AMD Radeon RX 8000 series 'RDNA 4' GPU spotted: 56 CUs and 16GB of memory

AMD's next-generation RDNA 4 GPU architecture has been teased again, with an RDNA 4-based Radeon RX 8000 series graphics card turning up on Geekbench.

2

The new graphics card in question is an RDNA 4-based GPU with 56 CUs (Compute Units) and 16GB of VRAM (which should be GDDR6 still, but faster 18Gbps GDDR6 memory chips). The GPU has a "GFX1201" device ID, which means we're looking at the Navi 48 SKU.

AMD's new Navi 48 "GFX1201" GPU is listed as 28 compute units, which results in 56 compute untis as the new RDNA 4 GPU architecture does things the same as RDNA 3: each GPU has a shader engine that features dual compute units, so we should see a doubling of the listed compute units (28 x 2 = 56).

The listed Radeon RX 8000 series "RDNA 4" graphics card had a GPU clock speed of 2.1GHz (2100MHz), which is lower than the current 2.5GHz to 2.6GHz+ overclocks on current RDNA 3-based Navi 3X graphics card. However, AMD is tapping AMD's latest process node, a newer RDNA 4 GPU architecture, and more... so we should expect some stellar performance-per-watt numbers from RDNA 4, and possibly lower GPU clocks to boot.

AMD's new RDNA 4-based Radeon RX 8000 series GPU that turned up on Geekbench 6 had 16GB of VRAM, so we're looking at a 256-bit memory interface for the highest-end RDNA 4 graphics card. We should see 12GB of GDDR6 memory on a 192-bit memory interface for the card/s under the flagship RDNA 4 chip.

AMD is expected to use faster 18Gbps GDDR6 memory chips on its new RDNA 4-based Radeon RX 8000 series GPUs, up from the 16Gbps GDDR6 chips on current Radeon RX 7000 series "RDNA 3" graphics cards.