Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[eek2121]

There is no realistic chance. In October last year Zen4 design wasn't even finished.

I was thinking a November/December release, not something "soon".

 

[uzzi38]

I was thinking a November/December release, not something "soon".

Even still.

Don't expect Z4 before Q2 next year at the earliest. Warhol getting cancelled is totally unrelated.

 

[scineram]

Well, Lisa said Zen 4 is 2022. So there’s that.

It isn’t the same interface. DDR5 alone will require changes, and if it also has PCIE 5? Yeah, no way. The new IOD will need a lot more bandwidth.

I doubt AM5 will have PCIe 5 at launch. Maybe with second generation.

 

[DrMrLordX]

There is no realistic chance. In October last year Zen4 design wasn't even finished.

And yet smart money has Genoa shipping to hyperscalars as we speak. Unless you are suggesting that's totally unrelated to AMD's ability to produce Raphael.

Don't expect Z4 before Q2 next year at the earliest.

Why Q2? Why not Q1?

 

[uzzi38]

And yet smart money has Genoa shipping to hyperscalars as we speak. Unless you are suggesting that's totally unrelated to AMD's ability to produce Raphael.

Wild hunch says those are ESes, not shipping anything final-ish silicon

Why Q2? Why not Q1?

Because 5nm capacity.

 

[DrMrLordX]

Wild hunch says those are ESes, not shipping anything final-ish silicon

Unless they're soliciting outside help on finalizing clocks/TDPs, I doubt it. They're either shipping QS or early commercial steppings.

Because 5nm capacity.

I wouldn't think that would be an issue by now. TSMC has had a year to address that issue, and up to this point practically nobody but Apple has been taking N5 wafers.

 

[eek2121]

Wild hunch says those are ESes, not shipping anything final-ish silicon


Because 5nm capacity.

I am not saying it happen, but I keep hearing AMD has quite a bit of 5nm allocation already. They must be doing something with it…

 

[xilli_fiberbit]

Wild hunch says those are ESes, not shipping anything final-ish silicon


Because 5nm capacity.

TSMC has already planned to double the capacity of 5nm for next year, but i think AMD won't move all this products on 5nm at the same time, they can stick on 7nm/6nm for console APU

 

[moinmoin]

they can stick on 7nm/6nm for console APU

That's not up to AMD anyway.

 

[scineram]

It is. All that IP is for N7.

 

[Ajay]

TSMC has already planned to double the capacity of 5nm for next year, but i think AMD won't move all this products on 5nm at the same time, they can stick on 7nm/6nm for console APU

The 'doubling' of Fab 18's production from the initial 60-70 KWM to 120 KWM should be complete by the end of Q2 2021 according to what I've read (so, wafer output will peak about 3 mos later). TSMC are building around the same wafer output (eventually) on N5 in the US, but HVM is due till 2024 (construction in AZ has already begun); wafer production will start out slow as they are not building GIGA Fabs. TSMC has said that they will be booking no new orders till 2H 2021. At the moment, none of the big players are getting all the wafers they want (supposedly, even Apple). Anyway, this is the best info I have from a few online sources (GizChina, Digitimes, etc.).

 

[Doug S]

At the moment, none of the big players are getting all the wafers they want (supposedly, even Apple). Anyway, this is the best info I have from a few online sources (GizChina, Digitimes, etc.).

Apple hasn't said anything about being supply constrained on A14 or M1, their only mention of supply constraints in the earnings call was iPad and Macbook - i.e. the two Apple products that use LCD displays and thus lines up perfectly with the claims that display driver shortages are an issue for them like everyone else. If they were supply constrained on N5 wafers they would have called out potential shortages on iPhone 12, iPad and all Macs not just Macbook.

So I'm skeptical that Apple isn't getting all the wafers they want from TSMC, because they had a record Q2 for iPhone, for iPad, and for Mac, which is can only be done if they get all the wafers they want.

 

[jpiniero]

AMD Zen4 "Raphael" Ryzen CPUs for desktops rumored to launch in Q4 2022 - VideoCardz.com

AMD Zen4 in late 2022? A new rumor suggests that AMD Zen4 architecture may launch later than expected. According to @sepeuwmjh who had previously leaked a part of the AMD roadmap, Zen4 architecture should now be expected in 2H 2022. This leaker has been especially active in relation to Raphael...

videocardz.com

Unsure if this is Fake News or not, but some dude is saying that Raphael isn't coming until Q4 next year. If you're saying why would they cancel Warhol if that is the case, my only theory is that they want the wafers for something else.

 

[DisEnchantment]

If there is something to surmise out of all the leaks for AMD lately, it is that the usual suspects have as much clue as everyone else regarding the roadmap of AMD's products ..... Zero

 

[Thibsie]

Hahaha it seems that way, indeed.

 

[Ajay]

If there is something to surmise out of all the leaks for AMD lately, it is that the usual suspects have as much clue as everyone else regarding the roadmap of AMD's products ..... Zero

It's almost like AMD is purposely screwing with the leakers. One way to make them viewed as unreliable.

 

[andermans]

I actually think most of the long time leakers have been fairly on point, but they have only leaked about server and mobile, and not really about desktop. I get the impression that there is a bunch of solid info about Genoa out there now, and also info about Rembrandt seems fairly solid (as far as leaks go anyway), but especially desktop seems like a mess.

Outside of that we have:

• Chips and Cheese doing their IPC article
• Redgamingtech basically talking about any rumor whatsoever
• Vegeta on twitter (which AFAIU started with a compilation of existing leaks)

Not sure what to make of Redgamingtech but most of the other leakers around the desktop roadmap seem to be new this cycle or am I missing something?

 

[LightningZ71]

Is it possible that AMD is realigning production a bit to embrace the market and supply realities that are present? Given the issues that the industry is having with various raw materials for production, is it possible that AMD is trying to just maximize APU production on existing nodes in the near term? Instead of putting resources into a mild refresh of the 5000/x desktop processors, they are maybe focusing on delivering more APUs to market? They are able to sell every 5000/X series desktop CPU above MSRP that they can make. There's no need to put anything there right now. They would be far better served by managing a shrink of Cezanne to N6 to squeeze out a few more dies per wafer and also providing a mild performance bump in the process. That way, they can have something to sell into the desktop sector in the lower and mid end market. The 5000G series APUs are still very competitive in the market. An N6 shrink that can squeeze out a tiny bit more performance would be even better.

I still think that they could make a killing by producing a smaller chip based on a half Lucienne, with maybe 8MB of L3 in a single CCX and 4 Vega CUs. It would satisfy a whole lot of the market, and they would get a ton of dies off of 7nm wafers.

 

[scineram]

AMD Zen4 "Raphael" Ryzen CPUs for desktops rumored to launch in Q4 2022 - VideoCardz.com

AMD Zen4 in late 2022? A new rumor suggests that AMD Zen4 architecture may launch later than expected. According to @sepeuwmjh who had previously leaked a part of the AMD roadmap, Zen4 architecture should now be expected in 2H 2022. This leaker has been especially active in relation to Raphael...

videocardz.com

Unsure if this is Fake News or not, but some dude is saying that Raphael isn't coming until Q4 next year. If you're saying why would they cancel Warhol if that is the case, my only theory is that they want the wafers for something else.

I doubt, hope at least, that Zen 4 will be not that late. But even so it might make a lot of sense to stretch out Vermeer lifetime and not introduce Warhol, whatever it is.

1. Rocker Lake is not very competitive, in the public eye.
2. Alder Lake perf and availability (including DDR5 supply) are questionable, might not be a big threat.
3. Vermeer supply still hasn't caught up, especially in high end.
4. By the time it does it might already be fall, closing in on the holiday season. When demand for everything shoots up cyclically, including consoles!
5. Lot of bitching about Renoir, then Cézanne being AWOL months after launch. Laptops are the big market, and I think AMD wants to make a big splash @ CES with Rembrandt but with much better supply this time. They have the mindshare now. That requires ramping already in Q4, if not Q3.

So Rembrandt it is!

 

[andermans]

Is it possible that AMD is realigning production a bit to embrace the market and supply realities that are present? Given the issues that the industry is having with various raw materials for production, is it possible that AMD is trying to just maximize APU production on existing nodes in the near term?

I get the impression that as far as wafer supplies goes, doing a refresh isn't really that costly. It just ends up mostly increasing demand because more people will want to upgrade.

Now while there are all kinds of scary timelines for how long the shortage will last, note that most of this is across the entire market. I expect that the high margin products will have significantly improved by Q4 2021. At that point, instead of trying to sell more low margin products (with reduced demand for high margin products due to Alder Lake and people who have already bought a zen3 CPU) the refresh might help them keep ASP high, maximizing the profit for their supply capacity.

Hence I don't think "they will sell all their supply anyway" is necessarily a good reason to argue they shouldn't release anything.

 

[yuri69]

It's been only 6 months since Zen 3 launched. The horrible availability makes it feel longer. Milan has launched very recently. There are no signs of any Zen 3 TR to be launched soon. Desktop non-XT models are also still lacking. So the Zen 3 IP hasn't been fully utilized yet.

An 18 months release cycle would put Zen 4 to Q2 2022. However, Zen 4 brings a brand new platform with DDR5/PCIe5. So delays are to be expected.

I guess we will have a "XT" refresh in Q3/4 2021 to refresh the media coverage.

 

[dnavas]

An 18 months release cycle would put Zen 4 to Q2 2022. However, Zen 4 brings a brand new platform with DDR5/PCIe5. So delays are to be expected.

I'm personally finding it increasingly difficult to justify investing in a pcie4 TR platform purchase, so I hate what this does for Zen4 TR, but it frankly doesn't make sense to ship a DDR5 platform prior to DDR5 being available in quantity. So yes, I'd expect later than sooner.

Unless [I add, somewhat self-servingly] you release on the platform that's already more expensive and might be more willing to absorb the cost. I don't expect AMD is doing this, but it seems like it would make some sense to put the TR chiplets after desktop, so that desktop can bake the design, but put the TR IO on the bleeding edge. You allow TR to benefit from the choice of the best chiplets, and allow the desktop to benefit from the better/cheaper supply chain of parts.

 

[Mopetar]

There's a bit of a chicken and egg problem where unless there are CPUs to use DDR5 what incentive is there to actually manufacture any memory chips and vice versa, but it makes a lot of sense for a high end product like Threadripper to come out first, even if there isn't a lot of DDR5 memory because it's a relatively small market (which makes it easier to get into without having to devote a lot of production towards) and a high-end market that will pay a premium for other components.

I think that it's pretty obvious that sTRX4 isn't going to be able to accommodate Zen 4, so I suspect that any Zen 3 TR parts are just an upgrade path for anyone who bought a Zen 2 Threadripper. It might make more sense for AMD to get a Zen 4 TR part out early to help push the DDR5 market forward so that there's more supply ready for consumer products in the future. I think they'd definitely want to be able to release APUs utilizing DDR5 as soon as possible if only because that would give a big performance uplift to the graphics which have been constrained by memory bandwidth.

Then Zen 3 product stack is so backed up right now that there's not a lot of incentive for them to rush Zen 4 consumer parts, especially if Intel doesn't have anything to counter with themselves.

 

[scannall]

I still think that they could make a killing by producing a smaller chip based on a half Lucienne, with maybe 8MB of L3 in a single CCX and 4 Vega CUs. It would satisfy a whole lot of the market, and they would get a ton of dies off of 7nm wafers.

The biggest shortage right now is substrates. And it takes the same sized substrate for a cheap $99 AM4 APU as it does for a 5950X. Assembly labor is also in shortage thanks to COVID. It really doesn't make sense right now to target the bottom of the market.

 

[mikk]

Realistically Zen4 wasn't expected before mid 2022 which makes late 2022 not that far fetched. In the old Warhol roadmap it looked like a full year after Warhol. Also I doubt it has something to do with Intels situation or 5nm capacity. I mean if it's capacity related they could launch a few limited high priced lower volume models. Also it's highly likely Intel is in a better place with ADL-S than currently, deliberately delaying Zen4 because Intel is too weak makes no sense for me either. My first question would be if the Zen4 design has been completed by now because in last October it wasn't. As for Zen3 AMD told us that it had been completed in August 2019, there is no such news for Zen4.