Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 72 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Joe NYC

Diamond Member
Jun 26, 2021
3,061
4,464
106
Why wouldn't anyone at AMD see an opportunity in that? Enthusiasts already pay quite a bit for the AM5 flagship CPU/mobo combo so they may not mind paying a bit more for a higher end chipset with quad channel RAM. AMD can keep 6,8 and 12 channel RAM for workstation and servers.
In their infinite wisdom, AMD marketing determined that "premium" mother board needs 28 USB ports rather than 4 memory channels.
 

Exist50

Platinum Member
Aug 18, 2016
2,452
3,105
136
It is not clear how it translates to desktop release schedule, but it seems that the dense CCD is not going to be lagging behind regular CCDs by very long.
Well if that is the cadence they hold to, it would not be surprising to see an 8+16 SKU as soon as that roadmap aligns. I think that's likely to become AMD's standard high end offering moving forward.
Volume could increase depending on how tangible the gains are.
It's sandwiched between the volumes of mainstream (2ch) client and lower end server (Sienna, 6ch). That's a very narrow product range.

Actually, if there was any desire to go in such a direction, they'd be better off using a physically compatible, but not electrically compatible version of the Sienna socket, just without all the pins used up. That might actually work.

But even if you solve the socket problem, they'd need a new die, which is a big deal. And if memory bandwidth is the only problem, they have other tools available to help, like V-cache and CXL memory expansion with bandwidth aggregation. Realistically, however, I think they'd have no problem doing 8+16 on a 2ch platform.
 
  • Like
Reactions: Tlh97 and Joe NYC

Joe NYC

Diamond Member
Jun 26, 2021
3,061
4,464
106
Well if that is the cadence they hold to, it would not be surprising to see an 8+16 SKU as soon as that roadmap aligns. I think that's likely to become AMD's standard high end offering moving forward.

8 core CCD with V-cache + 16 core CCD without could be a possible top end SKU.
 

Joe NYC

Diamond Member
Jun 26, 2021
3,061
4,464
106
Yeah, I think that's where their vision is heading. Just need to fix most of the clock speed regression with v-cache, and then there really wouldn't be any downside vs 8+8.
It seems AMD needs to overcome the challenges of dissipating the heat from the CPU better. In general CCDs and even more in CCDs with V-Cache.

Short of that, we might see flipping the chips, with big IO die / V-cache on the bottom and compute CCD on top.

RGT mentioned it in one of his videos. Not sure if that came from a real source with knowledge, or reading somebody's (maybe my :) ) posts on Twitter.

BTW, did anyone else noticed that random stuff that RGT guy reads on Twitter, he translates into "people telling him this".
 
  • Like
Reactions: Tlh97 and Exist50

Ajay

Lifer
Jan 8, 2001
16,094
8,111
136
8 core CCD with V-cache + 16 core CCD without could be a possible top end SKU.
Yeah, I think that's where their vision is heading. Just need to fix most of the clock speed regression with v-cache, and then there really wouldn't be any downside vs 8+8.

Not according to a recent TPU interview with David McAfee

How do you feel about hybrid CPU architectures? Does it make sense to bring the Zen 4c core to desktop?
I know that Mark Papermaster talked a lot of about different core types coming into our portfolio. I guess what I would say is that as we've looked at different core types there's probably two things that are overarching factors that we think about in terms of how they fit into the portfolio. One is the notion that P-Cores and E-Cores that the competition uses is not the approach that we plan on taking at all. Because I think the reality is that when you get to the point of having core types with different ISA capabilities or IPC or things like that, it makes it very complicated to ensure that the right workloads are scheduled on the right cores, consistently.

You know what I think about when we think about different core targeting is more the question of "what type of environment is this particular core going into?" Is it an environment that's power constrained like a notebook, or is it a power-unconstrained environment like a desktop. I think that those factors are going to drive first-and-foremost how we use different core types in our roadmap in the future and I think the benefits that you see of the cloud-optimized c-core that we've talked about is in the past is something that has a significant benefit in performance per watt that fits better in a power-constrained environment. Does that make its way into a desktop processor where you're power-unconstrained, I think that's a harder argument to make. We're constantly looking at different core types, how they might fit into our architectures in the future, but I think there's some more obvious places where different core types come in and bring an advantage much more quickly than in the desktop space.

 

Tuna-Fish

Golden Member
Mar 4, 2011
1,627
2,403
136
In their infinite wisdom, AMD marketing determined that "premium" mother board needs 28 USB ports rather than 4 memory channels.
To be clear, the cost to implement 28 USB ports is minimal compared to the cost of implementing 4 memory channels with DIMMs.

Wider memory interfaces are not common because they are expensive. Especially as the signal path requirements get more stringent with every DRAM generation, making the interface wider also gets proportionally more expensive each generation.

If you want more bandwidth, there is Threadripper. But until expandable/upgradeable main memory goes the way of the dodo, mass market consumer platforms will not get wider. I think Strix halo is AMD's trial balloon here: 256 bit LPDDR5X bus is reasonable in cost to implement, assuming that the DRAM chips are soldered on the same PCB as the CPU. If they are not, it gets expensive. Apple already started doing it, and now AMD is following suit.
 
Last edited:

A///

Diamond Member
Feb 24, 2017
4,351
3,160
136
Just FYI, MLID showed some road map where Turin Dense came ahead of regular Turin.

It is not clear how it translates to desktop release schedule, but it seems that the dense CCD is not going to be lagging behind regular CCDs by very long.
mlid needs to touch some grass.
 
  • Like
Reactions: Darkmont

Exist50

Platinum Member
Aug 18, 2016
2,452
3,105
136
It seems AMD needs to overcome the challenges of dissipating the heat from the CPU better. In general CCDs and even more in CCDs with V-Cache.
They'll figure it out eventually. Thermal density is definitely a big problem, especially on newer nodes, but that's the kind of raw engineering problem that engineers love. Maybe they'll flip the stacking or something else, but I doubt V-cache will always come with such a penalty.
Not according to a recent TPU interview with David McAfee
Eh, he kinda sidesteps it a little. And the raw technical merit of such a product is very clear. It's not going to happen for the Zen 4 gen, but Zen 5, Zen 6? I think that interview will not age particularly well, and is more targeted in the "selling the products we have today" sense.
 

A///

Diamond Member
Feb 24, 2017
4,351
3,160
136
They'll figure it out eventually. Thermal density is definitely a big problem, especially on newer nodes, but that's the kind of raw engineering problem that engineers love. Maybe they'll flip the stacking or something else, but I doubt V-cache will always come with such a penalty.

Eh, he kinda sidesteps it a little. And the raw technical merit of such a product is very clear. It's not going to happen for the Zen 4 gen, but Zen 5, Zen 6? I think that interview will not age particularly well, and is more targeted in the "selling the products we have today" sense.

May be something on am6 with zen 6.
???????????????????????????????????????????????????????????????????????????????????????
 

Exist50

Platinum Member
Aug 18, 2016
2,452
3,105
136
???????????????????????????????????????????????????????????????????????????????????????
I assume Zen 6 will reuse the AM5 socket. But even if they don't, they're not going to move the mainstream off 2ch for cost reasons, nor provision that many extra pins in the socket to support both. Just to steal a pinout from Igor's ARL leak, look how many pins even 2ch takes.

1689803427048.png
 

A///

Diamond Member
Feb 24, 2017
4,351
3,160
136
I assume Zen 6 will reuse the AM5 socket. But even if they don't, they're not going to move the mainstream off 2ch for cost reasons, nor provision that many extra pins in the socket to support both. Just to steal a pinout from Igor's ARL leak, look how many pins even 2ch takes.

View attachment 83277
this is all true and I agree with you with the grand exception of zen 6 being on am5. why i say this? amd is bound to take one big step and mess it all up. by then intel will e on their 2nd new ceo after pat, this ceo being a tuna fish or octopus,
 

Ajay

Lifer
Jan 8, 2001
16,094
8,111
136
Eh, he kinda sidesteps it a little. And the raw technical merit of such a product is very clear. It's not going to happen for the Zen 4 gen, but Zen 5, Zen 6? I think that interview will not age particularly well, and is more targeted in the "selling the products we have today" sense.
I think, by what he says, it *won't* be Zen5. Obviously, AMD didn't do it with Zen4 - so he's not talking about that. The point of 'E' cores is Area optimized cores vs performance optimized cores. Personally, I don't see the need for this on the desktop. The current CCD-IOD system offers the flexibility AMD needs for Desktop parts. # of active cores and frequency/voltage/power limits can be set by OEM in the BIOS to get the product that they need for desktop. Mobile is a different story. Servers may be a different story - depending on what customers want (AMD could very easily provide some selective CPUs with a mix of CCDs with performance and high density cores on the same package). I don't see the benefit of having one HP CCD and one HD CCD on the desktop. IMHO.
 
Last edited:

Joe NYC

Diamond Member
Jun 26, 2021
3,061
4,464
106
Not according to a recent TPU interview with David McAfee



There are 2 ways to interpret it - that hybrid means different instructions sets or different types of cores.

AMD is not going to pair CPUs that differ in the instructions they support. We will see about the other part / other interpretation.
 

Ajay

Lifer
Jan 8, 2001
16,094
8,111
136
There are 2 ways to interpret it - that hybrid means different instructions sets or different types of cores.

AMD is not going to pair CPUs that differ in the instructions they support. We will see about the other part / other interpretation.
He does say no to different ISAs or different IPCs. So, again, I’m confident that it’s a no for Zen5 and likely no for Zen6. But, as always, we shall see. All bets are off for AM6.
 

eek2121

Diamond Member
Aug 2, 2005
3,341
4,927
136
Considering where things stand:
- AM5 is dual channel
- Sienna platform is (likely) 6 memory channels
- Strix Halo is 4 internal memory channels

The chances of AMD releasing 4 memory channel socket are near zero.

What could be interesting, if Strix Halo can fit into AM5 socket, would be to have 4 internal memory channels and 2 external.
Disclosure: I have you on blocked (so your posts aren't shown unless I am suckered into clicking the show hidden button) because you at one point posted a ton of nonsense, but occasionally something somewhat intelligent wanders from your keyboard to this forum, and I usually find out by others NOT ignoring you..so here we go:

While I understand the argument, I'm gonna have to disagree here.

You know what happened when you plugged 2 memory sticks into a 1950X? You had dual channel memory. It wasn't like you gimped your machine by NOT having 4 sticks, 4 sticks just performed faster. Just like 2 sticks/dual channel on a 7950x will perform better than 1 stick/one channel. Your system won't refuse to boot, or even operate poorly, if you only have 1 stick of RAM. It will just operate more slowly. No. The REAL reason is that AMD doesn't want to add all those extra traces/wiring (which I suspect they can do without a socket change, but I could be mistaken) and the extra bandwidth, however, ignoring that....DDR5 IS ENOUGH for 24-32 cores. AMD would sooner sell me a bridge than convince me otherwise.

...even if bandwidth were an issue, AMD could save money and up core counts by releasing a 4hi+8lo chiplet and using that in place of the single chiplet offerings.

Atom in raptor lake is (mostly) just a 4 core cluster sharing a single 'big' core's resources. Food for thought.

...oh and in case some new person to this forum thinks I'm an Intel fan? Look at my post history. Also please note that my household consists of 7 AMD machines and zero Intel machines (fun fact: I also have 3 RISC-V machines and 7 non-mobile ARM machines...and a bunch of Apple products that aren't listed here)
 

A///

Diamond Member
Feb 24, 2017
4,351
3,160
136
...oh and in case some new person to this forum thinks I'm an Intel fan? Look at my post history. Also please note that my household consists of 7 AMD machines and zero Intel machines (fun fact: I also have 3 RISC-V machines and 7 non-mobile ARM machines...and a bunch of Apple products that aren't listed here)
I would like to think most of us are mature enough not to accuse anyone else being a fanboy. I don't think I have ever seen anyone accuse another person on here of being a fanboy or shill unless it was clear they were were here just to cause trouble as a new account.

I don't recall Joe ever spouting off nonsense but I also skim over most of his posts. This isn't for me to discuss but giving my bare minimum 2c here. I'd say more but I'm not feeling too well at the moment with the left side of my chest feeling peculiar.
 

moinmoin

Diamond Member
Jun 1, 2017
5,212
8,378
136
Considering where things stand:
- AM5 is dual channel
- Sienna platform is (likely) 6 memory channels
- Strix Halo is 4 internal memory channels

The chances of AMD releasing 4 memory channel socket are near zero.
I'd don't know if we ever got updates or rebukes on these rumours since but rumours were that the new Threadripper platforms would be 8 ch (Pro, then likely based on Genoa) and 4 ch (non-Pro, then likely based on Sienna).
 

eek2121

Diamond Member
Aug 2, 2005
3,341
4,927
136

I wonder if it occurred to the VideoCardz guy that the chip might be 12 “Zen 5” cores.

A bit of a ramble:

A lot of people are speculating that AMD will go hybrid on Zen 5, but IIRC Zen 5 work began prior to Intel publicly announcing their efforts. I am not saying this is the case, but I feel like AMD would “play it safe” by introducing a hybrid design with an existing architecture first. I will admit, however that there are some indications that AMD is pursuing a hybrid design for at least some chips.

I guess we will find out in January.

If they go hybrid for desktop, it will be interesting to see if they stick with chiplets since having multiple core designs/configs complicates the design. They would either have to move to 2 CCDs per chiplet or come up with another approach unless they are segregating server from the rest. Any thoughts on this?
 

Timorous

Golden Member
Oct 27, 2008
1,943
3,782
136

I wonder if it occurred to the VideoCardz guy that the chip might be 12 “Zen 5” cores.

A bit of a ramble:

A lot of people are speculating that AMD will go hybrid on Zen 5, but IIRC Zen 5 work began prior to Intel publicly announcing their efforts. I am not saying this is the case, but I feel like AMD would “play it safe” by introducing a hybrid design with an existing architecture first. I will admit, however that there are some indications that AMD is pursuing a hybrid design for at least some chips.

I guess we will find out in January.

If they go hybrid for desktop, it will be interesting to see if they stick with chiplets since having multiple core designs/configs complicates the design. They would either have to move to 2 CCDs per chiplet or come up with another approach unless they are segregating server from the rest. Any thoughts on this?

For desktop use dense cores and normal cores.

X950 could be 8 normal and 16 dense.
X900 could be 8 normal and 12 dense.
X800 could be 8 normal and 8 normal.
X600 could be 6 normal and 6 normal.

Obviously there can also be 3d variants as well.

This would be a large core count uplift but with Intel going 6p 8e on RPL-R and going with even more e-cores in future products I think AMD may have no real choice but to increase core counts something like the above to stay competitive in Mt and productivity workloads at each tier.
 

uzzi38

Platinum Member
Oct 16, 2019
2,746
6,637
146
@eek2121 Just FYI, but the 7540U ships in a configuration that screams the rumoured PHX2 specs, but currently ships with PHX1.

2 Zen 4 cores that can boost to advertised boost frequency, 4 Zen 4 cores that are capped to 3.8GHz for single thread loads, and when the entire chip is stressed they drop to 2.9GHz.

Seems fairly clear that the rumoured 2+4 design is almost certainly on point. So effectively the low end will migrate from 4 cores to a 2+4 design. Doesn't seem like a stretch to believe that 8 core designs will migrate to 4+8 really.

Only remaining question is: but what about gaming laptops. Will the 4+8 design be optimal, or will there need to be another pproach instead?
 

soresu

Diamond Member
Dec 19, 2014
3,808
3,129
136

I wonder if it occurred to the VideoCardz guy that the chip might be 12 “Zen 5” cores.

A bit of a ramble:

A lot of people are speculating that AMD will go hybrid on Zen 5, but IIRC Zen 5 work began prior to Intel publicly announcing their efforts. I am not saying this is the case, but I feel like AMD would “play it safe” by introducing a hybrid design with an existing architecture first. I will admit, however that there are some indications that AMD is pursuing a hybrid design for at least some chips.

I guess we will find out in January.

If they go hybrid for desktop, it will be interesting to see if they stick with chiplets since having multiple core designs/configs complicates the design. They would either have to move to 2 CCDs per chiplet or come up with another approach unless they are segregating server from the rest. Any thoughts on this?
I'm more interested in the rumored 'Sarlak' variant with up to 40 RDNA3 CUs.

With something that meaty I could probably just ditch my desktop for all but work related use cases.
 
  • Like
Reactions: Tlh97 and Joe NYC