Speculation: Ryzen 4000 series/Zen 3

[DrMrLordX]

I wonder if this will lead to extra efficiency or performance beyond HBM2.

HBM3 hasn't turned up anywhere yet (that I know of), but there's stuff like HBM2e. Should be cheaper/more efficienct.

 

[uzzi38]

There is no weirdness.

What happens is that in ALL PCIe 3.0 boards, the 5500XT works @ x8 PCIe 3.0 while with X570 boards it works @ x8 PCIe 4.0. The difference stems from the fact that boards with PCIe 3.0 have HALF the bandwidth of X570 boards, that have PCIe 4.0.

Effectively, when you place ANY PCIe 4.0 card in a non-PCIE 4.0 board, you're bottlenecking the card's true performance. It the case of the 5700(XT) card(s) you don't notice it much because of the VRAM present in the card but you may notice in SOME games if you use a 5500(XT) card.

You missed the point. No GPU with 5500XT levels of performance should be bottlenecking even Gen3 x8. Even a 2080Ti can only barely do so, and it only looses like 1-2% performance with x8.

 

[H T C]

You missed the point. No GPU with 5500XT levels of performance should be bottlenecking even Gen3 x8. Even a 2080Ti can only barely do so, and it only looses like 1-2% performance with x8.

But that card has WAY more VRAM, specially VS the 4GB 5500XT models so the need for PCIe communication in VRAM saturation cases doesn't happen as often and thus doesn't become relevant. Not so with the 5500XT 4GB models.

 

[uzzi38]

But that card has WAY more VRAM, specially VS the 4GB 5500XT models so the need for PCIe communication in VRAM saturation cases doesn't happen as often and thus doesn't become relevant. Not so with the 5500XT 4GB models.

And yet even the 8GB model of the 5500XT saw a difference in performance from the switch from Gen 3 to Gen 4.

 

[H T C]

And yet even the 8GB model of the 5500XT saw a difference in performance from the switch from Gen 3 to Gen 4.

Yes, but significantly smaller when compared to the 4GB models.

EDIT

This is off topic: there is a topic about this already so i suggest we continue the discussion there. Don't have the link, though

EDIT #2

Found it: https://forums.anandtech.com/thread...rippling-amds-rx-5500-xt-performance.2574515/

 

[uzzi38]

Yes, but significantly smaller when compared to the 4GB models.

EDIT

This is off topic: there is a topic about this already so i suggest we continue the discussion there. Don't have the link, though

But it shouldn't have seen one at all. See the problem?

In games where there was a noticable amount of CPU GPU communcation, it saw a difference even though it should have.

It's quite likely Navi 12 is aimed at workstations in an mGPU setup. Komachi doesn't think it's a mobile part, so considering the low clocks and HBM2 it's makes sense.

If that's what it's for, surely you can see how such an errata would need to be fixed, right?

 

[Veradun]

Nah, you can switch between the different modes in BIOS afaik. I think so anyway, I don't actually have one myself.

My Navi10 runs on a PCIe3 board with a PCIe3 CPU

 

[soresu]

HBM3 hasn't turned up anywhere yet (that I know of), but there's stuff like HBM2e. Should be cheaper/more efficienct.

I meant the fixed Navi 10 errata rather than future HBM derivatives/successors.

 

[Shivansps]

There is no weirdness.

What happens is that in ALL PCIe 3.0 boards, the 5500XT works @ x8 PCIe 3.0 while with X570 boards it works @ x8 PCIe 4.0. The difference stems from the fact that boards with PCIe 3.0 have HALF the bandwidth of X570 boards, that have PCIe 4.0.

Effectively, when you place ANY PCIe 4.0 card in a non-PCIE 4.0 board, you're bottlenecking the card's true performance. It the case of the 5700(XT) card(s) you don't notice it much because of the VRAM present in the card but you may notice in SOME games if you use a 5500(XT) card.

Well if thats the case them using older bios on AMD boards from before the PCI-E 4.0 was forcely removed by AMD is an option for 5500 users... but you will be limited to older agesa.

You missed the point. No GPU with 5500XT levels of performance should be bottlenecking even Gen3 x8. Even a 2080Ti can only barely do so, and it only looses like 1-2% performance with x8.

its about the VRAM, with 4GB you will run short these days, this means there are still textures loaded on system RAM that the driver need to move to VRAM and from VRAM ro RAM constanly... This is not gona be needed on a 8GB card. This may explain the rather large perf gap of 4GB to 8GB.

 

[Thunder 57]

AMD Zen 3 CPUs Allegedly Features 17% IPC Gain Over Zen 2 Architecture, Ryzen 4000 'Vermeer' CPUs to Feature Huge Performance Upgrade Over Ryzen 3000

AMD's next-generation Zen 3 architecture based Ryzen 4000 & EPYC Milan CPUs are rumored to feature a massive 17% IPC gain compared to Zen 2.

wccftech.com

17% IPC increase over Zen 2. They state 10-12% better integer performance and up to 50% better floating point. AVX 512?

 

[thigobr]

I really hope that's true! I have decided to skip Ryzen 3000 altogether due to 3950X delays and shortages but looking good for an improved 16 core Zen3 next year!

 

[burninatortech4]

I really hope that's true! I have decided to skip Ryzen 3000 altogether due to 3950X delays and shortages but looking good for an improved 16 core Zen3 next year!

I'm in the same boat. Can't justify a one generation upgrade. Waiting to see what the 4700x/4900x brings to the table late 2020.

 

[scannall]

I'm in the same boat. Can't justify a one generation upgrade. Waiting to see what the 4700x/4900x brings to the table late 2020.

This boat is getting pretty crowded! I couldn't justify upgrading my 2600X. Though I wanted to, just because.

 

[Elfear]

I couldn't wait and jumped from a 1700 to a 2700X to a 3900X. The past 7-8yrs were so boring on the CPU front that I couldn't justify spending money on an annual basis but the last couple years have really got me excited with the massive IPC and core count increases. I'll gladly jump to a 4000-series chip if IPC gains are really in the 15-17% range.

 

[DrMrLordX]

17% IPC increase over Zen 2. They state 10-12% better integer performance and up to 50% better floating point. AVX 512?

Maybe, maybe not. Correct me if I'm wrong, but Zen1/Zen+ were 2x128b FMAC and Zen2 is 2x256b FMAC, correct? Zen3 may just be 4x256b FMAC though I would think that would give more than +50% performance in fp. That would hint at AVX512 support, but that wouldn't absolutely be necessary.

 

[Saylick]

AMD Zen 3 CPUs Allegedly Features 17% IPC Gain Over Zen 2 Architecture, Ryzen 4000 'Vermeer' CPUs to Feature Huge Performance Upgrade Over Ryzen 3000

AMD's next-generation Zen 3 architecture based Ryzen 4000 & EPYC Milan CPUs are rumored to feature a massive 17% IPC gain compared to Zen 2.

wccftech.com

17% IPC increase over Zen 2. They state 10-12% better integer performance and up to 50% better floating point. AVX 512?

Or 50% more AVX2 FPUs.

 

[Thunder 57]

Or 50% more AVX2 FPUs.

I honestly don't expect AVX 512. If that leak is accurate though it may certainly be possible. I'd expect them to use the die space elsewhere though. All we can do is wait and see.

 

[dnavas]

I honestly don't expect AVX 512.

The API, sure, the vector width, not so much. Even after this rumor. Otherwise it wouldn't be "50%". "Up to 50%" sounds more like a third FMA. Interesting that they didn't aim for Cascade Lake parity, but not completely surprising for all the reasons quoted above.

 

[lobz]

I really hope that's true! I have decided to skip Ryzen 3000 altogether due to 3950X delays and shortages but looking good for an improved 16 core Zen3 next year!

It's almost sure that Ryzen 4000 will either come out @ holidays 2020 or 1H 2021.

 

[Thunder 57]

It's almost sure that Ryzen 4000 will either come out @ holidays 2020 or 1H 2021.

Almost certain or just what you expect? I'm thinking back to school (August) if they try to rush it, or more likely September/October.

 

[Thunder 57]

The API, sure, the vector width, not so much. Even after this rumor. Otherwise it wouldn't be "50%". "Up to 50%" sounds more like a third FMA. Interesting that they didn't aim for Cascade Lake parity, but not completely surprising for all the reasons quoted above.

So half-rate AVX 512 just like AVX with Zen 1? That sounds like a solid design choice.

 

[lobz]

Almost certain or just what you expect? I'm thinking back to school (August) if they try to rush it, or more likely September/October.

Zen 2 came out Q3 2019, consumer Zen 3 is practically impossible in Q3 2020, I'd say it's even improbable @ holidays 2020, but not impossible at least. There's also no reason for it, it's not like CML will be that huge of a threat on the desktop.

 

[moinmoin]

Maybe, maybe not. Correct me if I'm wrong, but Zen1/Zen+ were 2x128b FMAC and Zen2 is 2x256b FMAC, correct? Zen3 may just be 4x256b FMAC though I would think that would give more than +50% performance in fp. That would hint at AVX512 support, but that wouldn't absolutely be necessary.

Would be an odd intermediary step, but I could see Zen 3 with 3x256b FMAC and Zen 4 with 4x256b FMAC. The latter may need too much space to fit in 7nm+'s density improvement.

It's almost sure that Ryzen 4000 will either come out @ holidays 2020 or 1H 2021.

The pattern so far is a new Ryzen gen every 14 months.

 

[Thunder 57]

Zen 2 came out Q3 2019, consumer Zen 3 is practically impossible in Q3 2020, I'd say it's even improbable @ holidays 2020, but not impossible at least. There's also no reason for it, it's not like CML will be that huge of a threat on the desktop.

See what @moinmoin wrote. March '17. Refresh came April '18. Zen 2 came July '19. I expect the time between Zen+ and Zen 2 will be similar to that of Zen 2 and Zen 3. That puts it at September 2020. That's why I said August if they rush it for back to school sales, or more likely, September/October.

Also, I wouldn't be so sure that they "won't need it". The last time AMD got complacent was when they had K8 and got a bit lazy. We all know how badly that ended.

 

[lobz]

See what @moinmoin wrote. March '17. Refresh came April '18. Zen 2 came July '19. I expect the time between Zen+ and Zen 2 will be similar to that of Zen 2 and Zen 3. That puts it at September 2020. That's why I said August if they rush it for back to school sales, or more likely, September/October.

Also, I wouldn't be so sure that they "won't need it". The last time AMD got complacent was when they had K8 and got a bit lazy. We all know how badly that ended.

But you do realize, that the Ryzen 2000 series that came after 14 month, was not a new uarch, merely some optimizations plus a little bit better process? If I follow what @moinmoin said, Zen 3 would come 28 months after July 2019... of course it wont take that long, but I don't get you guys, maybe I'm dumb.