The company has been shipping a high-end CPU, the 4nm Ryzen 7040 Series, with AI technology since earlier this summer, and plans to roll their AI technology across more desktop / laptop Ryzen CPUs next year.
AMD CEO Lisa Su said, “Longer term, while we are still in the very early days of the new era of AI, it is clear that AI represents a multi-billion dollar growth opportunity for AMD across cloud, edge and an increasingly diverse number of intelligent end points.” Those end points came into the spotlight when Intel CEO Pat Gelsinger highlighted his company’s plans for the emerging “AI PC”, for which the upcoming Core Ultra chip will offer AI functionality and performance features.
But AMD beat Intel to the punch, by integrating the AI IP from the Xilinx Versal adaptive SoC into the Ryzen 7040 to get it to market faster with proven AI acceleration technology. It is essentially the same AI IP from the Xilinx ACAP reconfigurable product, fully integrated into the CPU in the- Ryzen 7040. It’s the first product to market that benefited from that joint IP, and a great example of how the two companies are executing post-acquisition.
While AMD has not yet provided any AI benchmarking such as those from MLCommons (there is no competition, to-date), the company is quick to assert the same benefits that Qualcomm has been touting for its Snapdragon ASIC used in Smartphones: dedicated performance, improved security/privacy and lower costs. AI on local devices can solve a myriad of problems, so long as the AI model isn’t too large. The AI feature is available across five SKUs at different clock rates and power levels.
AMD has also leveraged the Xilinx software stack to create a platform for quantizing and running AI models, including for generative AI. The Ryzen AI software platform allows developers to take machine learning models trained in PyTorch or TensorFlow and run them on select laptops powered by AMD Ryzen AI technology. The software intelligently optimizes tasks and workloads, freeing-up CPU and GPU resources and enabling optimal performance at low power.
Conclusions
AMD was able to leverage the assets from the Xilinx acquisition quickly, and brought out a chip combining the AI from Xilinx with the CPU speed of AMD. Intel is once again playing catch-up on the hardware, but seems to be applying more support, or at least more marketing, for the company’s AI solution partners with OneAPI.