In a recent earnings call, AMD revealed plans to beef up its EPYC processor lineup with Xilinx's FPGA-powered AI inference engine.
With the first of these new AI-powered processors scheduled to launch next year, the move shows the chip giant is hard at work integrating technologies from its €54bn acquisition of Xilinx into its existing products.
Still, recent AMD patents show that the company is working on various methods to connect AI accelerators to its processors, including using 3D chip stacking technology to do so.
AMD is not the first chipmaker to pair its processors with embedded FPGAs, as Intel attempted something similar after acquiring Altera in 2015. Despite the company announcing and even demonstrating a combined CPU + FPGA chip, the final product It ended up being more of an experiment that he hasn't revisited since.
FPGA-Infused Processors
Although AMD has yet to go into detail about its future FPGA-infused products, connecting Xilinx's silicon FPGAs to its processors will likely require some sophistication.
Instead of using standard PCIe lanes and a QPI interconnect to connect FGPA chips to a processor like Intel did, AMD's patents suggest the company is working on an accelerator port to do this in different ways.
AMD could use 3D stacking chip technology like it does with its Milan-X processors to deliver performance, power, and memory performance benefits. However, this approach can present thermal challenges that would reduce performance if the chip is placed too close to the compute dies.
Fortunately, we'll likely learn more about AMD's upcoming AI-powered processors at the company's upcoming Financial Analyst Day next month.
Through Tom's Hardware