Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

izaic3 · Apr 2, 2021

Alright, so we've had some leaks so far. I don't know if any of it's been confirmed yet, as it's pretty early, but here is what I've surmised so far (massive grain of salt of course):

If if turns out to have RDNA 2 and 12 CU, I could see iGPU performance potentially almost doubling over Cezanne.

If I've made any mistakes or gotten anything wrong, please let me know. I'd also love to hear more knowledgeable people weigh in on their expectations.

tamz_msc · Jan 4, 2022

Abwx said:
So if it s 4 cores why such differences since both SKUs have comparable performances when using 4C..?.

Because lighter games like CS:GO and Dota 2 stress only a few cores, meaning that within given TDP the cores can boost much higher, giving better performance?

Also, the MX450 GDDR5 is like 2x faster than the Iris Xe 96 EU when the latter is paired with DDR4.The heights of the bars on the slide are way off.

Abwx · Jan 4, 2022

tamz_msc said:
Because lighter games like CS:GO and Dota 2 stress only a few cores, meaning that within given TDP the cores can boost much higher, giving better performance?

Also, the MX450 GDDR5 is like 2x faster than the Iris Xe 96 EU when the latter is paired with DDR4.The heights of the bars on the slide are way off.

So there s enough cores on a 4C CPU, and since perfs are vastly different this cant be due to CPUs that perform the same given that they are not TDP limited to 15W, only remaining explanation is the GPU and BW, MX450 has a 64b bus with GDDR6 but RMB has a 128b bus.

Abwx · Jan 4, 2022

gdansk said:
How many TFLOPs is RMB iGP at peak then? 3.6ish?

2 flops/SP/cycle.

nicalandia · Jan 4, 2022

double post

inf64 · Jan 4, 2022

I doubt there is any significant IPC increase. 6800U has 4.7Ghz vs 4.4Ghz (5800U) for ST boost , that's ~7% higher clock for Cinebench 1T
6800U has 42% higher base clock as well so there is your 30% higher MT performance - it holds the MT clocks much longer at rated TDP.

It's going to be an awesome product and a worthy competitor to intel's top of the line H stuff. 16 great performing power efficient threads versus the 20T hybrid "abomination" . I know what I would pick. On top o that you get great iGPU that supports the new super resolution and on-demand eco mode when on battery, superb engineering.

leoneazzurro · Jan 4, 2022

Rembrandt is called "Zen3+" but it's probable that the improvements went in the power savings department, memory controller and sustained clocks@ iso power instead than a straight up IPC increase due to core improvements. But we will have surely more details when the products will actually launch.

nicalandia · Jan 4, 2022

inf64 said:
I doubt there is any significant IPC increase. 6800U has 4.7Ghz vs 4.4Ghz (5800U) for ST boost , that's ~7% higher clock for Cinebench 1T

There is a clear and substantial IPC increase even with half of the MB less than the Desktop 5800

Abwx · Jan 4, 2022

uzzi38 said:
There's your 15W numbers with footnotes, now stop complaining

View attachment 55457

View attachment 55458

So i was spot on in my post above, at 16% better perf at strictly same TDP.

With thoses numbers Intel s mobile ADL 6P + 8E advantage over previous Zen 3 gen is nullified.

Intel Announces 12th Gen Core Alder Lake: 22 New Desktop-S CPUs, 8 New Laptop-H CPUs

www.anandtech.com

Abwx · Jan 4, 2022

inf64 said:
I doubt there is any significant IPC increase. 6800U has 4.7Ghz vs 4.4Ghz (5800U) for ST boost , that's ~7% higher clock for Cinebench 1T
6800U has 42% higher base clock as well so there is your 30% higher MT performance - it holds the MT clocks much longer at rated TDP.

It's going to be an awesome product and a worthy competitor to intel's top of the line H stuff. 16 great performing power efficient threads versus the 20T hybrid "abomination" . I know what I would pick. On top o that you get great iGPU that supports the new super resolution and on-demand eco mode when on battery, superb engineering.

They state 11% better ST perf in CB R23, so that s not entirely due to frequency.

insertcarehere · Jan 4, 2022

nicalandia said:
There is a clear and substantial IPC increase even with half of the MB less than the Desktop 5800

View attachment 55468

A clear and substantial IPC increase...which AMD has not brought up, once, in their entire slide deck about Rembrandt, it's pretty clear how significant AMD considers the IPC increase to be....

nicalandia · Jan 4, 2022

insertcarehere said:
A clear and substantial IPC increase...which AMD has not brought up, once, in their entire slide deck about Rembrandt, it's pretty clear how significant AMD considers the IPC increase to be....

Who Cares? The numbers are there, just because they don't mentioned it. It does not mean it's not there.

Numbers don't lie, the 6800U Zen3+ even with half of the $ it's matches the full fat 32Mb of L3$

insertcarehere · Jan 4, 2022

nicalandia said:
Who Cares? The numbers are there, just because they don't mentioned it. It does not mean it's not there.

Numbers don't lie, the 6800U Zen3+ even with half of the $ it's ahead of the full fat 32Mb of L3$

View attachment 55469
View attachment 55470

A single acreenshot of an unknown benchmark for an unreleased processor still under NDA is your entire basis for a substantial increase in IPC?

Abwx · Jan 4, 2022

insertcarehere said:
A single acreenshot of an unknown benchmark for an unreleased processor still under NDA is your entire basis for a substantial increase in IPC?

Explain the ST perf difference in CBR23 posted above your post, or is CB an unknown benchmark.?

Btw, is it official enough.?.

nicalandia · Jan 4, 2022

insertcarehere said:
A single acreenshot of an unknown benchmark for an unreleased processor still under NDA is your entire basis for a substantial increase in IPC?

It's Cinebench R23 from CPU Monkey database.

Unless the 6800U is boosting to 4.9 Ghz there is no way a 4.7 Ghz Zen3 would have 11+ higher ST performance

Heck they even have the 12900KS CBR23 ST Score

Cinebench R23 (Single-Core) Lista de pruebas de CPU

www.cpu-monkey.com

Joe NYC · Jan 4, 2022

insertcarehere said:
Frankly, the greater concern is the 6800u Laptop's use of very aggressive LPDDR5 modules, which is 99% not going to translate to laptop OEMs, and thus put a damper to claimed gpu gains.

We will see what happens. From my casual reading on the subject, 6400 may end up being the most common speed for LPDDR5.

Hitman928 · Jan 4, 2022

nicalandia said:
Who Cares? The numbers are there, just because they don't mentioned it. It does not mean it's not there.

Numbers don't lie, the 6800U Zen3+ even with half of the $ it's matches the full fat 32Mb of L3$

View attachment 55472

Cinebench doesn't care about the extra L3 cache much.

According to AMD, the 6800u is 11% faster than the 5800u in Cinebench r23. The 6800u also has 6.8% faster single core boost clock. That comes out to ~4% more IPC for the 6800u but it could also potentially be influenced by a more sustained boost clock on the 6800u. We really don't know until we get some actual reviews.

insertcarehere · Jan 4, 2022

nicalandia said:
It's Cinebench R23 from CPU Monkey database.

Unless the 6800U is boosting to 4.9 Ghz there is no way a 4.7 Ghz Zen3 would have 11+ higher ST performance

Heck they even have the 12900KS CBR23 ST Score

Cinebench R23 (Single-Core) Lista de pruebas de CPU

Cinebench R23 (Single-Core) Lista de pruebas de CPU

www.cpu-monkey.com

And how does CPU monkey compile all these scores for unreleased CPUs? What's their source?

Joe NYC said:
We will see what happens. From my casual reading on the subject, 6400 may end up being the most common speed for LPDDR5.

Possibly, but my eye was more drawn to this in the footnotes for the test configuration:

AMD reference motherboard configured with 4x4GB LPDDR5-6400 (19-15-17),

Those seem like incredibly tight timings for anything LPDDR5 memory related...

inf64 · Jan 4, 2022

nicalandia said:
It's Cinebench R23 from CPU Monkey database.

Unless the 6800U is boosting to 4.9 Ghz there is no way a 4.7 Ghz Zen3 would have 11+ higher ST performance

Heck they even have the 12900KS CBR23 ST Score

Cinebench R23 (Single-Core) Lista de pruebas de CPU

Cinebench R23 (Single-Core) Lista de pruebas de CPU

www.cpu-monkey.com

It could be a better memory subsystem (DDR5), who knows. 6800U has much better score in Cinebench than 5980HS - around 8.5% higher at the same clock. That's a big difference to be honest. If I'd had to guess, it's something like sub 3% IPC ( less than Zen->Zen+) and the rest is faster memory access/tweaked memory controller (faster and more aggressive prefetching).

Abwx · Jan 4, 2022

Hitman928 said:
Cinebench doesn't care about the extra L3 cache much.

According to AMD, the 6800u is 11% faster than the 5800u in Cinebench r23. The 6800u also has 6.8% faster single core boost clock. That comes out to ~4% more IPC for the 6800u but it could also potentially be influenced by a more sustained boost clock on the 6800u. We really don't know until we get some actual reviews.

You think that AMD labs did use a very poor cooler on a third rate 5800U to downplay its scores.?.
So bad that it didnt even manage to dissipate the power in ST mode..?.

Edit :

Likely that they are comparing to the numbers they published when the 5800U was released.

majord · Jan 4, 2022

mikk said:
5800U DDR4-3200 22-22-22 vs 6800U LPDDR5-6400 19-15-17
5800U 15W vs 6800 28W
Zero about the Iris Xe device, probably DDR4

https://www.amd.com/en/products/ryzen-processors-laptop

But it's AMD, it's fine. If Intel does this the forum explodes.

Those aren't footnotes for the presentation though, just the statements on that product webpage , none of which mention Intel. So why would there be a system config for it?

The 25w vs 28w is scummy - just trying to hit that nice 2x figure for the wow factor.

Joe NYC · Jan 4, 2022

insertcarehere said:
Possibly, but my eye was more drawn to this in the footnotes for the test configuration:

Those seem like incredibly tight timings for anything LPDDR5 memory related...

I see, you were referring to timings, not the speed.

The favorable timings should help CPU more, high speed should help GPU more. So if timings are aggressive then the CPU performance could drop a notch, and if memory speed is common, than GPU performance should be on par in OEM laptops.

nicalandia · Jan 4, 2022

Hitman928 said:
Cinebench doesn't care about the extra L3 cache much.

Oh it does, look at this. That's nearly 5% IPC boost right there

IntelUser2000 · Jan 4, 2022

nicalandia said:
Oh it does, look at this. That's nearly 5% IPC boost right there

View attachment 55473

That site is CPUMonkey. They get results from prototype and other systems and don't do the tests on their own.

The slight differences such as above can be explained by different configurations. It's just to give you a ballpark idea and no more.

nicalandia · Jan 4, 2022

IntelUser2000 said:
That site is CPUMonkey. They get results from prototype and other systems and don't do the tests on their own.

The slight differences such as above can be explained by different configurations. It's just to give you a ballpark idea and no more.

I will get them from Anandtech then

andermans · Jan 4, 2022

inf64 said:
It could be a better memory subsystem (DDR5), who knows. 6800U has much better score in Cinebench than 5980HS - around 8.5% higher at the same clock. That's a big difference to be honest. If I'd had to guess, it's something like sub 3% IPC ( less than Zen->Zen+) and the rest is faster memory access/tweaked memory controller (faster and more aggressive prefetching).

Cinebench mostly doesn't care about the memory system since it does very well using decent caches. As such any changes in IPC are more likely to be core changes somehow.

Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

Member

Diamond Member

Lifer

Lifer

Diamond Member

Diamond Member

Senior member

Diamond Member

Lifer

Lifer

Senior member

Diamond Member

Senior member

Lifer

Diamond Member

Platinum Member

Diamond Member

Senior member

Diamond Member

Lifer

Senior member

Platinum Member

Diamond Member

Elite Member

Diamond Member

Member