Intel CEO Pat Gelsinger showcased a sample of the Xe Ponte Vecchio GPU, which is intended for use in high-performance computing (HPC) systems, as part of Intel’s Unleashed online event, which focused on current production and future plans.
This chip is the result of a major Intel project that kicked off with the arrival of Raja Koduri as Chief Architect for Intel Graphics. The report states that it took Intel engineers a little over two years to create a working prototype of a processor with a tiered design and more than one petaflops of processing power.
The product shown at the conference is reported to have a “tiled” design and consists of 47 semiconductor elements, which in total contain over 100 billion transistors.
Ponte Vecchio graphics accelerators will be used in conjunction with Xeon processors, codenamed Sapphire Rapids. Intel has already announced that the unveiled GPU will be used in the Aurora supercomputer, which is being developed for the US Department of Energy’s Argonne National Laboratory and is due for completion by the end of this year. It is assumed that the Aurora supercomputer will be able to provide computing power at the level of one exaflops.
According to available data, Intel Ponte Vecchio will compete with accelerators based on NVIDIA Hopper and AMD CDNA 2 architectures. All three architectures are expected to use chiplets, that is, several chips on a single substrate.
If you notice an error, select it with the mouse and press CTRL + ENTER.