X is no longer derived from mainline Cortex-A IP.
And it shows, X5 is a bloatmongus all in the name of keeping 1t CAGR alive.

I feel like the last couple of posts is people wanting that.

I don't think that's gonna happen, esp with how crazy expensive N2 is.

 

[Gideon]

I feel like the last couple of posts is people wanting that.

I don't think that's gonna happen, esp with how crazy expensive N2 is.

I agree it's unlikely to happen. But for 1-2 cores in a CCD the area bloat would be minimal, yet it could help ST workloads a bit.

Anyway, the only reason i even brought this up is to show how free of content MLIDs average videos are.

It doesn't even matter whether he makes them up himself or posts made up stuff by bored "leakers", his filler episodes are low quality specualtion anyone can conjure up (dragged to 20 minutes to get that sweet monetation)

I agree that every now and then he also gets legit info, but users want drama every week, thus the fillers.

 

[Gideon]

Well, this is MLID but we are talking about a core IP used in 2028-2029 (or later?) products. I wouldn't expect much details at this point.

Well yes. but keep in mind:

1. Zen 5 was initally mentioned in AMDs "Ryzen Processors one Year Later" video in april 2018 (6 years before release)
2. Mike Clark caused half the hype in his October 2021 interview to Anandtech (3 years before release and nearly a year before Zen 4)
3. The MLID's rare good leak (the one that was obvious was not a fake due to the very believable architcture slide) was leaked in September 2023

Now we only got official confirmation about Zen 6 in July 2024 and absolutely nothing about what's coming next. 2028 is only 3 years away.


EDIT: looking for info about "first zen 4 leaks" i found these threads:

Spoiler: Zen 4 'leaks' in 2020

Reddit post 1, Reddit post 2

So... what is this trump card? Here is the big reveal. "Infinity Fabric 2.0". I have no idea if this name will stick around for marketing. What it actually is is the culmination of all of our efforts since Zen 1 in architecting a solution to instruction starvation. The IMC this time around will be a 7nm chiplet to enable us to push the fabric clock as far as we absolutely can get. This is why our memory target is 6600MHz, because our real goal is an FCLK of 3300MHz. The package will be surrounded by 4 CCDs (of course 5nm, this is basically publicly known already too) with an absolutely gargantuan pool of L4 cache. Final cache amounts are a long ways off though so no information there. We have heavily leaned into L4 this time around as Radeon has done.

His source:

Source is me being an AMD employee

boy are these fun to read now. I should do it more often, going even further back

 

[Hulk]

20 years? Oh Kepler, it has been on death row since the 80s. 🤣

Apple has stagnated. They have relied on adding things like SME and higher clocks in order to push performance up.

They don’t currently have a server offering, and client sucks for some things like gaming. They are a node ahead and still struggle to keep up.

At any rate, my suspicion is Zen 6 will have a larger than normal performance increase due to node shrinks. I am guessing another 20-25% total improvement for single core. We shall see.

NOTE: that I did not use IPC here.

Apple needs to be careful. Yes, they have a golden goose that lays golden eggs, the iPhone, but these days a smartphone is pretty much a smartphone. Like their computer ecosystem, Apple has "East Berlin'ed" but even that could start to fail as people realize you can get a competing Andriod phone for half the price or less. It's not like it was 10 years ago when iPhones were demonstrably better than Andriod. All it took then was a quick stutterless fast scroll of busy webpage to show iPhone superiority. It's all the same today. Even my kids who have top of the line iPhones notice my Pixel 7 is "just like" thier iPhone when they need to use it. Also, I think people are weary of the "must upgrade" every year thing.

All of the doom and gloom "FAIL" talk about 5% IPC improvement is just that, talk. 5% is still an improvement and while some people will skip a generation all of the pre-built systems and mobile device will move to the next gen regardless. There will be a handful of DIY that will hold out for the next gen. 8% is the number I predict we'll see for Zen 6 over Zen 5. That's probably the high side though as far as an average is concerned.

 

[eek2121]

Most of the best software are written on Macs and hosted on Linux servers. Windows is terrible for development nowadays.

Hard disagree (and I use Linux anyway). Windows has WSL2 to run everything Linux at native speeds. Macs don’t run full blown visual studio, nor do they run a bunch of commercial CRMs and other software. None of the games I play are on the Mac. Linux runs more software than the Mac thanks to more ports and also WINE. WINE on Mac is much more limited due to a few reasons.

When I travel for client work, I take my PC. It has full compatibility with all the software I use.

Many of the companies I’ve worked with have transitioned away. Don’t get me wrong , Apple has its niche, but let’s not pretend they are bigger or more relevant than they are.

Regardless, this is a Zen 6 thread, not an Apple one.

 

[Glo.]

Hard disagree (and I use Linux anyway). Windows has WSL2 to run everything Linux at native speeds. Macs don’t run full blown visual studio, nor do they run a bunch of commercial CRMs and other software. None of the games I play are on the Mac. Linux runs more software than the Mac thanks to more ports and also WINE. WINE on Mac is much more limited due to a few reasons.

When I travel for client work, I take my PC. It has full compatibility with all the software I use.

Many of the companies I’ve worked with have transitioned away. Don’t get me wrong , Apple has its niche, but let’s not pretend they are bigger or more relevant than they are.

Regardless, this is a Zen 6 thread, not an Apple one.

Realistically, going Apple sillicon for macs made them useless for way more things, than the performance made it useful for.

Software>Hardware. And if you simply do not have software on your platform - your platform is useless.

 

[Joe NYC]

I have a (perhaps silly) hope that AMD is working on a radically new core and doesn't want the details to be public yet (thus also no patents).

I refuse to believe that AMD's long-term plan is to keep iterating Zen forever at the pace of "5-10% IPC improvements every 2 years," especially considering what is already public knowledge. The insiders must be much more aware of industry trends.

There are two main reasons supporting my belief:

1. Let's look at what the competition is doing:

• Qualcomm has a rather agressive roadmap that will definitely outpace "5-10% IPC in 2 years" designs for several generations to come (as they have plenty of low-hanging fruit to pick)
• Apple's M1 is already 5 years old, M4 shows that wide designs can achive good clock speeds. In low-power environments it has an almost silly fficeny edge when running flat out (4.4 vs 5.1 GHz)
• While cancelled, Intel's Royal Core was a significat redesign in the pipeline for Intel (rumored 2X IPC in some cases though with terrible area efficency). Now the same architects are developing something something likely equally radical at Ahead Computing
• While it may end up not delivering, Tenstorrent is planning to release a massive 16-wide clustered RISC-V design in 2-3 years

Some of these parts will not pan out, but most represent more aggressive approaches than slow iterations on existing cores every 20 months More importantly, none of these developments are new. Royal Core was a known entity for years before cancellation. It would be a terrible strategy for AMD to simply assume "Intel will screw up" and allow themselves to stagnate. Only the paranoid survive, and AMD needs to have a ground-up redesign in their pipeline as a contingency.

2. . Zen Architecture's Maturity

Despite considerable changes in Zen 5, the Zen microarchitecture is just getting long in the tooth. Here's how often AMD has started from scratch (first level indicates new designs, second level shows iterations):

• K5 - 1996
• K6 - 1997
• K7 (Athlon) - 1999
• K8 (Athlon 64) - 2003
  • K10 - 2007
• Bulldozer - 2011
  • Piledriver - 2012
  • Steamroller - 2014
  • Excavator - 2015
• Zen - 2017
  • Zen 2 - 2019
  • Zen 3 - 2020
  • Zen 4 - 2022
  • Zen 5 - 2024
  • Zen 6 - 2026

Granted, there are major low-level changes in Zen 3 and Zen 5 finally made considerable high-level improvements, but these are nowhere near the comprehensive redesigns seen with Bulldozer or Zen. Jim Keller often refers to these as "Clean Breaks" that managers typically heavily resist (as real engineering means real tradeoffs and new parts might regress in places). Nevertheless these breaks are essential for significant future improvements.

Of course AMD can continue iterating on Zen, but betting their entire CPU division's future solely on this approach would be strategically unsound.

Speaking of that ... where are the Zen 7 rumors?

We had information about Zen 4 and Zen 5 circulating 3-4 years before their launch. There is nothing substantial out there about Zen 7. Perhaps this absence is also telling?

Mike Clark already said that some of the features planned for Zen 5 design did not make it into Zen 5 design. Which implies that Zen 6 will mostly complete and fine tune further Zen 5 design. Plus new IOD / packaging. Plus 2 nodes jump.

And, as with the previous "odd" versions, Zen 7 will include bigger changes to the core. And as you said, AMD is very quiet about Zen 7.

 

[Gideon]

Realistically, going Apple sillicon for macs made them useless for way more things, than the performance made it useful for.

Software>Hardware. And if you simply do not have software on your platform - your platform is useless.

I disagree. Intel macbooks were practically unusable for any serious development work except maybe light frontend development. Just fire up Docker in the background and things turned into furnaces.

Some coworkers used to call extended software development on them "Eminem programming" (cause you know, palms are sweaty / mom's spaghetti)

On M1 Pro it was a night and day difference. It's way faster in real world tasks than even a Zen 3 laptop and you can develop anything on it you could previously (with the biggest exceptions being PC games or windows fat-client apps).

Yeah, there is plenty of old Enterprise software that won't run on it but 90% of it didn't run on Mac anyway.

Could you elaborate what software did you have in mind? 32-bit games, sure. but the Mac catalog was already miserable in that regard. Gaming on mac has always sucked.

P.S. Sorry for the OT.

 

[OneEng2]

Zen 3 and M1 were on par. 5950X was 9.15 and M1 Pro was 9.2 in SPECint 2017 1T.
Zen 5 and M4 are not on par. 9950X is 12.6 and M4 Pro is 13.6. Firestorm derivatives are pulling away.
I'm not the one arguing AMD should hit a wall. There is nothing fundamental. And if they deliver measly 1T gains again that's their own choice.

... and yet in so many real world applications, x86 dominates. Perhaps there is more to computation than SPECint?

Most of the best software are written on Macs and hosted on Linux servers. Windows is terrible for development nowadays.

Nonsense. What world do you live in?

 

[Gideon]

... and yet in so many real world applications, x86 dominates. Perhaps there is more to computation than SPECint?

What, where? In what natively compiled real world application x86 dominates?

I write software myself. And while I haven't written much in C lately. Be it Rust, Zig, Go, Java, or Node.js - every time I've compared Mac to x86, the real world performance of M processors tends to mirror the canned benches or be even slightly better (in part probably due to more memory-BW available).

Running any IDEs or development envs is usually faster. Running anything benchmarkable (server startup, unit/integration/e2e tests hot-reload , recompile cycles - comparable gen Mx processor is usually faster).

And let's not even mention browsers (and all the electron based apps). These reaaaaaaly like the fat an low-clocked cores and anything abusing javascript works noticabely faster than even on desktops. Oh and unlike x86 rivals, performance is the same even when unplugged.

I hate MacOS and I have no love for Apple as a company ... but come on. these are just delusional takes (usually by persons that have never tried to compare things on equal terms).

Look, i'd much rather have the old world back, where tuned desktops could run circles around laptops and Macs sucked. I'd rvery much like an alternative reality where AMD is ahead of everybody with a materialized 40% IPC bump Zen 5. Oh and I'd much rather use Linux rather than Macos. But you can't just live in la-la land and pretend that Apple's CPUs don't really have a ~30% IPC edge over their x86 rivals.

Not until you actually install and run some real world software on both a an recent mac and PC and show it lacking.

It's absolutely fine to hate it (and i agree, IMO MacOS and particularily iOS are terrible). But it's not ok to warp reality when it doesn't suit your world view.

 

[yuri69]

Well yes. but keep in mind:

1. Zen 5 was initally mentioned in AMDs "Ryzen Processors one Year Later" video in april 2018 (6 years before release)
2. Mike Clark caused half the hype in his October 2021 interview to Anandtech (3 years before release and nearly a year before Zen 4)
3. The MLID's rare good leak (the one that was obvious was not a fake due to the very believable architcture slide) was leaked in September 2023

Now we only got official confirmation about Zen 6 in July 2024 and absolutely nothing about what's coming next. 2028 is only 3 years away.

Great analysis, thanks.

1. The whole Ryzen Processors One Year Later presentation was performed to convey the message of AMD not being bankrupt but attempting to compete with a brand new family of processors. Therefore dropping the mention about a 5th gen of Zen being in the pipeline was certainly intentional.
2. In that that interview the Zen 5 topic was brought up by the interviewer as a direct continuation of that Ryzen Processors One Year Later presentation. So it was still based on the need to ensure the public the design pipeline was up-to-date.
3. That MLID's confidential 2023 slide was leaked just year before the release.

As for the roadmaps not presenting stuff like Zen 7... We actually most likely have more than 3 years till it launches. Let's do the maths:

• Zen 4 desktop launched in Sep 22
• Zen 5 desktop launched in Aug 24 - 23 months after Zen 4
• Zen 6 products are rumored to launch on end of 26/Q1 27 - thats more like 24+ months

Zen 7 is therefore likely to be launched on end of 28/early 29 - 3.5+ years.

With the slowing down launches it gets pretty riskly to show more than the directly suceeding gen.

Nonsense. What world do you live in?

So in your world most of the development is done on Windows and also the runtime is MS Windows. Okay, there surely are niches like that but hey. The wast majority of development has been centered around web technologies which are not really friendly for Windows.

 

[reaperrr3]

Anyway, the only reason i even brought this up is to show how free of content MLIDs average videos are.

It doesn't even matter whether he makes them up himself or posts made up stuff by bored "leakers", his filler episodes are low quality specualtion anyone can conjure up (dragged to 20 minutes to get that sweet monetation)

I agree that every now and then he also gets legit info, but users want drama every week, thus the fillers.

Well, he's not nearly as bad as RGT in both regards, at least.

He was mostly right about Zen5 and RDNA4, so his AMD sources seem to be rather reliable/legit, for the most part.
When he 'leaks' concrete AMD info, it's lately been either correct or at least closer to the truth than everyone else.

Now we only got official confirmation about Zen 6 in July 2024 and absolutely nothing about what's coming next. 2028 is only 3 years away.

All big companies from the PC space have effectively stopped giving any noteworthy roadmap outlook any further out than ~1-2 years.
Probably to avoid lawsuits and stock crashes over product delays and cancellations. Which are a normal part of R&D sometimes, but not everyone gets that, so it's probably just a "playing it safe" thing.
Besides, judging by recent AMD cadence (Zen4 and Zen5), I'd assume Zen7 is around 2 years after Zen6, which appears to be Q3/26 at the earliest, so Zen7 is probably more like 3.5 to 4 years away, rather than 3.

There's at least 1 "leak" from MLID from 10 months ago with some relatively concrete claims/rumors about Zen7:
- 2nm*
- won't be a complete grounds-up redesign, since AMD management wasn't too happy with the Zen5 problems and delays
- original plan was to make the variants (client, server, c-variants) more distinctly tailored to their target market, but engineers apparently aren't/weren't too happy about effectively having to design multiple uArch variants every 2 years, so who knows whether that's still the case
- one main focus of Zen7 is allegedly to reduce power consumption across the entire V/f curve

*my personal guess: this might possibly be outdated info, if the Zen6@2nm rumors are true; A16 for at least the server Zen7c CCD is certainly possible, considering how far out that is.

Plus 2 nodes jump.

Well, I wouldn't expect too much from that.

First of all, for Zen6c server it's actually not (the Turin CCDs are already N3E).

Second, TSMC's claimed perf/efficiency improvements vs. N5 suggest that N3E isn't that much better than N4P electrically, basically the improvement from N4P to N2(P?) would likely be only comparable to going from N5 to N3E or whereabouts.
In short, it may be 2 nodes on paper, but the electrical gains are more comparable to a single node jump before Moore's law died. Basically, it's just 2 half-node shrinks in terms of perf/efficiency gains, and not even that for SRAM density.

 

[Joe NYC]

• Zen 6 products are rumored to launch on end of 26/Q1 27 - thats more like 24+ months

In recent videos, MLID says he expects Zen 6 in H1 2026.

TSMC leak showed Zen 6 to be one of the early products for risk production of N2. First of those wafers start in late 2025. Which does not guarantee that Zen 6 will be on very first ones, but certainly makes mid 2026 feasible.

Zen 6 has 2 of its big issues already solved:
- it is a Zen 5 derivative, not breaking much new ground on the core itself
- IOD and packaging already (mostly) solved by Strix Halo. Only derivatives of it will be part of Zen 6