Next generation flagship GPU rumors: NVIDIA “Ada” GA102 with 21 Gbps memory, AMD’s “RDNA3” Navi 31 with 7 chipsets, Intel “Battlemage” with 20480 cores

Please note that this post is tagged as one rumor.

GeForce RTX 4090 vs Radeon RX 7900XT, or something else?

It’s pretty intriguing why leakers tend to share their information the same day. What is important to note is that both Kopite7kimi and Gryemon55 clearly have good sources, as their leaks have proven to be true many times.

NVIDIA RTX Ada GA102 GPU gets 21 Gbps memory?

Kopite reportedly has new information about the next generation RTX GPU flagship. Interestingly, it is no longer a given that the upcoming “Ada” GPU series are in fact called RTX 40. The reader now refers to them with ** 90 / ** 80, etc., just if they are called, e.g. for example, the RTX 50 series.

In any case, Kopite7kimi updated its message, which was posted on April Fool’s Day, as mentioned at the time it was not confirmed and probably still is not. What this new message says, however, is that the 4080 class GPU with a SKU number of 330 can actually be used for the 4090 series.

More importantly, the memory specifications were downgraded from 24 Gbps to 21 Gbps, which is actually the same speed as the RTX 3090 Ti. This upcoming SKU would have a TDP of 600W. This is probably the reason why some rumors suggested that the just released RTX 3090 Ti is a test vehicle for the next generation GPU series.

It is worth pointing out that by sharing the NVIDIA card numbers PG137, PG139 or PG141, this leak gains significantly more credibility. This is actually something that we can later confirm as we get closer to the launch.

AMD Navi 31 with 7 chipsets

It is now rumored that the upcoming Navi 31 GPU, the flagship processor in the RDNA3 and Radeon RX 7000 series, will contain a total of 7 chipsets.

AMD returned to monolithic designs with the Navi 21 GPU, after slightly less successful attempts to compete in advanced GPU space with complex and expensive HBM GPU designs such as the Fiji 10 and Vega 10/12.

Neither Greymon nor any other person is sure what the MCD chiplet is for anything, but it was long rumored to be a Memory Complex Die. Therefore, it may be where Infinity Cache or memory controllers are located. It is not completely confirmed whether these MCD chipsets will be located on top of the Graphics Complex Dies (GDC) or located elsewhere on the package.

In total, the Navi 31 GPU is expected to have 7 chipsets made using either 6nm or 5nm TSMC process technology.

INTEL ARC B-Series (Battlemage)

Some time ago, thanks to an Intel test driver that the company accidentally released, we learned that the company is working on a next-gen Xe-HPG architecture called Battlemage, internally referred to as Elasti. One of the configurations listed mentioned ‘ELG_X4’, which would involve a 4-tile configuration of the next generation graphics processor.

At present we do not know Battlemage’s specifications or how it will perform in relation to AMD and NVIDIA, but the company officially confirmed that it will target the higher “Enthusiast” level in the GPU market, later to be followed by “Celestial” which reaches as high as the “Ultra Enthusiast” segment. Battlemage will not be launched until 2023/2024, but there is a chance that it will be available as the RTX 4090 and RX 7900XT are still on the market.

According to Redfire’s analysis of the available data, a potential configuration with 4 tiles by Battlemage could have a total of 40 Xe-Cores and 320 Xe-Vector Engines per. This means that a 4-tile configuration would have 1280 Vector Engines (also known as Execution Units), which would basically mean 20480 FP32 cores. Of course, we do not know if “ELG_X4” really refers to the tile configuration, nor if such a part is planned at all. But just for the sake of speculation, this would really provide an interesting competition in the advanced space from Intel.

Next Generation Flagship GPU Comparison (RUMORED) GeForce RTX 4090 – Class Radeon RX 7900 – Class Arc B9XX – Class
Manufacturing node TSMC N5 TSMC N5 / N6 TBC
Architecture NVIDIA available AMD RDNA3 Intel Xe2-HPG “Battlemage”
GPU package Monolithic Multi-Chip Module (MCM) Multi-Chip Module (MCM)
Estimated GPU size ~ 600 mm² ~ 800 mm² TBC
Graphics die 1 2 GCD + 4 MCD + 1 IOD 4 tiles
GPU Mega Clusters 12 Graphics Processing Clusters (GPC) 2 × 3 Shader motors 10 Render Slices
GPU Super Clusters 72 Texture Processing Clusters (TPC) 2 × 30 RDNA Workgroups (WGP) 40 vehicle cores
GPU clusters 144 streaming multiprocessors (SM) 1280 Xe vector motors
FP32 cores 18432 CUDA colors 15360 Stream processors 20480 FP32 ALUs
GPU ur ~ 2.2 GHz ~ 2.5 GHz TBC
Memory type 24 GB GDDR6X TBC GB GDDR6 TBC
Memory & Bus 21 Gbps 384-bit TBC 256-bit Gbps TBC
Cache 96 MB (L2 Cache) 256 or 512 MB Infinity Cache TBC
Power consumption 600W (TDP) 500W (TBP) TBC
Release date Q3 / Q4 2022 Q3 / Q4 2022 2023-2024
Rumored achievement 2x GA102 2.5x NAVI 21 TBC

Source: @ kopite7kimi (NVIDIA), @ greymon55 (AMD), @ Redfire75369 (INTEL)

Leave a Comment