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!

 

[Timmah!]

Yeah, I'm aware of that - it's the same thing as fmax for 5950X. What I was trying to say is that 5.5Ghz for ST is a low number IMO, especially if they demonstrated multiple threads running at 5.5Ghz in a gaming workload. Whether they reach 5.7 or 5.8Ghz is anyone's guess, but the difference is really small (~1.5%).

I think i will be OK with 5,7GHz, lol.

Is that fmax value maximum limit for any overclocking? I mean, no 6 GHz possible, no matter what?

 

[inf64]

I think i will be OK with 5,7GHz, lol.

Is that fmax value maximum limit for any overclocking? I mean, no 6 GHz possible, no matter what?

I doubt it's a hard limit for OC conditions. It should be a hard limit for "stock"/out of the box operation, ie. if PBO is enabled and cooling is amazing (subzero).

 

[Karnak]

5.85mhz fmax means a 5.7-5.75ghz max boost clock fyi, the chip would never hit 5.85ghz.

That's most likely not true if we're assuming it'll work the same like it does since Zen3. The latter exceeds it's official SC boost e.g. the 4.9GHz for the 5950X while fmax is 5.05GHz. Run CB 1T on a 5950X and it'll boost all the way up to it's rated 5.05GHz fmax.

 

[eek2121]

That's most likely not true if we're assuming it'll work the same like it does since Zen3. The latter exceeds it's official SC boost e.g. the 4.9GHz for the 5950X while fmax is 5.05GHz. Run CB 1T on a 5950X and it'll boost all the way up to it's rated 5.05GHz fmax.

That would depend on cooling and any tweaks done to the system. My 5950x with a 280mm AIO does not hit 5050mhz at stock in Cinebench. It barely touches 4.9. If I tweak things a bit, it hits 5050mhz easily.

 

[eek2121]

I think i will be OK with 5,7GHz, lol.

Is that fmax value maximum limit for any overclocking? I mean, no 6 GHz possible, no matter what?

I’m fine with 5.5 even, as long as multicore performance is strong. It would save me from looking into buying a Threadripper.

 

[Makaveli]

That would depend on cooling and any tweaks done to the system. My 5950x with a 280mm AIO does not hit 5050mhz at stock in Cinebench. It barely touches 4.9. If I tweak things a bit, it hits 5050mhz easily.

Agreed i'm on a 360mm AIO and PBO on and do 5.05Ghz boost easily.

 

[DrMrLordX]

Anyone tell me how reliable Angstronomics is?

Just going off early impressions since that leaker/news source hasn't been around for long, but, not very reliable.

 

[Kepler_L2]

Just going off early impressions since that leaker/news source hasn't been around for long, but, not very reliable.

Skyjuice has pretty much a 100% track record so far.

 

[DrMrLordX]

Skyjuice has pretty much a 100% track record so far.

Angstronomics seems tilted to me. We'll see.

 

[deasd]

well..... I don't know what's going on there, especially it's Userbenchmark

AMD Ryzen 5 7600X 6 Core "Zen 4" CPU Sample Destroys The Intel Core i9-12900K, 20% Faster Single-Core Performance In Leaked Benchmark

AMD's Ryzen 5 7600X 6-Core "Zen 4" CPU has been benchmarked once again & it destroys Intel's Core i9-12900K in single-threaded tests.

wccftech.com

 

[igor_kavinski]

ST performance this strong, if true, could be the final nail in Intel's coffin.

 

[pakotlar]

well..... I don't know what's going on there, especially it's Userbenchmark

AMD Ryzen 5 7600X 6 Core "Zen 4" CPU Sample Destroys The Intel Core i9-12900K, 20% Faster Single-Core Performance In Leaked Benchmark

AMD's Ryzen 5 7600X 6-Core "Zen 4" CPU has been benchmarked once again & it destroys Intel's Core i9-12900K in single-threaded tests.

wccftech.com

That would be awesome. Hope it holds up, maybe a 7950x is in my future after all.

 

[pakotlar]

ST performance this strong, if true, could be the final nail in Intel's coffin.

Yeah if the ST results are anywhere near 55% uplift, Intel is utterly -redacted-. Even in MT, with 1.5-2x core count it won’t be able to make that up, probably.


Profanity is not allowed in the technical forums.

Daveybrat
AT Moderator

 

[pakotlar]

But if you choose DDR4 with 13900k, won't you loose the performance that will bring you in front of the 7950X?

And sure DDR5 is more expensive than DDR4 and if you have a system with 64GB or more I can understand the concern, but the total cost of a 32GB system is not going be significantly different choosing DDR5 or DDR4.

Yeah I pretty much agree with you, though DDR5 uplift will not be uniform, and I’m wondering how many customers out there are price sensitive enough for a few hundred extra on ddr5 to matter, if their workloads don’t get a big boost. We’ll see.

 

[pakotlar]

That is ridiculous. Zen 4 isn’t a new family, so we shouldn’t be expecting radical changes. The performance from SMT with much larger L2 and other improvements may be quite large and the MT clock speed is significantly improved. This should result in massive improvements for applications that can take advantage of 32 threads. Saying that this is “stagnation” is absolutely ridiculous. The IPC alone, without the clock speed bump, is likely more than we got from most intel generation during the monopoly.

32 threads is already niche for the consumer market anyway. Given the MT improvements + DDR5 to keep it all fed, I don’t think there is much of a real market for actual 24-core / 48-thread parts. I use a 24-core Epyc F-series part at work, and it is still often stuck at ~2% utilization (1/48) due to single thread portions of the code. I hope they have a new Threadripper level part, but the Threadripper level enthusiast market is tiny. It is good press among enthusiast, but not really much else. I suspect most of their Threadripper sales were OEM machines purchased by or for professionals as workstations.

The context of a word matters. “Core count stagnation” is exactly what has happened on AMD consumer parts, Ryzen 3-7 have 0 increase (3 gen).

There is a market. You admit this yourself, yet you call my post ridiculous.

Ridiculous.

 

[Makaveli]

well..... I don't know what's going on there, especially it's Userbenchmark

AMD Ryzen 5 7600X 6 Core "Zen 4" CPU Sample Destroys The Intel Core i9-12900K, 20% Faster Single-Core Performance In Leaked Benchmark

AMD's Ryzen 5 7600X 6-Core "Zen 4" CPU has been benchmarked once again & it destroys Intel's Core i9-12900K in single-threaded tests.

wccftech.com

My only concern is the source UB is utter trash and you cannot really trust anything from them.

Need another source and need to confirm the benchmark isn't using AVX 512

 

[Timmah!]

Agreed i'm on a 360mm AIO and PBO on and do 5.05Ghz boost easily.

Talking about AIO cooling, quick question, sorry for off-topic.

When i bought my current CPU (7940x), i got 360 AIO to cool it as well - my first AIO ever. Specifically, AlphaCool Eisbaer 360. I read good review for it beforehand, i liked the fact the radiator was copper and not aluminium, and it was not that expensive either (compared to more mainstream alternatives like H110s/115s. Krakens, etc..).
But ultimately i was kind of disapppointed with the performance. Not only it really could not cool even rather mild overclock (anything over 4,1 GHz OC all-core and some cores would go past 90C after few minutes), it was rather loud as hell.

Now my overclock was rather basic, surely it could have been tuned better to get better clocks at lower power, additionally i did not use the Eiswind fans from the package, but replaced them with supposedly superior Fractal Design ones (i think), but anyway, i have to wonder:

- would different brand of AIO fare better, so did i made a mistake with my choice...or nothing but full-blown custom loop could help cooling the CPU (that notoriously was not soldered), unless it was delidded
- did i made a mistake changing the fans? Or mounting the radiator up-top (had no choice there)?

Trying to figure out for my next build (probably 7950x) - which AIO to choose then, for the coolest and most silent operation possible. Recently saw some new 280mm and even 420mm AlphaCool ones, but as pointed above, not sure if they would be the right choice.

 

[lobz]

Can you provide some numbers ? So you are telling me that Zen4 will have +35% single core gain over Zen3 ?

CB R20 ST
5950X = 640~
12900k = 768~
13900K ES (chinese review) = 862~

Nobody implied here, that AMD will beat Intel in Cinebench. Or UserBenchmark. Or SiSoft Sandra.

 

[Asterox]

well..... I don't know what's going on there, especially it's Userbenchmark

AMD Ryzen 5 7600X 6 Core "Zen 4" CPU Sample Destroys The Intel Core i9-12900K, 20% Faster Single-Core Performance In Leaked Benchmark

AMD's Ryzen 5 7600X 6-Core "Zen 4" CPU has been benchmarked once again & it destroys Intel's Core i9-12900K in single-threaded tests.

wccftech.com

Zen 4/ 7-10% higher IPC+10% higher Singlecore CPU frequency, it is kinda of expected to have 20% higher singlecore score vs R5 5600X.

If that UB blah test loves L2 Cache, well that is very expected for Zen 4 with double L2 cache size.

 

[Saylick]

I don't trust Userbenchmark at all. Like others have said, we already kind of know what the upper end of the ST uplift should be.

5.75 GHz / 4.9 GHz * 1.08 IPC = 1.27 or 27% higher.

I think we can assume >20% with high confidence.

 

[Timmah!]

The context of a word matters. “Core count stagnation” is exactly what has happened on AMD consumer parts, Ryzen 3-7 have 0 increase (3 gen).

There is a market. You admit this yourself, yet you call my post ridiculous.

Ridiculous.

This. Market for 16core CPU compared to cheaper, more gaming oriented CPUs, is already "tiny", yet they are selling it. Unless its technically impossible to fit 3 zen4 dies onto AM5 due to size constraints, i am fairly sure if they made such CPU, it would sell no worse than 7950x is going to. Even at 1,5x higher price than 7950x. Half the people who now look at 7950x would be after it, all the enthusiasts, and so would the people looking for best workstation CPU, who got priced out by the new Threadripper price-hike. Additionally, AMD could claim un-contested performance crown in the client market and have that fabled HEDT CPU Intel wont, even if within client platform. 7950x is not going to be that, no matter what, if only because of the fact Raptor Lake is technically going to boast more cores, or the fact that former HEDT products did have more cores than that. If i owned 3960x or even 3970x myself, and was looking for CPU upgrade, and was not keen to pay over the top for TR5000s, 7950x probably does not move a needle for me. Hypothetical 24-core chip, significantly faster at single core and probably on par with 3970x in MT, just might.

 

[nicalandia]

I don't trust Userbenchmark at all.

I have seen for Years how Userbenchmark has changed their algorithm to benefit or paint a good light on Intel. Let see if they do that now.

From the Article.

"In terms of performance, while the UserBenchmark is known to be biased towards Intel CPUs, the chip AMD Ryzen 5 7600X still outperforms Intel's flagship Core i9-12900K in single-core tests. The Ryzen 5 7600X scores an impressive 243 points whereas the Intel Core i9-12900K scores an average of 200 points. This is a 20% higher single-threaded performance increase. Compared to the Ryzen 5 5600X, the Zen 4 chip offers a 55% performance boost in single-core tests"

 

[Saylick]

Seeing how they have been known to be a Intel Pro for many years. It's quite interesting how much of a jump on ST has AMD achieved.

That's part of it, but didn't Userbenchmark have some kind AMD review that read as if some angsty teenager wrote it? Maybe I'm misremembering who that was... but if it was Userbenchmark, then my point stands.

 

[Markfw]

I have seen for Years how Userbenchmark has changed their algorithm to benefit or paint a good light on Intel. Let see if they do that now.

From the Article.

"In terms of performance, while the UserBenchmark is known to be biased towards Intel CPUs, the chip AMD Ryzen 5 7600X still outperforms Intel's flagship Core i9-12900K in single-core tests. The Ryzen 5 7600X scores an impressive 243 points whereas the Intel Core i9-12900K scores an average of 200 points. This is a 20% higher single-threaded performance increase. Compared to the Ryzen 5 5600X, the Zen 4 chip offers a 55% performance boost in single-core tests"

While I would love to see this happen, I will not believe anything until real reviews come out. Too much clickbait out there.

Edit: BTW, its possible that their test uses the e-cores from the 12900k for the single core test ???? Just one possible explanation.

 

[maddie]

I have seen for Years how Userbenchmark has changed their algorithm to benefit or paint a good light on Intel. Let see if they do that now.

From the Article.

"In terms of performance, while the UserBenchmark is known to be biased towards Intel CPUs, the chip AMD Ryzen 5 7600X still outperforms Intel's flagship Core i9-12900K in single-core tests. The Ryzen 5 7600X scores an impressive 243 points whereas the Intel Core i9-12900K scores an average of 200 points. This is a 20% higher single-threaded performance increase. Compared to the Ryzen 5 5600X, the Zen 4 chip offers a 55% performance boost in single-core tests"

CPUID time again?