Intel announces Xeon Max server processors and data center GPUs – Computer – News

Intel announced Xeon Max, which are server processors known internally as Sapphire Rapids HBM. Additionally, the company is announcing the GPU Max series for the data center, and server GPUs codenamed Ponte Vecchio. Delivery of both products will begin in January.

The Xeon Max CPU Seriesformerly known as Sapphire Rapids HBM, has up to 56 performanceThe cores, which are distributed over four slab-like tiles. Processors do not get efficient electronic cores. Those squares are interconnected with what Intel Multi-Template Embedded Bridge called or emib. This is a type of small silicone bridge that acts as an interface. This CPU series has TDPs of up to 350W.

The CPU also gets 64GB of HBM2e memory spread over four 16GB stacks. This gives each core more than 1 GB of HBM2e memory at its disposal. Xeon Max also supports PCIe 5.0 and the CXL 1.1 I/O connectivity standard. Intel claims that the new CPU uses 68 percent less power than the AMD Milan-X lineup for the same performance, or 2.4 times faster when making climate models than AMD’s when using HBM memory.

The data center GPU Max series, formerly known as Ponte Vecchio, consists of three different cards that can be equipped with up to 128 Xe-HPC cores. Cards also consist of different types tiles And get up to 408MB L2 cache and up to 64MB L1 cache. Datacenter GPU cards also get The original accelerated ray tracing, which Intel says can speed up scientific visualizations and animation. By some financial standards, the cards will be 2.4 times faster than Nvidia’s A100 cards; In some physical simulations, Intel cards will be 1.5 times faster.

See also  Seamless docking in one go • TTM.nl

Intel’s most powerful data center GPU is the Max Series 1550, a 600W card with 128 Xe cores, 128GB HBM2e memory and 128 ray tracing units. The Max Series 1350 has a 450W TDP, 112 Xe and 96GB of memory. The 1100 GPU, in turn, has a 300W TDP, 56 Xe cores, and 48GB of memory. Cards can be linked together using Intel Xe Link bridges.

The Aurora supercomputer should, among other things, be equipped with Xeon Max products. It should be the first supercomputer that can achieve more than two exaflops in computing power. By the way, this supercomputer has been waiting for Intel server products for years; It was actually Aurora To be delivered in 2021. became this Last year postponed to 2022 Because the products announced today are delayed.

So the Aurora won’t hit two exaflops until next year, with the release of Xeon Max and Data Center GPU Max products. A supercomputer will have more than 10,000 blades. Each blade contains six Max series GPUs and two Xeon Max CPUs. Several US labs and Kyoto University are also acquiring supercomputers with the products announced today. The successor to the Max Series GPU should appear in 2024. These GPUs are now called Rialto Bridge.

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top