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

Page 39 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Vattila

Senior member
Oct 22, 2004
799
1,351
136
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).

Untitled2.png


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! :)
 
Last edited:
  • Like
Reactions: richardllewis_01

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
True. But they didn't feel the need to sell to smaller shops. There's more to the world than just hyperscalars.
I don't know how big Hetzner is (certainly not very small, but a far cry from all the big names usually mentioned) but it got Milan ready to use as one of its extensive cloud offers on "launch" day (press release). I imagine AMD has plenty such contacts, and the more its chips are in demand the more other shops will want to be in close contact with AMD as well.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
I think that if AMD is indeed releasing a Zen3+ (desktop/mobile) on something other than 5nm it is a smart play since, absent other factors, it gives them more capacity. They can use all of their 5nm allocation for EPYC.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
I am curious if AMD will stick with 7nm for lower end parts, 6nm for higher end, and 5nm for EPYC/RDNA 3.

EDIT: RDNA3 on 6nm, with lower end RDNA 2 parts on 7nm.
 

Justinus

Diamond Member
Oct 10, 2005
3,167
1,509
136
I am curious if AMD will stick with 7nm for lower end parts, 6nm for higher end, and 5nm for EPYC/RDNA 3.

EDIT: RDNA3 on 6nm, with lower end RDNA 2 parts on 7nm.

That doesn't make a lot of sense, it may not be trivial to have zen 4 chiplets designed for both 7nm and 5nm.

It also flies in the face of their entire strategy of having TSMC crank out chiplets by the truckload and then creatively binning them for EPYC, Threadripper, and Ryzen.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
AMD would likely be better served by having an N6 quad core APU with either a Vega 8 or an RDNA2 based iGPU with 16MB of L3 cache that can be shared between the CPU cores and the iGPU. It wouldn't be much larger than the current CPU chiplets, which should allow a lot of volume per wafer, and would nicely fill up the lower end of their stack.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
AMD would likely be better served by having an N6 quad core APU with either a Vega 8 or an RDNA2 based iGPU with 16MB of L3 cache that can be shared between the CPU cores and the iGPU. It wouldn't be much larger than the current CPU chiplets, which should allow a lot of volume per wafer, and would nicely fill up the lower end of their stack.

Van Gogh may still end up being made available for regular laptops even if the main purpose is embedded.
 

yuri69

Senior member
Jul 16, 2013
373
573
136
The problem with Warhol is it *might* be a "XT Part Deux". In that case Warhol would simply be nothing else than a 100MHz speed bump. That might be a pure marketing action requiring just another bin in the process. Things like that doesn't leak very well - things like yesterday's "Radeon Midnight Black".
 

exquisitechar

Senior member
Apr 18, 2017
655
862
136
The problem with Warhol is it *might* be a "XT Part Deux". In that case Warhol would simply be nothing else than a 100MHz speed bump. That might be a pure marketing action requiring just another bin in the process. Things like that doesn't leak very well - things like yesterday's "Radeon Midnight Black".
I'm pretty sure it's more substantial than that, more like Zen+. We'll see.
 

inf64

Diamond Member
Mar 11, 2011
3,685
3,957
136
The problem with Warhol is it *might* be a "XT Part Deux". In that case Warhol would simply be nothing else than a 100MHz speed bump. That might be a pure marketing action requiring just another bin in the process. Things like that doesn't leak very well - things like yesterday's "Radeon Midnight Black".
Or it might be a Zen3+ with ~4.5% IPC jump and 5% clock bump :)
 
  • Like
Reactions: spursindonesia

Racan

Golden Member
Sep 22, 2012
1,101
1,969
136
I'm not really enthusiastic about a Zen 3+, I don't see this thing faring well against Alder Lake... How late after Alder Lake will Zen 4 arrive on desktop?
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
That doesn't make a lot of sense, it may not be trivial to have zen 4 chiplets designed for both 7nm and 5nm.

It also flies in the face of their entire strategy of having TSMC crank out chiplets by the truckload and then creatively binning them for EPYC, Threadripper, and Ryzen.

I never stated Zen 4 would be on 7nm.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I don't know how big Hetzner is (certainly not very small, but a far cry from all the big names usually mentioned) but it got Milan ready to use as one of its extensive cloud offers on "launch" day (press release). I imagine AMD has plenty such contacts, and the more its chips are in demand the more other shops will want to be in close contact with AMD as well.

The hyperscalars likely had access to early Milan silicon last summer/fall or so.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
I'm not really enthusiastic about a Zen 3+, I don't see this thing faring well against Alder Lake... How late after Alder Lake will Zen 4 arrive on desktop?

It really depends on what they manage to do with Zen3+. If it's like the 3000XT processors where it's just a few percent more Mhz, then it's not going to compete as well. If, instead, they can get both a few hundred Mhz of all core and boost, and also improve IPC by a few percent, then it'll be a different story. Personally, I think that AMD needs to do what they can to improve the all-core clocks the most as it appears that Zen3 was a nice single thread boost, but wasn't quite as much of an improvement in multi-thread scenarios. Given that N6 isn't much of a density improvement, leaning to the clocks/power side of the curve may make the most sense.
 

Makaveli

Diamond Member
Feb 8, 2002
4,715
1,049
136
It really depends on what they manage to do with Zen3+. If it's like the 3000XT processors where it's just a few percent more Mhz, then it's not going to compete as well. If, instead, they can get both a few hundred Mhz of all core and boost, and also improve IPC by a few percent, then it'll be a different story. Personally, I think that AMD needs to do what they can to improve the all-core clocks the most as it appears that Zen3 was a nice single thread boost, but wasn't quite as much of an improvement in multi-thread scenarios. Given that N6 isn't much of a density improvement, leaning to the clocks/power side of the curve may make the most sense.

I thought the boost in all core clocks was pretty good for me going from a 3800X to 5800X

The Zen 2 chip would hit about 4.25Ghz on full load.

While the Zen 3 chip does about 4.55-4.6Ghz full load.
 
Last edited:

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
It really depends on what they manage to do with Zen3+. If it's like the 3000XT processors where it's just a few percent more Mhz, then it's not going to compete as well. If, instead, they can get both a few hundred Mhz of all core and boost, and also improve IPC by a few percent, then it'll be a different story. Personally, I think that AMD needs to do what they can to improve the all-core clocks the most as it appears that Zen3 was a nice single thread boost, but wasn't quite as much of an improvement in multi-thread scenarios. Given that N6 isn't much of a density improvement, leaning to the clocks/power side of the curve may make the most sense.

If it is indeed on 6nm, we will likely see a decent uplift. Likely improved multicore boost clocks, a minor IPC increase, etc. We may even see slightly higher single core boost clocks. Some of the rumors claim it will be DDR5, if that is indeed the case, we will see performance improved in memory constrained situations. I don't know if I believe that DDR5 rumor, but we will see.

Note that some people claim that the X570S leak indicates Warhol will land on AM4. I do not believe that will be the case. Zen 3 has been out on the market for less than 6 months. Given that we have a chip shortage, it is likely Warhol is shipping Q1 next year unless it really is a minor refresh.
 
  • Like
Reactions: lightmanek

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
Saying they might reserve 5nm for EPYC sort of implies other node will be used for non-EPYC...

If you read my post, you would see that I stated that other chips would be on 6nm. FYI nobody is expecting the next Ryzen release to have Zen 4. Pretty much everyone expects Zen 3+. Many sites are saying that Zen 3+ will be on 6nm.

The chip following Warhol would be a Zen 4 based Ryzen. EPYC would either stay on 5nm (since capacity would be improved by then) or move to 3nm.