Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

Page 7 - 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
61
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:

Geranium

Member
Apr 22, 2020
83
101
61
DDR5 is 64-bit per channel with 2x32 for brust mode. It is like HBM, a single channel is 1024-bit, but it is divided into 8 128-bit sub channel. Or like the GDDR6/6X, 32-bit single channel divided into 2x16-bit channel.

LPDDR5/5X is 16-bit per channel. If a APU supports both DDR5/LPDD5, then total memory channel width will be 128-bit. So it is Dual-channel for normal DDR and Octa-channel for LPDD/X.

On Renoir/Cezanne total memory width is 128-bit. When normal DDR4 is used it act as Dual channel, and when LPDD4/4X is used it act as Octa-channel. Same on Intel's cpu with both DD4/LPDDD4/4X.
 

izaic3

Member
Nov 19, 2019
61
96
61
With the announcement of stacked cache, do you think that Rembrandt could use that technology, or will Rembrandt be delivered before those kinds of chips come out?
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
With the announcement of stacked cache, do you think that Rembrandt could use that technology, or will Rembrandt be delivered before those kinds of chips come out?

Remembrandt uses RDNA2. RDNA2 is about Infinity Cache right?

Since they announced the cache for desktop CPUs I would be a bit surprised if Remembrandt doesn't have it since they can use the Infinity Cache marketing for it. And boost performance as well.
 

andermans

Member
Sep 11, 2020
151
153
76
Remembrandt uses RDNA2. RDNA2 is about Infinity Cache right?

Since they announced the cache for desktop CPUs I would be a bit surprised if Remembrandt doesn't have it since they can use the Infinity Cache marketing for it. And boost performance as well.

I think indications are that neither VanGogh nor Rembrandt have infinity cache (mostly gleaned from Linux patches for Yellow Carp which presumably is Rembrandt)
 

Gideon

Golden Member
Nov 27, 2007
1,608
3,573
136
I think indications are that neither VanGogh nor Rembrandt have infinity cache (mostly gleaned from Linux patches for Yellow Carp which presumably is Rembrandt)
The beauty of thks is even If it does have it, budget versions can be made without it as well
 

izaic3

Member
Nov 19, 2019
61
96
61





No links only or image only posts allowed. You must post your own personal commentary.


esquared
Anandtech Forum Director
 
Last edited by a moderator:
  • Like
Reactions: lightmanek

Tuna-Fish

Golden Member
Mar 4, 2011
1,324
1,462
136
Oh dear. @moinmoin was right, they are using fish names.

Can't wait to see Red Herring on a roadmap.

The color + fish codenames are from the Linux drivers. To get the drivers out on release they have to publish code well in advance, and they need some way to refer to gpus while preferably leaking the least information possible. For this reason, the drivers use their own codenames which are used nowhere else. Then enthusiasts try to connect those code names back to the other things. The news would appear that Yellow Carp is the GPU in Rembrandt.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
The color + fish codenames are from the Linux drivers. To get the drivers out on release they have to publish code well in advance, and they need some way to refer to gpus while preferably leaking the least information possible. For this reason, the drivers use their own codenames which are used nowhere else. Then enthusiasts try to connect those code names back to the other things. The news would appear that Yellow Carp is the GPU in Rembrandt.

In which APU are you the iGPU?

. . .

sorry, had to say it.

Seriously though I hope they don't get all the fish names muddled. I remember when people were calling Kaveri Refresh Godavari (and the name stuck!) when, according to The_Stilt, that was an internal name for something else and not Kaveri Refresh. It got mighty confusing.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136

Embedded rembrandt? The specs look great, if true. Polaris levels of performance on an iGPU.

EDIT: Depending on power budget, I estimate the iGPU will probably be somewhere around the RX560 in terms of performance.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
I wonder about die size. L3 is still 16MB. Do we know how big an RDNA 2 WGP and a Vega CU in Cézanne are?

7nm -> 6nm means the same chip made on 7nm. would be anywhere from 10-18% smaller. I suspect that 6nm costs less than 7nm thanks to less machine time, so AMD can likely afford to allow some growth.
 

andermans

Member
Sep 11, 2020
151
153
76
7nm -> 6nm means the same chip made on 7nm. would be anywhere from 10-18% smaller. I suspect that 6nm costs less than 7nm thanks to less machine time, so AMD can likely afford to allow some growth.

I thought it was possible to do 6nm using a 7nm design though? In which case there wouldn't be any space savings AFAIU.
 

tomatosummit

Member
Mar 21, 2019
184
177
116
Design rules are the same.

Doesn't mean using the same masks. Can't in any case, as one is for DUV and the other is for EUV.
New masks certainly but redesigning all the modules takes time and money.
There is nothing particularly new on die apart from pcie4 and I'd guess that's off the shelf IP while the zen3 ccx, rdna2 and the old soc components are all already designed and validated on n7.
Even without the shrink there's still performance, power and manufacturing improvements to be gained.