It’s The End of Road For Intel’s Ponte Vecchio GPU, Company Now Focused On Falcon Shores Development

Intel has ended the deployment of its Ponte Vecchio GPU, marking the end of the road for a chip that is considered to be a technical marvel.

Intel's Ponte Vecchio GPU was first revealed in 2019 and was the brainchild of Intel's Ex GPU guru, Raja Koduri. When it was announced, the chip was designed to power the next-generation exascale computing platforms but the company had to go through several hurdles to build this marvel of a chip that housed several chiplets on a single package. For reference, Intel's Ponte Vecchio GPU had a total of 47 tiles on a single package which included:

The Ponte Vecchio GPUs saw a house in the Intel Data Center Max GPU series and one of the lead products built on this exascale chip platform was the Aurora Supercomputer which did manage to break the exascale barrier but a little too late. The AMD-powered Frontier supercomputer not only managed to beat Intel-powered Aurora in the exascale race but currently sits at the number 1 spot and has higher peak efficiency than the Ponte Vecchio-powered system.

Intel did manage to break some AI performance records with the Aurora supercomputer thanks to its Xe hardware which includes dedicated AI accelerators but the company is now shifting the focus to its Gaudi accelerators with Gaudi 2 and Gaudi 3 now being the prime chips to serve the segment.

Talking to ServerTheHome, Intel confirmed that it won't be deploying further clusters with Ponte Vecchio GPUs. The company is going to continue to offer Ponte Vecchio in existing clusters but no new clusters will be made. For those who are interested in harnessing the HPC capabilities of the Ponte Vecchio GPUs, they will still be available on the Intel Developer Cloud but the company has reaffirmed that it is now focusing its attention on the next-generation Falcon Shores GPU.

As of last week, Intel Ponte Vecchio is moving into a new phase. Instead of hunting for new clusters, it is going to continue to be sold and filled into existing clusters. Likewise, the Intel Xe architecture is important to the company, so Intel is still going to continue developing the software behind Intel Xe as it moves ahead to Falcon Shores, hopefully next year.

via ServerTheHome

Intel was also expected to launch Rialto Bridge, an upgraded version of the Ponte Vecchio GPUs which was latter canned. And Falcon Shores was originally going to be a combination of x86 CPU and GPU architecture but that was dropped in favor of a GPU-only design. This design would've been similar to the AMD MI300A APU Accelerator which combines CDNA 3 GPUs with Zen 4 CPU cores in a singular package.

Based on what we are aware of, the Intel Falcon Shores GPU will utilize the best of both Gaudi and next-generation Xe GPU architectures and combine them in a single package. The GPU is scheduled for launch next year but we have to wait and see if the product actually ships on time or ends up facing a similar delay as Ponte Vecchio which came before it.