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

Page 140 - 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

Justinus

Diamond Member
Oct 10, 2005
3,165
1,507
136
"a few weeks" can mean any number of weeks.

few
/fyo͞o/

determiner · adjective

  1. 1.
    a small number of.
    "may I ask a few questions?"

Is 5+ a "small number"? It seems the definition of small is the problem here. In my experience, colloquially, a "few" means 3, 4, and sometimes 5. 5-6+ is several.
 

biostud

Lifer
Feb 27, 2003
18,173
4,648
136
If they are making a 5800x3d, it stands to reason that there's no logistical or technical reason that they couldn't make a 5950x3d. If we suppose that the limitations are around AMD choosing to pretest the Zen3 die to verify that all 8 cores are suitable for use, then both of those SKUs would be appropriate. If we suppose that there is a slight power hit for the L3 stacked die, then it stands to reason that they could produce a 5950x3d that has a slightly reduced max boost turbo and still hit power targets. The only stopping point that I can see is that AMD believes that they can't command a high enough price to make it worth it. With it being the likely best CPU ever released for a platform that has millions of installed motherboards that can support it, I can't see that being the case either. I believe that there will be a 5950x3d at some point. It makes too much sense.

With zen4 launching later this year and possibly a limited supply of 3D cache chiplets, I think AMD simply has prioritized the fastest gaming CPU and Threadripper/Epyc market. They really don't need to launch either a 5900x3D er 5950x3D to keep being relevant in their market segment.
 
  • Like
Reactions: scineram

nicalandia

Diamond Member
Jan 10, 2019
3,330
5,281
136
Lisa says, "All those Zen4 cores are running at 5GHz" during a halo infinite demo.

Good to know there is no frequency regression.

"AMD demoed a Ryzen 7000 ‘Raphael’ chip running Halo Infinite with all cores surpassing 5.0 GHz during the benchmark"

On an Early ES CPU? That's insane and the level of performance will be out of this world
 

Justinus

Diamond Member
Oct 10, 2005
3,165
1,507
136
Makes me wish AMD had an "Early Access" program for CPUs.

They most certainly do and we're not invited. I would jump at the opportunity, given I need to fully upgrade a real work machine that needs to be stable (with the 5950x setup) and I can have a new, in-test machine for everything else.
 
  • Like
Reactions: Tlh97

eek2121

Platinum Member
Aug 2, 2005
2,883
3,859
136
"AMD demoed a Ryzen 7000 ‘Raphael’ chip running Halo Infinite with all cores surpassing 5.0 GHz during the benchmark"

On an Early ES CPU? That's insane and the level of performance will be out of this world

One thought I had, they demoed this for Halo Infinite, which actually doesn't mean much. Halo Infinite is not exactly demanding on the CPU.
 

Mopetar

Diamond Member
Jan 31, 2011
7,794
5,897
136
Also remember that separate L3 cash has twice the density for some reason. (64MB on 32MB ZEN3)

I wouldn't think the density is really any different. The extra layer of cache just has more space to occupy so it's spread out over a larger area than the cache on the base die.

Lets see what they have to say on their "flawless roadmap execution", I hope someone from the media ask them that question.

Why waste a question when you already know the answer. It's the same answer you'd get from any company right now. Global pandemic and surrounding issues. It's a bit of a non-answer and who knows do what degree it's true, but what else would you expect?

If they launch in July then it's ~20 months since Zen 3. No one is going to care unless they're only shipping a small number of parts just before the end of the year. And really that's just consumers. If the bulk of Zen 4 is going to high-margin markets instead of desktop CPUs the investors aren't going to shed any tears.
 

Hitman928

Diamond Member
Apr 15, 2012
5,160
7,595
136
I wouldn't think the density is really any different. The extra layer of cache just has more space to occupy so it's spread out over a larger area than the cache on the base die.

No, the extra layer of cache is almost the same size as the L3 cache in the base layer. AMD said they are using different process options (or maybe cell library, or both, don't remember for sure) to get drastically increased cache density on the extra layer.
 
  • Like
Reactions: Tlh97

Mopetar

Diamond Member
Jan 31, 2011
7,794
5,897
136
No, the extra layer of cache is almost the same size as the L3 cache in the base layer. AMD said they are using different process options (or maybe cell library, or both, don't remember for sure) to get drastically increased cache density on the extra layer.

What's the catch then? SRAM doesn't shrink well as is, so either AMD had something that wasn't very good for their initial design, or it's not quite as simple as presented. You normally don't even get twice the density from a new node these days, so it seems a bit too good to be true.
 

jamescox

Senior member
Nov 11, 2009
637
1,103
136
The "idiots" are also customers.

Marketing as an "HEDT" chip and assuming only people who max out 100+ threads you'll lose a lot of potential customers.

Again, you consider 20 cores + HT to be "small" amount of threads? What about 30? 40?

You want a chip that's the fastest in all scenarios and that's what will happen.

We all know Intel is totally sucking in servers now. That's not the point.
Okay, I still don’t believe that Intel will be particularly competitive in HEDT in the near future. If you are buying a chip with more than about 10 cores, then I hope you are doing something other than gaming to justify it. Yes, those idiots are customers, but they are also a super tiny niche. AMD almost didn’t offer a HEDT product at all, so AMD isn’t going to expend too much energy chasing such idiots. They may still get them on the strength of their rebadged server parts though.

I also don’t believe that intel will win this fight based on low core count boost clocks. The number of applications where intel boost clocks will win vs. AMD boost clocks, in the context of you are already buying an high core count HEDT processor for something other than gaming, is super tiny.
 
  • Like
Reactions: Tlh97

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
I also don’t believe that intel will win this fight based on low core count boost clocks. The number of applications where intel boost clocks will win vs. AMD boost clocks, in the context of you are already buying an high core count HEDT processor for something other than gaming, is super tiny.

Not saying low core count MHz is only thing it matters. But that a 56 core SPR-X will win in all areas. Yes it'll use more power, but we all know that. 450W RTX 3090 Ti, 600W Turin, 1KW+ accelerators, etc, etc.
 

jamescox

Senior member
Nov 11, 2009
637
1,103
136
The workstation market (read: TR Pro) is somewhere in between. They want cores but they want frequency/per core performance too.
Depends on the application. For my workloads more cores the better. Boost clocks are almost entirely irrelevant. Any low core utilization apps that I run are gpu limited. This will be true more and more as software is optimized for the new baseline of most people having a lot of cpu cores available.
 

Hitman928

Diamond Member
Apr 15, 2012
5,160
7,595
136
What's the catch then? SRAM doesn't shrink well as is, so either AMD had something that wasn't very good for their initial design, or it's not quite as simple as presented. You normally don't even get twice the density from a new node these days, so it seems a bit too good to be true.

The catch is that you can only get this cache density increase on a die where the process and library are specifically good/optimized for caches. This works for AMD since the cache layer only has cache. You couldn't use the same process/library and build a competitive CPU.
 

DrMrLordX

Lifer
Apr 27, 2000
21,570
10,763
136
"AMD demoed a Ryzen 7000 ‘Raphael’ chip running Halo Infinite with all cores surpassing 5.0 GHz during the benchmark"

On an Early ES CPU? That's insane and the level of performance will be out of this world

It's not really an "early ES CPU". Zen4 in general is far along enough in development that they can ship QS and early retail silicon to hyperscalars . . . well, right now! Anyway you should not be surprised to see the CPU in good working order today. It is not for technical reasons that the market needs to wait well past April to buy this thing.
 
  • Like
Reactions: Tlh97 and dnavas

Justinus

Diamond Member
Oct 10, 2005
3,165
1,507
136
I don't know about you guys but I can't wait to unnecessarily upgrade my entire computer later this year!

I was hoping for a 5950X3D announcement so I could unnecessarily upgrade sooner. But now it'll have to wait until Zen 4. I suppose it isn't really unnecessary because the hand me down 5950X goes to my lady's work from home machine, although she doesn't really need 16 cores. The old i7 5960X hand me down she got in 2020 when I upgraded to the 5950x is really showing its age in the software she uses.

I am excited at the prospect of justifying a zen 4 upgrade though, and the 5950x will be so much faster with that single core boost and large cache she'll be thrilled.

I guess long story short - I am too. Absolutely.
 
  • Haha
Reactions: lightmanek

Hulk

Diamond Member
Oct 9, 1999
4,187
1,969
136
How much will the V-cache offset DDR5 gains?

What I mean is how much performance boost will we see in moving to V-cache with current DDR4 for memory bound applications?

How much without V-cache but with DDR5?

How much with V-cache and then adding DDR5?
 
  • Like
Reactions: lightmanek

itsmydamnation

Platinum Member
Feb 6, 2011
2,731
3,063
136
How much will the V-cache offset DDR5 gains?

What I mean is how much performance boost will we see in moving to V-cache with current DDR4 for memory bound applications?

How much without V-cache but with DDR5?

How much with V-cache and then adding DDR5?
I will say near 0 , ddr5 isn't any faster in access time , it allows more concurrency and more data throughput.

But a higher performing core will put more stress on the memory system anyway.
 
  • Like
Reactions: lightmanek