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]

Some prime performance numbers for 16T.

 

[deasd]

Wait what??
The 13400 Gets blasted to pieces here.

Micro-Star International Co., Ltd. MS-7D69 vs Micro-Star International Co., Ltd. MS-7D31 - Geekbench

View attachment 67912

wtf...... if this 13400 is normally working, looks like the i5s are utterly power constrainted and crushed by higher TDP Zen4......
the low base clock and cannot being OC are killing the whole i5 lineup.......

@nicalandia
To me it is pretty clear that these aren't final numbers for the 13400. Or are the clocks supposed to be that low? I would have expected performance to be more like a 12600.

even the 12600 doesn't perform any better than this 13400 above. +~10% MT maybe.

looks like 13400 has similar power limit of 12400 which leads to low MT perf.

 

[Hitman928]

Well, e-cores enabled and avx-512 disabled by default. The chances are, people who actually can use of avx are knowledgeable enough to find out about the switch and use it, if they need so.

This doesn't work when you have to support something as an actual business across the general population. This is more a modder mindset where there's no real repercussions if people break stuff because you're just a modder and don't support the actual software/hardware. At most, I could see Intel no longer fusing it off to allow modders to do something like this, but they wouldn't officially support it and would probably put out advisories against it, including your warranty being voided if you do it. I doubt it, but that's the most I could see happen.

 

[inf64]

@nicalandia
To me it is pretty clear that these aren't final numbers for the 13400. Or are the clocks supposed to be that low? I would have expected performance to be more like a 12600.

13400K should be roughly on par with 7600X in MT workloads (~same level as 12600K) and ~10% behind 7600X in ST workloads.

 

[Abwx]

Geekbench at least tests real-world workloads. The same thing can't be said about SiSoft Sandra. Hyping up Sandra arithmetic scores - it's the same thing as the Intel camp hyping up CPU-Z scores.

Sandra predicted accurately the perf uplift from the FX8350 to Zen 1 and yet those are two radically different uarch, assuming of course that you look at the relevant benches, so looking at the same benches we know how Zen 4 based SKUs will perform respectively to Zen 3 counterparts...

As for CPU Z you should know that the 1.78 version was the most recent when Zen 1 was released, but it happened that its score were so good that it got updated under the pretense that Zen 1 was taking advantage of a bug (lol) in the bench to get high scores.

The subsequent 1.79 release saw Zen 1 perfs tanking, and looking at Zen 4 scores it s obvious that at the time they exagerately favoured Intel uarch in this "revision" and the following ones...

 

[nicalandia]

13400K should be roughly on par with 7600X in MT workloads (~same level as 12600K) and ~10% behind 7600X in ST workloads.

Not at the Lock Speed of 4.5 Ghz

 

[inf64]

Not at the Lock Speed of 4.5 Ghz

I'm talking about the overall MT and ST performance of 13400 part, as per my own estimate here: bit.ly/3PEY8cM

BTW the wprime numbers for 7700X look to be without AVX512?

 

[Timmah!]

This doesn't work when you have to support something as an actual business across the general population. This is more a modder mindset where there's no real repercussions if people break stuff because you're just a modder and don't support the actual software/hardware. At most, I could see Intel no longer fusing it off to allow modders to do something like this, but they wouldn't officially support it and would probably put out advisories against it, including your warranty being voided if you do it. I doubt it, but that's the most I could see happen.

Not being officially supported might be good enough. Anyway not sure why you think anything has to be broken - turning one thing on would turn the other one off, so there is never a situation where its crashing because of AVX-512 code being forced on e-core.
That the switch is in BIOS and that requires above average computer knowledge to access is IMO non-issue, since general population does not need AVX-512 in the first place, and whoever does, is very very likely to be capable of accessing that switch themselves.

 

[nicalandia]

BTW the wprime numbers for 7700X look to be without AVX512?

It's an Apples to Apples comparison(Speed @4ghz, 16T, AVX-512) so AVX-512 is on as opposed to the 12900K that had to be enabled.

 

[inf64]

It's an Apples to Apples comparison(Speed @4ghz, 16T, AVX-512) so AVX-512 is on as opposed to the 12900K that had to be enabled.

The odd thing to me is that 12900K has the "AVX-512 enabled" while 7700X has no such thing in the result line. That's why it stood out to me.

 

[Hitman928]

Not being officially supported might be good enough. Anyway not sure why you think anything has to be broken - turning one thing on would turn the other one off, so there is never a situation where its crashing because of AVX-512 code being forced on e-core.
That the switch is in BIOS and that requires above average computer knowledge to access is IMO non-issue, since general population does not need AVX-512 in the first place, and whoever does, is very very likely to be capable of accessing that switch themselves.

The original context that I was replying to was turning e-cores on and off from within Windows so stuff breaking is a huge concern. Putting the switch in the bios would at least fix that issue but then you still have a lot of headaches if it is officially supported (and some even if it's not) because if you give people an avenue to mess something up or have an experience that differs from what they think was advertised, they will take that avenue and it will come back to bite you.

 

[Kaluan]

PS3 emulator will FLYYYYYY on Zen 4.

Dolphin Emulator as well, but not for the same reasons, check this out:

https://twitter.com/x/status/1567129743077638145

@nicalandia
To me it is pretty clear that these aren't final numbers for the 13400. Or are the clocks supposed to be that low? I would have expected performance to be more like a 12600.

Low? It's a (up to) 4,6GHz P core (GB ST lines up very well to 12400's low 1600s) and 3,3GHz E cores (lower than any 12th gen). With a PL2 of around 117W.

I'm talking about the overall MT and ST performance of 13400 part, as per my own estimate here: bit.ly/3PEY8cM

You estimated a PL2=117W (or around that) Alder Lake to be within 2% of a PL2=150W Alder Lake? Really? The clocks on their own should be a dead giveaway that 13400 is not = to 12600K, let alone the much more modest PL2. 7600X will eat it for breakfast. Especially in lightly-threaded tasks and gaming.


I see no one is looking at the most obvious culprit of all from that GB run. 13400 seems to be running in single channel mode (RAM reported as 8GB). But GB isn't the most trustworthy in reporting system specs, so who knows.
Either way, I don't think this SKU will crack 10K MT points, even under a more optimal setup. People expecting it to cost the same as 12400 should prepare for disappointment, in about 3 months or so.

 

[biostud]

The odd thing to me is that 12900K has the "AVX-512 enabled" while 7700X has no such thing in the result line. That's why it stood out to me.

I guess that is to tell that it runs with an older bios that has not disabled AVX512 on ADL.

 

[nicalandia]

The clocks on their own should be a dead giveaway that 13400 is not = to 12600K, let alone the much more modes PL2. 7600X will eat it for breakfast.

Whoever thinks a budget locked 13400 without any type of fancy boost technology(as the 12400 was) is on the same tier level as the 7600X is just not beign honest with themselves.

 

[inf64]

You estimated a PL2=117W (or around that) Alder Lake to be within 2% of a PL2=150W Alder Lake? Really? The clocks on their own should be a dead giveaway that 13400 is not = to 12600K, let alone the much more modes PL2. 7600X will eat it for breakfast.

Well I gathered the info online, so whatever I found is in the excel sheet. I made an error (clearly) as I assumed that 13400 and 13100 are Raptor Cove based.

 

[nicalandia]

I made an error (clearly) as I assumed that 13400 and 13100 are Raptor Cove based.

Only the i5 13400 will be Raptor Cove based. the i3 will be based on Golden Cove

 

[inf64]

Only the i5 13400 will be Raptor Cove based. the i3 will be based on Golden Cove

If it's Raptor Cove it will most likely perform very closely to 12600K if this spec is correct.

 

[Rigg]

Wait what??
The 13400 Gets blasted to pieces here.

Micro-Star International Co., Ltd. MS-7D69 vs Micro-Star International Co., Ltd. MS-7D31 - Geekbench

View attachment 67912

Looks like I may have severely overestimated the 13400's performance but the multi thread is clearly hampered by power limits. The fact of the matter is we don't really have a direct competitor for the 7600x. A power limit unlocked 13400 is the best we can do. I should have been more clear, but I wasn't suggesting the 13400 would compete in multi-thread at stock 65w power limits.

 

[nicalandia]

If it's Raptor Cove it will most likely perform very closely to 12600K if this spec is correct.

There is no IF... Just look at the L2$ Size.


But It's a locked CPU without any boosting technology that has a TDP Cap of 65 Watts.


Now can We please get past this topic of how the 13400 Low budget CPU can't compete with the 7600X which AMD Is Clearly pitting against the 12900K(In games) and get back at how the 7700X is peforming against the 12900K In WPrime?...

 

[inf64]

There is no IF... Just look at the L2$ Size.

But It's a locked CPU without any boosting technology that has a TDP Cap of 65 Watts.

The TDP cap will surely limit it unless there is a way to circumvent the power limit in BIOS. It's odd that intel will hand out the "perf. crown" to 7600X so easily (and AMD has Zen 3 parts that will compete in this performance bracket as well). If 7600X is pushing the power to the limit then intel can claim better power efficiency there, so not all is lost I guess.

 

[inf64]

Ryzen 7000 breaking world records on AIO cooling : https://videocardz.com/newz/amd-ryzen-9-7950x-breaks-four-hwbot-world-records-with-aio-cooler

"Using either AORUS X670E Xtreme, ASUS ROG Crosshair X670E Extreme, ASRock’s X670E Taichi or MSI X670E MEG ACE motherboards and Corsair H115i liquid cooler AIO cooler, both overclockers managed to break four records: Cinebench R15, R20, R23 and 7-zip. "

40500pts @ 5.40 GHz (all core) in R23 with around 226W to 244W max power

I wonder if these chips have cold bug and whether they will scale with sub zero temps. If they scale well with LN cooling, then the limit is how much power can the board supply to it.

 

[nicalandia]

Ryzen 7000 breaking world records on AIO cooling : https://videocardz.com/newz/amd-ryzen-9-7950x-breaks-four-hwbot-world-records-with-aio-cooler

Sweet Jesus..!

 

[nicalandia]

40500pts @ 5.40 GHz (all core) in R23 with around 226W to 244W max power

I wonder if these chips have cold bug and whether they will scale with sub zero temps. If they scale well with LN cooling, then the limit is how much power can the board supply to it.

This guy will be posting numbers soon.

 

[Saylick]

Ryzen 7000 breaking world records on AIO cooling : https://videocardz.com/newz/amd-ryzen-9-7950x-breaks-four-hwbot-world-records-with-aio-cooler

"Using either AORUS X670E Xtreme, ASUS ROG Crosshair X670E Extreme, ASRock’s X670E Taichi or MSI X670E MEG ACE motherboards and Corsair H115i liquid cooler AIO cooler, both overclockers managed to break four records: Cinebench R15, R20, R23 and 7-zip. "

40500pts @ 5.40 GHz (all core) in R23 with around 226W to 244W max power

I wonder if these chips have cold bug and whether they will scale with sub zero temps. If they scale well with LN cooling, then the limit is how much power can the board supply to it.

VERY impressive considering 226W - 244W isn't that much higher than AMD's max socket power of 1.35 * 170W = 230W. It literally does look like the limitations to hitting 5.4 GHz all core just comes down to keeping it cool.

 

[igor_kavinski]

It literally does look like the limitations to hitting 5.4 GHz all core just comes down to keeping it cool.

Great. We can get max performance from it when we take it with us to Mars.