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

Page 462 - 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
821
1,457
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

BorisTheBlade82

Senior member
May 1, 2020
707
1,130
136
"Severe"? In a strange, undefined metrics?

How about this metrics: Penalty for guessing wrong. In case of 7950x3d, the penalty is nearly zero. In case of Alder Lake, Raptor Lake? This is where you can use the word severe.
Maybe you misunderstood me. The penalty for a wrong decision is much less, when compared to Big.little. But I think it is much harder to implement a robust scheduling approach when the best resource depends on the specific workload and you do not want to use and maintain explicit Allowedlists. In-flight-analysation might be an option, but sometimes the past only can tell you so much about the future.
 

ondma

Diamond Member
Mar 18, 2018
3,308
1,692
136
The simple fact that the slide from AMD compared the 7xxx x3D chips to the 5800x3D instead of the 7700x or 13900k is telling, IMO. A 7800x3D will most likely still be the fastest gaming processor, but I dont think it is a home run like the 5800x3D was. The slide from AMD showed about a 20% improvement vs 5800x3D, with one outlier at 30%. Those are undoubtedly among the best case scenarios. So my guess would be that overall, about 10% or so faster than the 13900k, with the gap perhaps closer with a well tuned 13900k with the fastest memory available. Of course it will also be quite variable from game to game.
 

Mopetar

Diamond Member
Jan 31, 2011
8,487
7,726
136
To be fair, whether ones consider 20 percent performance difference big or small, its rather subjective. Perhaps to him is 20 percent "trading blows" - after all its IMO safe to say that in majority of cases those 20 percent would not be very noticeable.

If you use such a loose definition, then a 13400F also trades blows with a 13900K since it'll be within that same 20%. I don't think anyone would ever say that with a straight face.

Trading blows implies that neither wins on average (within some margin of error) even if they may have some titles that each does better in. It's a back and forth fight at that point, not one guy getting punched in the head until they fall over.

Sure, most people don't need much more than 13400F levels of performance because they're not using a card where the CPU will matter. Unless you're playing something like Factorio where the CPU is more important.

But if we're talking about too-end gaming where people are going to use a 4090, then even a 5% edge is massive.
 

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,177
136
5ghz boost for 7800X3D while disappointing is still 500mhz over the 5800X3D and according to TPW the 13900K is only 6.2% faster than the 5800X3D at 1080p with an RTX 4090 in their 53 game benchmark. https://www.techpowerup.com/review/rtx-4090-53-games-core-i9-13900k-vs-ryzen-7-5800x3d/2.html

You have taken 1 variable, between 5800x3d and 7800x3d - clock speed.

There are other variables, each one accounts for something:
- double L2 cache size
- lower L3 latency
- higher L3 bandwidth
- Zen 4 core vs. Zen 3 core
 

JM Popaleetus

Senior member
Oct 1, 2010
375
47
91
heatware.com
1672944360423.png

Using the worst % examples given in the conference for gaming, the 7950X3D will be 13% faster than the 13900K.

The 7800X3D at 21% faster than the 5800X3D, makes it 10% faster than the 13900K and relatively equal to the 7950X3D in gaming. This is probably why they didn't want to compare the 7000-series directly, as a chart with no noticeable difference can be easily exploited by YouTube/Twitter/forum trolls.

Both will offer equally the fastest gaming performance. Productivity and pricing is where it becomes interesting in regards to both company's entire product stack.

The 7900X3D will probably be the best all around chip if you're okay with whatever MSRP is.

I imagine the 13900KS will be 5-10% better than the K.
 
Last edited:

Mopetar

Diamond Member
Jan 31, 2011
8,487
7,726
136
You have taken 1 variable, between 5800x3d and 7800x3d - clock speed.

There are other variables, each one accounts for something:
- double L2 cache size
- lower L3 latency
- higher L3 bandwidth
- Zen 4 core vs. Zen 3 core

That's all true, but the point is that a 5800X3D at 7800X3D clock speed would already beat the 13900K in that particular suite of games. It's just to establish a lower performance bar.
 
  • Like
Reactions: Joe NYC

Asterox

Golden Member
May 15, 2012
1,058
1,864
136
Buyers of AMD Ryzen 65W 7000s; it will hit “retail”l on 10th of January. :D Cannot wait!

One 7600 is mine!😁

Realistically, the "120%" of PC users are actually waiting for even cheaper AM5 motherboards. :grinning:

Current situation, $160 is too much for the cheapest AM5 board if you want to build a "cheap but very good PC".So another month of waiting, and only then it will be much easier when you have AM5 motherboards for $80-100$.
 

Hitman928

Diamond Member
Apr 15, 2012
6,695
12,370
136
View attachment 74004

Using the worst % examples given in the conference for gaming, the 7950X3D will be 13% faster than the 13900K.

The 7800X3D at 21% faster than the 5800X3D, makes it 10% faster than the 13900K and relatively equal to the 7950X3D in gaming. This is probably why they didn't want to compare the 7000-series directly as a chart with no noticeable difference can be exploited.

Both will offer equally the fastest gaming performance. Productivity and pricing is where it becomes interesting in regards to both company's entire product stack.

The 7900X3D will probably be the best all around chip if you're okay with whatever MSRP is.

If you check TPU’s 50(ish) game test, the 5800X3D is only around 5% behind the 13900k.
 
  • Like
Reactions: Kaluan

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,177
136
View attachment 74004

Using the worst % examples given in the conference, for gaming the 7950X3D will be 13% faster than the 13900K. The 7800X3D at 21% faster than the 5800X3D, makes it 10% faster than the 13900K and probably equal to the KS in gaming.

Both will offer equally the fastest gaming performance. Productivity and pricing is where it becomes interesting.

If money is no object, the 7900X will probably be the best all around chip.

But, but, but...

How come there are no tuning benchmarks? How about tuned Raptor Lake with DDR-5 10,000 that can light up 2 seconds of BIOS boot screen?

We need more input on this from:

1672945662460.png
 
  • Like
Reactions: ftt

JM Popaleetus

Senior member
Oct 1, 2010
375
47
91
heatware.com
Here you go:
Cool.

Doesn't change my prediction: 7800X3D will be roughly equal to the 7950X3D in gaming. And both X3Ds will be roughly 10-15% on average faster than the 13900K in 1080p gaming. The 13900KS may trade blows (when cranked up to 11 with fast RAM).

EDIT: Clarified my grammar.
 
Last edited:

Joe NYC

Diamond Member
Jun 26, 2021
3,647
5,177
136
I believe that the threads must be primarily assigned to die with cache for game threads and to bare die with other application, this does not seem difficult to do for me, but I know nothing about inner workings of an operating system, I wonder if AMD is working on some Windows update with Microsoft.

I think you are right.

If the scheduler evicts all of the non-game threads from the V-Cache CCD, then with those threads, all of the cashed memory those threads used will also be evicted.

Which would leave all of the L3 in V-Cache CCD for the use by the game.
 

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,882
3,311
146
I said it may trade blows in perfect circumstances, power limits be damned.

Nice word twisting though.

"with the 13900KS trading blows (when cranked up to 11 with fast RAM)."

Ah yes, word twisting. Could you clarify how what you said means "in perfect circumstances"? Sure seems to me like you're implicating it will trade blows only in some games and overall be slower, except that's not what you said.
 
  • Like
Reactions: Det0x

inf64

Diamond Member
Mar 11, 2011
3,884
4,692
136
I said it may trade blows in perfect circumstances, power limits be damned.

Nice word twisting though.
Well you did state that you expect the 13900KS to end up 10-15% faster than 13900K which makes little sense, to be honest. You might get ~4% clock boost with KS and that's it. Performance doesn't scale perfectly with clocks so 4% is the absolute best case scenario.
 
  • Like
Reactions: ftt and Det0x

JM Popaleetus

Senior member
Oct 1, 2010
375
47
91
heatware.com
"with the 13900KS trading blows (when cranked up to 11 with fast RAM)."

Ah yes, word twisting. Could you clarify how what you said means "in perfect circumstances"? Sure seems to me like you're implicating it will trade blows only in some games and overall be slower, except that's not what you said.
when cranked up to 11 with fast RAM

You even bolded, underlined, and italicized it. In the perfect nutty tuned circumstances that @Joe NYC mentioned.
 

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,882
3,311
146
when cranked up to 11 with fast RAM

You even bolded, underlined, and italicized it. In the perfect nutty tuned circumstances that @Joe NYC mentioned.

lol ok.

Since you want to move the goalposts so far away it's into high end enthusiast tuning territory, that claim is completely meaningless in the first place unless you're going to compare it to a heavily tuned Zen 4 3D chip. Which is enabled this time around.

You basically said nothing.

Heavily tuned processor is faster than stock processor, news at 11
 

JM Popaleetus

Senior member
Oct 1, 2010
375
47
91
heatware.com
Well you did state that you expect the 13900KS to end up 10-15% faster
I actually said I expect the 13900KS to be able to trade blows when cranked up to 11 with fast RAM as highlighted above.

I think tuned, with bleeding edge memory, power be damned, Raptor Lake may be able to throw a few punches back. Nothing that would make it a clear W.
 

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,882
3,311
146
I actually said I expect the 13900KS to be able to trade blows when cranked up to 11 with fast RAM as highlighted above.

I think tuned, with bleeding edge memory, power be damned, Raptor Lake may be able to throw a few punches back. Nothing that would make it a clear W.

Just quit now.

JM Popaleetus said:
roughly 10-15% on average faster than the 13900K in 1080p
 

PJVol

Senior member
May 25, 2020
854
838
136
So they don't seem to have solved the V/F problem for 3D stacked CCD, basically the same 400MHz deficit as with the 5800x3d.
And given the most of the Zen3->Zen4 performance uplift came from higher clock speeds, this doesn't bode well for 7800x3d. Let's see what they squeeze out of the "hybrid approach" in 2CCD models.
All in all, dissapointing so far.

Capping the 7700X at 5000mhz might give us an idea of 7800x3d performance in non-gaming scenarios.
 
Last edited:
  • Like
Reactions: Timmah!