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!

 

[Saylick]

I trust MLID, he has been spot on multiple times(except when He giving false info to discredit him)

Except when he is giving false info to discredit him? Like he intentionally puts out fake info to soil his own reputation?

He's also been so quick to pump out a video only a few hours after big leaks are posted to twitter by other leakers. Whether it's because he's blatantly ripping off any and every grain of information he can to profit or the inside leakers are leaking to multiple outlets is the question.

I feel like that's more true of RedGamingTech, who I'll admit has a nugget or two every so often, but the vast majority of his videos are regurgitations from Twitter.

Half of MILD's videos are Broken Silicon podcasts where he either has his brother or a guest on the channel. If they do discuss something that was mentioned on Twitter, it will be more of a "let's discuss the news" and less of a "leak".

 

[nicalandia]

Except when he is giving false info to discredit him? Like he intentionally puts out fake info to soil his own reputation?

Are you kidding me? I guess kids now a days prefer this type of "Tech Tips" right?

 

[Hail The Brain Slug]

Except when he is giving false info to discredit him? Like he intentionally puts out fake info to soil his own reputation?


I feel like that's more true of RedGamingTech, who I'll admit has a nugget or two every so often, but the vast majority of his videos are regurgitations from Twitter.

Half of MILD's videos are Broken Silicon podcasts where he either has his brother or a guest on the channel. If they do discuss something that was mentioned on Twitter, it will be more of a "let's discuss the news" and less of a "leak".

I have watched a few of these in realtime - twitter leaker posts some leak, and then 2-3 hours later MLID video comes out where he says exactly the same thing, except "my sources sent this to me a while ago and I was sitting on it, not sure if or when I should talk about it".

The reason why he does what he does is to make money. He will say whatever it takes to do that.

His videos can be entertaining to watch and frequently contain an amalgamation of leaks that I frequently have seen already from other sources. There's nothing wrong with that, he provides a younger audience attractive format to give big leak collections. What I find dubious are his claims about where he got the information, how long he's had it, and what he's willing to say to defend these questionable claims.

 

[igor_kavinski]

What I find dubious are his claims about where he got the information, how long he's had it, and what he's willing to say to defend these questionable claims.

Sounds like a complete douche.

 

[Saylick]

I have watched a few of these in realtime - twitter leaker posts some leak, and then 2-3 hours later MLID video comes out where he says exactly the same thing, except "my sources sent this to me a while ago and I was sitting on it, not sure if or when I should talk about it".

The reason why he does what he does is to make money. He will say whatever it takes to do that.

His videos can be entertaining to watch and frequently contain an amalgamation of leaks that I frequently have seen already from other sources. There's nothing wrong with that, he provides a younger audience attractive format to give big leak collections. What I find dubious are his claims about where he got the information, how long he's had it, and what he's willing to say to defend these questionable claims.

Fair enough. Not that I'm defending him, but how else should he go about it? Let's give him the benefit of the doubt and assume that his sources told him the same thing as the Twitter leaker, but at the time he was not 100% sure and wanted to sit on it for more corroboration. Should he just never mention that he had info when the Twitter leakers drop their Tweets? Or should he just lower his standards and post as soon as his sources give him the info to beat any potential Twitter leakers?

Of course, the simplest answer is usually the correct one, i.e. he doesn't have sources and is just cashing in on Twitter, but if he really did have them, wouldn't the above situation be a lose-lose?

 

[Timmah!]

Doesn't look right. According to TPU the 7700X is ~39% faster in CS:GO than the 5800X3D so a 7800X3D with only +23% would be way slower.

That doesn't make sense unless it's clocked at something like 4.x GHz. Which I doubt.

ofc it does not make sense, its almost 100 percent fake.

 

[Hail The Brain Slug]

Fair enough. Not that I'm defending him, but how else should he go about it? Let's give him the benefit of the doubt and assume that his sources told him the same thing as the Twitter leaker, but at the time he was not 100% sure and wanted to sit on it for more corroboration. Should he just never mention that he had info when the Twitter leakers drop their Tweets? Or should he just lower his standards and post as soon as his sources give him the info to beat any potential Twitter leakers?

Of course, the simplest answer is usually the correct one, i.e. he doesn't have sources and is just cashing in on Twitter, but if he really did have them, wouldn't the above situation be a lose-lose?

There was an incident regarding RDNA3 that someone here with a few others sent him completely falsified leak details and he made a video about it without much of a delay at all. The details were hard to believe, and in some cases totally made up. If he was really thoughtful with regards to holding onto leaks sent to him, he should have been able to easily see it was not reliable information and that he shouldn't run with it as if it were.

He frequently does things that sabotage his own credibility.

 

[Karnak]

Pretty nice clock speeds right there for the 7900X3D and 7950X3D.

 

[Hail The Brain Slug]

Pretty nice clock speeds right there for the 7900X3D and 7950X3D.

IF this leaked slide is legitimate, it demonstrates credibility to the rumor that they would only stack cache on the primary CCD of dual CCD SKU's.

However, those TDP's really leave me scratching my head. Like, really scratching hard.

 

[Abwx]

7950X@142W is at 95% of the stock performance in apps, so at 120W TDP = 162W PPT a 7950X3D should be at 97% of the former s apps stock perfs, in games the 7950X is well below 162W, so thoses 120W TDPs have nothing exceptional.

 

[RnR_au]

However, those TDP's really leave me scratching my head. Like, really scratching hard.

Less work for the io die?

 

[Thunder 57]

IF this leaked slide is legitimate, it demonstrates credibility to the rumor that they would only stack cache on the primary CCD of dual CCD SKU's.

However, those TDP's really leave me scratching my head. Like, really scratching hard.

Look at the 5800X3D. Maybe one CCD is limited to 5.0/5.1GHz because of the vcache?

 

[eek2121]

AMD may actually sell me another expensive chip! We will see.

 

[turtile]

IF this leaked slide is legitimate, it demonstrates credibility to the rumor that they would only stack cache on the primary CCD of dual CCD SKU's.

However, those TDP's really leave me scratching my head. Like, really scratching hard.

It will also be interesting to see how the cache works with 6 cores. Maybe it's possible for them to do 8+4?

 

[Saylick]

Look at the 5800X3D. Maybe one CCD is limited to 5.0/5.1GHz because of the vcache?

That would explain why the 7800X3D tops out at 5 GHz, while the other SKUs are listed as having higher single core turbos. The max clock for the dual CCDs would come from the non-V-cache CCD while the one that has the V-cache tops out at 5 GHz.

 

[lopri]

At same Speed

View attachment 73920

That is even more inaccurate. Zen 3 to Zen 4 less than 10%?

edit: same frequency, or same power?

 

[Saylick]

Twitter is buzzing with leaks:

https://twitter.com/x/status/1610796604541538304

 

[Exist50]

I trust MLID, he has been spot on multiple times(except when He giving false info to discredit him)

You're joking, right? He's been hilariously far off base, especially with AMD leaks. Remember his Zen4 IPC projections?

 

[lopri]

Oh wow, so that is how it’s going to work. Disable one CCD and give the whole 3D cache to the other CCD (for gaming, I suppose)

Brilliant for 2 x CCD chips, but what about single CCD chips?

 

[Exist50]

AMD is Sandbagging again, The 5800X3D Speed is 4.5 Ghz so they are clocking the 7800X3D at 4.5 Ghz. At 5.5 Ghz tha'ts an additional 22% speed boost

Why would they sandbag in their own announcement presentation? Especially given some of their recent marketing antics. They certainly don't say anything in the fine print.

If those numbers are correct, then some of the early predictions were (unsurprisingly) wildly over-optimistic.

 

[Saylick]

Oh wow, so that is how it’s going to work. Disable one CCD and give the whole 3D cache to the other CCD (for gaming, I suppose)

Brilliant for 2 x CCD chips, but what about single CCD chips?

I think for gaming it will just use the "gaming" CCD, i.e. the one with the V-cache, and background tasks will use the other CCD, which has no V-cache, sort of like pseudo P cores and E cores.

For the single CCD V-cache variant, the 7800X3D, it will behave like... well, a faster 5800X3D.

 

[Hail The Brain Slug]

It will also be interesting to see how the cache works with 6 cores. Maybe it's possible for them to do 8+4?

It's still a requirement that each CCD has a matching core count. Doesn't matter which cores, but it won't work with asymmetrical counts.

 

[lopri]

Do we know if this cache is 5nm or 6nm?

And all 3 of them at 120W, that cannot be right. With lower boost clock for 7800X3D..

 

[poke01]

However, those TDP's really leave me scratching my head. Like, really scratching hard.

Why so? No snark intended. Just want to know why.

 

[lopri]

One theory I can come up with is that they are coupling lesser CCD with V-cache and leaving higher-performing CCD on its own. Sounds conter-intuitive at first but I think that might make more sense since whichever CCD may not be able to clock high because of V-cache.