Speculation: Ryzen 4000 series/Zen 3

[Ajay]

What this really shows is that it only takes one idiot to ruin decades of work. Easy to destroy, difficult to build is the lesson. Strangely enough, Intel is facing the same one fool nightmare.

IDK wrt Intel. Seems like Intel has lost it's touch since Grove retired. Krzanich just brought all the incompetence to a head (well, plus adding his own special stupid sauce). I guess Barrett did a pretty good job on the process side.

 

[moinmoin]

Strangely enough, Intel is facing the same one fool nightmare.

And Intel really didn't waste time getting one such on board after AMD had one. AMD is quite lucky in that regard.

 

[Panino Manino]

I just don't see a lot of room for improvement for AMD's APUs on the desktop. I think it was D8auer that produced the youtube video that demonstrated a whole slew of overclocking results for a 2400G with different memory sticks at different timings. When pushed from 1500 to 1600 Mhz, he saw very little performance increase. When pushing the RAM past 3200 to 3400 and trying to tighten the timings, he was also seeing very little improvement. For the APU in Renoir to exceed the TFlops in Vega 11 at 1.6Ghz, it needs to be North of 2.0Ghz. So, even if we manage to see the desktop APUs pushing DDR4 at 4266 on two channels, at just 2.1Ghz, its only really addressing the VRAM bandwidth starvation problem by roughly 30% over VEGA 11 with 3200 DDR4 and not making any significant gains on actual iGPU core video computation throughput.

I'm going to be quite shocked if desktop Renoir exceeds Picasso performance by more than 20% when the performance limit is entirely iGPU related. On titles that are CPU starved, or that were hitting thermal/power walls on Picasso, Renoir should show significant improvements.

Renoir don't need to match Picasso in TF, Navi performs better than Vega anyway.
The Navi IGP just needs more memory bandwidth. With DDR5 can't expect a good improvement?

 

[maddie]

Well, i dont think anybody considers Vega at 2.1Ghz stock on an APU as not suprising. When we discussed this a few months ago we were talking about 1.7Ghz with 2.0Ghz being possible with OC. Now we are talking about 2.5Ghz as possible with OC, how this is not a suprise to you?

And im sorry, but wharever AMD did to Vega they did not announce it, the only thing we know is that they clock higher, any sustancial change to the arch would have caused at minimum a name change to "Vega 2" or something to draw attention to it, not sweep it under the rub.

Where am I saying 'substantial change to the architecture'? Don't you understand that sometimes a single critical pathway can undermine everything else? Maybe the clock signals weren't propagating correctly to all areas preventing the ramping of frequency, maybe whatever.

Very narrow thinking in my opinion.

 

[RetroZombie]

Hector is still available.

Do you think he might join raja koduri and jim keller at intel ?

 

[Thibsie]

AMD is now a world class semiconductor manufacturer again (albeit, a bit on the small side compared to rivals). I think their odds of shooting themselves in the foot have greatly diminished compared to the past.

I agree with you. It's just that people will stay cautious given AMD's history.

 

[Valantar]

Renoir don't need to match Picasso in TF, Navi performs better than Vega anyway.
The Navi IGP just needs more memory bandwidth. With DDR5 can't expect a good improvement?

Renoir is Vega, not Navi.

 

[DrMrLordX]

Do you think he might join raja koduri and jim keller at intel ?

Intel has likely learned that Hector is more useful to them working for a competitor.

 

[Shivansps]

I just don't see a lot of room for improvement for AMD's APUs on the desktop. I think it was D8auer that produced the youtube video that demonstrated a whole slew of overclocking results for a 2400G with different memory sticks at different timings. When pushed from 1500 to 1600 Mhz, he saw very little performance increase. When pushing the RAM past 3200 to 3400 and trying to tighten the timings, he was also seeing very little improvement. For the APU in Renoir to exceed the TFlops in Vega 11 at 1.6Ghz, it needs to be North of 2.0Ghz. So, even if we manage to see the desktop APUs pushing DDR4 at 4266 on two channels, at just 2.1Ghz, its only really addressing the VRAM bandwidth starvation problem by roughly 30% over VEGA 11 with 3200 DDR4 and not making any significant gains on actual iGPU core video computation throughput.

I'm going to be quite shocked if desktop Renoir exceeds Picasso performance by more than 20% when the performance limit is entirely iGPU related. On titles that are CPU starved, or that were hitting thermal/power walls on Picasso, Renoir should show significant improvements.

Remember you still need to compare Picassos to similar products, Vega 8@2.1Ghz is, most likely, a Ryzen 7 product. Probably the 4700G pictured here due to the code number:

AMD Ryzen 7 4700G "Renoir" Desktop Processor Pictured

Here is the first picture of the AMD Ryzen 7 4700G, the company's upcoming socket AM4 APU based on the 7 nm "Renoir" silicon, courtesy of VideoCardz. The picture reveals a standard-looking socket AM4 chip with commercial name and OPN markings (100-000000146), matching the Igor's Lab OPN code...

www.techpowerup.com

With one of the Ryzen 7 revealed, the 4400G could be the #143 Vega 7 @ 1.9Ghz... it should perform better than Vega 11@1.4Ghz, but not sure by how much.

 

[NostaSeronx]

On the next IOD... it might be on sub-12FDX. Which isn't so far a full node shrink.

12FDX => 84Cx/56Mx
Sub-12FDX => 64Cx/48Mx {Numbers from secret project}


{Date of slides and location of venue: 2 October 2019, Dresden}

Which I believe is going to be 10nm node in the second part and the mislabeled sub-10nm node in the first part. A lot of stuff is weird in the first part.

Q3 2022 is when "10FDX" is expected, so if it comes out sooner GloFo will be taking Malta's advanced node badge and giving it to Dresden. By the way, 64Cx is the critical density in which FDSOI becomes faster than any smaller than 64Cx FinFET by a stable slope.

 

[RetroZombie]

Intel has likely learned that Hector is more useful to them working for a competitor.

But he already have a good resume in selling fabs that don't work properly, he could help them doing a fire sale to some asian or arab country so they can get rid of those beautiful 10nm facilities.

 

[DrMrLordX]

But he already have a good resume in selling fabs that don't work properly, he could help them doing a fire sale to some asian or arab country so they can get rid of those beautiful 10nm facilities.

Hmmmm. You might be on to something.

 

[uzzi38]

Remember you still need to compare Picassos to similar products, Vega 8@2.1Ghz is, most likely, a Ryzen 7 product. Probably the 4700G pictured here due to the code number:

AMD Ryzen 7 4700G "Renoir" Desktop Processor Pictured

Here is the first picture of the AMD Ryzen 7 4700G, the company's upcoming socket AM4 APU based on the 7 nm "Renoir" silicon, courtesy of VideoCardz. The picture reveals a standard-looking socket AM4 chip with commercial name and OPN markings (100-000000146), matching the Igor's Lab OPN code...

www.techpowerup.com

With one of the Ryzen 7 revealed, the 4400G could be the #143 Vega 7 @ 1.9Ghz... it should perform better than Vega 11@1.4Ghz, but not sure by how much.

At stock I'd imagine about on par. You'll be able to overclock these I'm certain though, I wouldn't be surprised if a 2.3GHz OC on the iGPU's is standard for Renoir.

 

[Ajay]

One of the fast hexacores would be awesome for my wife, if she lets me rebuild her ancient system (well, depending on CPU cost).

 

[Shivansps]

At stock I'd imagine about on par. You'll be able to overclock these I'm certain though, I wouldn't be surprised if a 2.3GHz OC on the iGPU's is standard for Renoir.

Thats for sure, Vega 11 can get to 1750mhz max, while these Renoirs may be able to hit 2.5ghz. Another advantage for Renoir is memory support, getting anything over DDR4-3200 to work with Picasso for more than a few benchmarks is really difficult, in contrast these Renoirs may be able to hit DDR4-4266 on 1:1 fabric... not even Matisse can do that.

Still, it would need to be investigated, currently Vega IGP overclock is limited by the fact that power delivery is tied up to SoC memory controller, and motherboard vSoC VRMs... even really good motherboards dont invest too much into vSoC power delivery, from my expenrience overcloking Raven and Picassos, no matter if its a good motherboard you will always end up overloading the vSoC power phase OR with a vSoC that is so high that it may kill the mem controller. When i had a 3200G on my Gigabyte B350 Gaming overcloking Vega 8 anywere past 1500mhz whiout touching the voltage already made the vSoC vrm to go 120°C and thermal toggle, and every Gigabyte motherboard without a "top" vrm heat spreader are likely to share this issue.

Not sure how Renoir will handle that and im not really sure how much you can push vSoC in Renoir. In fact the VRM issue on Picasso is already bad enoght that it may be a valid reason not to allow 300/400 series support for Renoir APUs.

 

[Valantar]

I'm hoping there will be some nice ITX B550 boards with 3 or more phase SoC VRMs with some decent power phases. Or just two doubled ones I guess. You don't need fancy cooling if the load is spread across more hardware after all.

As for frequencies, I think 2.5GHz sounds way too optimistic. Even if this ships at 2.1, that's a really, really high clock for a GPU, and hitting 2.5 would be borderline miraculous. 2.3 sounds likely as an OC if they ship at 2.1 though.

Also, for those of you trying to compare FLOPS and such, remember that this will carry over the decoupled Fclock of mobile Renoir, so it should be able to run DDR4->4000 if you are willing to pay for it or spend some hours tweaking some E-die. That would be a significant boost over 3200 on Picasso (yes, I've seen people hit 3400 on those, but it's rare and rarely stable over time).

 

[DisEnchantment]

Well, some new OPNs showed up for Vermeer/Ryzen4000
Slight clock bumps, 200 MHz Base while Boost remain the same which if AMD learnt from its fiasco with Zen2 should be a more sustainable boost.

Name: Vermeer (VMR)
Family: 19h
Models: 20h-2Fh
CPUID: 0xa20f00

OPN 1: 100-000000063-07_46/40_N
OPN 2: 100-000000063-08_46/40_Y
OPN 3: 100-000000063-23_44/38_N
Revision: A0
Cores: 8
Threads: 16

OPN 1: 100-000000059-14_46/37_Y
OPN 2: 100-000000059-15_46/37_N
Revision: A0
Cores: 16
Threads: 32

BTW these are A0 samples, things could still change(improve?)

New details on Ryzen 4000 “Renoir” desktop APUs and Ryzen “Vermeer” CPUs have appeared? The image is getting denser | igor'sLAB

Let's start with the new "Renoir" desktop APUs, about which I had already reported in detail in the article " All technical data of AMD's new "Renoir" desktop APU portfolio for AM4 at a glance". Here…

www.igorslab.de

One interesting observation by Igor

As far as the 500 chipset and backward compatibility of the Ryzen 4000 series is concerned, some OEMs are arguing with regard to the Renior APUs with details like the increased FCLK, which would at least make stable operation on certain older boards more difficult. For the Ryzen 4000 “Vermeer” CPUs with the changed architecture, it is said that it is not so much the chipset itself but the current motherboard layout that is the reason why the upcoming CPUs will not be able to run on the old hardware in this way. Nobody could (or wanted to) say whether there have been any major changes in the pin assignment, the general external circuitry or the new power supply specifications.

This could impact latency positively. (aside of the downside of the lack of BC for non X570 boards)

 

[DrMrLordX]

if AMD learnt from its fiasco with Zen2 should be a more sustainable boost.

That is my suspicion. Not much extra max boost, but higher sustainable MT clocks.

 

[inf64]

It's going to be a beautiful autumn for PC enthusiasts, a real generational jumps with new CPU and GPU parts from AMD. I'm ready and waiting with x570 and 1600AF to tide me over

PS Those base/boost clocks seem to be near final, that's what I heard.

 

[uzzi38]

Slight clock bumps, 200 MHz Base while Boost remain the same which if AMD learnt from its fiasco with Zen2 should be a more sustainable boost.

I mean, they certainly did with Renoir. Renoir can actually peak above the rated boost clocks by 100mhz on all the R7 and up skus. I'm really hoping they do the same for Vermeer

 

[Gideon]

It's going to be a beautiful autumn for PC enthusiasts, a real generational jumps with new CPU and GPU parts from AMD. I'm ready and waiting with x570 and 1600AF to tide me over

PS Those base/boost clocks seem to be near final, that's what I heard.

Probably, it's A0 silicon (and Renoir for instance shipped with A1). They might get maybe 100 Mhz more out of boost clocks with A1, but I wouldn't count on it. Doesn't really matter as well. If the rumored IPC gains are true and the clocks are actually sustainably reachable (unlike top-end zen2) it's still considerable upgrade.

 

[Gideon]

Slight clock bumps, 200 MHz Base while Boost remain the same which if AMD learnt from its fiasco with Zen2 should be a more sustainable boost.

The base clock increase should really improve MT results when coupled with the IPC increase.

I mean, they certainly did with Renoir. Renoir can actually peak above the rated boost clocks by 100mhz on all the R7 and up skus. I'm really hoping they do the same for Vermeer

I hope so too. IMO AMD should bring back the XFR moniker from Zen 1 era (+100 Mhz opportunitic "bonus" non-guaranteed boost). It was quite easy to understand and also useful for describing AMDs boost behaviour.

Those Renoir FCLK improvements make me hopeful of better memory latency as well.

 

[uzzi38]

If the rumored IPC gains are true and the clocks are actually sustainably reachable (unlike top-end zen2) it's still considerable upgrade.

If both are true we get another generation of 6 core chips performing like 8 core chips in multi-core workloads.

Maybe more if the rumours undershoot like the Samsung cat suggested.

Thank you AMD.

 

[PhonakV30]

4.6Ghz , sustained boost speed ?

 

[Asterox]

4.6Ghz , sustained boost speed ?

Well, it is very logical 4.6ghz is singlecore boost no doubt.

In reality you dont need higher CPU frequencies. Or simple, if IPC goes up it is much harder to hit higher CPU frequencies.

5ghz matra or hunt for 5ghz, it is a thing of the past.

All core turbo, with 8/16 Zen 3 you can expect 4.3ghz with no problem.

Zen 3

- 15% higher IPC

- 8 Core CCX=much lower Latency