Intel Announced That Arctic Sound M Series Data Center GPU Accelerator Card Will Be Available In The Third Quarter

take 6 minutes to read
Home News Main article

Intel has just disclosed more details about the Arctic sound M series data center GPU, which is scheduled to be launched in the third quarter of 2022 It is reported that the new accelerator card focuses on a series of professional workloads such as cloud games, media processors and delivery, virtual desktop infrastructure, and thrust It is based on Xe HPG GPU core, adopts PCIe 4.0 interface, and provides two SKUs.

The flagship SKU is based on a single acm-g10 GPU core, with a thermal design power consumption of 150W and is committed to achieving peak performance. The other is based on dual acm-g11 GPU core, with 75W TDP, which is mainly for high-density multi-purpose workload.

Intel claims that Arctic sound GPU can provide 30 + 1080p streams, 40 + game streams, up to 62 virtualization features, and 150 AI tops.

In terms of functions, the new data center GPU is also quite novel and comprehensive, such as hardware level AV1 codec and built-in Xmx AI accelerator.

Unfortunately, Intel has not disclosed a more precise specification except for 4-way Xe media engine + 32 Xe cores + light tracing units.

As a reference, the acceleration card SKU designed by 75W has 16 Xe cores - the GPU still seems to be based on the full specification, but operates at a lower clock frequency to meet the TDP requirements.

It was reported earlier that the Intel Arctic sound-M GPU encoded with AV1 can reduce the bit rate loss of the data center by 30%. Now, the company confirms that the new chip will support 8-way @ 4K / 30-way @ 1080p video streaming.

Partners including Netflix, Amazon, Google and Microsoft can apply this potential to their game streaming servers.

It can be seen that compared with the older AVC coding, the open source AV1 has a 30% higher efficiency advantage and can be used as a powerful alternative to AVC / hevc coding.

If all goes well, Intel Arctic sound M series GPU accelerator cards will be available in the third quarter of 2022.

In addition, blue team has promised to provide more than 15 system designs from industry-leading partners, including Cisco, H3C, HPE and Dell( http://www.anrdoezrs.net/links/9019719/type/dlg/sid//https://www.dell.com/zh -cn/shop/deals)(DELL, Inspur and supermicro.

Finally, these accelerators are not only compatible with PCIe 4.0 bus, but also run around a complete software stack. With the support of a set of open standard APIs, it can be applied quite flexibly.

[figure] GM And Red Hat Jointly Develop A New Open Source Vehicle System Based On Linux
« Prev 05-11
Musk Talked About Tesla's Needs, Electric Vehicle Start-ups And Scooters In An Interview
Next » 05-11