You can take other Ryzen
AI-related workloads are always resource-intensive, but due to the fact that this is a very young field, sometimes it’s enough just to optimize the software or find some new way, and everything starts to work much faster. For example, the Radeon RX 7900 XTX, which is Stable Diffusion, is now an order of magnitude faster than before . And a new experiment shows that a $100 processor can run AI much faster than you might think.
Enthusiast Ttiotech was able to “turn” a Ryzen 5 4600G that can be bought for less than $100 into a graphics card with 16GB of memory. Of course, not in the literal sense.
“Turn” Ryzen 5 4600G into a 16GB graphics card for faster AI performance.
The bottom line is that the same 16 GB of RAM can be allocated to this APU through the BIOS settings of the motherboard, and memory is very important for AI tasks. Some boards may allow you to allocate an APU and 32 GB of RAM, but there are no such tests yet. You also need a Linux system and AMD ROCm processor support. For example, you can use more modern Ryzen G.
Stable Diffusion tests have shown that, properly configured, the processor can generate a 50-step 512×512 image in about 1 minute 50 seconds, which is better than some of the more powerful CPUs.
Of course, the Ryzen 5 4600G will not overtake monstrous video cards for hundreds and thousands of dollars, but still it costs several times less.