Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

[izaic3]

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):

• Zen 3 "enhanced"
• DDR5 and LPDDR5 support
• PCIE4
• 6nm
• CVML
• Navi 2/RDNA/RDNA 2
• 12 CU
• Socket FP7/AM5
• Possible USB4
• 2021(unlikely) -2022 (more likely) release date

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.

 

[Abwx]

Nah, look at the Witcher 3 result. That's the only one which is actually GPU-limited, rest are all CPU-limited.

Then all other games are using 8C if it s the CPU making all that difference...

So you are saying that HP ProBook 635 Aero G8 with R7 5800U was actually set to 25W even though they stated in

Spoiler: footnotes

1. Based on testing by AMD as of 12/14/2021. CPU performance evaluated with a geomean of 9 multi-threaded content creation and CPU tests. GPU performance evaluated with 3DMark® Time Spy. System configuration for Ryzen™ 7 5800U CPU/GPU performance: HP ProBook 635 Aero G8 configured with 2x8GB DDR4-3200 (22-22-22), Windows® 11 22000.282, Samsung 980 Pro 1TB SSD, 15W nominal processor TDP, GPU driver 27.20.21026, BIOS T83. System configuration for Ryzen™ 7 6800U CPU/GPU performance: AMD reference motherboard configured with 4x4GB LPDDR5-6400 (19-15-17), Windows® 11 22000.282, Samsung 980 Pro 1TB SSD, 28W nominal processor TDP, GPU driver 30.0, BIOS TRM0081D. RMB-13

that It was only at 15W, right? OK, so where is your proof? I guess It was just too much of a bother searching for It.

Unlike you, I found a review of HP ProBook 635 Aero G8 with R7 5800U and yes, the score in CB R23 is 9576 points, which would mean It was set to 25W.
now the question is If they have a typo in their footnotes or they manually set TDP to 15W.

15W nominal doesnt mean that it s the maximum power, or does it..?.

As said if the difference is due to a 15W to 28W comparison, as you say, then it means that :

N6 bring 0% perf/watt improvement vs N7.

Zen3+ is exactly the same uarch as Zen3, without any efficency improvement and with the same MT IPC.

FTR IPC in ST has improved by 5.6% in CB R23 wich mean that it likely improved in MT, and i doubt that N6 bring no perf/watt improvement.

 

[TESKATLIPOKA]

15W nominal doesnt mean that it s the maximum power, or does it..?.

As said if the difference is due to a 15W to 28W comparison, as you say, then it means that :

N6 bring 0% perf/watt improvement vs N7.

Zen3+ is exactly the same uarch as Zen3, without any efficency improvement and with the same MT IPC.

FTR IPC in ST has improved by 5.6% in CB R23 wich mean that it likely improved in MT, and i doubt that N6 bring no perf/watt improvement.

They wrote for R7 6800U -> 28W nominal processor TDP. Do you think It had higher TDP set as R7 5800U?

Let's be honest, a 30% improvement in CB R23 is a lot, considering the core did see only minor improvement in IPC(~6%), so this performance would have to come mostly from higher frequency, but 6nm is not much better than 7nm process, yet It would allow ~20% higher frequency during CB? Very hard to believe.

 

[Hitman928]

Again, the footnotes from AMD themselves say:


How does the above somehow translate into internal testing with actual silicon? Especially when the performance footnotes with testing are significantly more specific.

With modern toolsets, engineering estimates have gotten pretty darn accurate most of the time. With that said, I agree that you should always wait for actual silicon results.

 

[Abwx]

Ok, but a 30% higher CB R23 score is a lot, considering the core did see only minor improvement in IPC and and 6nm is also not much better than 7nm process.
This would mean R7 6800U could have a score of ~12450 points in CB R23. That's more than what R9 5900HX achieved in Techspot review.

They surely played with :
7% better perf/watt of the process.
5% better MT IPC.
4% higher frequency thanks to 10% efficency improvement of the design itself.
3-4% higher frequency thanks to 12.5% higher TDP (28W vs 25W, wich provide theorically 5% but 1.25W of those 3W are consumed by the higher IPC).

That s 20% improvement with few means if we look at all the time that was devoted to this chip, after all it s still Zen3 based and at this time we would had rather expected Zen4 in a few months.

They did obviously put quite a lot of work in RMB, and i doubt that it was just to stick a 5800U with a RDNA2 GPU and a few other features.

 

[leoneazzurro]

They surely played with :
7% better perf/watt of the process.
5% better MT IPC.
4% higher frequency thanks to 10% efficency improvement of the design itself.
3-4% higher frequency thanks to 12.5% higher TDP (28W vs 25W, wich provide theorically 5% but 1.25W of those 3W are consumed by the higher IPC).

That s 20% improvement with few means if we look at all the time that was devoted to this chip, after all it s still Zen3 based and at this time we would had rather expected Zen4 in a few months.

They did obviously put quite a lot of work in RMB, and i doubt that it was just to stick a 5800U with a RDNA2 GPU and a few other features.

6800U also has much better base clocks vs 5800U, which may mean that it has better sustained clocks in multicore scenarios.

 

[TESKATLIPOKA]

6800U also has much better base clocks vs 5800U, which may mean that it has better sustained clocks in multicore scenarios.

I didn't notice the much higher base clockspeed. Thanks
R7 6800U 2.7GHz(base) -> 4.7GHz(boost)
R7 5800U 1.9GHz(base) -> 4.4GHz(boost)
That's ~42% higher base clockspeed.

 

[Abwx]

6800U also has much better base clocks vs 5800U, which may mean that it has better sustained clocks in multicore scenarios.

Methink that RMB clocks about 15% higher at same TDP, the numbers you posted seems clock comparisons at 15W for the 5800U and 28W for the 6800U, hence the big frequency difference.

Assuming this 5800U is actually at 1.9@15W then at 28W it would clock at something like 2.4GHz.

 

[TESKATLIPOKA]

Methink that RMB clocks about 15% higher at same TDP, the numbers you posted seems clock comparisons at 15W for the 5800U and 28W for the 6800U, hence the big frequency difference.

Assuming this 5800U is actually at 1.9@15W then at 28W it would clock at something like 2.4GHz.

45W Rembrandt H has pretty much the same base clock as 45W Cezanne H.

 

[xilli_fiberbit]

Another exciting aspect of Rembrandt is that the upcoming AMD mobile APUs will be the first microprocessors to feature a Microsoft Pluton security processor that was launched back in 2020. Back then, AMD had stated that "Future Ryzen Products" would be using Pluton and it looks like the feature will finally debut with Rembrandt.

 

[Abwx]

45W Rembrandt H has pretty much the same base clock as 45W Cezanne H.

And same base clock as the 35W part, so what is your conclusion..?.

 

[TESKATLIPOKA]

Let's wait for reviews.

 

[eek2121]

So you are saying that HP ProBook 635 Aero G8 with R7 5800U was actually set to 25W even though they stated in footnotes:


that It was only at 15W, right?
OK, so where is your proof? I guess It was just too much of a bother searching for It.

Unlike you, I found a review of HP ProBook 635 Aero G8 with R7 5800U and yes, the score in CB R23 is 9576 points, which would mean It was set to 25W.
Now the question is If they have a typo in their footnotes or they manually set TDP to 15W.

"nominal"

That does not mean the chips were running at 15W/28W.

 

[andermans]

I didn't notice the much higher base clockspeed. Thanks
R7 6800U 2.7GHz(base) -> 4.7GHz(boost)
R7 5800U 1.9GHz(base) -> 4.4GHz(boost)
That's ~42% higher base clockspeed.

I think that is likely because it looks like the default TDP has switched from 15W to 28W. So even though the cTDP ranges are pretty similar, I'm guessing the base clock is at that default TDP.

 

[leoneazzurro]

I think that is likely because it looks like the default TDP has switched from 15W to 28W. So even though the cTDP ranges are pretty similar, I'm guessing the base clock is at that default TDP.

In reality the cTDP changed from 10-25W for 5800U to 15-28W for the 6800U. So only a little higher.

 

[eek2121]

The 6980HS is an amazing chip. 35W, 3.3/5.0 GHz and the GPU clock is still 2.4 GHz with 12 CUs. Hopefully we will see more laptops centered around that chip.

EDIT: By comparison, the 5980HS was 35, had a 3.0/4.8 GHz boost, 8 CUs @ 2.1 GHz

Unsure why people are saying that AMD is being misleading, clearly there is improvement.

 

[andermans]

The 6980HS is an amazing chip. 35W, 3.3/5.0 GHz and the GPU clock is still 2.4 GHz with 12 CUs. Hopefully we will see more laptops centered around that chip.

EDIT: By comparison, the 5980HS was 35, had a 3.0/4.8 GHz boost, 8 CUs @ 2.1 GHz

Unsure why people are saying that AMD is being misleading, clearly there is improvement.

I think the misleading part is mostly the graph where they showed a +30% improvement 5800U->6800U by doing the 15W->28W trick. I think an improvement is reasonable, but that improvement is just not realistic as it is not like for like.

If we compare 5980HS->6980HS clock speeds that is a more reasonable (given what changed) improvement of 4-10% in clock speed.

(To clarify: it is totally ok to perform above expectations, it just sucks all the hype out if that is due to testing methodology and not the product)

 

[moinmoin]

I thought AMD learnt the lesson with cezanne and U models but I was wrong.

At least Barcelo is not part of the 6000 series so one can't accidentally get a last gen chip looking at the model number unlike with Lucienne.

 

[eek2121]

[Q


I think the misleading part is mostly the graph where they showed a +30% improvement 5800U->6800U by doing the 15W->28W trick. I think an improvement is reasonable, but that improvement is just not realistic as it is not like for like.

If we compare 5980HS->6980HS clock speeds that is a more reasonable (given what changed) improvement of 4-10% in clock speed.

What I've been trying to tell people is the use of the word "nominal" means that they may have NOT in fact making that comparison. They were listing the nominal TDP values for each CPU.

Anyways,

Zephyrus G14 specs: ROG Zephyrus G14 (2022) (asus.com)

Curious they list the boost clock at 4.9, probably a typo.

6980HS
Radeon 6800S
32GB DDR5 4800
16:10 Display (2560x1600) @ 120hz
1 TB PCIE 4.0 SSD
Wifi 6E
USB-C Charger (!!)
1.65 Kg (3.64 lbs)

We will have to see how the 6800S performs, but NOW we are getting somewhere with lightweight desktop replacements.

 

[lightmanek]

p

Is it an HP Pavilion Aero? Because that's one of the three or four laptops with the 5800U that is available in my country.

Yes, Pavilion Aero.

 

[tamz_msc]

Then all other games are using 8C if it s the CPU making all that difference...

CPU bound doesn't always mean core count bound. Have you actually played the games in question?

 

[uzzi38]

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

 

[The Hardcard]

What I've been trying to tell people is the use of the word "nominal" means that they may have NOT in fact making that comparison. They were listing the nominal TDP values for each CPU.

Anyways,

Zephyrus G14 specs: ROG Zephyrus G14 (2022) (asus.com)

Curious they list the boost clock at 4.9, probably a typo.

6980HS…

The machines in your link feature the 6900HS, hence the 4.9 boost clock.

 

[eek2121]

The machines in your link feature the 6900HS, hence the 4.9 boost clock.

Ah, too fast to post I guess

 

[Abwx]

CPU bound doesn't always mean core count bound. Have you actually played the games in question?

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

 

[gdansk]

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