Info 64MB V-Cache on 5XXX Zen3 Average +15% in Games

[Kedas]

Well we know now how they will bridge the long wait to Zen4 on AM5 Q4 2022.
Production start for V-cache is end this year so too early for Zen4 so this is certainly coming to AM4.
+15% Lisa said is "like an entire architectural generation"

 

[JoeRambo]

I wonder if the possible 4 stack Epyc devices will still be at 2TB/s or even faster

The 2TB/s number is for desktop dual chiplet CPU - aka what you get from 2 chiplets @ desktop chip clocks.
EPYC has 8 chiplets and lower clocks and accordingly a different cumulative L3 bandwidth, i think 5-6TB/s total. Someone with Aida64 and 64C Epyc Z3 could check.

 

[Timorous]

While I do think there is a danger of AMD becoming complacent (the last time in 2005ish they wanted to make hay rather than invest in their fabs, and then Conroe happened), there might be logic with having both an AM4 and an AM5 release next year.
Unlike Alder Lake, we have had no indication that AM5 can use both DDR4 and DDR5, and DDR5 prices are a huge risk factor.

This is something I said a while ago when the Warhol rumors were floating around.

It makes a lot of sense for AMD to keep both AM4 and AM5 around for a while. Especially on the budget end with the 4c 12nm Zen3 parts that are rumoured.

 

[DrMrLordX]

Zen3D AM4 Q1 2022
Zen4 AM5 Q2 2022

What's the issue?

To quote myself

I think you need to look at the bigger picture, especially if this product is being released in lieu of something else.

This is something I said a while ago when the Warhol rumors were floating around.

It makes a lot of sense for AMD to keep both AM4 and AM5 around for a while. Especially on the budget end with the 4c 12nm Zen3 parts that are rumoured.

That would be fantastic if they actually did that. If.

 

[LightningZ71]

I fw l that the 5900x3d and 5950x3d will be relevant for a very long time forward. With that massive L3 cache (up to 16MB per core for the 5900x), and 12-16 cores, it stands to reason that they will have a similar life to the Broadwell chips that had the L4 cache.

 

[yuri69]

Zen3D AM4 Q1 2022
Zen4 AM5 Q2 2022

What's the issue?

Zen3D is going to production in late 2021. AM5 boards are aiming for a Q2 2022 release. Zen4 is coming "late 2022" with the 3D filling the gap.

* Zen3D AM4 - Q1-Q2 2022
* Rembrandt AM5 - Q2 2022
* Zen 4 AM5 - Q3-Q4 2022

 

[jpiniero]

See above comment to KompuKare. Q2 2022 should be the latest time at which AMD would release Raphael(Zen4) according to their 15-18 month cadence.

They are still doing Zen 3+, it just seems that the CPU only product is getting Zen 3 vcache instead.

 

[DrMrLordX]

Zen4 is coming "late 2022" with the 3D filling the gap.

It's great that AMD wants to release Zen3 + vcache but that's no reason to delay AM5 + Raphael by 2-3 quarters.

 

[yuri69]

It's great that AMD wants to release Zen3 + vcache but that's no reason to delay AM5 + Raphael by 2-3 quarters.

How about TSMC 5nm - tests, validation, or initial capacity might affect this launch. Plus there might be an advanced packaging employed this time. Additionally, work on Zen 4 was fully hit by working remote - loss of efficiency, etc.

Anyway, AM5 itself doesn't seem to be affected since Rembrandt seems to be the spearhead in Q2 2022. We just need to wait for a AM5 BIOS dump in order to see Raphael alive.

 

[DrMrLordX]

How about TSMC 5nm - tests, validation, or initial capacity might affect this launch.

TSMC has been shipping N5 for some time now. AMD had working N7 silicon (Vega20) in late 2018. I got my Radeon VII in early 2019. Genoa is already pretty far along as well. Obviously we aren't going to see it officially launched like Milan anytime soon, but hyperscalars should have it before 2021 is out if they haven't already gotten some samples.

Anyway, AM5 itself doesn't seem to be affected since Rembrandt seems to be the spearhead in Q2 2022.

That would be really, really strange if AMD shipped desktop Rembrandt before Raphael.

 

[jamescox]

This will leave one last upgrade to the am4 socket. I think the best thing would be to watch ebay 2024 and after for a cheap one.

I suspect these will be expensive when released. More then the Ryzen 5000 series, AMD has no reason not to push the price up.

They are not operating in a complete vacuum. I haven’t been paying much attention to intel since they have been leaking all kinds of stuff in an attempt to stay relevant, or at least still in the news cycle. If they finally have 10 nm chips with good yields, then they should be competitive with desktop Ryzen parts, so I think AMD needs zen 3D and they need it to be relatively cheap. The price will depend on what Intel actually manages to deliver and on general availability at the time. Intel probably isn’t going to have very good competition for Epyc for a while yet though. They can’t compete with Epyc using a monolithic die, so it depends on when they have die stacked packages available.

 

[Joe NYC]

If they are the leading customer for this, surely TSMC determines the timeline?
It's not like anyone other than TSMC can package this, however much we might want this yesterday.

AMD showed one Zen 3D on May 31st. If they make 1 per day from May 31st, they will have higher launch Zen 3D on September 9 (101 chips) than Intel had of Pentium 1 GHz - 100 CPUS - when Pentium 1 GHz launched.

 

[A///]

It's great that AMD wants to release Zen3 + vcache but that's no reason to delay AM5 + Raphael by 2-3 quarters.

Especially when they serve two sets of buyers. Those wanting to extend their current hardware shelf life and those wanting cutting edge. AMD could easily keep the current sales price for Zen 3, Zen 3 V-Cache and Zen 4's debut prices and laugh all the way to the bank. That's what I'd do if I were AMD, but I'm not so meh. Knowing AMD's cheesy antics, I wouldn't be shocked if they release Zen 4 on May 4th or on May 5th for a 5-5 release.

Zen 4 is on a completely different node. Zen 3 and Zen 3 V-Cache are the same chips and only bonded and packaged differently AFAIK, which to be fair is little because I was never keen on buying. Based on what little I know it should work out perfectly fine for AMD?

 

[DrMrLordX]

Based on what little I know it should work out perfectly fine for AMD?

The only legit reason I can think of for AMD to pin all their hopes on "Zen3d" or what have you until Q4 2022 or some other ridiculously late date is that AMD may be planning on committing all their 5nm wafers to Genoa + Bergamo (at least for a few quarters). While we're all paying attention to the consumer desktop side of things, what's really happening is a matchup between Genoa and Sapphire Rapids, which is a matchup Intel does not want. Sapphire Rapids will be hard-pressed to eclipse Milan, but Genoa? Not a chance. Meanwhile, us retail plebs will get the N7/N6 scraps left over.

 

[Topweasel]

The only legit reason I can think of for AMD to pin all their hopes on "Zen3d" or what have you until Q4 2022 or some other ridiculously late date is that AMD may be planning on committing all their 5nm wafers to Genoa + Bergamo (at least for a few quarters). While we're all paying attention to the consumer desktop side of things, what's really happening is a matchup between Genoa and Sapphire Rapids, which is a matchup Intel does not want. Sapphire Rapids will be hard-pressed to eclipse Milan, but Genoa? Not a chance. Meanwhile, us retail plebs will get the N7/N6 scraps left over.

Remember how late (6+ Months) Zen 2 came to market after Rome. They might be willing to take an extended time under ADL lead (assuming there really is one) and reduced ASP on the desktop market, to ramp up Genoa, get seeds out, and probably give DDR5 some substantial manufacturing time before launching AM5. Honestly I am not sure we are really ready for DDR5 in the desktop. If they launched anytime soon, DDR5 would feel like Rambus in 2001.

 

[DrMrLordX]

Remember how late (6+ Months) Zen 2 came to market after Rome. They might be willing to take an extended time under ADL lead (assuming there really is one) and reduced ASP on the desktop market, to ramp up Genoa, get seeds out, and probably give DDR5 some substantial manufacturing time before launching AM5. Honestly I am not sure we are really ready for DDR5 in the desktop. If they launched anytime soon, DDR5 would feel like Rambus in 2001.

DDR5 should be ready by Q1 2022 as long as the motherboard OEMs can deliver.

 

[moinmoin]

DDR5 should be ready by Q1 2022 as long as the motherboard OEMs can deliver.

In unlimited quantity?

 

[DrMrLordX]

In unlimited quantity?

Nah. But there should be enough available for the early adopters. The only lagging indicator is that Intel doesn't have a server/workstation platform using it yet, so you don't get any of the lovely early-market green generic DIMMs on eBay like you did at the DDR3 and DDR4 "launch" (it was not a singular event).

 

[A///]

The only legit reason I can think of for AMD to pin all their hopes on "Zen3d" or what have you until Q4 2022 or some other ridiculously late date is that AMD may be planning on committing all their 5nm wafers to Genoa + Bergamo (at least for a few quarters). While we're all paying attention to the consumer desktop side of things, what's really happening is a matchup between Genoa and Sapphire Rapids, which is a matchup Intel does not want. Sapphire Rapids will be hard-pressed to eclipse Milan, but Genoa? Not a chance. Meanwhile, us retail plebs will get the N7/N6 scraps left over.

Don't think it'll be Q4. AMD has said 2H22, but don't be surprised if they launch it earlier to trip Intel up. N6 would be Rembrandt, would it not? Extra N7 or other flavor of 7nm @ TSMC will be taken up by the IOD as there have been strong rumors and suggestions that AMD will solely rely on TSMC for the entire package including the IOD for Zen 4 and future Zens. It's going to be interesting to see how the two companies pull this off considering the shortages will last until 2023 if not midway of that year.

Your Genoa/Bergamo suggestion makes sense because they'd have a head start and have chiplets that didn't pass Epyc standards be up for binning on Ryzen products or future Threadrippers. That said, I wouldn't be surprised if AMD begins production in a few months or close to the end of the year. Or even now. The upcoming fall and winter is going to prove difficult given COVID is still rampant and the ramp down and or pause for Lunar New Year.

I'd say we'll get hardware sometime in summer if my May the 4th or 5/5 prediction falls flat on its ass.

Do we know what AMD's MCM consumer cards will utilize? As far as consoles go, there was an article posted online a few weeks back that Sony has already designed a better PS5 housing for the digital mode, and I suspect the optical disk model will get a redesign sooner or later, too. This won't affect internals much, but I suspect the refresh of these consoles including what Microsoft is offering will be offered on N6. But these will come out in 2-3 years. Enough time to not hamper production. I do not see them getting a refresh to 5nm due to costs involved of a new design. Though given the rapid improvement we may see in the near future, I wouldn't be surprised if this generation of consoles suffer from a short lifespan before the next generation is out.


April-June release goes with AMD's 12-18 month cadence. Late '22 wouldn't make sense. The V-cache is going to affect a small amount of processors per AMD's words of on the higher end of their product line. As far as I'm concerned, these are essentially XT like refreshes that have zero bearing on their cadence.

I'm not too worried about Alder Lake, Meteor Lake, Rapotor lake or their HEDT ambitions over the next few years. Let them prove themselves for a few generations with their hybrid system. Even Intel's latest mental departure involving their naming scheme has left a bad taste in people's mouths.

 

[eek2121]

I fw l that the 5900x3d and 5950x3d will be relevant for a very long time forward. With that massive L3 cache (up to 16MB per core for the 5900x), and 12-16 cores, it stands to reason that they will have a similar life to the Broadwell chips that had the L4 cache.

I mean, so will the 5950X.

AMD showed one Zen 3D on May 31st. If they make 1 per day from May 31st, they will have higher launch Zen 3D on September 9 (101 chips) than Intel had of Pentium 1 GHz - 100 CPUS - when Pentium 1 GHz launched.

You mean the RECALLED One? I still remember those days. AMD beat Intel to 1 GHz. Intel had to launch a new stepping after their initial 1ghz chip didn’t work.

DDR5 should be ready by Q1 2022 as long as the motherboard OEMs can deliver.

DDR5 was available on Amazon for a bit. It even had an add to cart button (I almost bought it for fun). Then it sold out. I suspect in the next 2 months we will begin to see it in stock.

 

[Thunder 57]

...You mean the RECALLED One? I still remember those days. AMD beat Intel to 1 GHz. Intel had to launch a new stepping after their initial 1ghz chip didn’t work...

Nah, that was the 1.13GHz that was recalled. Tomshardware exposed it and was punished by Intel for it, despite other review sits confirms Intel was wrong. I mean, they did end up recalling the damn thing after all. One of Intel's numerous crappy moves.

Thankfully their more aggressive moves have stopped in favor of more comical ones like the chilled water demonstration. Still, the damage has been done. For example I was saying how I liked that we were replacing Dell laptops with HP's which happened to be Elitebooks with 3500U's. Mostly because I think Dell has gone downhill in a big way but also because I thought it was nice to see AMD in a more premium product that was being purchased large scale. My co-worker (an older guy) responded by saying they weren't as good because they came with "radon CPU's". Yes, Radon as in the element, not even Radeon. I just let it go.

For the longest time AMD laptop chips were pretty crappy, typical in terms of battery life. If you aren't going to keep up with tech though, you shouldn't be working in IT.

 

[A///]

You mean the RECALLED One? I still remember those days. AMD beat Intel to 1 GHz. Intel had to launch a new stepping after their initial 1ghz chip didn’t work.

Wait a dang second. I remember this. That was the Pentium 3 incident, wasn't it?

Nah, that was the 1.13GHz that was recalled. Tomshardware exposed it and was punished by Intel for it, despite other review sits confirms Intel was wrong. I mean, they did end up recalling the damn thing after all. One of Intel's numerous crappy moves.

Isn't that what he's referring to? Toms Hardware did get punished, but I remember Kyle Bennett ended up saving the day by offering his own chip to be tested and confirm. I think after that, everything snowballed and Intel couldn't keep denying it.

 

[Thunder 57]

...Isn't that what he's referring to? Toms Hardware did get punished, but I remember Kyle Bennett ended up saving the day by offering his own chip to be tested and confirm. I think after that, everything snowballed and Intel couldn't keep denying it.

Well pretty much yeah. Except it was the 1.13GHz model that was recalled, not the 1GHz. Small detail, not trying to nitpick just wanted to clarify in case someone tried to go search for info on it.

AMD did beat Intel to 1GHz by a matter of a week or two maybe, but the Pentium 3 model was a paper launch whereas you could get the Athlon 1 GHz.

Kyle Bennett, there's a name I haven't heard in awhile. He was right up there with Anand. Anandtech was involved in the recalled P3 as well. With so many reputable sources confirming the problem Intel had no choice but to go into damage control mode.

 

[Joe NYC]

April-June release goes with AMD's 12-18 month cadence. Late '22 wouldn't make sense. The V-cache is going to affect a small amount of processors per AMD's words of on the higher end of their product line. As far as I'm concerned, these are essentially XT like refreshes that have zero bearing on their cadence.

There may have been a non-trivial amount of effort on Zen 3d with V-cache. We know that there is a new stepping of Zen 3. We don't know what it entails. It could be from trivial, to gain of some frequency all the way to moving Zen3 (and Zen 3d to 6nm)

Given the likely tightness of 5nm capacity in 2022, I thin Zen 3 will be around for some time in desktop. We may see a full line-up of Zen 3d desktop products.

I'm not too worried about Alder Lake, Meteor Lake, Rapotor lake or their HEDT ambitions over the next few years. Let them prove themselves for a few generations with their hybrid system. Even Intel's latest mental departure involving their naming scheme has left a bad taste in people's mouths.

Well, I am worried. Alder lake will almost certainly edge regular Zen 3. AMD will need Zen 3D to keep desktop crown.

Going forward to Meteor Lake, Intel is the only company that has on its public roadmap a standalone GPU chiplet / tile. Ever since ATI acquisition, AMD has talked about integration a lot, has done a just rudimentary effort to have integrated graphics of low end quality integrated on some CPUs. But never a game changing solution, that would replace, at least on the mobile side, the external GPU.

And Intel is doing it with Meteor Lake.

 

[Joe NYC]

I fw l that the 5900x3d and 5950x3d will be relevant for a very long time forward. With that massive L3 cache (up to 16MB per core for the 5900x), and 12-16 cores, it stands to reason that they will have a similar life to the Broadwell chips that had the L4 cache.

I agree. I think many people underestimate the 3D V-Cache as the first implementation of 3D stacking. It will completely redefine the industry and how higher end chips are put together and how much the performance can scale with additional silicon.

Before chiplets, massive chips led to a dead end in manufacturability cost and optimization.

Before 3D, partitioned chiplets are leading to a dead end with the limits bandwidth, latency and power overhead of their links.

It is possible that the 2D horizontal interconnects and 2.5D inteconnects will all get replaced with vertical 3D connections using hybrid bonding.

Something like this for Epyc CPU, that has 8 of these CCDs. They could all be attached in vertical fashion to an elongated IO die.



Based on Intel's presentation and timing of Intel's Feveros Direct, TSMC and AMD have approx. 3 year lead on Intel with this technology. I think in these 3 years, AMD will drive Intel crazy with it, extending 3D one step at the time, just as it looks like Intel can catch up.

And in these 3 years, they can possibly eclipse Intel in size and scale.

 

[Joe NYC]

You mean the RECALLED One? I still remember those days. AMD beat Intel to 1 GHz. Intel had to launch a new stepping after their initial 1ghz chip didn’t work.

Yeah, Intel, with help of Dell, tried to fake it, that they had 1 GHz Pentium. And those were highly cherry picked parts. Someone leaked from Dell that entire Dell organization got only about ~25 of them...