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"

 

[Timorous]

Yeah, I don't fully buy the "you might actually lose a little performance" when going with dual chiplets, the same logic could have been applied to 5800X vs. 5900X at the time of launch, and the 5900X won or traded blows with 5800X in gaming. That being said, the extra cache on both parts would probably help the single die product more, essentially matching the dual die product in most games.

I do believe them when they say that normal consumer apps won't see much of a benefit from 3D cache though, which reinforces their decision to go for single chiplet only on the consumer side.

The 5800X3D lost 200Mhz turbo clock. A 5900X3D hitting the same TDP would need to lose more, especially if they went with weaker CCDs to not impact Milan-X shipments. While it is not true in the general sense it could be with the added Milan-X caveats.

 

[coercitiv]

The 5800X3D lost 200Mhz turbo clock. A 5900X3D hitting the same TDP would need to lose more, especially if they went with weaker CCDs to not impact Milan-X shipments.

Milan-X was something AMD reps could not talk about on that stream, so I was addressing their reasons as they were presented. If 5800X3D turns out to have negligible performance gains in consumer apps other than games, their decision to stay with just one chiplet is justified no matter the Milan-X demand. On the other hand, if any real-world productivity workloads see major uplifts... 😱

 

[Hans Gruber]

Regardless of what people think about V-cache (probably a letdown). AMD is now behind Intel in every aspect with power consumption being the only AMD advantage. This doesn't include the 5950x. Considering the clock regression and the 3-4 month time table before launch. Intel already has more CPU's slated for release before the 5800x3d.

I can see it now, 15% gaming gains in 1080p but lower computing power in every other aspect compared to a 5800x. This is another Zen 2 XT end of life release. The XT had lower operating voltage and a higher single core clock vs other Zen 2 parts but no real performance gains to justify the price premium.

Significant price reductions in Zen 3 is their best option before Zen 4. With Zen 4 they will have a double boost in performance (IPC gains and 5nm silicon).

 

[DrMrLordX]

AMD is now behind Intel in every aspect

. . . really? Milan is behind IceLake-SP? No. Genoa won't be behind Sapphire Rapids, either.

 

[Markfw]

. . . really? Milan is behind IceLake-SP? No. Genoa won't be behind Sapphire Rapids, either.

Yes, in server, so far behind its not funny.

Also, HEDT for Intel does not exist.

And mobile ? Not an expert there, but I know there are some pretty good chips out there for AMD.

Even ahead in desktop ? Maybe if you use Intel default power settings and smoke the PSU's, and don't count multi-threaded.

Hans is delusional.

 

[igor_kavinski]

Hans is delusional.

He is frustrated with AMD's strategy of not trying to expand their desktop market share. The fastest CPU with the best IPC right now is 12900K and soon, it will be the 12900KS. AMD is just sitting still and letting Intel count their bills. We want a bloodbath. But AMD is doing the sensible thing and increasing profits for their shareholders. Good for them but bad for those of us who want excitement in the PC space.

 

[maddie]

Regardless of what people think about V-cache (probably a letdown). AMD is now behind Intel in every aspect with power consumption being the only AMD advantage. This doesn't include the 5950x. Considering the clock regression and the 3-4 month time table before launch. Intel already has more CPU's slated for release before the 5800x3d.

I can see it now, 15% gaming gains in 1080p but lower computing power in every other aspect compared to a 5800x. This is another Zen 2 XT end of life release. The XT had lower operating voltage and a higher single core clock vs other Zen 2 parts but no real performance gains to justify the price premium.

Significant price reductions in Zen 3 is their best option before Zen 4. With Zen 4 they will have a double boost in performance (IPC gains and 5nm silicon).

The best for you & us is either company leapfrogging the other one every new release.

 

[Markfw]

He is frustrated with AMD's strategy of not trying to expand their desktop market share. The fastest CPU with the best IPC right now is 12900K and soon, it will be the 12900KS. AMD is just sitting still and letting Intel count their bills. We want a bloodbath. But AMD is doing the sensible thing and increasing profits for their shareholders. Good for them but bad for those of us who want excitement in the PC space.

But even just for desktop, Intel is not a clear winner. Default power usage to too high (ridiculous on the 12900k), multi-threaded is still not in their court, and windows and other scheduler issues are still around. And in ALL other areas, Intel is the looser, especially in server where they are like 3-4 generations behind. The only CLEAR win for Alder lake is gaming, and that may change soon.

Thus saying "AMD is now behind Intel in every aspect" is so totally wrong, that he is delusional.

 

[Mopetar]

AMD is just sitting still and letting Intel count their bills. We want a bloodbath. But AMD is doing the sensible thing and increasing profits for their shareholders. Good for them but bad for those of us who want excitement in the PC space.

What is AMD supposed to do that wouldn't count as sitting still? You can't magically launch a product months before it's actually ready and anything coming out in the next year is so far along in development that there's nothing that can be done to significantly change how it will perform.

This argument that AMD is sitting on its hands and doing nothing is ridiculous. Maybe someone thinks they should cut prices, but I doubt that would do much good. Retailers aren't going to sell for less than they can get and even low-end CPUs several generations old are often being sold above the original MSRP.

How is this not the most exciting time in recent memory for CPUs? AMD overtook Intel with Zen 3, Intel has released a new design with a big.LITTLE approach, Apple has their own SoC that offers desktop class performance, and we're soon going to have a CPU with stacked dies. What does excitement even mean to you that your post implies a lack of it?

 

[Markfw]

He is frustrated with AMD's strategy of not trying to expand their desktop market share. The fastest CPU with the best IPC right now is 12900K and soon, it will be the 12900KS. AMD is just sitting still and letting Intel count their bills. We want a bloodbath. But AMD is doing the sensible thing and increasing profits for their shareholders. Good for them but bad for those of us who want excitement in the PC space.

Those 2 have horrible efficiency. While barely faster than the 5950x, they use almost twice the power. They may technically be the fastest, but I will still take AMD at this point

 

[igor_kavinski]

What does excitement even mean to you that your post implies a lack of it?

I would be excited if AMD had a desktop 6nm Zen3+ die (with V-cache and RDNA2 iGPU) by end of January. I guess they thought that Intel couldn't possibly top Zen 3's performance so they got complacent instead of being ready to fire at a moment's notice. But I agree that whatever they are doing now makes financial sense. I just don't agree with their strategy as an enthusiast. If I had the kind of history that AMD has with Intel, I wouldn't let Intel get up. I would keep landing punch after punch.

 

[Markfw]

I would be excited if AMD had a desktop 6nm Zen3+ die (with V-cache and RDNA2 iGPU) by end of January. I guess they thought that Intel couldn't possibly top Zen 3's performance so they got complacent instead of being ready to fire at a moment's notice. But I agree that whatever they are doing now makes financial sense. I just don't agree with their strategy as an enthusiast. If I had the kind of history that AMD has with Intel, I wouldn't let Intel get up. I would keep landing punch after punch.

In server, they are doing exactly that. Rome beats anything they have, but they also have Milan, Milan-X and now coming (or here in sampling) Genoa ! I mean that is a army that is working to topple the giant in the data center. Once that is done, desktop is almost an afterthought.

 

[eek2121]

But even just for desktop, Intel is not a clear winner. Default power usage to too high (ridiculous on the 12900k), multi-threaded is still not in their court, and windows and other scheduler issues are still around. And in ALL other areas, Intel is the looser, especially in server where they are like 3-4 generations behind. The only CLEAR win for Alder lake is gaming, and that may change soon.

Thus saying "AMD is now behind Intel in every aspect" is so totally wrong, that he is delusional.

Yeah it cracks me up that people say AMD is behind. Compare the 5900X or the 5950X with the 12900k and you will see AMD is the perf/watt king and they still win many benchmarks. AMD’s chips are more than a year old and Intel just now caught up! Zen 4 is likely going to put Intel behind again until Arrow Lake minimum.

AMD could release higher clocked chips if they felt threatened, but Zen 3 is still a top seller, so why bother?

 

[LightningZ71]

While I admit to cheerleading for the 5900x3d and 5950x3d, seeing the clock regressions for the 5800x3d and figuring that it's power related, I can understand why AMD would choose to not release the 12+ core parts fir destop. Base and sustained all core clocks alreasy take a hit when those two parts are kept firmly in the stock power envelope. It makes a lot more sense to keep the 3d vcache to the 5800x. That said, Warhol, or whtever the rumored 6nm shrink was supposed to be called, seemed like a perfect fit for the 5900+ parts. The power draw and thermal load reductions alone would have had a significant impact on the sustained performance of highly threaded MT loads on those parts. Unfortunately, I think that the limitations of DDR4 at JEDEC settings would have jeavily handicapped many of those tasks that can't fit in the L3. It might have been a cinebench monster, but not been much better elsewhere.

 

[Markfw]

The thing is (IMO) that AMD is still ahead of Intel in everything but gaming. This will counter that. Now in single thread, they sometimes also have a slight lead, but not always. This puts AMD at least on par if not ahead of Intel, until Zen4 hits. I don't know enough to say how much better it will really be, but 25% seems to be the consensus.

 

[Thibsie]

Yeah it cracks me up that people say AMD is behind. Compare the 5900X or the 5950X with the 12900k and you will see AMD is the perf/watt king and they still win many benchmarks. AMD’s chips are more than a year old and Intel just now caught up! Zen 4 is likely going to put Intel behind again until Arrow Lake minimum.

AMD could release higher clocked chips if they felt threatened, but Zen 3 is still a top seller, so why bother?

Yah but you have to feed that youtube and twitter chanels somehow huh ?

 

[Mopetar]

I would be excited if AMD had a desktop 6nm Zen3+ die (with V-cache and RDNA2 iGPU) by end of January. I guess they thought that Intel couldn't possibly top Zen 3's performance so they got complacent instead of being ready to fire at a moment's notice. But I agree that whatever they are doing now makes financial sense. I just don't agree with their strategy as an enthusiast. If I had the kind of history that AMD has with Intel, I wouldn't let Intel get up. I would keep landing punch after punch.

I don't think that's a reasonable position to take for a number of reasons. Why should AMD plan to have a half generation desktop part when it would be replaced by Zen 4 as few as 5 months later? Should they not have Zen 3D since that would be three different products landing within ~6 months of each other?

You also act as though AMD who only pulled ahead of Intel with Zen 3 wouldn't believe that Intel could beat them. Can you point to anyone that's been behind a competitor for the majority of the last decade that would assume the competition would stay beaten once down? What leads you to believe that AMD actually thinks they way you're describing?

Your idea of excitement is ludicrous. I don't even think that you believe it yourself but have to come up with some ridiculous explanation to back up your claim that AMD isn't bringing the "excitement" you demand. CPUs are products that are in development for several years. It's not possible to "fire at a moment's notice" when they have to spend months for silicon to be made a packaged, never mind that the rest of the platform has to be there as well.

Why did it take so long for Intel to fire back when Zen 3 came out? It was a year before before they had anything to answer AMD. Why weren't they, a company many times the size of AMD with far more resources at their disposal, able to bend the laws of reality to excite you? If AMD not having something to beat Intel until sometime this spring is unbearable, I can't imagine your anguish at a year without excitement. However were you able to survive?

 

[igor_kavinski]

@Mopetar

You are taking my desire to see something exciting soon, a bit too seriously. To be honest, I didn't expect Alder Lake to get ahead of Zen 3. But they managed it and this should have heated things up. It hasn't, unfortunately. AMD's response seems to be lukewarm. They don't seem concerned and are fine to let Intel bask in the sun a bit. They are NOT being a fierce competitor. Intel cooking up 12900KS constitutes competition. 5800X3D seems like a half-hearted fulfillment of a promise and only because everyone waited in dismay for the 5900X3D for Christmas that AMD initially showed off. AMD could have used the same mobile Zen3+ die to at least have G-series line-up for desktop. They have deliberately slowed down or they are having some unforeseen manufacturing snags.

 

[coercitiv]

They are NOT being a fierce competitor.

Now that's a joke. Every Zen3 chiplet AMD sells in the server market and every APU they sell in the laptop market cuts deep in Intel's profit margins. DEEP. Just because you're witnessing a "cold" war on the DYI front doesn't mean the competition is any less fierce.

5800X3D seems like a half-hearted fulfillment of a promise and only because everyone waited in dismay for the 5900X3D for Christmas that AMD initially showed off.

You realize that AMD is NOT a certain forum user who continuously spread this false rumor that they were going to deliver Zen3D before the end of the year? You do acknowledge that all AMD ever said was they were beginning production in Q3 2021?

AMD could have used the same mobile Zen3+ die to at least have G-series line-up for desktop.

I'm not gonna take this seriously.

 

[igor_kavinski]

Fine. AMD can do no wrong. Happy?

 

[IronLynx]

The 5800X3D lost 200Mhz turbo clock.

But Lisa could be tempering our expectations? Would be impossible having two X3D models(like 5800 and 5890) one with 105 and the other with 125TDP reaching 5.0/5.1 ST?

 

[Timorous]

Fine. AMD can do no wrong. Happy?

B350 and X370 owners would disagree with that.

 

[Mopetar]

You are taking my desire to see something exciting soon, a bit too seriously.

Not really, I just think you had an awful hot take with sloppy reasoning that makes no sense. We're getting Zen 3D in the next few months. Is it seriously diminished that much just because it's not RIGHT NOW! instead of a few months from now?

To be honest, I didn't expect Alder Lake to get ahead of Zen 3.

If you didn't even expect it to beat Zen 3, then why do you expect AMD to have taken it more seriously even though there's still nothing that they could have done about it even if they had?

But they managed it and this should have heated things up. It hasn't, unfortunately. AMD's response seems to be lukewarm. They don't seem concerned and are fine to let Intel bask in the sun a bit.

What should they be doing? Telling all of the board partners that the schedule they gave them last year needs to be pushed up 5 months so that AM5 platforms can hit the market for a small number of Ryzen 6000 CPUs that get sold as desktop parts. Should they tell the notebook manufacturers that they'll have to do with half of their original orders so that they can sell some desktop parts? You're basically demanding that AMD alienate everyone else that they work with and for what? A desktop part that they would be replacing in ~6 months? Outside of people happy with an APU that won't want Zen 4, who's going to buy that?

They are NOT being a fierce competitor.

Even though Intel overtook AMD with Alder Lake, AMD hasn't been this competitive with Intel in decades. What does this even mean? How is the current state not fierce competition when it hasn't been this competitive in such a long time? You keep saying words as though you're making some profound statement, but they're just completely hollow because there's no logic or sense behind them.

Maybe you're expending AMD to cut prices. They probably could, but if they're still selling through their inventory at a good enough rate, why should they? Lowering MSRP by $50 or $100 won't change the products they have, and as the GPU market shows, MSRP is an illusion. If AMD cut the MSRP by $100 across the Zen 3 lineup the reality would be that retailers would pocket most of that $100.

Intel cooking up 12900KS constitutes competition.

What do you even mean by cooking up? They were always going to release a better binned part, just like both companies always do.

5800X3D seems like a half-hearted fulfillment of a promise and only because everyone waited in dismay for the 5900X3D for Christmas that AMD initially showed off.

So the 5800X3D is bad, yet Intel had to resort to "cooking up" the 12900KS?

Why was anyone expecting a product to release at Christmas if it was never announced to be available at that time? I've only seen one other poster here think that it was coming out (or needed to come out) in that period of time and I think they're dead wrong on their reasoning.

The 5800X3D is among other things a pipecleaner product for AMD to gain experience working with TSMC's die-stacking process. Most of it is probably being sold into the server market where they can get far greater margins, but it makes an okay halo product and a nice swansong for AM4.

AMD could have used the same mobile Zen3+ die to at least have G-series line-up for desktop. They have deliberately slowed down or they are having some unforeseen manufacturing snags.

As I pointed out earlier they'd need to have AM5 motherboards available and would need to take away supplies from notebook manufacturers.

You seem to have formed your conclusion first and have only tried to work backwards and argue it after the fact. That's why all of your points are ridiculous and don't even stand up to basic scrutiny. If it's a bad hot take, just leave it at that. It's not like internet forums aren't full of them and I've had enough of my own as well, but you can't make a bad hot take look good any more than you can polish a turd. And much like anyone you might see polishing a turd, they're going to look ridiculous for trying to do so.

 

[KompuKare]

Milan-X was something AMD reps could not talk about on that stream, so I was addressing their reasons as they were presented. If 5800X3D turns out to have negligible performance gains in consumer apps other than games, their decision to stay with just one chiplet is justified no matter the Milan-X demand. On the other hand, if any real-world productivity workloads see major uplifts... 😱

I would imagine that file compression would benefit greatly from more cache. Having lots of space to keep the dictionaries for solid archives should make a huge difference.

Other tasks might be compilation, and perhaps certain database loads. Problem might be that a lot of these tasks have been written and optimised specifically to run well without so much cache and they - along with things like the jvm or .NET vm - would really need to aware of so much cache to properly benefit.

 

[Makaveli]

But Lisa could be tempering our expectations? Would be impossible having two X3D models(like 5800 and 5890) one with 105 and the other with 125TDP reaching 5.0/5.1 ST?

Don't see the point in that.

5800X3D @ 105 TDP PPT 142
5890X3D @ 125 TDP PPT 142