An update to the AMD ROCm software has confirmed the potential number of compute units for AMD’s upcoming Navi 32 and Navi 33 GPUs. These GPUs are based on the RDNA 3 architecture and are expected to be used in lower-end RX 7000 video cards.

The number of compute units in AMD’s upcoming RDNA 3 GPUs are in the source code for the new AMD ROCm update, Known Twitter user Kepler_L2. This shows that the Navi 32 GPU has a maximum of 60 compute units, as can be seen in AMD ROCm repository on GitHub. This amounts to 3840 stream handlers. The Navi 33 chip should have a capacity of 32 cu in, which is equivalent to 2048 stream processors.

Earlier this year, there were rumors about Navi 32 and Navi 33 GPUs. Angstronomics, which is likely to share information about new AMD products, Male in August That two gpus get 60 and 32 cu’s respectively. Now it seems that AMD has indirectly confirmed it. Angstronomics also stated that the Navi 32 GPU has an infinity cache of 64MB and a 256-bit memory bus. The Navi 33 chip will have 32MB of cache and a 128-bit memory bus. This specification is not listed in the ROCm update.

AMD announced its first RDNA 3 desktop video cards earlier this year. The Video cards AMD Radeon RX 7900 XT and XTX Get the Navi 31 GPU, the highest positioned RDNA 3 chip ever released by AMD. The GPU has a maximum of 96 computational units and 6144 stream processors. AMD has not yet announced any lower-end video cards with a Navi 32 or 33 GPU. It is expected that these video cards will appear from next year.

It is not yet known for which video cards the Navi 32 and Navi 33 chipsets will be used. In the current generation, the Radeon RX 6900 and RX 6800 desktop graphics cards have Navi 21 GPU. The RX 6700 series has a Navi 22 GPU. The lower positioned Radeon RX 6600 video cards have a Navi 23 chip.