Intel showcases four-tile Xe-HP GPU with 42 teraflops FP32 compute

AnandTech reports Intel's Raja Koduri showed off a four-tile Xe-HP GPU at this week's Hot Chips 2020 conference. The chief architect of the chip giant briefly held up this four-tile model, which is the largest configuration that will be offered. At last week's Intel Architecture Day, Koduri showed a chip with a one-tile configuration.

The four-tile Xe-HP GPU promises FP32 computing power of 42 teraflops. Intel also claims that thanks to the use of tensor cores, this Xe-HP GPU can achieve "petaflop scale computing."

Assuming it makes it to market, a multi-tiled GPU – essentially multiple GPUs in a single package – would be a major accomplishment for Intel. GPUs are notoriously bandwidth-hungry due to the need to shovel data around between cores, caches, and command frontends, which makes them non-trivial to split up in a chiplet/tiled fashion. Even if Intel can only use this kind of multi-tile scalability for compute workloads, that would have a significant impact on what kind of performance a single GPU package can attain, and how future servers might be built.