Beyond that, the chipmaker said architecture work has begun on Celestial, its third-gen CPU, that will appeal to the ultra-enthusiast segment. Intel also said it is planning a new service to deliver access to Intel Arc GPUs for an “always-accessible, low-latency computing experience.” The service, currently called Project Endgame, should be available later this year. All told, Intel expects to ship more than 4 million discrete GPUs in 2022. Intel previously revealed that first-gen Arc consumer graphics products, code-named Alchemist, will have hardware-based ray tracing, use artificial intelligence to power super sampling, and have support for DirectX 12 Ultimate. Alchemist products are based on Intel’s Xe microarchitecture. The new chips are just one segment of Intel’s Accelerated Computing Systems and Graphics Group (AXG), which is on track to deliver more than $1 billion in revenue in 2022, the company said. AXG’s three segments together will approach $10 billion of revenue for Intel by 2026. The AXG group also includes supercomputing systems, as well as Intel’s new custom compute group, which is working on a Bitcoin mining accelerator. The Custom Compute Group, Intel said Thursday, will also work on tailored products for emerging workloads such as supercomputing at the edge, premium infotainment for cars and immersive displays. Within supercomputing, Intel on Thursday said Ponte Vecchio – its HPC GPU – outperforms the competition by up to 2.6x on key benchmarks used for complex financial services applications. Intel also promised a new architecture, called Falcon Shores, that will bring x86 and X GPU together into a single socket. Targeted for 2024, it’s projected to deliver benefits of more than 5x performance-per-watt, 5x compute density, 5x memory capacity, and bandwidth improvements. Intel also unveiled Arctic Sound-M, a media supercomputer GPU for the data center, which supports hardware-accelerated AV1 encode.
