Over and over, I’ve heard PC enthusiasts, gamers, and bloggers alike ask the same question: “What is Intel doing?” AMD has seen an explosion in its market share with its Ryzen chips. Intel wears the crown with the Core i9-9900K, but AMD has gone from an also-ran to a true competitor against both Intel and Nvidia. What Intel’s been doing is working on its Xe GPU architecture, which it plans to use not just to make integrated and discrete consumer graphics cards, but also to push back at Nvidia’s invasion of the data center with its Ponte Vecchio GPU.
Intel gave up the details on Ponte Vecchio earlier this week, so here’s what we know. Ponte Vecchio will be manufactured using Intel’s 7-nm process. For reference, that Core i9-9900K is manufactured with the company’s 14-nm process.
The GPU will be a multi-core module design, much like Nvidia’s next-gen design is rumored to be. The design uses Intel’s Foveros chip-stacking technology, which lets Intel stack the processor dies and lets the chiplets of a multi-chip module communicate more efficiently. It will also use Intel’s Embedded Multi-Die Interconnect Bridge (EMIB) tech to join the chips together.
A unified base architecture
Ponte Vecchio will use the same base architecture as Xe chips, as will any other GPU Intel makes within this generation. The company will then branch out into sub-architectures. Anandtech describes the possibilities: “ultra-mobile parts of the product stack might focus on small die size… whereas a compute product might have high double-precision performance and run high-performance libraries.” Xe is the base, and Ponte Vecchio is one branch of that.
Intel said in an earnings call last month that the company is on track to launch 7nm products and a “datacenter-focused discrete GPU in 2021,” and that’s pretty much what Ponte Vecchio sounds like.
With Intel diving into the GPU game, it’s going to be very interesting to see all the different ways the chipmaker develops its new architecture and how that changes the playing field.