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]

At one point of another Locuza is going to use the Better Rez Zen4 Die Shot that I did(can someone please find a High Rez one so I can send it to Locuza) but this is my best attempt at it.

Zen4 compared at scale with Zen3 and Golden Cove, Look at the double L1 and Double 256 FP Registers(Huge)



They are at scale. Zen4 is 71mm2 for the CCD and Zen3 is 74mm2. Source AMD.

 

[Thunder 57]

True, but look at what frequency his system actually runs when he sets It manually.
At 88W, he has comparable or a bit higher frequency than stock 5950x at 142W.

That is quite impressive and perhaps some tuning was involved. I was just talking about a general trend though. For example, see this image:

The 5600X is the most efficient of the Zen 3 CPU's out of the box.

 

[xilli_fiberbit]

In CPU-Z 2.02 version, there are some unannounced SKU :

https://twitter.com/x/status/1564593491010695169

 

[Gideon]

I know that, as I said though, about 1 out of 5 Alder Lake apparently won't do DDR5-6000. Almost 100% Alder Lake can do DDR5-5600. 100% of Raptor Lake should do DDR5-5600 though, since it is rated for that speed. If it can't it's a defective RMA.

We have no idea what % of Zen 4 will do DDR5-6000 though.

Just as a reminder, how many of you Zen 3 owners are running DDR4-4000, hm?

You are so grasping at straws. The slide was about overclocking (and pretty accurate as many 3xxx series CPUs also couldn't do 3800 MHz).

The point is AMD always lists JEDEC support on those pages:

5950X: https://www.amd.com/en/products/cpu/amd-ryzen-9-5950x
7950X: https://www.amd.com/en/products/cpu/amd-ryzen-9-7950x

So you're telling me no XMP memory kit runs over 3200 MT/s on Zen 3?

 

[rainy]

True, but look at what frequency his system actually runs when he sets It manually.
At 88W, he has comparable or a bit higher frequency than stock 5950x at 142W.

He clearly won a silicon lottery, his 5950x is an excellent sample.

 

[Rigg]

There are AMD and Intel specific RAM settings (timings) that can lead to significant performance boosts.

What are those specific primary, sub, or tertiary timings that give a significant advantage to one over the other? How do the specific timings that provide your claimed significant performance boost manifest themselves in actual performance data?

 

[shady28]

I don't get the focus on the EXPO timings. This is nothing but XMP with AMD's branding. Were AMD CPU's previously handicapped when using (intel specific) XMP timings? I'm failing to understand how testing both CPU's on the exact same 6000 Mt/s CL32 DDR5 kit is some kind of grand conspiracy to misrepresent 12900k performance.

C'mon man! If you only get 10% more 1080p gaming performance (using a 3090ti) from 100% more memory speed (3200-6400), how much difference is a 7% jump (6000-6400) going to make? Show me the benchmarks. The copium is strong with you.

All you're really telling me here is that you don't know anything about RAM, especially as regards memory sub timings and Intel-specific XMP vs AMD specific AMP / EXPO.

 

[Det0x]

He clearly won a silicon lottery, his 5950x is an excellent sample.

Well i did say earlier in this thread that i had tried a few different 5950x's
(dont affect stock PBO results much tho)

https://imgur.com/a/6ovsyya

As binning data above show, its about ~100mhz difference between a good and a bad Zen3 CPU (CCD) at same voltage.

 

[DrMrLordX]

What are those specific primary, sub, or tertiary timings that give a significant advantage to one over the other? How do the specific timings that provide your claimed significant performance boost manifest themselves in actual performance data?

One example is/was trfc values. AMD memory controllers typically want much lower trfc/trfc2/trfc4 values than Intel memory controllers at a given clockspeed and given primary timings. XMP has clasically set screwed up trfc values on AMD CPUs. Then it'll set funky voltages to try to get everything to work . . .

All you're really telling me here is that you don't know anything about RAM, especially as regards memory sub timings and Intel-specific XMP vs AMD specific AMP / EXPO.

That wasn't very nice.

 

[CakeMonster]

In CPU-Z 2.02 version, there are some unannounced SKU :

https://twitter.com/x/status/1564593491010695169

No 7900X..? Its in the changelog so this is weird (or edited/fake?)

 

[shady28]

You are so grasping at straws. The slide was about overclocking (and pretty accurate as many 3xxx series CPUs also couldn't do 3800 MHz).

The point is AMD always lists JEDEC support on those pages:

5950X: https://www.amd.com/en/products/cpu/amd-ryzen-9-5950x
7950X: https://www.amd.com/en/products/cpu/amd-ryzen-9-7950x

So you're telling me no XMP memory kit runs over 3200 MT/s on Zen 3?

I have no idea how you came up with that lame statement other than making straw men.

I clearly said that it was not guaranteed to run beyond DDR5-5200 and that is a fact.

The ignorant poster that I was responding to implied that it could, because AMD said that Zen 4 ran optimally with DDR5-6000. That is not the same thing, pointed out yet again by the fact that Zen 3 optimal speed is DDR4-4000 yet well under half can even hit that.

Your links prove my point.




And to refer back to what I actually said, as opposed to your total mis-representation

 

[inf64]

Here is an excel sheet I created that compares GB5 ST performance and IPC for Zen 3 (5600X), Golden Cove (12400) and Zen 4 (7600X):

GB5 Zen IPC table.xlsx

Sheet1 ST comparison,AES included,AES excluded <a href="https://browser.geekbench.com/v5/cpu/15896861">https://browser.geekbench.com/v5/cpu/15896861</a>,Core,IPC,Core,IPC 5600x; ST clock=>,4.723,Zen 3,1.00,Zen 3,1.00 Golden Cove,1.11,Golden Cove,1.12 <a href="https://browser.geekbench.com/v5/cpu...

docs.google.com

 

[Abwx]

I have no idea how you came up with that lame statement other than making straw men.

I clearly said that it was not guaranteed to run beyond DDR5-5200 and that is a fact.

The ignorant poster that I was responding to implied that it could, because AMD said that Zen 4 ran optimally with DDR5-6000. That is not the same thing, pointed out yet again by the fact that Zen 3 optimal speed is DDR4-4000 yet well under half can even hit that.

Your links prove my point.

View attachment 66843


And to refer back to what I actually said, as opposed to your total mis-representation


View attachment 66844

Intel rate the 12900K max ram speed at 4800MT, yet AMD used 6000MT for this CPU in their comparisons, so they outspecced the 12900K by much more than the 7950X wich is rated at 5200MT.

Anyway we ll have benches from Computerbase, they use max specced RAM, guess that the result will be more fair than AMD s own comparisons.

 

[shady28]

Intel rate the 12900K max ram speed at 4800MT, yet AMD used 6000MT for this CPU in their comparisons, so they outspecced the 12900K by much more than the 7950X wich is rated at 5200MT.

Anyway we ll have benches from Computerbase, they use max specced RAM, guess that the result will be more fair than AMD s own comparisons.

Their entire presentation is polluted by the fact that they effectively ran overclocked systems, arbitrarily stopping at DDR5-6000 using AMD specific performance enhancing timings.

To be blunt I thought that was plainly obvious once the system specs were revealed.

Intel could just as well present Alder Lake at DDR5-7800. They could say, "This is optimal for Alder Lake. Good Luck!"

 

[maddie]

I'm 100% in the same situation.

That user does not provide more info backing it up except being confident, so its not worth anything even if they could theoretically be Lisa herself. Even semi famous 'leakers' have been caught BS'ing.

One important factor is that for the 1st time in ages, AMD will be able to probably produce everything customers want. Would they want to offer, say a $899 7950X3D? I think yes.

 

[Rigg]

One example is/was trfc values. AMD memory controllers typically want much lower trfc/trfc2/trfc4 values than Intel memory controllers at a given clockspeed and given primary timings. XMP has clasically set screwed up trfc values on AMD CPUs. Then it'll set funky voltages to try to get everything to work . . .

I won't argue with you there. I think my b-die kit had trfc at 700 something with xmp. I got it down to 280. It didn't exactly equate to any massive performance gains though.

 

[Thunder 57]

Their entire presentation is polluted by the fact that they effectively ran overclocked systems, arbitrarily stopping at DDR5-6000 using AMD specific performance enhancing timings.

To be blunt I thought that was plainly obvious once the system specs were revealed.

Intel could just as well present Alder Lake at DDR5-7800. They could say, "This is optimal for Alder Lake. Good Luck!"

You really are fighting a losing battle. Why don't you just come out and say "mehhh AMD te suxx, Intel gud" already?

 

[nicalandia]

As you can see AMD has devoted a lot of resources on the AVX512 Implementation, even if it's 2 256 FP Registers, they doubled on that and they doubled L2

 

[Gideon]

That is not the same thing, pointed out yet again by the fact that Zen 3 optimal speed is DDR4-4000 yet well under half can even hit that.

...

Your links prove my point.

View attachment 66843

My point was that the connectivity chart on the Ryzen 5950X page also only lists "up to 3200 MHz", yet as you know and mentioned above, it can do at least 3600 MT/s - 3800 MT/s on the vast majority of them:


AMD also confirmed that that memory makers will release at least 15 different EXPO kits at launch running as high as 6000 MT/s - 6400 MT/s.

So what you are saying is that AMD will let memory manufacturers release special EXPO kits, that will not run at all on, let's say half of the released CPUs? And you don't mean only in 1:2 Infinity Fabric mode, etc, but simply won't boot at all when enabling their default EXPO overclock profiles?

I personally find it very hard to believe (to put it lightly).

 

[Abwx]

Their entire presentation is polluted by the fact that they effectively ran overclocked systems, arbitrarily stopping at DDR5-6000 using AMD specific performance enhancing timings.

To be blunt I thought that was plainly obvious once the system specs were revealed.

Intel could just as well present Alder Lake at DDR5-7800. They could say, "This is optimal for Alder Lake. Good Luck!"

View attachment 66845

Higher frequencies than 6000MT have been displayed with ESs, AMD used 6000MT for their CPU because thay are confident that it will work on about all MBs and CPUs.

 

[Rigg]

All you're really telling me here is that you don't know anything about RAM, especially as regards memory sub timings and Intel-specific XMP vs AMD specific AMP / EXPO.

Yep nothing at all



Pipe down on the insults, provide some meaningful data, or STFU.

 

[Det0x]

Higher frequencies than 6000MT have been displayed with ESs, AMD used 6000MT for their CPU because thay are confident that it will work on about all MBs and CPUs.

Yes, like here from almost 1 month ago

https://twitter.com/x/status/1554396893920522245

 

[moinmoin]

You said:

But as Zen 3 already has shown if faced with the question of supplying low margin mass market or high margin specialist markets AMD will pick the latter, happily leaving the former to Intel which is still more competent at creating true mass market products in high quantity.

I replied:

I'm going to disagree here.

I suggest that the only reason that AMD went for margin at the expense of marketshare was their constrained production capacity. Once any business can produce at will they will be forced to go for volume. There is no other option if you want to grow revenue.

In other words there is no longer any "either or" choice for AMD.

I see. I feel you are too optimistic about the production capacity AMD has available on N5. AMD just stated that there should be no supply issue in this Zen 4 consumer launch, nothing else. 7600x-7950x all target the enthusiasts DIY desktop market, which is very likely the smallest market AMD targets nowadays, with server, mobile and OEM/ODM desktop markets all being significantly bigger. All of those markets currently suffer on limited supply to different degrees. That AMD planned ahead a launch for the enthusiasts DIY desktop market with a quantity where they are confident that there won't be supply constraints and state as much doesn't mean AMD has the quantity yet to eradicate all supply constraints in all markets and on top of that can push volume in lower margin higher volume markets they purposefully don't target at the moment.

 

[HurleyBird]

The huge perf/watt advantage that Zen 4 has over Zen 3 is going to make Dragon Range and Phoenix unbeatable, even versus Meteor Lake.

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.

 

[In2Photos]

I see. I feel you are too optimistic about the production capacity AMD has available on N5. AMD just stated that there should be no supply issue in this Zen 4 consumer launch, nothing else. 7600x-7950x all target the enthusiasts DIY desktop market, which is very likely the smallest market AMD targets nowadays, with server, mobile and OEM/ODM desktop markets all being significantly bigger. All of those markets currently suffer on limited supply to different degrees. That AMD planned ahead a launch for the enthusiasts DIY desktop market with a quantity where they are confident that there won't be supply constraints and state as much doesn't mean AMD has the quantity yet to eradicate all supply constraints in all markets and on top of that can push volume in lower margin higher volume markets they purposefully don't target at the moment.

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?