Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

Page 64 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

izaic3

Member
Nov 19, 2019
61
96
91
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.
 
Last edited:

Heartbreaker

Diamond Member
Apr 3, 2006
4,228
5,228
136
A direct comparision with Intel.


Results are just what everyone could expect.

Except not that direct. It's only the Intel 4P core version, not the 6P core version, and it's running DDR4 3200 on the Intel and DDR5 6400 on the AMD.

So a ways off from a real direct comparison.

But really it's an embarrassment of riches with CPUs today. Both Intel and AMD (and Apple) have great CPU cores. AMD still owns Intel on GPU.
 

ryanjagtap

Member
Sep 25, 2021
108
127
96
Just a thought/question.
Will there be any further optimizations for the RDNA2 igpus on the silicon level? As we can see from several of the videos, the igpu needs at least 30W for maximum performance while gaming which is not possible to sustain on an APU of 35-45W (Can they lower the power needed to reach and sustain such clocks on either the igpu or the core?). But from what was shown on AMD FAD, Phoenix point and onwards APUs will be using RDNA3 and RDNA3+. The RDNA2 igpu in Raphael is also for troubleshooting purposes, so will RDNA2 igpu will be only for these generations? It's feels very different after being on Vega igpu for such a long time. [Well there is Mendocino, but it is also rumored to have 2 RDNA2 CU as on the Raphael IO die]
 

Thibsie

Senior member
Apr 25, 2017
749
801
136
love these reviews and rembrandt performance is excellent but it's a mute point because 6800U's are virtuslly unavailable and may not be for under $1,000 anytime soon if ever.

No. It isn't moot. It allows you to mesure how it US better than 5800U and decide if you wait or not.
In any case, Intel is a big NO in a laptop for now. No way.

Btw, usual laptop offerings with 5800U are drying up, whjich means 6800U isn't that far.
 

Kaluan

Senior member
Jan 4, 2022
500
1,071
96
Except not that direct. It's only the Intel 4P core version, not the 6P core version, and it's running DDR4 3200 on the Intel and DDR5 6400 on the AMD.

So a ways off from a real direct comparison.

But really it's an embarrassment of riches with CPUs today. Both Intel and AMD (and Apple) have great CPU cores. AMD still owns Intel on GPU.
I'd argue the 2 chips are not part of the same class of chips to begin with...
1. 6000U (15-28W) and Alder Lake-P (28-64W) are not in the same power class/bracket, Alder Lake-U (2P cores, 9W-29W/15-55W) would be a better fit versus 6000U if you ask me (but there are no designs featuring these chips yet lol). 6000HS seem like a much better fit to ADL-P. Just as 6000H and HX are a better comparison fit vs Alder Lake-H. Both 6000U and 12xxU seem to target ultra portables, 12xxP is part of a more odd bracket IMHO, more like a successor to 113xxH Tiger Lake "H35" chips rather than 11xxG "UP3" ones.
2. There are no i7-1280P laptop design out AFAIK anyway, so even if they wanted to test that for whatever reason, HU couldn't get such a thing to bench. And at any rate, HU test at fixed TDP... which let me remind people, is something that does not bode well for Alder Lake, so 6 or 4 Performance cores wouldn't make much of a difference when TDP is 28W. It wouldn't surprise me if 1280P would perform worse than 1260P and 1270P in some tasks at fixed TDP. As Intel mobile SKUs sometimes do (1165G7 being sometimes faster than 1185G7, 12700H being sometimes faster than 12900HK etc)

The DDR4 bit is fair I suppose, but other than iGPU gaming (where 28W 1260P+DDR4 seems to get bested by 28W 1165G7+DDR4 lol), I don't really see all that much relevance. It's not like DDR5 12900K was magically a lot faster than DDR4 12900K/5950X either.
But I do like apples to apples tests myself (where they actually make sense).


That being said, I'd love to see a Ryzen 6000 (whatever 8C model) tested with both DDR5-4800 and LPDDR5-6400, both productivity and graphics/gaming. For gaming I think there was one of the early iGPU reviews that showed relatively minor gains going to LPDDR5-6400. But I'm more curious about latency sensitive productivity tasks and LPDDR5 timings and such, how Zen3+ handles them versus the typical DDR5-4800 kits.
 
Last edited:
  • Like
Reactions: Tlh97

Mopetar

Diamond Member
Jan 31, 2011
7,842
5,994
136
More, way more. Including handling video codecs...

I'm assuming that they'll include at least a minimal set of hardware decoding/encoding blocks, but has it been confirmed? With Navi 24 they skimped on that front more than some people liked. The IO die probably needs to be a certain size just for all of the physical interfaces it needs to have, but it's doubtful they'd make the chip any bigger than it needs to be as 6nm is not cheap.
 

ahimsa42

Senior member
Jul 16, 2016
225
194
116
No. It isn't moot. It allows you to mesure how it US better than 5800U and decide if you wait or not.
In any case, Intel is a big NO in a laptop for now. No way.

Btw, usual laptop offerings with 5800U are drying up, whjich means 6800U isn't that far.
drying up? they never had widespread availability in the first place-at least not in the US. my point was that it doesn't really matter how much better they are if most people cannot even buy them.
 

moinmoin

Diamond Member
Jun 1, 2017
4,952
7,666
136
so will RDNA2 igpu will be only for these generations? It's feels very different after being on Vega igpu for such a long time.
This may be AMD with its new riches, capable of exchanging blocks every gen instead only every couple gens.

I'm assuming that they'll include at least a minimal set of hardware decoding/encoding blocks, but has it been confirmed?
Not confirmed, but Mendocino appears to have the complete media block (including AV1 etc.) and guesses are that the same iGPU block is being reused on the IOD (since both are on N6). And better use in mobile as part of Dragon Range, as jpiniero mentioned.
 

JasonLD

Senior member
Aug 22, 2017
485
445
136
There are no i7-1280P laptop design out AFAIK anyway, so even if they wanted to test that for whatever reason, HU couldn't get such a thing to bench.

Well.....There are quite a few laptops out there with 1280p already.


 

Thibsie

Senior member
Apr 25, 2017
749
801
136
I'm assuming that they'll include at least a minimal set of hardware decoding/encoding blocks, but has it been confirmed? With Navi 24 they skimped on that front more than some people liked. The IO die probably needs to be a certain size just for all of the physical interfaces it needs to have, but it's doubtful they'd make the chip any bigger than it needs to be as 6nm is not cheap.

That was on an GPU supposed to be used with an APU in a laptop. Not that stupid. Using this on a PCIe though...
 

ahimsa42

Senior member
Jul 16, 2016
225
194
116
That never stopped anyone on testing GPUs although the situation was even worse.
correct but the gpu situation was due to crypto & companies were churning them out but could not come close to meeting the demand because miners across the globe were snatching them up far quicker than they could be produced.

when it comes to laptop APU's though, it seems that AMD relagates the average consumer to an afterthought at best. the vexxing part is that AMD flat out said that 6800's would be available from many vendors. this may still turn out to be true but now they are already hyping phoenix so if Dec comes along and rembrndt ends up being like zen3 cezzane as far as availability, i think many people are not going to bother with them.
 

Mopetar

Diamond Member
Jan 31, 2011
7,842
5,994
136
That was on an GPU supposed to be used with an APU in a laptop. Not that stupid. Using this on a PCIe though...

That's all well and good but they sold it as a stand alone GPU all the same.

It's not a massive deal for most consumers. The CPU can always handle a software implementation and an 8+C CPU won't suffer for doing so.

It's really just a matter of battery life since dedicated hardware will always be able to handle a job at less power. If you're plugged in it hardly matters, but on battery it can.
 

trivik12

Senior member
Jan 26, 2006
260
217
116
Well.....There are quite a few laptops out there with 1280p already.


Those are some crazy expensive laptops. But MSI with 32GB ram is $1300 and available at Costco with 2 year warranty(plus another 2 if you buy with Costco credit card).

 

mikk

Diamond Member
May 15, 2012
4,141
2,154
136
I'd argue the 2 chips are not part of the same class of chips to begin with...
1. 6000U (15-28W) and Alder Lake-P (28-64W) are not in the same power class/bracket, Alder Lake-U (2P cores, 9W-29W/15-55W) would be a better fit versus 6000U if you ask me (but there are no designs featuring these chips yet lol). 6000HS seem like a much better fit to ADL-P. Just as 6000H and HX are a better comparison fit vs Alder Lake-H. Both 6000U and 12xxU seem to target ultra portables, 12xxP is part of a more odd bracket IMHO, more like a successor to 113xxH Tiger Lake "H35" chips rather than 11xxG "UP3" ones.


It depends on the device and your numbers are inaccurate (short burst Intel vs sustained power AMD). Most 1260p devices are 30W sustained or lower. Alder Lake-U class is cheaper than 6800U, it's certainly not the same class. 1260p is targeted for ultra portables just as 6800U. Go to Dell, Lenovo or Asus and check it out. Unlike Rembrandt-U there are lots of devices listed or out.

2. There are no i7-1280P laptop design out AFAIK anyway, so even if they wanted to test that for whatever reason, HU couldn't get such a thing to bench. And at any rate, HU test at fixed TDP... which let me remind people, is something that does not bode well for Alder Lake, so 6 or 4 Performance cores wouldn't make much of a difference when TDP is 28W. It wouldn't surprise me if 1280P would perform worse than 1260P and 1270P in some tasks at fixed TDP. As Intel mobile SKUs sometimes do (1165G7 being sometimes faster than 1185G7, 12700H being sometimes faster than 12900HK etc)

6800U isn't out either and of course 6 cores makes a big difference. Lowering the clock speed is the key, a wider lower clocking CPU is usually much more efficient. 1185G7 and 1165G7 have the same amount of cores, this is a really bad comparison. However ADL-P isn't really great because of the shared IA voltage rail for P+E cores and generally speaking 4+8 is too low against 8 big cores for a real MT advantage, same amount of threads in the end. Also I guess Intel 7 is not as good as TSMC 6nm. So there are several factors to consider.
 

uzzi38

Platinum Member
Oct 16, 2019
2,635
5,983
146
generally speaking 4+8 is too low against 8 big cores for a real MT advantage, same amount of threads in the end. Also I guess Intel 7 is not as good as TSMC 6nm. So there are several factors to consider.

Even if we had 6 P core SKUs tested the results wouldn't be as good as you might think.

At 25 and 30W difference between 2+8 and 4+8 is only ~10%


EDIT: Turns out there is a chart of it after all: https://media.discordapp.net/attachments/682675911488700419/988008049581101106/unknown.png
 
Last edited: