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!

 

[A///]

Someone please correct me, but Zen3 has quite high voltage in single core boost. Could be just reading single core boost voltage?

Although, your speculation has some legs.

That's PBO, I think. Normal boosting is within spec.

 

[DisEnchantment]

Someone please correct me, but Zen3 has quite high voltage in single core boost. Could be just reading single core boost voltage?

That is transient VID, what comes out of the LDO is the actual voltage fed to the logic gates via the PDN.
But indeed you are right not sure what is being measured here.

But you can already see a weakness here.
By the time the VID request is handled by the VRM circuitry, lots of ms would have elapsed.
Having an IVR is a necessity for the super fast boost behavior (which for a 5GHz CPU has to be really really fast)

 

[randomhero]

That is transient VID, what comes out of the LDO is the actual voltage fed to the logic gates via the PDN.
But indeed you are right not sure what is being measured here.

But you can already see a weakness here.
By the time the VID request is handled by the VRM circuitry, lots of ms would have elapsed.
Having an IVR is a necessity for the super fast boost behavior (which for a 5GHz CPU has to be really really fast)

Zen3+ already went "mental" with control circuitry for clock gating, transient voltages, etc.
Zen4 probably(definitely) got further.IVR are probably necessity on smaller nodes.

 

[lightmanek]

Can't it be 6Nm Rembrandt core for AM5?

 

[gdansk]

Can't it be 6Nm Rembrandt core for AM5?

Good point. That could be. Rembrandt seems to report around that voltage.

 

[A///]

32 of those small cores is about equal to 8 big cores, so thats the equivalent of 16 strong cores. AMD has 16 pretty strong cores, and I know This redwood/crestmont will be stronger than alder lake, but against Zen 4 ? Or Zen 5 ? I don't know that Intels troubles are over or dominance will be theirs any time soon.

Edit: This IS the Zen 3 5800X3D thread..... back on topic ?

Bringing it here if it's ok. Yes, every 4 cores is equal to 1 big core, but I'm really curious how Intel would have performed if it managed to put 16 big cores instead of a 50/50 split for Alderlake. The Gracemont cores are no slouch. I got to play with a 12700 build I made for someone a month ago and was impressed with the gaming performance. Productivity is no Zen 2 or 3, but it's still good. Easier to cool and manage vs the 12900K.

Your point about RL/Meteor going up against Zen 4 and being uncertain is spot on. I actually hadn't considered the performance we'd see with Zen 4. I've heard some.... outlandish figures but I went with a conservative 10% generational overall uplift which in hindsight is a poor choice given AMD's past performance uplifts the past few years.


I know you and I have spoken to some length on processors off the boards but it does make me a wee bit nervous when it comes to the yin and yang of these two upcoming processor groups. If AMD delivers a whalloping on Intel I'll also agree with you and say Intel may not be out of the woods yet. It really hinges on Meteor Lake's performance unless Gelsinger was blowing hot smoke up everyone's tooter.

 

[Markfw]

For a minute I was confused, yes, this is where it belongs ! And based on my experience with ADL, I will not venture any hard guesses until things come out.

But I will say that Intel's troubles may continue for quite some time, if not indefinitely. For Gelsinger to say AMD is in the rear view mirror never to get in the headlights ever again was really stupid. (It was him that said that, right ??) Especially how far behind they are on the server front, where the big boys play, and the real money is !

 

[A///]

For a minute I was confused, yes, this is where it belongs ! And based on my experience with ADL, I will not venture any hard guesses until things come out.

But I will say that Intel's troubles may continue for quite some time, if not indefinitely. For Gelsinger to say AMD is in the rear view mirror never to get in the headlights ever again was really stupid. (It was him that said that, right ??) Especially how far behind they are on the server front, where the big boys play, and the real money is !

Yes,, that was him. I was at work when I saw that come up in my newsfeed. I thought it was a misquote but he really said that.

Borrowing from sister site Tom's. It's, as you say, a pretty ballsy statement to make given the extenuating circumstances of Intel's current position and possible fallback they may have to rely on. The only time I remember either company getting a full lead on was maybe AMD when they hit the 1 Ghz barrier before Intel, their x2, but I think what really stands out is C2D dropping. I'm sure Meteor will be better than Zen 2 and Zen 3, but will it be better than Zen 4? That's the big question. Zen 4 will likely beat Raptor Lake, but if it manages to beat the next product coming a few months later or heckin heck, the next product after that while being out for over a year then.... Gelsinger is gonna go the way of Tiny Tim?

Intel CEO Gelsinger Says AMD Is "In the Rear-View Mirror" After Alder Lake

The war of words escalates.

www.tomshardware.com

However, he also took the opportunity to say that AMD is now firmly behind Intel in the consumer market and "in the rearview mirror." As you can see in the video below, that's also how Gelsinger expects the technology race to continue to play out under his leadership.

But yes, DC ASP counts more than low end or mid range processors for the most part. DC and mobile really. Though I think Intel fumbling on their Altera purchase may provide another way of AMD screwing down the clamps.

 

[Markfw]

Yes,, that was him. I was at work when I saw that come up in my newsfeed. I thought it was a misquote but he really said that.

Borrowing from sister site Tom's. It's, as you say, a pretty ballsy statement to make given the extenuating circumstances of Intel's current position and possible fallback they may have to rely on. The only time I remember either company getting a full lead on was maybe AMD when they hit the 1 Ghz barrier before Intel, their x2, but I think what really stands out is C2D dropping. I'm sure Meteor will be better than Zen 2 and Zen 3, but will it be better than Zen 4? That's the big question. Zen 4 will likely beat Raptor Lake, but if it manages to beat the next product coming a few months later or heckin heck, the next product after that while being out for over a year then.... Gelsinger is gonna go the way of Tiny Tim?

Intel CEO Gelsinger Says AMD Is "In the Rear-View Mirror" After Alder Lake

The war of words escalates.

www.tomshardware.com

But yes, DC ASP counts more than low end or mid range processors for the most part. DC and mobile really. Though I think Intel fumbling on their Altera purchase may provide another way of AMD screwing down the clamps.

And I think it was in the Zen3D thread that I summed up where I thought Intel stood, and was basically agreed with (so far)

Desktop - I thought overall AMD slightly ahead someone said behind, and thats really questionable either way.
Gaming ? a tie, but AMD takes less power and costs less and runs cooler. So it really wins IMO
Desptop productivity ? AMD

Server and HEDT - unquestionably AMD

Laptops, again split decision.

So overall they are not ahead, but behind. Again, stupid Gelsinger.

 

[A///]

And I think it was in the Zen3D thread that I summed up where I thought Intel stood, and was basically agreed with (so far)

Desktop - I thought overall AMD slightly ahead someone said behind, and thats really questionable either way.
Gaming ? a tie, but AMD takes less power and costs less and runs cooler. So it really wins IMO
Desptop productivity ? AMD

Server and HEDT - unquestionably AMD

Laptops, again split decision.

So overall they are not ahead, but behind. Again, stupid Gelsinger.

Laptops are going to be fierce when Zen 5 lands if anything spoken about Zen 5 is true. Everything else I agree, that may have been me who agreed with you point blank actually. Gaming is a tie, but in some cases the 12900K seems to have better lows, but against a 5900X. There's been some neat chit chat on Zen 4 memory speed and latency which should further push it up the gaming pole, but everything else should go up. It's really a matter of price. Zen 4 won't have the luxury of offering DDR4 as a RAM choice. This may be cost prohibitive for a lot of people, especially as Zen 3 was generally more expensive than prior generations..

Given the pandemic, possible artificial shortages by companies looking to make a quick buck and the general unknown about the next 5 months when it comes to disease, supply routes and more product scalping, it's a major burden on the DIY builders

 

[jpiniero]

I'm sure Meteor will be better than Zen 2 and Zen 3, but will it be better than Zen 4?

Meteor is mobile only, and even there a sideshow at best. Arrow Lake could be a problem for AMD but probably only if it does end up being at TSMC.

 

[A///]

Meteor is mobile only, and even there a sideshow at best. Arrow Lake could be a problem but for AMD probably only if it does end up being at TSMC.

I'm not placing faith in anything Graymon says ever again.

 

[gdansk]

I'm not placing faith in anything Graymon says ever again.

What was the particular trigger? Probably a good idea anyway

 

[A///]

What was the particular trigger? Probably a good idea anyway

There's been a few times where I knew better than follow him or Komachi. My take is that if Intel can't deliver Meteorlake outside a mobile product, while mobile is important, than AMD have free reign over Intel for over a year simply due to how Intel would operate in that moment and that only Arrowlake-s in Late 2024 would come out. This makes no sense when you compare it to what Gelsinger said about AMD being in the rear view mirror.

if so, how much mobile is he expecting to sell that year?

 

[jpiniero]

What was the particular trigger? Probably a good idea anyway

The TSMC talk, I'm guessing.

if so, how much mobile is he expecting to sell that year?

Tons. Corporate types just can't get enough of Intel. Intel gained desktop and laptop market share last year even.

 

[A///]

The TSMC talk, I'm guessing.



Tons. Corporate types just can't get enough of Intel. Intel gained desktop and laptop market share last year even.

Only because they could produce more, and IIRC they were stockpiling at one point. I don't work in corporate/enterprise IT so I dunno the turnover rate of old laptops to new laptops, but I still think Intel is shooting themselves in the foot if they want to come back with a mild gamer win and then let it drop flatter than Will Smith's current career trajectory for nearly 2 years.

 

[jpiniero]

Only because they could produce more

That's kind of important if you think about it.

 

[A///]

That's kind of important if you think about it.

Yeah, but leaving your desktop open for 2 years doesn't exactly say hey AMD is behind us now yay, especially when they haven't got their DC or HEDT in line at all. Also, this is assuming the mobile processor is any good compared to what AMD will have.

 

[deasd]

Raphael ES running Bonic but I don't understand the result

https://twitter.com/x/status/1515785903297417218

Computer 925206

milkyway.cs.rpi.edu

Application details for host 925206

milkyway.cs.rpi.edu

All tasks for computer 925206

milkyway.cs.rpi.edu

 

[igor_kavinski]

Computer 925206

milkyway.cs.rpi.edu

Application details for host 925206

milkyway.cs.rpi.edu

All tasks for computer 925206

milkyway.cs.rpi.edu

i7-7700K: Computer 925218 (rpi.edu)
Application details for host 925218 (rpi.edu)

That's possibly a very lowclocked sample. A lowly 7th gen Core i7 is managing 10+ GFlops.

Never mind that.

3900X: Computer 925213 (rpi.edu)
Application details for host 925213 (rpi.edu)

11.46 GFlops for 3900X. Doesn't seem like a good benchmark.

 

[biostud]

And I think it was in the Zen3D thread that I summed up where I thought Intel stood, and was basically agreed with (so far)

Desktop - I thought overall AMD slightly ahead someone said behind, and thats really questionable either way.
Gaming ? a tie, but AMD takes less power and costs less and runs cooler. So it really wins IMO
Desptop productivity ? AMD

Server and HEDT - unquestionably AMD

Laptops, again split decision.

So overall they are not ahead, but behind. Again, stupid Gelsinger.

AMD just went around the globe and are now appearing in the rear view mirror once again

 

[igor_kavinski]

It will hurt Intel more if AMD attacks from the rear.

 

[eek2121]

Allow me to place my tinfoil hat on and profess my fortune telling abilities, my fine gentleman. I suspect you may be right but it may also be a July launch. I only say that while it may not fit the perceived timelines leaked a half year ago because of how AMD loves to play with their product name numbers and dates.

Though at this moment I can't remember why Zen 2 came out in July 2019.

7nm! 7/7! 😂

This year it’s Ryzen 7000! Available on 7/7! 🙃

Zen 3 Vcore is much lesser than that. Very strange to have such high Vcore considering part of N5 efficiency is improving operation at lower voltages.
According to TSMC, N4X is officially their only leading edge node that is designed to operate beyond 1.2V.

It is possible they replaced the LDO with an IVR which could explain such high Vcores.
They would need to do that at some point, because lowering voltage means they need to drive up current to maintain switching performance.
This has the effect of reducing VDD pads because higher voltage is delivered in fewer pads and supply is kept very close to the logic resulting in lower I2R losses.

Zen 2 and above have 4 LDOs on both sides of the L3 which take around 3-4 mm2 of area.
But using N5 real estate to make an IVR is too wasteful ... unless there is really a fanout layer underneath the die where they can use RDLs to weave the needed inductors.
Nevertheless very interesting if true.

AMD is using a custom version of N5(p?) FYI.

 

[DrMrLordX]

This year it’s Ryzen 7000! Available on 7/7! 🙃

We should be so lucky. Sadly, I think they are not going to be able to launch before August, no matter how much they want to repeat the 7/7 launch.

 

[igor_kavinski]

We should be so lucky. Sadly, I think they are not going to be able to launch before August, no matter how much they want to repeat the 7/7 launch.

Paper launch!