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!

 

[CakeMonster]

RGT did say take it with some salt, as he gets daily changes to price rumors.

That's frankly laughable excuse. We really should stop paying attention to the absolute crap-tier of 'leakers', that are 90% likely just making stuff up from the daily Twitter chatter we can all read.

 

[CakeMonster]

The only ounce of credibility Userbenchmark/CPUPro had left before was for believers of its conspiracy theories already. Not sure those believers suddenly see the pure inanity of all of CPUPro's writings.

Why are they relevant though? I never come across them myself, I only see them referenced in forums etc. Do they rank high on Google? (that would be damning of the search algorithm) Is it in any way considered a decent benchmark despite their attitude (probably not if they just alter the formula on a whim)?

 

[igor_kavinski]

They do have some technical knowledge to be able to develop that benchmark. However, they have got to be shills to appear to be so illogically retarded.

 

[moinmoin]

Why are they relevant though? I never come across them myself, I only see them referenced in forums etc. Do they rank high on Google? (that would be damning of the search algorithm) Is it in any way considered a decent benchmark despite their attitude (probably not if they just alter the formula on a whim)?

They rank high on Google (which really should be changed considering all the malicious tweaking), and aside Geekbench it seems to be a fairly popular place for early benches of unreleased CPUs.

 

[jpiniero]

FWIW, if Intel follows through on their 20% price hike, assuming they are talking about Raptor Lake versus Alder Lake, this is what it would look like:

13900K $699
13700K $499
13600K $349

That seems... realistic? for AMD to copy or maybe even a bit higher. This would assume there would be a 7800X3D between the 7700X and 7900X and also a 7950X.

 

[pakotlar]

FWIW, if Intel follows through on their 20% price hike, assuming they are talking about Raptor Lake versus Alder Lake, this is what it would look like:

13900K $699
13700K $499
13600K $349

That seems... realistic? for AMD to copy or maybe even a bit higher. This would assume there would be a 7800X3D between the 7700X and 7900X and also a 7950X.

Yeah, Intel put downward pressure on 5950x by pricing 12900K at 550 IIRC, now things go back to pre-competitive Intel days. Shows how important competition is.

7950x probably ends up at 700 as well, below 5950x

 

[Kedas]

FYI they changed the words on the website
recent launch -> recent announcement (August 4)

AMD Meet the Experts - An Exclusive Look at Partner Products: A Showcase of The New AM5 Motherboards

explore.amd.com

Not sure if this is an announcement that still has to come before Friday or if they refer to old announcements.
Maybe they just wrote it down wrong the first time or the launch date moved after writing it. (nothing wrong with that though, since there is no official launch date)

 

[deasd]

UB edited their comment, emphasized the 7600X ES is GOLDEN SAMPLE

https://twitter.com/x/status/1553881638991548416

 

[lobz]

all out flat earth believers 😂

 

[ondma]

UB edited their comment, emphasized the 7600X ES is GOLDEN SAMPLE

https://twitter.com/x/status/1553881638991548416

Whole thing is fishy though. >50% increase in single thread performance just doesnt seem possible, unless the weighting of the benchmark has been changed. It the ES was only running at 4.9 ghz that means the data is even "fishier" because the gain would have to be almost entirely from IPC.

 

[pakotlar]

UB edited their comment, emphasized the 7600X ES is GOLDEN SAMPLE

https://twitter.com/x/status/1553881638991548416

LOL. So sore

 

[DrMrLordX]

Whole thing is fishy though. >50% increase in single thread performance just doesnt seem possible,

1). Userbenchmark is trash
2). Zen3 (and older Zen versions) were likely gimped by Userbenchmark while Zen4 is not (necessarily). If an older uarch is crippled then it would be easy to make seemingly-impossible gains by switching to a uarch that isn't crippled.

There likely won't be any ST gains like this moving from Zen3->Zen4 in benchmarks that aren't hot garbage like Userbenchmark.

 

[moinmoin]

Whole thing is fishy though.

Indeed, the benchmark, the website and its author all themselves are fishy.

Zen3 (and older Zen versions) were likely gimped by Userbenchmark while Zen4 is not (necessarily). If an older uarch is crippled then it would be easy to make seemingly-impossible gains by switching to a uarch that isn't crippled.

Userbenchmark keeps changing the weighting to make some Intel chip always rank first in the end. The math behind it very likely is a fragile house of cards that crashes whenever a new chip appears. Considering how quick CPUPro is to respond with their pure inane nonsense I can imagine data like this being leaked from within AMD just for all the entertainment it causes. It's supposedly only about the 7600X at a mere 4.9 GHz after all, lol.

 

[DrMrLordX]

@moinmoin

Yup! Exactly. They have to update their software to make it score artificially low on Zen4.

 

[biostud]

I don't know if it is possible but if you could somehow make a benchmark that strongly benefits from 1mb L2 cache, while tanking when only 512kb is present, then it could explain the results. (or maybe it is just a crappy benchmark)

 

[maddie]

Is userbenchmark a one person operation & why are they so angry and personal? Ex employee?

 

[igor_kavinski]

why are they so angry and personal?

Got burned by Bulldozer probably.

 

[CakeMonster]

Got burned by Bulldozer probably.

Literally? Did he take one to the knee or something? His bitterness sounds like some really personal business.

 

[inf64]

Maybe he lost a lot of money on AMD stock back in the days, who knows.

 

[DrMrLordX]

Is userbenchmark a one person operation & why are they so angry and personal? Ex employee?

Probably taking money under the table from Intel.

 

[Saylick]

Maybe he lost a lot of money on AMD stock back in the days, who knows.

Maybe he lost a lot of money on Intel stock these days, who knows.

 

[Mopetar]

Probably taking money under the table from Intel.

Intel has hired some shady "independent research" firms in the past, so it is a possibility. The prominent Google placement isn't cheap, so it almost seems like someone is bankrolling the site.

It would be far funnier if it were just some crazed fanboy with an axe to grind though. After the recent financials I can imagine the bean counters at Intel looking for costs to cut and someone informing an account that "Actually we don't pay that guy and no one here knows why he does it." but why turn down the free marketing when it's just another review sample to send out.

 

[igor_kavinski]

https://www.similarweb.com/website/maxon.net/#traffic

https://www.similarweb.com/website/cpuid.com/#traffic

https://www.similarweb.com/website/geekbench.com/#traffic

https://www.similarweb.com/website/3dmark.com/#traffic

https://www.similarweb.com/website/userbenchmark.com/#traffic

Weird. Guess those 9 million visitors don't know jack about benchmarks. This guy is just lucky and can afford to be a prick.

 

[darkswordsman17]

Is userbenchmark a one person operation & why are they so angry and personal? Ex employee?

There's like a dozen posters on here that I feel like it could be.

Intel has hired some shady "independent research" firms in the past, so it is a possibility. The prominent Google placement isn't cheap, so it almost seems like someone is bankrolling the site.

It would be far funnier if it were just some crazed fanboy with an axe to grind though. After the recent financials I can imagine the bean counters at Intel looking for costs to cut and someone informing an account that "Actually we don't pay that guy and no one here knows why he does it." but why turn down the free marketing when it's just another review sample to send out.

Intel did hire several tech journalists a few years back to be mouthpieces. I have a strong hunch this might be what they've gotten out of it, with the particular bitterness towards social media and the like being particularly revealing.

And all the new CEO has done is throw out megaplans that I'm not sure are actually coming to fruition and I'm not sure stockholders realize how expensive that stuff will be.

 

[DrMrLordX]

Intel also bankrolled that "security company" that published Ryzenfall.