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!

 

[eek2121]

The rumored pricing does not surprise me. Raptor Lake has not dropped yet, so AMD will be the undisputed performance king until it does. If Raptor Lake ends up being faster, I expect prices will change.

 

[exquisitechar]

If you think that 60% better perf/watt is not that large then you ll be spot on for your expectations...

You are saying that the 16 core Raphael at 230W will have 60% better perf/watt than the 13900K at 253W (since that's what I'm comparing)? Explain to me how that can be achieved. You do know that the 13900K was leaked to score 35000-36000 in Cinebench R23 at 253W...what exactly do you think the 7950X's score will be at 230W?

 

[Abwx]

You are saying that the 16 core Raphael at 230W will have 60% better perf/watt than the 13900K at 253W (since that's what I'm comparing)? Explain to me how that can be achieved. You do know that the 13900K was leaked to score 35000-36000 in Cinebench R23 at 253W...what exactly do you think the 7950X's score will be at 230W?

35000 pts@153W, anything higher for score and power will be out of specs and thus will require XFR/PBO being enabled.

Dunno how it will exactly do at 230W but for sure perf/Watt would be lower than the 5950X and perfs hardly 15% better than at said 153W.

 

[Exist50]

Seems to me that those who talk of 230W comsumption are just aknowledging sheer stupidity and unability to understand the most basic maths.

We literally got confirmation from AMD's own chief of marketing about that number. Denying it at this point is just silly.

I'm sure there are SKUs/TDPs at which the numbers AMD gave hold true. But not even the most optimistic reading of those numbers gives you 2x vs Raptor Lake. Nor do both numbers have to be simultaneously true for the same chip at the same TDP.

 

[exquisitechar]

35000 pts@153W, anything higher for score and power will be out of specs and thus will require XFR/PBO being enabled.

Robert Hallock disagrees.

Dunno how it will exactly do at 230W but for sure perf/Watt would be lower than the 5950X and perfs hardly 15% better than at said 153W.

Exactly. I wasn't sure why a lot of people were expecting stellar perf/watt from a stock 7950X (or whatever the top 16 core part will be called), but it's just that you were oblivious about the 230W PPT...

 

[Abwx]

We literally got confirmation from AMD's own chief of marketing about that number. Denying it at this point is just silly.

I'm sure there are SKUs/TDPs at which the numbers AMD gave hold true. But not even the most optimistic reading of those numbers gives you 2x vs Raptor Lake. Nor do both numbers have to be simultaneously true for the same chip at the same TDP.

AMD s representative in a forum is of no consequence if he make a blunder, but the pic below would have very costly legal consequence if the stated numbers are made up, i remind you that this was disclosed at their financial day.

Btw, how do you understand those numbers.?..
Are you only paying attention to their meaning.?.

But not even the most optimistic reading of those numbers gives you 2x vs Raptor Lake. Nor do both numbers have to be simultaneously true for the same chip at the same TDP.

5950X is within 142W in Cinebench and score 26196 pts,
12900K score 4% better but at 240W.

Hence 5950X has 1.625 x better perf/Watt than the 12900K and 7950X will have 2.03 x the perf/Watt of a 12900K, the 13900K wont be enough to just get back to the 5950X/12900K ratio.

 

[maddie]

CPUs on 15th but the mobos on 27th?
That and pricing basically being copy/pasted 5000 launch pricing makes this fail the common sense test IMO.

But then again, WCCFTech and failing common sense go together most of the time. This is why I didn't link their article earlier, they get more than enough traffic from the troll farms they sustain in their comment sections anyway.

This was in a post of theirs.

"Both the 15th September and the 27th September dates were reported by us first, so in the case AMD decides to proceed with sales on either of the two dates, it will be a confirmation of what we reported."

 

[Saylick]

We got a leak from WCCftech regarding pricing (bad news). According to the leak, AMD will be asking 450USD for 16T part (very bad decision on their part, if true).

Another leak regarding the launch of CPUs and motherboards: https://wccftech.com/amd-ryzen-7000...otherboards-27th-september-b650-10th-october/

Welp, was thinking about the possibility of going straight to AM5 but now I'm not sure it's worth it, especially since I'm targeting 4K anyways where CPU performance isn't as critical.

Perhaps AMD know Intel have to raise prices to keep revenues afloat so they are intentionally setting them high right off the bat. No one wins in a price war.

Or, this could be yet another jebait. WCCFTech could have been fed wrong info. Wouldn't be the first time, and they certainly don't vet their info to begin with.

 

[inf64]

Raichu posted this on twitter:

https://twitter.com/x/status/1564233283046875136

Supposedly 32T Zen 4 CPUz MT score (15645pts). For comparison, 13900K QS got around 16.5K.

 

[Exist50]

AMD s representative in a forum is of no consequence if he make a blunder, but the pic below would have very costly legal consequence if the stated numbers are made up, i remind you that this was disclosed at their financial day.

Dude, it's AMD's own chief of marketing, clarifying a previous mistake at an official presentation. While we can't completely dismiss the possibility that he's messed up twice, if you're going to outright ignore his statements, then the presentation is no more reliable.

5950X is within 142W in Cinebench and score 26196 pts,
12900K score 4% better but at 240W.

Hence 5950X has 1.625 x better perf/Watt than the 12900K and 7950X will have 2.03 x the perf/Watt of a 12900K, the 13900K wont be enough to just get back to the 5950X/12900K ratio.

I'll point out that even taking those numbers at face value, that's still not 2x what we can reasonably project from the 13900k.

 

[DrMrLordX]

Plenty of AMD loyalists won't mind paying that, to support AM5. Value conscious buyers have to look at Intel or AM4. AMD knows the initial production batch will sell out quickly. They will leverage that to get the maximum profits they can.

If I recall, the 5800X was the only Vermeer CPU that didn't sell out. For awhile you couldn't get anything but a 5800X. The 7700X may hold the same dubious distinction this time around.

 

[Joe NYC]

Welp, was thinking about the possibility of going straight to AM5 but now I'm not sure it's worth it, especially since I'm targeting 4K anyways where CPU performance isn't as critical.

Perhaps AMD know Intel have to raise prices to keep revenues afloat so they are intentionally setting them high right off the bat. No one wins in a price war.

Or, this could be yet another jebait. WCCFTech could have been fed wrong info. Wouldn't be the first time, and they certainly don't vet their info to begin with.

It's probably the best to wait a couple of month after the initial launch and see where things settle. 5800x3d dropped $30 below MSRP within 1st month and now it is $50 off. And this is the newest processor.

 

[Exist50]

The rumored pricing does not surprise me. Raptor Lake has not dropped yet, so AMD will be the undisputed performance king until it does. If Raptor Lake ends up being faster, I expect prices will change.

Looks like there's only going to be about a month between the two. While not unheard of, it would be surprising to see a price drop so soon. And realistically, we should be able to ballpark Raptor Lake pricing at around the same per-tier as Alder Lake, maybe plus a smidge. I think AMD's probably factored that into their numbers.

 

[Joe NYC]

If I recall, the 5800X was the only Vermeer CPU that didn't sell out. For awhile you couldn't get anything but a 5800X. The 7700X may hold the same dubious distinction this time around.

I recall someone did some back of the envelope math, and 5800x was the most profitable SKU for AMD.

 

[Joe NYC]

Looks like there's only going to be about a month between the two. While not unheard of, it would be surprising to see a price drop so soon. And realistically, we should be able to ballpark Raptor Lake pricing at around the same per-tier as Alder Lake, maybe plus a smidge. I think AMD's probably factored that into their numbers.

Alder Lake pricing offered some bargains against Zen 3.

With Intel signaling intention to increase pricing, the most logical course of action is for Intel to raise prices on those "bargains"

 

[Kocicak]

Tuesday 1AM in Europe is BAD NEWS !

They should leak the presentation... now?

 

[LightningZ71]

Is it possible that AMD is holding a speculated 7800X as a 10 core / 20 Thread release at a later date to better match up with the Intel 6+8/20 thread product in the market? While I don't think that N5 yields are so bad that there will be a lot of CCDs that have three under-performing from spec cores, it may be worth it from a marketing perspective. A 2 CCD, 5 cores each, product would have a lot of cache per core (a bit over 7MB if including L2) to keep performance rather high.

 

[jpiniero]

Is it possible that AMD is holding a speculated 7800X as a 10 core / 20 Thread release at a later date to better match up with the Intel 6+8/20 thread product in the market? While I don't think that N5 yields are so bad that there will be a lot of CCDs that have three under-performing from spec cores, it may be worth it from a marketing perspective. A 2 CCD, 5 cores each, product would have a lot of cache per core (a bit over 7MB if including L2) to keep performance rather high.

I think there will be a 7800X3D and they just decided that having a 7800X and 7800X3D was too confusing.

 

[gdansk]

I'm pretty confident Zen 4 will have great performance per watt in Genoa.
But in desktop I think it'll look inefficient in the top SKUs. But so does ADL/RPL and that's what they're dealing with.

Just my guess.

 

[Abwx]

Dude, it's AMD's own chief of marketing, clarifying a previous mistake at an official presentation. While we can't completely dismiss the possibility that he's messed up twice, if you're going to outright ignore his statements, then the presentation is no more reliable.

I'll point out that even taking those numbers at face value, that's still not 2x what we can reasonably project from the 13900k.

I simplified things by assuming that the 5950X use the full 142W PPT, this way it s the worst case figure for AMD, but if you want actual numbers let s get there.

According to Computerbase the 5950X doesnt exhaust the full PPT and is at 132W for 26196 pts, the 12900K scoring 27281@240W.

5950X/12900K real perf watt is hence 1.748 x.
AMD state >1.25 x for the 7950X, so let s take the minimal number above this one, that is 1.26 x to heir on the cautious side of things.

1.748 x 1.26 = 2.2 x is the 7950X/12900K perf/watt ratio.

13900K/12900K perf/watt ratio is expected at 1.25x, hence 7950X will be 1.76 x more efficient than a 13900K.
Not quite 2 x but still close, and as you can notice about the same as the 5950X/12900K.

That being said AMD start with a N5P process that has 2 x the perf/watt of Zen 3 s N7 and likely also of Intel 7, the fact that they have only 1.25x for Zen 4 is indicative that they traded most of the efficency for higher clocks.

 

[gdansk]

I simplified things by assuming that the 5950X use the full 142W PPT, this way it s the worst case figure for AMD, but if you want actual numbers let s get there.

According to Computerbase the 5950X doesnt exhaust the full PPT and is at 132W for 26196 pts, the 12900K scoring 27281@240W.

5950X/12900K real perf watt is hence 1.748 x.
AMD state >1.25 x for the 7950X, so let s take the minimal number above this one, that is 1.26 x to heir on the cautious side of things.

1.748 x 1.26 = 2.2 x is the 7950X/12900K perf/watt ratio.

13900K/12900K perf/watt ratio is expected at 1.25x, hence 7950X will be 1.76 x more efficient than a 13900K.
Not quite 2 x but still close, and as you can notice about the same as the 5950X/12900K.

That being said AMD start with a N5P process that has 2 x the perf/watt of Zen 3 s N7 and likely also of Intel 7, the fact that they have only 1.25x for Zen 4 is indicative that they traded most of the efficency for higher clocks.

See your napkin math would make some sense if AMD was running Zen 4 cores at 8-9W per core. I'm sure they would look very efficient. But it sounds like they'll be running them closer to 14W per core for some marginal increases to deal with RPL's extra cores.

 

[Exist50]

Alder Lake pricing offered some bargains against Zen 3.

With Intel signaling intention to increase pricing, the most logical course of action is for Intel to raise prices on those "bargains"

True, true. But unequal price increases could do some weird things to their lineup, so I'm curious what we'll end up with.

According to Computerbase the 5950X doesnt exhaust the full PPT and is at 132W for 26196 pts, the 12900K scoring 27281@240W.

I'll point out that if you equalize for the performance gap using cubic scaling for power, that cuts down on the 12900k's power by something like 10%.

13900K/12900K perf/watt ratio is expected at 1.25x, hence 7950X will be 1.76 x more efficient than a 13900K.

I think you're illustrating my point. With the best possible interpretation of the existing numbers in AMD's favor, the gap is still a third short of the claimed ratio.

 

[Abwx]

See your napkin math would make some sense if AMD was running Zen 4 cores at 8-9W per core. I'm sure they would look very efficient. But it sounds like they'll be running them closer to 14W per core for some marginal increases to deal with RPL's extra cores.

So you re stating that the 7950X will have way lower perf/watt than a 5950X, that s your "maths", or rather your sound...

For once and for all :

1.35 x the perf at 1.25 x the perf/watt mean 1.35x the perf at 1.35/25 = 1.08x the power.

Is that so difficult to understand..?.

 

[Kaluan]

This was in a post of theirs.

"Both the 15th September and the 27th September dates were reported by us first, so in the case AMD decides to proceed with sales on either of the two dates, it will be a confirmation of what we reported."

Double-dipping FTW? Christ, WCCFTech are so bad... 😂

How come there's no video premier announcement up on AMD YT channel yet? Sometimes they have the video announcement 2-3 days in advance but now it's less than 8 hours away and it's still not up. 🤔

 

[DrMrLordX]

I recall someone dis some back of the envelope math, and 5800x was the most profitable SKU for AMD.

Makes sense. They did (eventually) sell them, and they sold at a relatively high price for only one CCD.