In the age of artificial intelligence, demand for GPUs like the H100 has skyrocketed, making it difficult for the average consumer to get hold of one. However, A.J Reddit-Beuser He created a unique solution by converting an AMD Zen 2-based Ryzen 5 4600G “Renoir” APU to a 16GB GPU and then using it for AI workloads on Linux. You know the saying: modern problems need modern solutions.
Older AMD Ryzen APUs provide decent performance in AI workloads after switching to GPUs
Before diving into rebuilding an APU, let’s take a look at the AMD Ryzen 5 4600G. The Ryzen 5 4600G is considered one of the best APUs out there, after it was eventually replaced by its Cezanne counterpart. It has a 6C/12T configuration with Radeon Vega iGPU with seven CUs (computing units). To explain how you hit the 16GB VRAM mark, you need to note that APUs support “shared memory” where you can allocate 50% of your RAM capacity to your APU. In this case, this Reddit user had 32 GB of DDR4 memory, half of which went to the processor.
The next big hurdle is actually running AI workloads on a Ryzen APU. If you have a desktop GPU, you can use AMD’s ROCm (Radeon Open Compute) platform to run AI applications on Linux. However, in the case of iGPUs, third-party packages allow ROCm to run on the APU, which was used here. Using ROCm will solve most of your problems as you can now run any type of AI application from Tensorflow to PyTorch.
In a detailed video, a Reddit user shared his interesting experience and claimed that the Ryzen 5 4600G can handle all kinds of AI workloads. However, it only showed stable diffusion tests and to our surprise the APU was able to produce a 512 x 512 image in around 1 minute 50 seconds. This is an important event for APUs and we believe it can be a hacking attack if used properly.
While you can’t compare the Ryzen 5 4600G with an AI GPU like the H100, this test shows the untapped potential of the Ryzen APU. Imagine that instead of the Ryzen 5 4600G, the latest APUs from Phoenix are used here, which are significantly more powerful. However, offering such an AI feat for $95 is certainly commendable and hopefully gives the “average consumer” a way to satisfy their AI needs.
News Source: Tom Device
