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!

 

[nicalandia]

I wouldn't expect anything approaching the 75% we see in the slide deck.

You will when you realize that it's a 16C/32T CPU.

 

[inf64]

I went through 5800X3D launch slides (by AMD) and compared with Ryzen 7000 launch slides for gaming. I assumed 7950X will boost to ~5.4Ghz in games. 5900X boosts to around 4.6Ghz while 5800X3D boosts to ~4.4Ghz.

CS GO => 7950X should win with at least 13% margin
a tie between 5900X and 5800X3D


13% win for 7950X vs 5950X



Shadow of the Tomb Raider => 7950X should win with at least 24% margin
9% win for 5800X3D vs 5900X


35% win for 7950X vs 5950X


Watchdogs Legion => most likely a slight win by about 7% for 7950X
36% win for 5800X3D vs 5900X


24% win for Zen 4 vs Zen 3 at fixed clock (16T @ 4GHz)



Far Cry 6 => most likely a slight win by about ~5-6% for 7950X
24% win for 5800X3D vs 5900X


12% win for Zen 4 vs Zen 3 at fixed clock (16T @ 4GHz)

 

[nicalandia]

I went through 5800X3D launch slides (by AMD) and compared with Ryzen 7000 launch slides for gaming.

I was trying to do that, but you did a fine Job.

 

[AtenRa]

- "7600X will have to compete against i7 137xx" WAT... seriously... WHAT?

Current prices from newegg

Core i7 12700F = $315
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Alder Lake= $539

Core i7 13700K = $400
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Raptor Lake = $624

----------------------------------------
Ryzen 7600X = $300
B650 = $125 (lets assume they will be available at launch)
32GB DDR5 6000 CL40 = $220

Total for 6 core ZEN4 = $645

Im sorry im not giving this money for a six core in the end of 2022 start of 2023.

 

[moinmoin]

I'm sure Phoenix will be amazing, but people are reading way too much into that perf@65w slide. First, Rembrandt is already very efficient on 6nm. Second, a large, chunk of Raphael's efficiency over Vermeer at lower TDPs is certainly down to a more efficient 6nm IOD, leaving more budget for the cores by lowering static power draw.

A ~30% more efficient Phoenix (vs. Rembrandt) may be doable, maybe even a bit more, but I wouldn't expect anything approaching the 75% we see in the slide deck.

The mobile chips are where all the I/O uncore improvements were worked on, and it's only with Zen 4 that that finally feeds back into desktop/server. So obviously in that area Dragon Range and Phoenix will see less improvement over Rembrandt than Raphael sees over Vermeer. But the core improvements, higher performance at lower power usage, likely will still be significant. For Phoenix we likely see further core optimizations shared with Bergamo.

How many consumers do you think watched that launch yesterday? I'm gonna say none. It was all enthusiasts and media. So why mention anything about supply of desktop products unless you are trying to reach the intended audience?

Not sure who you target those questions at, personally I find them uninteresting. I'd say the press is mainly about reassuring the DIY sales channels.

 

[inf64]

Current prices from newegg

Core i7 12700F = $315
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Alder Lake= $539

Core i7 13700K = $400
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Raptor Lake = $624

----------------------------------------
Ryzen 7600X = $300
B650 = $125 (lets assume they will be available at launch)
32GB DDR5 6000 CL40 = $220

Total for 6 core ZEN4 = $645

Im sorry im not giving this money for a six core in the end of 2022 start of 2023.

DDR4 with AL or Raptor Lake will be noticeably inferior in gaming to 7600X. Also, where did you get 400$ price tag for 13700K ?

 

[exquisitechar]

I went through 5800X3D launch slides (by AMD) and compared with Ryzen 7000 launch slides for gaming. I assumed 7950X will boost to ~5.4Ghz in games. 5900X boosts to around 4.6Ghz while 5800X3D boosts to ~4.4Ghz.

CS GO => 7950X should win with at least 13% margin
a tie between 5900X and 5800X3D
View attachment 66856

13% win for 7950X vs 5950X
View attachment 66857


Shadow of the Tomb Raider => 7950X should win with at least 24% margin
9% win for 5800X3D vs 5900X
View attachment 66859

35% win for 7950X vs 5950X
View attachment 66858

Watchdogs Legion => most likely a slight win by about 7% for 7950X
36% win for 5800X3D vs 5900X
View attachment 66861

24% win for Zen 4 vs Zen 3 at fixed clock (16T @ 4GHz)
View attachment 66860


Far Cry 6 => most likely a slight win by about ~5-6% for 7950X
24% win for 5800X3D vs 5900X
View attachment 66862

12% win for Zen 4 vs Zen 3 at fixed clock (16T @ 4GHz)
View attachment 66863

That's pretty good, but I wonder how much more performance can be wrung out of it compared to the configuration AMD used. My guess is, not all that much.

 

[inf64]

That's pretty good, but I wonder how much more performance can be wrung out of it compared to the configuration AMD used. My guess is, not all that much.

I suppose with good cooling (higher boost as the chips are limited by temperature) and very fast/low latency DDR5, there could be additional 5-10% on the table. Probably not worth it Vs stock settings running AMD Expo memory (DDR5-6000).

 

[FangBLade]

Current prices from newegg

Core i7 12700F = $315
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Alder Lake= $539

Core i7 13700K = $400
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Raptor Lake = $624

----------------------------------------
Ryzen 7600X = $300
B650 = $125 (lets assume they will be available at launch)
32GB DDR5 6000 CL40 = $220

Total for 6 core ZEN4 = $645

Im sorry im not giving this money for a six core in the end of 2022 start of 2023.

DDR4 for Intel? Then Zen 4 lead will be even bigger, and RPL platform is dead end, while with AMD you have options for upgrade, Zen 5 will be completely new architecture so huge IPC is expected, combined with third generation v-cache, Zen 4 to Zen 5 v-cache will be worth it, especially if you buy now 6 core, and later you want more cores but not from the same architecture, and Zen 5 will raise amount of cores in mainstream.

 

[nicalandia]

Lets wait and see how those two Massive FP Registers perform. AMD invested dearly on them


Deleted Twitt by source

 

[itsmydamnation]

You can always use high clocked memory to run even tighter timing @ 6000. I find the arguments about stopping at 6000 being a big deal pretty disingenuous.

 

[inf64]

Lets wait and see how those two Massive FP Registers perform. AMD invested dearly on them

https://twitter.com/x/status/15

Tweet is gone already

 

[itsmydamnation]

Lets wait and see how those two Massive FP Registers perform. AMD invested dearly on them

https://twitter.com/x/status/15

your link is bad

 

[carrotmania]

Current prices from newegg

Core i7 12700F = $315
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Alder Lake= $539

Core i7 13700K = $400
GIGABYTE B660M AORUS PRO AX = $128
32GB DDR4 3600 CL16 = $96

Total for Raptor Lake = $624

----------------------------------------
Ryzen 7600X = $300
B650 = $125 (lets assume they will be available at launch)
32GB DDR5 6000 CL40 = $220

Total for 6 core ZEN4 = $645

Im sorry im not giving this money for a six core in the end of 2022 start of 2023.

With Intel rumoured to increase prices by 20%, a 12700K was $409, so a 13700K should be ~$500.

7600X will be about the same performance as a 13700 with DDR4, which you're actually paying ~$720 for. Then in 15-18 months, IF Intel get something out before 2025, it'll be:

Core i7 14700K = $500
GIGABYTE B760M AORUS PRO AX = $128
32GB DDR5 6000 CL40 = $220

ANOTHER $720 for a midrange system. Or get a 8950X that's 50% (ST) faster than a 7950X (AMD are reportedly REALLY excited about Zen5) and spend that $20 on, you know, yourself. You're worth it.

 

[AtenRa]

DDR4 with AL or Raptor Lake will be noticeably inferior in gaming to 7600X. Also, where did you get 400$ price tag for 13700K ?

Raptor Lake with DDR4 3600 CL16 will be equal in gaming performance with 7600X, dont forget that RL will have an increase in gaming performance over AL.
Also 13700K will completely annihilate 7600X in MT.

As for the price of 13700K, I expect to be the same as 12700K at $399. There is a possibility to increase at $449 but for now ill stick to $399.

We will know in a few weeks time.

 

[HurleyBird]

If the 7700X had been priced $20 lower (0.95x current price) and the 7600X $30 lower (0.9x current price), I think that would have removed most of the price angst without actually lowering ASPs that much.

7700X $379.99
7600X $269.99

Even though it isn't a huge difference, to me that feels quite a bit nicer.

 

[gdansk]

If the 7700X had been priced $20 lower (0.95x current price) and the 7600X $30 lower (0.9x current price), I think that would have removed most of the price angst without actually lowering ASPs that much.

7700X $379.99
7600X $269.99

Even though it isn't a huge difference, to me that feels quite a bit nicer.

AMD adjusts prices often. So they'll drop if Intel launches something competitive.

But here is how it should be in my mind:
$200 7600 65w
$250 7600x 105w day 1
$300 7700 65w
$350 7700x 105w day 1
< 7800x3d fits somewhere in here >
$450 7900 105w
$500 7900x 170w day 1
$650 7950 105w
$700 7950x 170w day 1

But that wouldn't increase ASPs so it's a no go. Maybe AMD can fix their uncompetitive low end when they can add a $1000 24 core part at the top

 

[FangBLade]

Our dear neutral friend have positive prediction about AMD, as usual: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...lake-discussion-threads.2606448/post-40833688

 

[Frenetic Pony]

The mobile chips are where all the I/O uncore improvements were worked on, and it's only with Zen 4 that that finally feeds back into desktop/server. So obviously in that area Dragon Range and Phoenix will see less improvement over Rembrandt than Raphael sees over Vermeer. But the core improvements, higher performance at lower power usage, likely will still be significant. For Phoenix we likely see further core optimizations shared with Bergamo.


Not sure who you target those questions at, personally I find them uninteresting. I'd say the press is mainly about reassuring the DIY sales channels.

The core efficiency yes, though we can't take the power efficiency as a guess here as mobile Zen 4 uses a different process node (N4P?). 6000 matches up with M2 for performance, but at 15 vs 10 watts. It'll be interesting to see what the Zen 4 mobile and RDNA3 will bring.

 

[inf64]

I was thinking about this slide:


We know that 5950X with 230W PPT limit scores ~30K in R23 MT while running @ 4.4Ghz (thanks to Det0x!).
This implies that 7950X with the same 230W PPT limit should score 1.35 x 30000~=40500 pts. This score is true for 7950X that is running an all core boost of around 5.45Ghz!

So if the 7950X is let lose to rampage up to its theoretical PPT limit, it should run compute heavy workloads at almost 5.5GHz, which is insane. For reference, 5950X *can* boost all cores up to 4.2Ghz in non-compute heavy workloads (in R23 MT test it boosts all cores to 3.85Ghz or so). No wonder AMD cited ~48% uplift Vs 5950X in some workloads, the chip is boosting to ~5.5GHz most likely.

Edit:

Going to the first chart:
5950X PPT limited to 88W runs R23 MT at ~2.4Ghz and scores ~16210pts. (thanks again to Det0x!).
The 74% gain at the same 65W TDP/88W PPT implies that 7950X should run the same test at around ~3.8Ghz on all 16 cores and scoring ~28000 pts. Crazy stuff.

Compare this efficiency with 12900K that scores ~26-27K points while running at the 241W limit and you get a crazy ~2.8X better perf./watt. Am I missing something here?

 

[gdansk]

I was thinking about this slide:
View attachment 66867

We know that 5950X with 230W PPT limit scores ~30K in R23 MT while running @ 4.4Ghz (thanks to Det0x!).
This implies that 7950X with the same 230W PPT limit should score 1.35 x 30000~=40500 pts. This score is true for 7950X that is running an all core boost of around 5.45Ghz!

So if the 7950X is let lose to rampage up to its theoretical PPT limit, it should run compute heavy workloads at almost 5.5GHz, which is insane. For reference, 5950X *can* boost all cores up to 4.2Ghz in non-compute heavy workloads (in R23 MT test it boosts all cores to 3.85Ghz or so). No wonder AMD cited ~48% uplift Vs 5950X in some workloads, the chip is boosting to ~5.5GHz most likely.

I think it'll be lower.
1. Det0x always has great results. Might be his cooling setup.
2. It's still only ~14W per core. 5.5GHz is very high for that. Might be possible but I doubt it outside of nearly empty loops.

 

[jpiniero]

AMD adjusts prices often. So they'll drop if Intel launches something competitive.

These are the competitive prices. AMD's costs are way higher. Margins are going to be worse as it is.

They'll fill out the AM5 product stack with Rembrandt (probably w/o an IGP) at some point. Maybe the timing of that depends on how long it takes for them to accumulate enough busted IGP dies.

 

[Joe NYC]

Compare this efficiency with 12900K that scores ~26-27K points while running at the 241W limit and you get a crazy ~2.8X better perf./watt. Am I missing something here?

I was commenting about it yesterday, from slightly different angle and potential implications to Genoa vs. Sapphire Rapids

Page 332 - Discussion - Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

Page 332 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[jpiniero]

I was commenting about it yesterday, from slightly different angle and potential implications to Genoa vs. Sapphire Rapids

Yeah if AMD fixed the power draw of the IO die... that's going to be an even bigger slaughter.

 

[Abwx]

I was thinking about this slide:
View attachment 66867

We know that 5950X with 230W PPT limit scores ~30K in R23 MT while running @ 4.4Ghz (thanks to Det0x!).
This implies that 7950X with the same 230W PPT limit should score 1.35 x 30000~=40500 pts. This score is true for 7950X that is running an all core boost of around 5.45Ghz!

So if the 7950X is let lose to rampage up to its theoretical PPT limit, it should run compute heavy workloads at almost 5.5GHz, which is insane. For reference, 5950X *can* boost all cores up to 4.2Ghz in non-compute heavy workloads (in R23 MT test it boosts all cores to 3.85Ghz or so). No wonder AMD cited ~48% uplift Vs 5950X in some workloads, the chip is boosting to ~5.5GHz most likely.

Edit:

Going to the first chart:
5950X PPT limited to 88W runs R23 MT at ~2.4Ghz and scores ~16210pts. (thanks again to Det0x!).
The 74% gain at the same 65W TDP/88W PPT implies that 7950X should run the same test at around ~3.8Ghz on all 16 cores and scoring ~28000 pts. Crazy stuff.

Compare this efficiency with 12900K that scores ~26-27K points while running at the 241W limit and you get a crazy ~2.8X better perf./watt. Am I missing something here?

At a strict 65W the 7950X should perform about 50% better than the 5950X, this 1.74x is surely erroneous, moreover if it s at 88W PPT.

At 105W, wich is surely 142W PPT, the number is right, and spot on with AMD s slide at their financial day.

Then the comparison at 170W, and surely at 230W PPT or so, is made with an overclocked 5950X and the 32% figure is quite good for the 7950X considering the high, and forcibliy power inneficient, frequencies to reach that perf lead.