AMD has confirmed its next-gen RDNA 4 GPUs through ROCm, which includes Navi 44 “GFX1200” and Navi 48 “GFX1201” SKUs for the Radeon RX 8000 series. Although none of this is massive news, AMD ROCm Github has recently confirmed both the Navi 44 and Navi 48 GPUs based on RDNA 4 graphics architecture along with their respective graphics IDs. Here, as expected, the AMD Navi 44 turns out to be codenamed “GFX1200,” while the Navi 48 ends up being known as the “GFX1201” SKU.
More About AMD Navi 44 GFX1200 and Navi 48 GFX1201
While the ROCm GitHub is light on these new AMD 3D V-Cache dies, a round of rumors from earlier in the month claimed that Navi 48 will be the faster chip of the two and is expected to find a place in higher-end products. However, unlike the Navi 31 (RDNA 3) or Navi 21 (RDNA 2) series, which were reportedly discontinued in favor of more mainstream chips targeting the mass market, the Navi 48 GPUs are expected to offer strong performance-to-price ratios for gaming enthusiasts.
This means that the AMD Navi 48 GPUs will offer performance closer to the Navi 31 GPUs but at a much more attractive price point, which of course is beneficial for gamers. Consider the AMD Radeon RX 7900 GRE, which at present costs just approximately $549 US after the fresh prices were revealed for an entry-level RDNA 3 SKU. If AMD can, for example, deliver better performance than the 7900 GRE at a price of around $400 to $500 US, then that will be an excellent scenario for gamers looking to pick up some powerful GPUs at good prices.
Another thing to keep in mind is how AMD’s RDNA 4 GPU architecture will also include a brand new design, so there could be optimizations within ray tracing titles and the arrival of things like AI-infused FSR upscaling technology (which would compete against NVIDIA‘s DLSS). Even though FSR 3 is looking to be a pretty attractive proposition for games, using AI acceleration could help alleviate some of its limitations and would allow it to produce a better image in scenarios where it’s currently losing out against DLSS.
Although many details about the RDNA 4 GPUs are unknown, it is expected that they will have GDDR7 memory in the first phase with speeds of 28 / 32Gbps and a capacity of approximately 16Gb (2 GB VRAM per module).
However, since these are only samples and not full GPU dies with faster HBM clocked at 2.048 Gbps/pin memory – it is suspected that AMD might go the cheaper route by employing faster GDDR6 dies initially. There are rumors of the RDNA 4 chips also being smaller and monolithic, which would mean that AMD has (after just one year) given up on their new MCD design.
