Recently, Intel held an Intel on Industrial Innovation Summit (Intel vision), during which it announced a number of progress made in chips, software and services, showing a comprehensive and powerful strength of technology and ecosystem integration. At the summit, Intel publicly introduced a new data center GPU graphics card, code named Arctic sound-M (ats-m), a single GPU solution for multimedia transcoding, visual graphics processing, cloud games and cloud reasoning**
Like arc a series in the consumer game market, it adopts Xe architecture, DG2 alchemist core , integrates up to 32 Xe cores, 32 optical tracking units, 4 Xe media engines, Xmx AI acceleration unit and AV1 hardware codec. It is also the first independent GPU that Intel supports AV1 codec in the field of data center**
In terms of video transcoding, it can synchronously process more than 30 1080p video streams , while in terms of cloud games, it can meet more than 40 gamers at the same time, and supports 62 virtualization functions and AI reasoning computing power of 150 trillion operations per second (150 tops).
Ats-m adopts the concept of full openness. Developers can use the open software stack supported by oneapi to carry out relevant design work, and the adaptation of system and software is also advancing rapidly.
It will be matched with the fourth generation Xeon scalable processor code sapphire rapids, and each node supports two ats-m and two sapphire rapids.
Ats-m has two different forms of design. One is a single chip, which is oriented to the workload requiring peak performance, and the power consumption is 150W; The second is dual chip, which is oriented to high-density and multi-purpose workload, and the power consumption is only 75W
*Intel ats-m graphics card will be launched in the third quarter , and has obtained more than 15 designs from * Dell, ultramicro, Cisco, Huihe, Inspur, Xinhua and other partners.
At the same time, Intel also further introduced the Xe HPC architecture, the first high-performance computing accelerator code named Ponte Vecchio, and the supercomputer "Aurora" based on it and sapphire rapids Xeon scalable processor.
The supercomputing is jointly built by Intel, Huiyu, the U.S. Department of energy and Aragon National Laboratory. The newly disclosed computing power has exceeded 2eflops, that is, 20 billion times per second**
Ponte Vecchio
When it comes to sapphire rapids Xeon processor, Intel recently revealed that it will integrate accelerators for different purposes such as AI, encryption, network and data center , optimize for various types of rapidly growing workloads, and meet the needs of massive data processing.
It will also integrate and package hbm2e high bandwidth memory for the first time , up to four, with a total capacity of 64GB, but this is optional.
Sapphire rapids has started shipping the first batch of models
Sapphire Rapids
Sapphire rapids and four hbm2e memory on the edge
