hot fries Intel says its data center-focused Flex-series GPUs, codenamed Arctic Sound, are finally ready, with computer makers expected to start shipping systems in the coming months.
While the cards share a similar Xe core architecture to Intel’s HPC and AI-focused Ponte Vecchio GPUs, detailed in detail at this week’s Hot Chips virtual event, the chipmaker’s Flex-series GPUs target a group very different demographic: game and media streaming applications where streaming density exceeds computational power.
The cards themselves are closer to Intel’s recently announced Arc family of discrete GPUs, but have been tuned for use in data center environments. The cards feature up to four Xe media engines and 32 Xe cores and ray tracing units, while AI acceleration is achieved using the card’s XMX matrix math processors.
Flex series GPUs are available in two form factors: the 75W Flex 140 with 12GB of memory and the 150W Flex 170 with 16GB. Both feature a PCIe 4.0 interface and are passively cooled, single-slot GPUs. However, the lower TDP offering is half-height, allowing it to be deployed in high-density systems with up to 10 cards.
According to Intel, the Flex 140 offers five times the media transcoding performance and twice the decoding performance of Nvidia’s A10 GPUs, and can support up to 36 simultaneous streams of 1080p at 60fps or eight streams of 4K at 60fps. And, at least for streaming applications, this scales linearly, with a 10-card host capable of delivering 360 simultaneous streams at that resolution, or 80 4K streams using the popular HEVC H.265 format, it claims.
Intel’s big feature with this generation is native support for AV1 encoding, which the chipmaker says is the first to offer in a data center form factor.
The royalty-free codec was developed under the Alliance for Open Media by several large streaming media companies, including Amazon, Netflix, and Google, to name a few. The format promises to offer 30 percent bandwidth savings compared to the already space-efficient HEVC format.
The chipmaker claims that this streaming density significantly reduces operating cost for Android game streaming and media streaming workloads, as well as applicability for AI streaming and the metaverse.
Speaking of cloud gaming, Intel claims that a single Flex 170 can deliver 68 game streams at 720p 30fps and is validated on nearly 90 popular Android titles at launch.
However, for AI and HPC applications, Intel Flex GPUs fall short. According to Intel, the Flex 140 offers a maximum performance of 8 teraflops in FP32 calculations, putting it on par with Nvidia’s $450 A2000 MSRP. Meanwhile, Intel’s Flex 170 offers roughly twice the performance at 16 teraflops of FP32, roughly half the performance Nvidia’s A10 claims at 31.2 teraflops.
The performance figures here suggest that the Flex 140 is simply a dual-die version of Intel’s A50 workstation GPU announced earlier this month. That card boasted a maximum performance of 4.8 teraflops of FP32 in a half-height, dual-slot, actively-cooled form factor.
Intel Flex Series GPUs will be available through more than 15 OEM partners including Cisco, Dell Tech, Hewlett Packard Enterprise, Supermicro, Inspur and H3C in the coming months. ®