• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

AMD’s Flagship Navi 31 (512mb cache) GPU Based on Next-Gen RDNA 3 Architecture Has Reportedly Been Taped Out

tusharngf

Member
AMD ready their next-generation RDNA 3 lineup and it looks like the flagship Navi 31 GPU for the upcoming Radeon RX lineup has been taped out today. A tweet from leaker, Greymon55, suggests that AMD has taped out its flagship graphics chip for 2022.



AMD-Radeon-RX-7900-XT-Big-Navi-31-GPU-With-RDNA-3-Architecture-Block-Diagram-1480x1077.png

A preliminary block diagram of AMD's next-gen RDNA 3 based Navi 31 GPU that will power the flagship Radeon RX 7900 XT graphics card.



The Navi 31 GPU configuration shown here features two GCD's (Graphics Core Die) and a single MCD (Multi-Cache Die). Each GCD has 3 Shader Engines (6 in total) and each Shader Engine has 2 Shader Arrays (2 per SE / 6 per GCD / 12 in total). Each Shader Array is composed of 5 WGPs (10 per SE / 30 per GCD / 60 in total) and each WGP features 8 SIMD32 units with 32 ALUs (40 SIMD32 per SA / 80 per SE / 240 per GCD / 480 in total). These SIMD32 units combine to make up 7,680 cores per GCD and 15,360 cores in total.
The Navi 31 (RDNA 3) MCD will be linked to the dual GCD's via a next-generation Infinity Fabric interconnect and feature 256-512 MB of Infinity Cache. Each GPU should also feature 4 memory connect links (32-bit). That's a total of 8 32-bit memory controllers for a 256-bit bus interface.

AMD RDNA 3 Navi 3X GPU Configurations (Preliminary)​

GPU NameNavi 21Navi 33Navi 32Navi 31
GPU Process7nm6nm5nm/6nm5nm/6nm
GPU PackageMonolithicMonolithicMCMMCM
Shader Engines424 (2 per GCD)6 (3 per GCD)
GPU WGPs402040 (20 per GCD)60 (30 per GCD)
SPs Per WGP128256256256
Compute Units (Per Die)804080
160 (Total)
120
240 (Total)
Cores (Per Die)5120512051207689
Cores (Total)512051201024015360
Memory Bus256-bit128-bit192-bit256-bit
Memory TypeGDDR6GDDR6GDDR6GDDR6
Infinity Cache128 MB256 MB384 MB512 MB
Flagship SKURadeon RX 6900 XTXRadeon RX 7700 XT?Radeon RX 7800 XT?Radeon RX 7900 XT?
TBP330W~200W~300W~400W
LaunchQ4 2020Q4 2021?Q4 2021?Q4 2021?

Source: AMD's Flagship Navi 31 GPU Based on Next-Gen RDNA 3 Architecture Has Reportedly Been Taped Out (wccftech.com)
 
Last edited:

Sentenza

Member
Can't wait to read about its performances at launch and then wait 16 more months for it to be available on the market.
 
i think it will be more around 1800

in the face of those who said "it's difficult to distribute the load on more cu's" hahaha fuckin insane 240 cu's
It still is difficult.

It’s one of many reasons why MCM designs have been neglected in the past.

Significant changes and upgrades have to be made to the command processors and geometry engines to ensure workloads are being distributed to the CU efficiently. Then there’s also the issue of overhead and latency when the chiplets are communicating and transferring data to one another.

There’s actually patents for task scheduling units by AMD which line up with RDNA 3, which I’m guessing is one of their solutions to this problem.
 

Lethal01

Member
Will PC gamers even get to take advantage of any of this compute performance considering consoles exist?
Yes, if consoles didn't exist the games would still be made to be run on significantly older PC hardware, nobody is making pc games exclusive to the 3090 and even in years a game requiring a 2070 will be an exception.
 

martino

Member
i think it will be more around 1800

in the face of those who said "it's difficult to distribute the load on more cu's" hahaha fuckin insane 240 cu's
it's true with current engine and situation
it will become less and less the more they use and replace fixed function with tailored and more optimized shader works.
Will we see a threshold point about it this gen ? i'm less confident.
 
Last edited:

winjer

Gold Member
Will PC gamers even get to take advantage of any of this compute performance considering consoles exist?

Of course they can. PC can run at higher settings than consoles.
On PC you can run all games at true 4K. Or higher if one desires.
Or you can run games at high refresh rates. 144hz displays are very common on PC. And some go to 240hz.
Then there is ray-tracing. There are more games on PC that use ray-tracing. And most use much higher quality settings than on consoles.
And then there is the usual stuff, higher resolution shadows, volumetric fog and illumination. Increased LODs. Higher use of tessellation.
And the list goes on and on.
 
Last edited:

MonarchJT

Banned
it's true with current engine and situation
it will become less and less the more they use and replace fixed function with tailored and more optimized shader works.
Will we see a threshold point about it this gen ? i'm less confident.
it's just impossible to think to evolve GPUs perf without growing the number of cu's, at least according to the technology and the working pipeline of today's 3d graphics. We know well that the discussions that have arisen about it are if precisely to defend a certain fast & narrow design as someone likes to call it ... a design that cannot be adopted after a certain performance ceiling (it would take two power supplies just to power the GPU)
 
Last edited:

MikeM

Member
Hopefully supply isn't the shit show it is now. I want to actually get my hands on a 7800XT without having to pay an absurd amount of money. Maybe I can finally build a gaming pc!
 

MonarchJT

Banned
i don't see the direction GPU are going on energy consumption last for long with energy sobriety that is to come ( i give it 2 gen max)....
M1max having the same performance as a 2080 in the benchmarks (if that's true) bodes well ... you know how much that chip consumes?
 

PaintTinJr

Member
M1max having the same performance as a 2080 in the benchmarks (if that's true) bodes well ... you know how much that chip consumes?
At 400watts and 61TF/s vs 90watts and 10TF/s won't both products be in the same efficiency ballpark for watt/TF?
 
Last edited:

Lethal01

Member
61tf is enough to do 4k60 ray tracing at current gen graphics fidelity :)

Because raytracing is used extremely limitedly because of how bad the current cards, especially AMD's are.



Of course they can. PC can run at higher settings than consoles.
On PC you can run all games at true 4K. Or higher if one desires.
Or you can run games at high refresh rates. 144hz displays are very common on PC. And some go to 240hz.
Then there is ray-tracing. There are more games on PC that use ray-tracing. And most use much higher quality settings than on consoles.
And then there is the usual stuff, higher resolution shadows, volumetric fog and illumination. Increased LODs. Higher use of tessellation.
And the list goes on and on.

Look at the next gen version for Metro Exodus, Yes you can get a much better experience when playing on PC but if you have to make sure it runs on something without raytacing you are still held back by old hardware.
 

martino

Member
M1max having the same performance as a 2080 in the benchmarks (if that's true) bodes well ... you know how much that chip consumes?
i expect power consumption going up for both and it's not a good trend.
but even if it's not the case i expect regulations to hit gpu directed to mainstream public with current high end level during this decade.
 
Last edited:
ps5 outdated hardware now then lol.

can't wait to get one of these cards in my PC.

Keep it coming, I want 4K 120FPS with raytracing
pathetic. 8K 360fps is where it's at.

If you can afford one of these have fun finding them at a normal retail price.

1NGyBgS.gif
anyone who can afford one won't mind spending more.

i got a 3090 which was meant to be £1,400. actually paid £2,100.
 
Last edited:

winjer

Gold Member
Look at the next gen version for Metro Exodus, Yes you can get a much better experience when playing on PC but if you have to make sure it runs on something without raytacing you are still held back by old hardware.
What does that old hardware matter, on a thread about future hardware?
 
Top Bottom