Speculation: Ryzen 4000 series/Zen 3

Page 180 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Thunder 57

Platinum Member
Aug 19, 2007
2,647
3,706
136
Sorry for joke post, but I need to lower tensions here :D

You are all missing the glaring hint AMD is sending us, Zen 3 unveil is on 8th of October, clearly it will be manufactured on Samsung 8nm process same as RTX 30xx series.
Radeon RX6000 unveil is on 28th ... yes, you guessed it right! GloFo 28nm chip is coming!

:Hides behind sofa:

Gotta do something with GloFo to satisfy the WSA :D
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
As of October 8th Matisse will be 15 months old.
Actually, it is 25 months old from the RTG sampling date.
Spec-wise it is identical to the 7/7/2019-AMD Ryzen™ 7 3800X.

Ax Stepping => Late 2018
Bx Stepping => Mid 2019

Summit Ridge for example was a mid-2016 sampling processor for Q4 HVM. In which it did Doom at E3 2016.

Zen ->
Leapfrog design teams -> Zen2(one team) + Zen3(another team)
Based on some factors the Zen3 team was pushed to Zen5. Which lead to another team to do Zen3, which also happened to be the Zen4 team. Through those factors Zen4's node is also Zen3's node.

If one would look up the people doing Zen3/Zen4. They would find out they have only done single node architectures in the format of next-gen microarchitecture(AVX) -> refined microarchitecture(AVX2).

14nm -> 7nm -> 5nm(To TSMC's Bugatti race track) is equivalent to 45nm -> 32nm -> 28nm(GloFo's bumpy road), keeping cadence relative to the foundry is very important.
32nm=130CPP -> 28nm=114CP
vs
7nm=57CPP -> 5nm=57CPP&51CPP.

5FFstdcell.png
7nm+ and 5nm seem to have same design flows and design rules. Most cases if 57nm is used, it is potentially 1:1 with 7nm+.

Setting up for some weird play from TSMC:
7nm family equals: N7, N6(w/ N7 RTO), N7+, N5(w/ N7+ RTO)
5nm family equals: N5(w/ 51CPP), N4(NTO & RTO)
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Okay, I'm gonna need you to back all the way off with the Zenception.

It's making my head hurt.
Initially:
Sunnyvale Cores: Bulldozer -> Zen -> Zen3
Austin Cores: Zen2 -> Zen5
Boston Cores: Zen4

Sometime ago in 2018:
Sunnyvale Cores: Zen -> Zen5
Austin Cores: Zen2 -> Zen5
Boston Cores: Zen3 -> Zen4 <== Both cores are N5 at Boston.

There is also Israel Cores, India Cores, and Colorado Cores coming back.

Zen1/Zen2 => Cyclone Era (but with 4-wide FPU & SMT2)
Zen3/Zen4 => Vulcan/Triton Era (but with Centaur CNS-style ports & no SMT4 & still 4-wide FPU)

Zen/Zen2 = ~20 macros
Zen3/Zen4 = <15 macros, ~14 tiles/blocks

HMC Fin => 18%+ current drive at a given leakage over non-HMC Fin and improved aging effects should fit in N5's 57CPP lib and be RTO'able of N7+.


Q3 2018 risk production -> Q3 2019 Zen3 complete // Four quarter (Q4/Q1/Q2/Q3) complete <== 7nm+ // Q3 2014 14LPP risk => Zen design complete in Q3 2015 || Inline with Zen
or
Q1 2019 risk production -> Q3 2019 Zen3 complete // Two quarter (Q2/Q3) complete <== 5nm // Q2 2017 N7 risk => Zen2 design complete in Q4 2017 || Inline with Zen2

2015 => 14nm
2017 => 7nm
2019 => 5nm

Which might explain 2019 [5 nm] on this page: https://www.amd.com/en/technologies/introduction-to-semiconductors

With 2017 bringing cost per yielded mm2 for the 7nm data point and 2019 for the 5nm data point.
Also with anandtech link: https://forums.anandtech.com/threads/speculation-ryzen-3000-series.2558009/page-48#post-39706900

Earliest I can find for 5nm: https://www.dvhardware.net/article71210.html => hires https://www.planet3dnow.de/cms/48524-amd-semicon-west-praesentation/nggallery/page/1
https://forums.anandtech.com/thread...ng-with-chiplets-warning-many-images.2577325/ <== exploiting 7nm technology slide
Then there is this one:
https://www.techdesignforums.com/blog/2020/07/08/vlsi-amd-chiplet-processors/

Oof:
Neoverse N2 = 2x128b SVE => 128-cores in 2021
Neoverse V1 = 2x256b SVE => 96-cores in 2020
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
As of October 8th Matisse will be 15 months old.

I know. It's a bit depressing actually.

You're confusing OEMs such as Dell and HP with AIBs. I don't see the point in dropping $500-800 every 12 months for a 10% increase in performance, when I can wait 14-17 months and get maybe 17-20% increase in overall performance, or even more if major ground was broken.

I don't recall mentioning OEMs or AIBs. And there's nothing forcing anyone to buy once every generation of product, either. Nor is there any indication that waiting a few more months gets you any more performance.

Is Zen3 going to be any better released in October of this year instead of July? I'm guessing "no". AMD is probably holding it off because:

Motherboards weren't ready, especially since "the community" cried foul when 4xx-series chipsets weren't going to support it
Matisse, despite being a bit old now, is still selling quite well
Related to the above, Intel hasn't released anything new that would require them to sell something better than Matisse on the desktop, and may not until next year (barring TigerLake-H becoming a desktop part)

You could hardly call Zen+ an amazing feat by any stretch of the imagination.

It was actually pretty good - much better than the XT launch. The top-end Pinnacle Ridge picked up maybe 300 MHz on the 1800x, and it had a small IPC gain to boot. That didn't translate into major victories for the 2700x - according to AT Bench, in CBR15 MT, the 2700x was only 8% faster, while it was ~11% faster in Blender. Not bad for a chip that was cheaper at launch than the 1800x. In contrast, the 3900XT cost more than the street price for the 3900X, and had about 0% improvement in MT workloads like CBR20 vs the 3900X.

I've said it before, and I'll say it again - if AMD can pull off another Zen+ with Warhol then it looks like it might be a solid product. Assuming it even exists, which is not guaranteed.

Would you prefer them to be on a strict 12 month cadence with much less YoY growth and charging the same price as they do now

Why not? Not only would it be great for consumers (shocking, I know), but it would also force everyone to take them seriously, especially if Intel found themselves facing Zen3 right now and Zen4 in May-July 2021. Zen3 in July 2020 would have been a stunning blow to Intel. They have no desktop product to launch until maybe January at the earliest, fewer people would be looking at TigerLake-H (or really anything else), and on top of all that, there'd be the spectre of something newer and better hitting in July of next year.

Or would you prefer them to take their time and launch a great product every 14-18 months?

I'm reasonably certain they can launch a great product more frequently than that. Hell 14 months wouldn't be so bad.

Right in line with their cadence.

All I can say at this point is:


The XT chips didn't garner favorability with people.

But they still did it. And it probably made them some extra cash by pushing prices back up to last year's level on what was essentially the same product.
 

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
I don't recall mentioning OEMs or AIBs. And there's nothing forcing anyone to buy once every generation of product, either. Nor is there any indication that waiting a few more months gets you any more performance.

Is Zen3 going to be any better released in October of this year instead of July? I'm guessing "no". AMD is probably holding it off because:

Motherboards weren't ready, especially since "the community" cried foul when 4xx-series chipsets weren't going to support it
Matisse, despite being a bit old now, is still selling quite well
Related to the above, Intel hasn't released anything new that would require them to sell something better than Matisse on the desktop, and may not until next year (barring TigerLake-H becoming a desktop part)
Matisse is going to sell regardless of whether it's release today, a month ago, or four months from now. Matisse won't get an immediate price drop. There are people on 300 series boards who can upgrade to Matisse who are locked out of Zen3. There are people with 2700Xs whose boards can support Ryzen 9, so they'll buy that. There are people with 3600s who don't want to go to Zen 3 but will pickup a cheaper Ryzen 9. The whole motherboards weren't ready is a bit of BS. It doesn't take months to revamp a BIOS. This is board partners preferring to sell new hardware, and not necessary a new useless chipset. And also motherboard partners not having enough storage on the mobos to support multiple generations, as well as electrical components possibly not being up to spec. Get those people on 300 and 400 series boards to purchase a X570 or B550 board. It's still money for them, and their expenditure will be minimal. You want AMD to release every 12 months and then deal with software issues for 60-120 days as a result like every release they've rushed in the past, right?
 

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
It was actually pretty good - much better than the XT launch. The top-end Pinnacle Ridge picked up maybe 300 MHz on the 1800x, and it had a small IPC gain to boot. That didn't translate into major victories for the 2700x - according to AT Bench, in CBR15 MT, the 2700x was only 8% faster, while it was ~11% faster in Blender. Not bad for a chip that was cheaper at launch than the 1800x. In contrast, the 3900XT cost more than the street price for the 3900X, and had about 0% improvement in MT workloads like CBR20 vs the 3900X.

I've said it before, and I'll say it again - if AMD can pull off another Zen+ with Warhol then it looks like it might be a solid product. Assuming it even exists, which is not guaranteed.
Zen+ fixed issues present in Zen. While Zen was under dev for a long time, they still rushed it to a point. Zen+ brought minimal IPC improvement and better clocks, as you pointed out. You're implying here that Zen3 will be a flawed product and Warhol is now not a stepping stone towards DDR5 but a fix product meant to allay issues present in the upcoming Zen3 processors. Which is it? You're also avoiding the costs issue I brought up.

XT was launched at the same prices the non XT prices were launched at. You're infatuated with the non XT and XT lineup, specifically the 3900.
 
  • Like
Reactions: Tlh97

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Why not? Not only would it be great for consumers (shocking, I know), but it would also force everyone to take them seriously, especially if Intel found themselves facing Zen3 right now and Zen4 in May-July 2021. Zen3 in July 2020 would have been a stunning blow to Intel. They have no desktop product to launch until maybe January at the earliest, fewer people would be looking at TigerLake-H (or really anything else), and on top of all that, there'd be the spectre of something newer and better hitting in July of next year.
Again, would you prefer a new product every 12 months with say 7-10% IPC increase and a few hundred Mhz in clock speeds at an ASP for $385 for consumer retail channels or would you prefer waiting 15 months to get a product that's nearly bug free knowing how AMD is when they rush a product, a processor that costs $385 but has a nice 20-25% IPC increase and maybe 400 Mhz increase in all core? You're advocating what Intel was doing for years. Releasing a product every 12 months, sometimes less, with minimal improvement for the same or increased sales price.

Intel doesn't have anything now. Cometlake is great for gamers and maybe people who spend all day long in CAD/CAM and PS. Frequency loving programs. Intel's MT is a joke. Rocketlake, if rumors are true, is set at 8 cores with a one core boost set at 5.5 according to various and questionable sources. Intel's RKL won't be an answer to Zen3. Alderlake is not a legitimate answer to whatever Zen we'll be on by then. Assuming Intel can launch Alderlake S on time.

Your proposed upgrade path makes no sense if you're an every other year buyer or even a yearly upgrader. You spend way more for far less. People take AMD CPUs seriously, unless they're a complete idiot. Or your initials are FP and secretly own many AMD products in your home and post about when drunk.

People don't take AMD GPUs seriously. You will never win the mindshare game when it comes to people preferring Intel first unless they're technically inclined. You would need over a decade of Intel being 2nd as they have been for the last few years. Who knows. The way Intel is operating, they may be there in half a decade. Sure, they'll still be printing money, but their mindshare among regular consumers will be void. Intel vaguely hinted at this during their quarter report. The consumer market isn't as profitable as DC-enterprise. Ironically, AMD is hitting them there, along with mobile.
 
Last edited:
  • Like
Reactions: Tlh97

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
I'm reasonably certain they can launch a great product more frequently than that. Hell 14 months wouldn't be so bad.
Like if they released this week? My only theory here is that AMD taped out a long time ago and has been stockpiling processors so they don't run into immediate supply issues. You don't gobble up every canceled wafer you can just to tape out at the very last moment and repeat your 2019 mistakes of underestimating the take rate of your products and acting surprised when the market demand is greater than supply.

Yeah, who knew a 12 core for $500 would outsell milk when the nearest Intel product was over double the price.

Giving it some more thought: Warhol is probably a shrink of Zen 3 to 5nm. It is also likely legit.

Take it with a grain of salt. I'm not suggesting AMD is anyone's friend but their bank account, but they're quickly moving into Intel territory with such a move. They win financially, but people will be pissed off. If this is the case, then Zen3 will see a healthy price hike over Zen2 launch MSRP prices.

If Zen3+/Warhol is merely a shrink to 5nm without any improvements apart from energy efficiency and works on all the boards that Zen3 is capable of working on, then I don't see why it wouldn't be real. If it's anything significant, such as USB4, DDR5, et al. then I doubt AMD won't take the chance to release two different chipsets, pocketing money, pissing off consumers, and helping AIB partners make more money.

If I were to pull an Intel with AMD here as AMD, I'd release Warhol with a 5nm shrink and DDR5 support. However, PCIe5 and USB4 will go onto Zen4, among other new exclusive features.


All I can say at this point is:

I don't get the meme?
But they still did it. And it probably made them some extra cash by pushing prices back up to last year's level on what was essentially the same product.

There were some people stating on here their 3600XT performed better than their old 3600. I remember one guy made a long post about his wife's XT being better than the 3600X that was in there in terms of all core. Keeping itself at the advertised all core for a long time without fancy gadgetry. I think in this instance, the 3900XT was a turd.
 
Last edited:
  • Like
Reactions: Tlh97

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
In all seriousness, The Ryzen 3000C series will likely fill that role along with other budget chips. I wouldn’t put it past AMD to drum up new 14nm products for different segments in the future.
Wouldn't mind a high performing, efficient and 'cool' processor for a small htpc with hevc and av1 baked in.
 

Thunder 57

Platinum Member
Aug 19, 2007
2,647
3,706
136
In all seriousness, The Ryzen 3000C series will likely fill that role along with other budget chips. I wouldn’t put it past AMD to drum up new 14nm products for different segments in the future.

I didn't even know those existed until about an hour ago. Clever move IMO.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136

You have to realize that porting Zen 3 isn’t a bad thing. We would likely see some combination of increased core counts, increased clocks, a larger cache, DDR5 and possibly PCIE5 support, etc.

By sticking to a proven core design they can focus on optimizing the shrink itself and work out the kinks in other areas like DDR5.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Giving it some more thought: Warhol is probably a shrink of Zen 3 to 5nm. It is also likely legit.
Warhol is more likely River Hawk.

Matisse -> Vermeer -> Raphael // Receding in Eras
Dali/Pollock -> Warhol // Proceeding Birth Dates

Warhol doesn't follow the refresh naming scheme: Matisse -> Matisse or Renoir -> Lucienne.

River Hawk was known for sometime, so Dali for Raven2/Banded Kestral and Warhol for River Hawk.
You have to realize that porting Zen 3 isn’t a bad thing.
It is more likely Zen4 is the port.
Zen3 => N5 with 7nm+ pitches (7nm+ = 57CPP/38M1/36Mx => 5nm = 57CPP/38M1/36Mx) // Same leakage higher current(Vdd same), lower leakage same current(Vdd lower), reduced capacitance/resistance since the BEOL is 7nm+'s, even more reduced mask count because 14 EUV layers compared to 4 or none, and higher yields long term five years out(for Socket SP4 - Snowy Owl successors:EPYC 3000 = Planned product availability extends up to 10 years, providing customers with a long-lifecycle support roadmap.).

Zen4 => N5 with 5nm pitches (5nm 51CPP/34M1/28Mx) // Designed around the negatives of the 5nm BEOL.

Vermeer(old socket) straight to Raphael(new socket).
 
Last edited:

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
You have to realize that porting Zen 3 isn’t a bad thing. We would likely see some combination of increased core counts, increased clocks, a larger cache, DDR5 and possibly PCIE5 support, etc.

By sticking to a proven core design they can focus on optimizing the shrink itself and work out the kinks in other areas like DDR5.
What has one got to do with the other? If it's optimizations they seek, then they should take a 12-18 month cadence with Zen3+ if it's real. It took between 12 and 13 months to deliver Zen+. We should see a similar time period for your suggestion and not a mid 2021 refresh. However, that would in light of following Zen+'s footsteps, put Raphael firmly in late 2022 if not mid 2023. I'm not sure if I'm being clear with my replies.

Zen3 is on some flavor of 7nm EUV.
Zen3+ "Warhol" AM4+
Zen4 is on AM5

For the sake of discussion, if Zen3+ is truly an optimization ring as you put it then going by a 13 month cycle like Zen to Zen+...

Zen3: October 10th 2020 launch (2 day post announcement)
Zen3+: November 10th 2021 launch.

Then take the typical lifecycle for a major release from AMD, so 15 months?

February 3rd 2023 for 5nm/5nm+ Raphael AKA Zen4. This doesn't work with the MebiuW "leak." You'll likely see increased core counts with Zen4, not Zen3+. If Zen3+ on a shrink is offering your optimizations, you don't blow that with increasing core counts which you would then tame through a lengthy development cycle. Or if you do, you're looking at the latter end of a cadence cycle. If your goal is to reduce overhead complexity with Zen4 and AM5, you don't bring forth DDR5 to Zen3+ to reduce later headaches, and then increase core counts and other end user features. You're simply transferring complexity over to a halfway generational refresh.

You're not going to fit everything on an AM4 pinout. You'd need a new socket. Which is fine, AMD promised until/through 2020 for AM4 support.

If AMD releases a AM4+ socket for a Warhol refresh on 5nm utilizing DDR5 in 2021, then that points to a new motherboard. USB4 devices come out late this year or 1H21. It may be too early to include USB4 support then, but fine in 2022 for Raphael. DDR5 seems to be the only likely candidate, except RAM prices will be high vs. the outgoing generation.


Either way, the customers are footing the extra costs in the end. And dare I say, with impending new feature sets we may see AMD abandon PGA for LGA which would increase board costs. AMD's constant influential growth would secure them negotiation power over AIB partners while Intel is in the backseat screaming their heads off about gaming and the 4800U.


The problem I have with this leak vs. official media is the leak doesn't add up. Nor do any of the contrived theories here including some of my own. Either AMD has finite resources they must use conservatively on products they know will work and sell well or they have infinite resources allowing them to have a five prong attack on Intel. Don't even get me started on the weird GPU leaks that make no sense and seem as if AMD is living in 2012.
 
Last edited:
  • Like
Reactions: Tlh97

dnavas

Senior member
Feb 25, 2017
355
190
116
If AMD releases a AM4+ socket for a Warhol refresh on 5nm utilizing DDR5 in 2021, then that points to a new motherboard. USB4 devices come out late this year or 1H21.
....[snip]....
The problem I have with this leak vs. official media is the leak doesn't add up. Nor do any of the contrived theories here including some of my own.

I'm not sure what the driving force behind having DDR5 compatibility is. If it were PCIe5 or USB4 support so that board/device makers had something to test against, that would make some sense, but DDR5? It's just going to be a money sink, no?
 
  • Like
Reactions: Tlh97 and KompuKare

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
SEV in use at Google. Previously SUSE was providing this service downstream.
SEV is almost done being upstreamed. SEV-SNP is being worked upon downstream.

SME(Zen)--> SEV(Zen2)-->SEV-SNP(Zen3)-->?(Zen4)

MPI kfd bits also being upstreamed.
amdgpu support for TMZ is also in. Looks like the whole circle is getting bigger and covering all the neccessary bits.
Found no more fitting thread, VMware finally updated its vSphere (with 7 Update 1) to support SEV and SEV-ES:
Choice quote:
"There is considerable upside to the way AMD has designed SEV-ES. First, workloads running on a supported guest OS gain these deep protections without modifications to the applications themselves. This means that you don’t have to wait for a commercial application vendor to recompile their application to support this technology. Second, this technology isn’t all or nothing, even on a single host. You can enable it for certain workloads, leave it disabled for others, and they all coexist peacefully. That flexibility means that enabling and deploying this technology can be done at your own pace. Last, it’s easy to enable this for one or 10000 VMs by using a PowerCLI cmdlet."
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Either way, the customers are footing the extra costs in the end. And dare I say, with impending new feature sets we may see AMD abandon PGA for LGA which would increase board costs. AMD's constant influential growth would secure them negotiation power over AIB partners while Intel is in the backseat screaming their heads off about gaming and the 4800U.

Easiest explanation is to drive AM5 board sales. If anything, the board makers would be the ones that would be driving Zen 3 AM4 to be short lived.
 

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Easiest explanation is to drive AM5 board sales. If anything, the board makers would be the ones that would be driving Zen 3 AM4 to be short lived.

Who said anything about AM5? If it's AM5 then it's AM5 and that launches in 2022. This would be AM4+. New pinout to support a few new features. DDR5 was his reasoning. DDR5 has yet to make headway in datacenter use. What good is there to introduce DDR5 5-6 months after launching Zen3? High cost, low speeds? New one generation motherboard before it's abandoned for AM5?

We already know 2020 was the end of AM4. Don't make things complicated. Either AMD is going to jump the gun to release AM5 with this supposed "Warhol" or they're going to release an interim board with an AM4+ socket. We know AM4 is dead in the water after 2020. AMD is not going to pull a "Got ya!" and release this rumored processor on AM4 and face the ire of their partners. This would benefit AIB partners in 0 ways.

If AMD is launching AM5 on this refresh, "Warhol," then it sees fit for it to be an AM5 board. Because why not? They'll be the first platform to offer consumer DDR5, PCIe5 and USB4. No point in introducing an interim socket when all you're doing is bumping up all that work to keep AM5 X770 featureless, if X670 goes to "Warhol."
 

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Get ahead of Alder Lake maybe?
This is assuming Intel is releasing it on time in 2H21, pointing towards a October-November release based on their prior history. Again, presuming they can ship it on time. This is a 10nm product using a big.little approach. Two things Intel hasn't done at mass volume:

1. Deliver 10nm
2. History delivering big.little processors

In that case, just release Raphael at the end of 2021 on AM5. If Warhol is a stationary bug release like Zen+, then its timeline is nearly a year from Zen3's release. If it's a feature release and comes in 6 months, then you're still looking at 14-16 months from Zen3's release, well after Alder Lake S to reclaim their title. Assuming Alder Lake is somehow a competent processor and isn't mired in issues.

If I had to take a stab at all of this, Warhol isn't what you guys think it is. It's a placeholder for Raphael to be bumped up if AMD gets wind of Intel coming out with something very good. That gap between Alder Lake S and Raphael may cause a lot of AMD=>Intel defections if Intel release a really good product, that excels in ST and MT workloads.

It's the only logical explanation, and I wish I thought of it earlier. It wasn't until you brought up Alder Lake S that I thought of that scenario. However, that also implies AMD has Genoa ES up and running on whatever DDR5 ECC modules they can get their hands on.

Warhol doesn't even fit the pattern of all the painter names they use. Warhol was an American pop culture painter.
 
Last edited:

jamescox

Senior member
Nov 11, 2009
637
1,103
136
Like if they released this week? My only theory here is that AMD taped out a long time ago and has been stockpiling processors so they don't run into immediate supply issues. You don't gobble up every canceled wafer you can just to tape out at the very last moment and repeat your 2019 mistakes of underestimating the take rate of your products and acting surprised when the market demand is greater than supply.

Yeah, who knew a 12 core for $500 would outsell milk when the nearest Intel product was over double the price.



Take it with a grain of salt. I'm not suggesting AMD is anyone's friend but their bank account, but they're quickly moving into Intel territory with such a move. They win financially, but people will be pissed off. If this is the case, then Zen3 will see a healthy price hike over Zen2 launch MSRP prices.

If Zen3+/Warhol is merely a shrink to 5nm without any improvements apart from energy efficiency and works on all the boards that Zen3 is capable of working on, then I don't see why it wouldn't be real. If it's anything significant, such as USB4, DDR5, et al. then I doubt AMD won't take the chance to release two different chipsets, pocketing money, pissing off consumers, and helping AIB partners make more money.

If I were to pull an Intel with AMD here as AMD, I'd release Warhol with a 5nm shrink and DDR5 support. However, PCIe5 and USB4 will go onto Zen4, among other new exclusive features.




I don't get the meme?


There were some people stating on here their 3600XT performed better than their old 3600. I remember one guy made a long post about his wife's XT being better than the 3600X that was in there in terms of all core. Keeping itself at the advertised all core for a long time without fancy gadgetry. I think in this instance, the 3900XT was a turd.

Some double negatives in there which make it a little hard to follow. Anyway, I don’t see why AMD launching AM5 with Zen 3 is a bad thing. People who buy Zen 3 on AM4 aren’t going to rush out and swap their board + memory + cpu some months later. It is only recently that upgrading has become actually worthwhile. A lot of people were still using very old cpus prior to Zen 2 due to Intel monopoly stagnation. I still use an i7-920 Linux box with 24 GB of memory for some things. The board makers probably don’t care as much about having a new socket frequently the way intel was going. People are upgrading more frequently anyway, due to AMD pushing the technology, which drives board sales without unnecessary socket changes.

Some early adopters coming from older systems may want to go for DDR5. It would probably be a low volume product due to the expense of DDR5 and possibly not much performance increase over AM4 Zen 3. Being a low volume product, is there really that much point to doing a 5 nm shrink? If they had other uses for the die, then maybe. Epyc doesn’t get updated as much since they have longer term production obligations. I don’t think there was a Zen+ Epyc. Epyc will probably stay the same die until Zen 4, so I don’t see where else they would use 5 nm Zen 3.

It seems like a good idea to get some boards out there prior to Zen 4 though, which is the much bigger launch. AM4 boards were already around for the A-series APUs prior to Zen launch. I don’t think there will be an AM4+; that would be pointless. I would expect it to be a fully featured AM5, except possibly missing pci-e 5. I don’t know if we will see pci-e 5 in the desktop market for quite some time. It just isn’t needed.

Given the timing, I wouldn’t be surprised if it is the same Zen 3 die launching October 8. Things are going to get weird with Zen 4 since the Epyc processors are likely use stacked die on an interposer. Interposers are expensive though, so are they going to have a tiny interposer for Ryzen parts or will they make two different versions of the chip for use on an interposer and in an MCM? They aren’t really compatible and I don’t think one die could be designed for both. AMD has a lot more resources now, so they could do both, I guess. I was thinking that they would just make a more powerful 8 core monolithic die for the desktop and only high end parts (over 8 core) would be interposers and chiplets. That brings up the question of what products would use the Warhol IO die. If they have a Zen 4 chiplet made for MCM then they could continue to use it. It could still be used as a chipset also. I guess the Warhol IO die could be some odd in-between thing with some memory controller and IO chiplets on a little interposer. That may make sense to do before going full interposer for everything that isn’t monolithic. The interposers don’t take much design work, especially if they are passive interposers. They are metal interconnect layers and TSVs, no transistors.

I am am not convinced that Warhol exist either, but it seems like if it does exists, it just Zen 3 with new IO die. This would be more like a chipset upgrade than a new cpu. I don’t think I have ever seen a Zen 3+ mentioned and there doesn’t seem to be that much of a reason to shrink the Zen 3 cpu die to 5 nm. I think the first 5 nm will be a mobile APU. AMD may have changed their mobile APU schedule due to 5 nm availability and high demand for mobile parts due to lack of competition from Intel.