RDNA3’s new flagship AMD Radeon RX 7900 XT exposure: nearly 100 TFLOPs, four times that of the 6900XT

IT House April 30 news, according to the whistleblower @Greymon55, the AMD Radeon RX 7900 XT is proven to be equipped with a new RDNA 3 core “Navi 31” GPU. According to the latest information, AMD’s new generation of graphics cards is more powerful than previously expected.

He said that the flagship RDNA 3 graphics card is equipped with the strongest Navi 31 GPU, with a main frequency of more than 3 GHz and FP32 performance of 92 TFLOPS, which is 28% higher than the previously claimed 75 TFLOPS. This figure is even four times that of RDNA2’s flagship card, the RX 6900XT (Navi 21).

Greymon55 previously stated that there have been major changes in the specifications of the new generation of GPUs, but he did not explain what those changes were. With this calculation, if AMD were to keep 120 Workgroup Processors (WGPs, each of which can hold two CUs) and 15360 Stream Processors, its single-precision computing performance would require a 2995 MHz GPU frequency, so looking at it this way, AMD should be similar to NVIDIA and its Ada architecture, and will introduce a different 32-bit instruction matrix for the new generation of graphics cards, which will greatly improve floating-point performance.

From his sources, AMD is currently working on at least four new GPUs, which have been confirmed via ID leaks as follows (source: @Kepler_L2):

• GFX (1100) – probably Navi 31 GPU

• GFX (1101) – probably Navi 32 GPU

• GFX (1102) – probably Navi 33 GPU

• (GFX 1103) – probably Phoenix APU

While there isn’t much to tell from these IDs alone, they do indicate that AMD is bringing initial support for certain features to the new generation of GPUs, and testing will soon begin to improve performance.

Also, contrary to previous rumors, AMD may still reserve compute units for its RDNA3 (GFX11) architecture. @Kepler_L2 recently found out that AMD made changes to their GPU drivers, confirming this:

According to the latest rumors from Greymon55, the new generation GPU seems to have been directly boosted from 2.5 GHz to 3 GHz or even higher. This means that the next-generation flagship GPU will have 15,360 stream processors and will be able to deliver nearly 100.0 TFLOPs of FP32 computing performance.

It is worth mentioning that the Ada Lovelace GPU architecture of the NVIDIA GeForce RTX 40 series has also been revealed, and it seems that it will also usher in a major update. The leaker expects Ada Lovelace’s flagship graphics card to be the first consumer-grade product to break the 100 TFLOPs barrier, and also to offer high clock speeds and performance similar to AMD’s RDNA 3. Of course, this may be due to the 4N process from TSMC. After all, the current graphics card is based on Samsung’s 8nm, and the process improvement across several levels must also bring about skyrocketing performance.

IT House has learned that the AMD Radeon RX 7000 series is expected to be launched at the end of this year, and a new generation of flagship graphics cards will inevitably be released at that time, so the Navi 31 GPU, which is most likely to be used for the RX 7900 XT-class graphics card, will have more tests in the near future. And information leakage, so stay tuned.