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!

 

[Abwx]

That 7700x is pushing very high multithreaded clocks in this test, compared to the 7600x. You can see this from the delta between 1t and nt.

Without knowing the single core frequency that s a wild guess...

Otherwise the 7700X perform 31% better with 33% more cores, that s indicative that they are clocked about identically for MT.

 

[StinkyPinky]

4 More e cores. So basically a 12900K/S

Ah. Not so impressive then.

 

[Zucker2k]

Without knowing the single core frequency that s a wild guess...

Otherwise the 7700X perform 31% better with 33% more cores, that s indicative that they are clocked about identically for MT.

True, I should've paid more attention

 

[Abwx]

First of all, AMD wouldn't give out nearly accurate performance estimate of a product more than a quarter prior to its release!

Either one know what reading is or either he turn a blind eye...

That was already posted, btw...

Notebook-Roadmap: AMD versieht Zen-4- und Zen-5-APUs mit RDNA-3(+)-Grafik

In AMD-Notebooks werden auch im Jahr 2023 und 2024 neue APUs eingesetzt, die mit Zen 4 und Zen 5 schnelle CPU-Kerne plus RDNA-3-GPU bieten.

www.computerbase.de

 

[igor_kavinski]

What has changed?

Intel is providing more value to desktop gamers at more affordable price points.

 

[deasd]

Those wondering why we have not seen CPU-Z leaks featuring Ryzen 7000 CPUs yet, is actually simple. In its current state, Ryzen CPUs reportedly have problem fishing the full benchmark. This might be solved soon as AMD releases an updated BIOS, however for now there are CPU-Z scores available.
AMD 7000-series CPUs either do not run the benchmark in full (stuck at 80%) or show lower performance gains compared to Cinebench. In any case, it is probably better to wait for someone to verify this issue with the latest BIOS or using different samples.

This is also what I thought. Let's hope this is not a cache related bug like Raptorlake ES1.

 

[pakotlar]

This is also what I thought. Let's hope this is not a cache related bug like Raptorlake ES1.

What cache related bug?

 

[nicalandia]

What cache related bug?

Intel i9-13900K ES1 and ES3 really can't perform properly

https://twitter.com/x/status/1542003665975709696

 

[deasd]

What cache related bug?

The ES RaptorCove's IPC down 10% and gaming perf down quite a bit due to unknown reason.
There was a speculation that it's L2 cache issue to blame.

 

[Markfw]

The ES RaptorCove's IPC down 10% and gaming perf down quite a bit due to unknown reason.
There was a speculation that it's L2 cache issue to blame.

Why are we talking about Raptor cove in a Zen 4 thread ?

 

[nicalandia]

Why are we talking about Raptor cove in a Zen 4 thread ?

I believe he was mentioning because at the time the ES sample was buggy so the 7950X might be buggy by way of being ES as well.

I mean at this stage we are dealing with ES and QS samples with buggy Bios and AMD/Intel Scrambling to see who looks the least worst at release date(more so AMD since it's a new platform)

 

[Rigg]

AMD sales & marketing has a real challenge ahead of them. At least if we assume the performance and SKU leaks are mostly accurate for upcoming consumer desktop.

When Intel opted to split i5's into 2 different core configs, as opposed to AMD's approach with 2 different R9 core configs, they made AMD's pricing look really bad when 12th gen launched. Especially in the mid range. Had AMD been prepared for this (which they should have been) they could have dropped prices to current levels on Vermeer shortly before Alder Lake released. Instead they spent an entire review cycle looking overpriced.

I fear they might make an even worse pricing mistake this time around. They don't have the luxury of being a year ahead of Intel for this launch. If they're priced too high, and Intel undercuts them again, they're going to look really dumb cutting prices 2 months after launch.

A zen4 6/8c R5/R7 logically can't compete with a RPL 6+8/8+8 i5/i7. It's plausible the 7700x gets beat by the 13600k. While R 3/5/7/9 & i 3/5/7/9 are largely arbitrary designations, and price to performance ultimately reigns supreme, an i5 beating an R7 looks REALLY bad from a marketing perspective. Even more so when the products are launching in such a close succession.

I personally think they should shift the SKU's down a tier and not launch another 950x until they can put more than 16 cores on consumer desktop. It looks like this is highly unlikely to happen though. We'll more likely see AMD's sales and marketing get jebaited by Intel's again.

 

[nicalandia]

It's plausible the 7700x gets beat by the 13600k.

What makes you think 6P-8E 13600K Will beat 8 Big cores in anything worth measuring? All the rage will be gaming(AMD will take the lead there), most apps don't scale past 8 cores...

 

[PJVol]

That 7700x is pushing very high multithreaded clocks in this test, compared to the 7600x. You can see this from the delta between 1t and nt.

There's nothing unusual. 1T/nT delta is basically the same as in 5000 series.
That's because the ST performance mostly depends on process and packaging variations, and ~ the same troughout the whole product line if we're not to take into account the restriction of boost clocks for the product segmentation.

 

[tamz_msc]

What makes you think 6P-8E 13600K Will beat 8 Big cores in anything worth measuring? All the rage will be gaming(AMD will take the lead there), most apps don't scale past 8 cores...

7700X would lose to the 13600K in Cinebench if the leaked scores are correct.

 

[Rigg]

What makes you think 6P-8E 13600K Will beat 8 Big cores in anything worth measuring? All the rage will be gaming(AMD will take the lead there), most apps don't scale past 8 cores...

I didn't say I thought it would beat it. I said it was plausible. It's possible AMD wins gaming. It's also possible they trade blows with Intel or lose. I don't think the gaming win is a foregone conclusion without vcache. Intel will almost certainly win adobe apps because of quick sync alone. That's worth measuring for people doing creative work. Either way if the 2 SKU's are close in performance the Intel looks better from a marketing perspective. That was my entire point.

 

[Kaluan]

AMD sales & marketing has a real challenge ahead of them. At least if we assume the performance and SKU leaks are mostly accurate for upcoming consumer desktop.

When Intel opted to split i5's into 2 different core configs, as opposed to AMD's approach with 2 different R9 core configs, they made AMD's pricing look really bad when 12th gen launched. Especially in the mid range. Had AMD been prepared for this (which they should have been) they could have dropped prices to current levels on Vermeer shortly before Alder Lake released. Instead they spent an entire review cycle looking overpriced.

I fear they might make an even worse pricing mistake this time around. They don't have the luxury of being a year ahead of Intel for this launch. If they're priced too high, and Intel undercuts them again, they're going to look really dumb cutting prices 2 months after launch.

A zen4 6/8c R5/R7 logically can't compete with a RPL 6+8/8+8 i5/i7. It's plausible the 7700x gets beat by the 13600k. While R 3/5/7/9 & i 3/5/7/9 are largely arbitrary designations, and price to performance ultimately reigns supreme, an i5 beating an R7 looks REALLY bad from a marketing perspective. Even more so when the products are launching in such a close succession.

I personally think they should shift the SKU's down a tier and not launch another 950x until they can put more than 16 cores on consumer desktop. It looks like this is highly unlikely to happen though. We'll more likely see AMD's sales and marketing get jebaited by Intel's again.

Ryzen 5000 outsold 12th gen in DIY markets even with launch/near launch prices, for months and months, the ADL buzz barely made a dent in Ryzen 5000 sales, it's only since spring that prices have come down a lot (or the MSRP were readjusted and more budget SKUs arrived, if you will), up to the point where Ryzen 5000 beats 12th gen in both overall performance and price, aside from not needing expensive mobos to run and not being no way near as power hungry as Intel's PL2/unlocked insanity.

I think AMD knew full well what they were doing (which is, bring in as much revenue as possible and maintaining the "we're not a budget option anymore" without harming the brand by looking too greedy). Mindshare does wonders to sales, and AMD for a few years now, has stared gaining and harnessing it, and harnessing it well they have.

We already know 13th gen won't be launching with prices lower than 12th gen, prices which in and of themselves, were no real issue for AMD to begin with. So what exactly is the worry here?
Neither 7700X or 7800X will cost *$450 this gen, hell I'm pretty sure 7600X and 7700X will undercut Intel hard, considering their cheapest 13th gen (13600K) until next year, will cost up to $350, it's likely it's Intel who won't have a good 'next gen budget option' this time. One also should not underestimate the "new platform" effect, particularly since it's AMD, who have created this thing I like to call "platform mindshare", which I suspect they'll cash in on with AM5 even harder.

* - "7800X3D" probably will tho.

What makes you think 6P-8E 13600K Will beat 8 Big cores in anything worth measuring? All the rage will be gaming(AMD will take the lead there), most apps don't scale past 8 cores...

When the slew of Alder Lake leaks started pouring, particularly Cinebench ones, people used to say the same thing "omg Intel i5 destroys Ryzen 5800X in R23!". Fast forward to the actual reviews (places that don't just run GB and CB for 1m and that do meta-results) and whoops, what do we see? 5800X edges out the 12600K in productivity. But either way, none of that matters, to me at least, price/performance and efficiency is where it's at.

 

[nicalandia]

I didn't say I thought it would beat it. I said it was plausible. It's possible AMD wins gaming. It's also possible they trade blows with Intel or lose. I don't think the gaming win is a foregone conclusion without vcache. Intel will almost certainly win adobe apps because of quick sync alone. That's worth measuring for people doing creative work. Either way if the 2 SKU's are close in performance the Intel looks better from a marketing perspective. That was my entire point.

The 13600K is as far as MT and ST goes a 12700K, the 5800X was very competitive with that CPU both in gaming and application performance.








The 7700X will be a tier higher than the 13600K

 

[Karnak]

7700X would lose to the 13600K in Cinebench if the leaked scores are correct.

Right. In Cinebench. Same goes already for 5800X vs. 12600K (the latter is like 10% ahead) but overall the 5800X is slightly faster in MT apps. Cinebench is probably the worst case for AMD.

 

[igor_kavinski]

One also should not underestimate the "new platform" effect, particularly since it's AMD, who have created this thing I like to call "platform mindshare", which I suspect they'll cash in on with AM5 even harder.

Seems you are the inventor of this term. Better write a formal paper and get the credit or someone else will.

 

[Rigg]

Ryzen 5000 outsold 12th gen in DIY markets even with launch/near launch prices, for months and months, the ADL buzz barely made a dent in Ryzen 5000 sales, it's only since spring that prices have come down a lot (or the MSRP were readjusted and more budget SKUs arrived, if you will), up to the point where Ryzen 5000 beats 12th gen in both overall performance and price, aside from not needing expensive mobos to run and not being no way near as power hungry as Intel's PL2/unlocked insanity.

I think AMD knew full well what they were doing (which is, bring in as much revenue as possible and maintaining the "we're not a budget option anymore" without harming the brand by looking too greedy). Mindshare does wonders to sales, and AMD for a few years now, has stared gaining and harnessing it, and harnessing it well they have.

We already know 13th gen won't be launching with prices lower than 12th gen, prices which in and of themselves, were no real issue for AMD to begin with. So what exactly is the worry here?
Neither 7700X or 7800X will cost *$450 this gen, hell I'm pretty sure 7600X and 7700X will undercut Intel hard, considering their cheapest 13th gen (13600K) until next year, will cost up to $350, it's likely it's Intel who won't have a good 'next gen budget option' this time. One also should not underestimate the "new platform" effect, particularly since it's AMD, who have created this thing I like to call "platform mindshare", which I suspect they'll cash in on with AM5 even harder.

* - "7800X3D" probably will tho.

The only pricing rumors I've read thus far are Intel RPL is +20% ADL MSRP, and AMD Raphael is the same MSRP as Vermeer. If i5 beats or even closely competes with R7 that's bad for AMD's mind share.

 

[Saylick]

The only pricing rumors I've read thus far are Intel RPL is +20% ADL MSRP, and AMD Raphael is the same MSRP as Vermeer. If i5 beats or even closely competes with R7 that's bad for AMD's mind share.

I believe it was up to 20% higher prices for client chips, not a flat +20% across the board. As for AMD, I think the yield and thus margins on the N5 CCDs will be pretty good still so they can likely afford to not raise the price all that much, if any at all, over what Vermeer launched at.

 

[Timmah!]

Either one know what reading is or either he turn a blind eye...

That was already posted, btw...

Notebook-Roadmap: AMD versieht Zen-4- und Zen-5-APUs mit RDNA-3(+)-Grafik

In AMD-Notebooks werden auch im Jahr 2023 und 2024 neue APUs eingesetzt, die mit Zen 4 und Zen 5 schnelle CPU-Kerne plus RDNA-3-GPU bieten.

www.computerbase.de

I really want you to be right, in other words 7950x to score 35k at 150W and be an absolute monster chip, as i want it to get myself, but if you are basing those numbers off that slide, it probably wont end well.

I think the more likely approach here is to take 5950x, multiply its stock CB score by 1,35 (should be about 35k?), find what clocks it needs to have to achieve that score (5GHz+?), look at its power consumption at those clocks and multiply it 0,75x. And that would be your 7950x power-draw at 1,35x more perf than 5950x.
I dont know how much power it draws at 5GHz, but found video on utube, where the package power at 4,85 is apparently 290W. 290 x 0,75 = 217, thus slightly under mentioned 230W PPT.

I am generally crap at math, so i might be completely off.

 

[Joe NYC]

unboxing of Asus X670-E Hero and Gene, if youre interested. In Czech language though. I dont think there is anything new particular being said. But some nice close-up shots.

Ha! Who knew I would one day be able to put to use my (weak) knowledge of Czech language, to watch computer hardware review video.

 

[nicalandia]

Gaming Benchmarks are due at any moment now(basing this on the past info released by the source extreme player)