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!

 

[Mopetar]

No, that is wrong comparison or why compare i5 13400 vs R7 7700.

The logical comparison is i5 13400 vs R5 7600X, regardless of the price difference and TDP difference.

- Intel 10 CPU Cores+SMT vs AMD 6 CPU Cores+SMT

Acer Predator PO5-650 vs Gigabyte Technology Co., Ltd. X670 AORUS ELITE AX - Geekbench

I'm not sure why your particular comparison is any more logical. Logical comparisons are core for core (or as close as possible given the nature of the different approaches used), price for price, or power-based comparisons. Anything else (at least to the extent that it doesn't try to compare like for like) is just arbitrary.

 

[Asterox]

I'm not sure why your particular comparison is any more logical. Logical comparisons are core for core (or as close as possible given the nature of the different approaches used), price for price, or power-based comparisons. Anything else (at least to the extent that it doesn't try to compare like for like) is just arbitrary.

Well, it is simple R5 vs i5K or non K+certain amount of E Cores=not for gaming.

If you go back in history, let the old i5 buyers complain that the old i5 was only a 4/4 CPU.

 

[nicalandia]

Well, it is simple R5 vs i5K or non K+certain amount of E Cores=not for gaming.

If you go back in history, let the old i5 buyers complain that the old i5 was only a 4/4 CPU.

Look, the R5 vs i5 is no longer relevant here since Intel is Spamming those R5 With e cores, so how about this, Thread vs Thread at a the same power envelope like the 13900K vs 7950X?

So 16T vs 16T 65W vs 65W




AMD Ryzen 7 7700 Non-X “Zen 4” CPU Benchmark Leaks Out, 5.3 GHz Clocks at 65W & Around 10% Slower Than 7700X

AMD Ryzen 7 7700 Non-X "Zen 4" CPU Benchmark Leaks Out, 5.3 GHz Clocks at 65W & Around 10% Slower Than 7700X

The first benchmarks of AMD's Non-X Ryzen 7000 CPUs, starting with the Ryzen 7 7700, have leaked out in the Geekbench 5 database.

wccftech.com

 

[Harry_Wild]

Look, the R5 vs i5 is no longer relevant here since Intel is Spamming those R5 With e cores, so how about this, Thread vs Thread at a the same power envelope like the 13900K vs 7950X?

So 16T vs 16T 65W vs 65W

View attachment 69858


AMD Ryzen 7 7700 Non-X “Zen 4” CPU Benchmark Leaks Out, 5.3 GHz Clocks at 65W & Around 10% Slower Than 7700X

AMD Ryzen 7 7700 Non-X "Zen 4" CPU Benchmark Leaks Out, 5.3 GHz Clocks at 65W & Around 10% Slower Than 7700X

The first benchmarks of AMD's Non-X Ryzen 7000 CPUs, starting with the Ryzen 7 7700, have leaked out in the Geekbench 5 database.

wccftech.com

It is a sure buy for me at 65W, the 7700! Will AMD release a 7050 too? If so, buy the 7050 65W CPU!

 

[DrMrLordX]

That's not the reason I would have said there's a chance we may not see a clock regression (mind you I'm not ruling it out - I don't know at all if it will or it won't)

Zen 4 actually has seperate clock domains for the Core + L1 and L2 + L3 (+ future V-Cache).

That is a good point; the only reason I would think of the undervolt/binning side of things is that adding v-cache is going to make it slightly harder to cool a hypothetical 7950X3D (which is already hobbled by the thick IHS). By how much I'm not sure. Perhaps they can do some die thinning.

 

[Timmah!]

I can fully appreciate wanting more performance for a lower cost, but justification is a simple matter of mathematics.

If an extra $2,500 spread over several years of ownership doesn't make you more productive in a way that let's you get more work done so as to justify the cost, then you clearly don't need the additional resources.

A 13900K or a 7950X is loads better than what you could have bought for that price five years ago. Prior to the launch of the original Zen CPUs, an 8-core CPU was HEDT and cost $1,000 or more. Now 16 is consumer desktop.

Have you perhaps considered that the people who do buy the professional chips can get more work done which allows them to be more competitive on price and push the people using less efficient equipment out of the market? Sure it sucks if you're the grandma knitting socks by hand that's been driven out by the new mill running power looms, but everyone buying just wants less expensive socks.

Assuming you are using it for work, you can probably even write off the costs as a business expense. If it's a cutthroat competitive market, the question isn't whether you can afford to, it's can you afford not to.

This is very simplified. If you do stuff like archviz rendering, you dont do that particular thing 24/7. There is lot of other work to be done, not just doing final renders, that CPU like Threadripper would do faster. I am small freelancer, not rendering farm. After say 2 years of usage of TR i doubt i would be able to say TR made more money than slower CPU would, or if yes, how much. I would be however able to tell, that when it came to doing the actual rendering, it was faster and more convenient, and it saved my time. Which i could then dedicate to something else than waiting for rendering to be finished, even if that something was not necessarily money-making related. Does that mean i did not need it then?

And yeah, CPUs were slower in the past. You kind of expect them to get faster, if they want your money for them.

 

[moinmoin]

This is very simplified.

Of course it is. Everybody's individual work process is unique in the end.

Does that mean i did not need it then?

Did it increase your productivity (by whatever mean, be it more happiness due to less waiting time or whatever)? If yes I'd say it may have been worth it.

 

[Saylick]

Probably fake but the latest Twitter gossip:

https://twitter.com/x/status/1569313159353335809

EDIT: I'm going to call it now that it's 100% a fake. I looked closer at the images and there's definitely some photo editing going on... for example:


Clearly some image manipulation / pasting going on there.

Also, for the CPU-Z image, I'm pretty sure it's not supposed to say "RPL" in the Revision box...

 

[RnR_au]

And to add to the post above, this thread has some interesting implications of MLID's leak of AMD using N6 for the vcache;

https://twitter.com/x/status/1585376201774071808

Edit: I don't think TSMC gained a 30% decrease in SRAM density going from N7 to N6. Which makes the N6-for-vcache leak of MLID's iffy in my opinion. Although maybe AMD got more density gains out of the libraries than their first effort in the 5800X3D.

 

[Det0x]

New leak
AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores have been spotted

AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores have been spotted - VideoCardz.com

AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores Another unreleased AMD processors shows up early. AMD Ryzen 7 7800X could feature 2 more cores than predecessor, according to the new leak. This X-series SKU has just been spotted on Geekbench along with Gigabyte B650 AORUS...

videocardz.com

The 7800X processor scored 2097, and 16163 points in single and multi-core Geekbench tests. In the same test, Ryzen 3 7300X scores 1984 and 7682 points respectively. Earlier today, we got a glimpse of Ryzen 7 7700 performance, meaning we know have data on three unreleased AMD processors.

The score achieved by Ryzen 7 7800X is not as high as expected because Ryzen 7 7700X already scores higher in single-core test, according to Geekbench data. Thanks to 20 threads, the 7800X is indeed faster in multi-core test by around 15%.

 

[Det0x]

EDIT: I'm going to call it now that it's 100% a fake. I looked closer at the images and there's definitely some photo editing going on... for example:
View attachment 69901
Clearly some image manipulation / pasting going on there.

Also, for the CPU-Z image, I'm pretty sure it's not supposed to say "RPL" in the Revision box...

This is my 100% legit screenshot:

Full screenshot:


*edit*
Also my CPU-Z screenshot

It does say "RPL" in the Revision box

 

[nicalandia]

20C/40T in coming?

Gigabyte Technology Co., Ltd. B650 AORUS ELITE AX - Geekbench

Benchmark results for a Gigabyte Technology Co., Ltd. B650 AORUS ELITE AX with an AMD Ryzen 7 7800X processor.

browser.geekbench.com

10C CCD x 2 should be doable

 

[Saylick]

20C/40T in coming?

Gigabyte Technology Co., Ltd. B650 AORUS ELITE AX - Geekbench

Benchmark results for a Gigabyte Technology Co., Ltd. B650 AORUS ELITE AX with an AMD Ryzen 7 7800X processor.

browser.geekbench.com

View attachment 69909

10C CCD x 2 should be doable

Are you suggesting there's a 10C CCD? I find that hard to believe. Not saying it's not possible, but that it's an odd choice of core count for a separate CCD design, which requires another multi-million dollar mask set.

The 7800X, as shown above, is clearly two CCDs with 5 cores on each. Still an odd number of cores on each CCD but it's all on a ring bus so it doesn't have to be an even amount.

 

[nicalandia]

Are you suggesting there's a 10C CCD? I find that hard to believe. Not saying it's not possible, but that it's an odd choice of core count for a separate CCD design, which requires another multi-million dollar mask set.

The 7800X, as shown above, is clearly two CCDs with 5 cores on each. Still an odd number of cores on each CCD but it's all on a ring bus so it doesn't have to be an even amount.

Would be the first time for AMD producing an odd number CCD.

 

[Saylick]

Would be the first time for AMD producing an odd number CCD.

I would expect something like a 12-core CCD if anything. 10 cores seems too close to the 8 core CCD to justify another tapeout. Might as well go larger and cut down.

 

[jpiniero]

Probally 5+5.

 

[nicalandia]

Probally 5+5.

To compete against the 13600K? I could see that. The 7900X does well against the 13700K

 

[Hail The Brain Slug]

It's definitely 5+5, the L3 cache specifies it's 2x32MB, aka 2 CCD's. Zen4 still has the limitation where dual CCD configuration has to have equal core counts in each CCD.

 

[Joe NYC]

20C/40T in coming?

Gigabyte Technology Co., Ltd. B650 AORUS ELITE AX - Geekbench

Benchmark results for a Gigabyte Technology Co., Ltd. B650 AORUS ELITE AX with an AMD Ryzen 7 7800X processor.

browser.geekbench.com

View attachment 69909

10C CCD x 2 should be doable

It says 32MB x 2 so 2 CCDx.

I think this is a dumb idea for an SKU.

AMD just needs to concentrate on V-Cache and bring it to market ASAP, and create new SKUs by adding V-Cache to more products (7700/7800x, 7900x, 7950x, even 7600x)

 

[Asterox]

New leak
AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores have been spotted

AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores have been spotted - VideoCardz.com

AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores Another unreleased AMD processors shows up early. AMD Ryzen 7 7800X could feature 2 more cores than predecessor, according to the new leak. This X-series SKU has just been spotted on Geekbench along with Gigabyte B650 AORUS...

videocardz.com

View attachment 69903
The 7800X processor scored 2097, and 16163 points in single and multi-core Geekbench tests. In the same test, Ryzen 3 7300X scores 1984 and 7682 points respectively. Earlier today, we got a glimpse of Ryzen 7 7700 performance, meaning we know have data on three unreleased AMD processors.

The score achieved by Ryzen 7 7800X is not as high as expected because Ryzen 7 7700X already scores higher in single-core test, according to Geekbench data. Thanks to 20 threads, the 7800X is indeed faster in multi-core test by around 15%.

View attachment 69904

Geekbench, R3 7300X is quite close to "old 6/12 R5 5600X".

For gaming, even 4/8 Zen 4 we can expect that in most cases it will be faster vs R5 5600X.

AM5 socket definitely needs a much cheaper CPU than current 300$ for R5 7600X.

200$ for R3 7300X, hm very likely this is what we can expect.

 

[maddie]

Geekbench, R3 7300X is quite close to "old 6/12 R5 5600X".

For gaming, even 4/8 Zen 4 we can expect that in most cases it will be faster vs R5 5600X.

AM5 socket definitely needs a much cheaper CPU than current 300$ for R5 7600X.

200$ for R3 7300X, hm very likely this is what we can expect.

Who would buy that for $200 over an i3-13100 with DDR4, unless MB & DDR5 drop significantly?

 

[eek2121]

It's definitely 5+5, the L3 cache specifies it's 2x32MB, aka 2 CCD's. Zen4 still has the limitation where dual CCD configuration has to have equal core counts in each CCD.

Yes, it is 5+5. I will be shocked if they don’t charge a premium. There is no way they have dies with large enough defects to kill off 3 cores unless they have some kind of edge optimization going on. I suspected AMD might try something like this, but it will likely be $400-$500. They will blow my mind if they sell it for less. As it is, margins on this chip are going to be significantly lower than the 7700X (again, unless they charge a premium)

Good on AMD though.

EDIT: if these chips are being tested now, CES will see them launch most likely, which surprises even me. I guess they are more worried about midrange than they would lime to admit. Cheers to competition! 🥂

 

[Harry_Wild]

The differences in prices is going be only $150 differences between the 7800X and the 7050X! Might as well go with the top of the line for that amount!

 

[alexruiz]

New leak
AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores have been spotted

AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores have been spotted - VideoCardz.com

AMD Ryzen 7 7800X with 10 cores and Ryzen 3 7300X with 4 cores Another unreleased AMD processors shows up early. AMD Ryzen 7 7800X could feature 2 more cores than predecessor, according to the new leak. This X-series SKU has just been spotted on Geekbench along with Gigabyte B650 AORUS...

videocardz.com

View attachment 69903
The 7800X processor scored 2097, and 16163 points in single and multi-core Geekbench tests. In the same test, Ryzen 3 7300X scores 1984 and 7682 points respectively. Earlier today, we got a glimpse of Ryzen 7 7700 performance, meaning we know have data on three unreleased AMD processors.

The score achieved by Ryzen 7 7800X is not as high as expected because Ryzen 7 7700X already scores higher in single-core test, according to Geekbench data. Thanks to 20 threads, the 7800X is indeed faster in multi-core test by around 15%.

View attachment 69904

7800x sounds great
7300x, well, I don't think Zen 4 should be in 4C/8T iteration, at least not yet with AM4 still alive and well.

 

[RnR_au]

7300x, well, I don't think Zen 4 should be in 4C/8T iteration, at least not yet with AM4 still alive and well.

Perhaps for budget build possibilities. But if the reputed manufacturing cost for a 7950X is somewhere in the region of $69, I would argue that a price drop would be better than a 4 core sku for boosting diy adoption.