AMD AI and compute platform ROCm, The open source competitor to Nvidia CUDA and now also offers Windows support the Radeon RX 9060 and 9060 XT for the first time in the latest version 7.0.2.
AMD continues to advance artificial intelligence (“AI”) and machine learning (“ML”) with its open-source compute platform ROCm, which is intended to act as a CUDA competitor, in the latest version 7.0.2 and now supports all RDNA-4 graphics cards with the Navi-4x GPUs for the first time. In addition to the Radeon RX 9070 XT, the Radeon RX 9070 GRE and the Radeon RX 9070 with Navi 48 (“GFX1201”), product support for Navi 44 (“GFX1200”) has been included for the first time.
Supported video cards and accelerators with ROCm 7.0.2
AMD wants to penetrate even more into the consumer sector with ROCm (“Radeon Open Compute Platform”), which is why more and more Radeon graphics cards are now to be supported in addition to the professional Radeon Pro, including the new Radeon AI Pro R9700, and Instinct accelerators. For the graphic architectures, it now looks like this:
- RDNA 2: ✔
- RDNA 3: ✔
- RDNA 3.5: ✔
- RDNA 4: ✔
- CDNA 1: ✔
- CDNA 2: ✔
- CDNA 3: ✔
- CDNA 4: ✔
Further information is provided by the AMD Developer Cloud as well as the very detailed GPU Architecture Documentation and the official website.
In direct comparison to ROCm 6.x, ROCm 7.x should achieve 3.1 to 3.8 times the performance in AI inferencing and be on average three times as fast in AI training as the manufacturer benchmarks of the new compute platform attest in advance.
AMD ROCm (“Radeon Open Compute Platform”) combines drivers, runtime environments, programming models and interfaces, libraries, compilers and tools into an all-in-one compute platform and is intended to compete with Nvidia CUDA.
Your opinion is in demand!
How do you feel about this issue? The PCGH editorial team is pleased about your well-founded opinion in the comments to this message.
