Intel Launches Data Center GPU Flex Series for AI, Gaming, and Video


Intel officially unveiled its Data Center GPU Flex series on Wednesday — although it began shipping some of its Arctic Sound-M discrete processing units for data centers about a month ago. The new graphics cards are based on the company’s Arc Alchemist graphics processors and are targeted at a variety of data center applications, and will be available from various Intel partners as the products reach required maturity levels.

The Intel Data Center GPU Flex family of graphics cards includes two base offerings: one aimed at performance-demanding workloads and another designed for ultra-dense deployments. The single-chip Flex Series 170 series is based on an ACM-G10 GPU with up to 32 Xe cores (equivalent to up to 4,096 stream processors) and 16 GB of memory and is intended for workloads requiring performance maximum. The dual-chip Flex Series 140 board packs two ACM-G11 GPUs with 16 Xe cores and 12 GB of memory and is targeted at high-density machines.

Intel Data Center GPU Flex Specifications

GPU Flex 170 Data Center Flex 140 GPU Data Center
GPUs ACM-G10 2x ACM-G11
Xe Cores 32 16
Execution units 512 256
Stream processors 4096 2048
Ray Tracing Units 32 16
XMX engines 512 256
Media Engines 4 2
GPU base clock 1950MHz 1600MHz
Maximum dynamic GPU clock 2050MHz 1950MHz
Memory 16 GB of GDDR6 memory 12 GB GDDR6 (2x 6 GB)
Memory bus 256 bit 192 bit (2x 96 bit)
Memory bandwidth 576 GB/s 336 GB/s
Host Interface PCIe 4.0 x16 PCIe 4.0 x8
PTB/PDT 150W 75W

Cards and GPUs are powered by Intel’s Xe-HPG architecture – designed primarily for gaming, but which can also serve a wide variety of data center workloads, including artificial intelligence, gaming Android and Windows cloud, video transcoding (with HEVC, AV1, AVC and VP9 codecs) and virtual desktop infrastructure (VDI) applications. For now, Intel is positioning these cards for Android cloud gaming as well as media transcoding workloads. AI, VDI and Windows cloud gaming will only be supported “when the product is fully mature”.


(Image credit: Intel)

Intel’s Flex 170 series (with an ACM-G10 GPU) can handle transcoding of up to eight simultaneous 4K video streams, over 30 1080p streams, and rendering of 68 720p30 game streams. As a bonus, Intel’s high-end Arc Alchemist GPUs also support XMX instructions so they can accelerate AI inference workloads.

In contrast, Intel’s dual-chip Flex 140 series (with two ACM-G11 GPUs) meets the industry’s one-second delay requirement while providing real-time 8Kp60 transcoding to AV1 and HEVC formats. /H.265 with HDR. It can also handle 46 720p30 game streams (select titles only) on a single card and up to 216 game streams in a multi-GPU setup (we’re speculating about five cards and 10 GPUs – but that’s a guess) .

The server-grade Intel Data Center GPU Flex family of graphics cards is fully supported by the company’s contemporary application programming interfaces and tools, including oneAPI, OpenVINO, oneVPL and VTune Profiler.

Servers powered by Intel’s Data GPU Flex card family will be available from Dell, HPE, H3C, Inspur, Lenovo and Supermicro.


About Author

Comments are closed.