Xe-HPC, Intel DG2 GPUs based on gaming Xe-HPG architecture are now sampling

According to Intel, Xe-HPC (known as Ponte Vecchio) has powered on and it is currently being validated. We have seen this high-performance accelerator being teased by Raja Koduri and Pat Gelsinger before, so Intel is definitely expecting it to deliver. This HPC-focused accelerator is set to launch next year at best.

Ponte Vecchio will be available in at least three form factors, Intel today confirmed, including OAM (Open Accelerator Module) and in x4 Sub-system configuration. The former form factor will also be used by AMD Instinct MI200, which could be a potential competitor for Ponte Vecchio-based products.

 

Intel Ponte Vecchio OAM version, Source: Adreas Schilling

 

Today Intel revealed that its Xe-HPG series (known as Arctic Sound) is already present in its DevCloud, although the company has not stated if this means a physical presence or actual availability to customers.

More importantly, though, Intel confirmed that its DG2 family of GPUs for discrete gaming series is now sampling. While Intel did not confirm if its Xe-HPG series will indeed launch this year, it is a good indicator that the product is already being evaluated by partners. Hopefully, this will also mean we should be seeing DG2 performance leaks in the coming weeks.

 

Intel shows Ponte Vecchio up close

Intel’s supercomputing processor already up and running at Intel labs.

 

Intel CEO Pat Gelsinger holding Ponte Vecchio, Source: Intel

During the Intel Unleashed event dedicated to manufacturing and roadmap updates, Pat Gelsinger the CEO of Intel has also talked about the company’s Xe-HPC graphics processor, the Ponte Vecchio. This chip is a big project at Intel ever since Raja Koduri was hired as Chief Architect at Intel Graphics division.

In just over two years the engineers at Intel were able to deliver a working silicon featuring a multi-tile design and over petaflop of compute power. Today Intel confirmed that the processors feature 47 tiles in total. The company also confirmed that all tiles feature over 100 billion transistors.

Intel Ponte Vecchio is an advanced supercomputing accelerator that will be used in tandem with future Xeon architecture codenamed Sapphire Rapids. Intel already announced that the first supercomputer to feature Ponte Vecchio is Aurora. This supercomputer for Argonne National Laboratory is now expected to completed by the end of 2021. It will offer around 1 exaFLOPs of computing power.

 

Intel Ponte Vecchio will compete with NVIDIA Hopper and AMD CDNA2 based accelerators. All three architectures are expected to feature chiplet designs.

 

Intel Xe Architecture Update, Source: Intel