Intel® Data Center GPU Flex Series is flexible, robust, and the industry's most open GPU solution for the intelligent visual cloud. The GPU will support a diverse range of workloads in the industry starting with media streaming and cloud gaming, followed by support for AI visual inference and virtual desktop Infrastructure workloads. It supports an open, standards-based software stack optimized for density and quality with critical server capabilities for high reliability, availability, and scalability. This helps reduce the need for data centers to use disparate solutions and manage heterogenous or proprietary environments.
Open Architecture
The Intel Flex Series GPU supports an open, flexible, standards-based software stack together with oneAPI so developers can build high-performance, cross-architecture applications and solutions. This helps organizations reduce the complexity, cost, and time requirements to bring new solutions to market, enabling engineers and programmers to innovate instead of maintaining code.
Built-In AV1 Encode
Services built on the royalty-free open-source AV1 codec mean lowering operational expenses while providing higher video quality. Advanced video coding (AVC), High Efficiency Video Coding (HEVC), and VP9 support also comes standard with the Intel Data Center GPU.
The Power of More
5x
Media transcode throughput at half the power of the competition. Intel® Data Center GPU Flex 140 compared to NVIDIA A10 HEVC 1080p601.
2x
Decode throughput at half the power of the competition. Intel® Data Center GPU Flex 140 compared to NVIDIA A10 across HEVC, AV1, AVC, VP91.
Up to
68
720p30 on select game streams. Single Intel® Data Center GPU Flex 1701.
5x
Media transcode throughput at half the power of the competition. Intel® Data Center GPU Flex 140 compared to NVIDIA A10 HEVC 1080p601.
2x
Decode throughput at half the power of the competition. Intel® Data Center GPU Flex 140 compared to NVIDIA A10 across HEVC, AV1, AVC, VP91.
Up to
68
720p30 on select game streams. Single Intel® Data Center GPU Flex 1701.
Media Processing and Delivery
Today, video makes up over 80% of global Internet traffic2, and viewers expect broadcast quality with higher resolutions across a growing variety of devices. At the same time, new 5G networks are dramatically increasing the throughput for mobile customers.
All of this means that media providers must find new ways to optimize total cost of ownership while meeting subscriber demands for more sophisticated content. The Intel Flex Series GPU helps by increasing the density of streams that can be supported per server—without compromising quality.
1st
First Hardware-Based AV1 Encoder in a GPU
Built-In AV1 Encoder
AV1 is a royalty-free video coding standard created by the Alliance for Open Media, which includes Google, Amazon, Apple, Intel, Cisco, Meta, Microsoft, Netflix, and Tencent. AV1 provides significantly improved compression efficiency over AVC and HEVC.3 This results in more than a 30% bandwidth improvement which can help enable significant Total Cost of Ownership savings.1
5X Better Than the Competition
The Intel® Data Center GPU Flex 140 provides 5X media transcode throughput at half the power compared to NVIDIA A10 on HEVC 1080p60. And 2X decode throughput at half the power of NVIDIA A10 across HEVC, AV1, AVC, VP9.1
Lower Total Cost of Ownership
The Intel Flex Series GPU lets providers serve more subscribers with a smaller data center footprint, reducing the costs associated with equipment and facilities. Plus, high-performance per watt helps reduce the total cost of ownership (TCO).
Improved AVC and HEVC Support
In addition to its forward-looking AV1 implementation, the Intel Flex Series GPU also supports existing HEVC, AVC, and VP9 codecs.
Low Level Controls
With low level controls over encoder and rate controls provided by the oneAPI Video Processing Library (oneVPL), end users can finely tune their encoder configuration to best meet their needs. Industry experts may also implement their own rate controls to combine customer expertise with Intel® hardware innovation.
Cloud Gaming
With the global cloud gaming market expected to reach $3.2B by 2026, growing 43.2% from 2021 through 2026, service providers need to optimize TCO while meeting subscriber demands for superior experiences. The Intel Flex Series GPU increases the density per server without compromising quality. Serve a growing number of subscribers with a smaller data center footprint that reduces equipment and facility costs for both Windows Cloud Gaming and Android Cloud Gaming. High-performance per watt helps drive TCO down further.
There are two SKUs available that customers can deploy based on their solution needs. The Intel® Data Center GPU Flex 140 is a low-profile 75W option with two 8 Xe-core GPUs. The 170† is a 150W option featuring a single 32 Xe-core GPU for higher peak performance.
AI Visual Inference†
The growth in security and site monitoring cameras, live broadcast streams, and OTT video is driving demand for AI Visual Inference. The Intel Data Center GPU Flex Series is ideal for these workloads, including Smart City, Library indexing and compliance, and AI-guided video enhancement workloads. Intel’s oneAPI toolkit supports the most common AI frameworks—TensorFlow* and PyTorch*—and the optimization of AI algorithms from other frameworks through the OpenVINO™ toolkit.
The Intel Data Center GPU Flex Series allows you to choose the power level that suits your needs—from simple object detection to complex AI.
Basic AI Models
A low power, 75W adapter with a smaller form factor, the Intel® Data Center GPU Flex 140 provides more video decode/encode engines than the Intel Server GPU. This GPU is ideal for pipelines that involve lighter AI models, such as simple object detection. Due to the higher number of video engines, the Flex Series 140 GPU can support a high number of streams.
Complex AI Models
A high-power, 150W adapter, the Intel® Data Center GPU Flex 170† is optimal for pipelines that include more complex AI models, such as multiple object detection or multiple classification models. With its AI compute power, the Intel® Data Center GPU Flex 170 avoids becoming AI bound. It can support the same number of video streams per adapter as the compute requirements increase.
Virtual Desktop Infrastructure (VDI)†
Virtual Desktop Infrastructure (VDI) is a mature market with an accelerated move toward the adoption of discrete graphics, as modern OSes and applications increasingly rely on GPU and media hardware acceleration. High density knowledge workers will account for the largest growth of the VDI segment3, necessitating the need for high performing streaming throughput and end-to-end delivery.
No Licensing Fees
Intel provides free virtual GPU software license to customers for lowering their total cost of ownership in VDI deployments. Furthermore, this can act as a catalyst to accelerate the GPU adoption rate in VDI deployments where graphics & encode accelerations are desired/preferred but can often be cost prohibitive.
Flexible vGPU Management
The Intel® Data Center GPU Flex Series supports multiple vGPU configuration & scheduling options for VDI solutions to meet different customer workload and QoS requirements—such as Linux† Kernel-based Virtual Machine (KVM) open-source virtualization technology and VMware ESXi**. It also supports both VMware Horizon† and Citrix DaaS† on top. The Flex Series accelerators have strong ecosystem support. The top two industry leading desktop & application virtualization solutions, VMware Horizon and Citrix DaaS, will be supported.
Flexible Performance
The Intel® Data Center GPU Flex 140 accelerator has two GPUs on a single card, supporting heterogeneous vGPU profiles. Having fewer virtual machines per GPU contributes to more predictable quality of service, shorter GPU scheduling queues, and strong performance isolation across machines. Smaller physical cards and fewer virtual machines means you can cover a variety of user needs more economically.
High Density Virtual Desktop
The Intel® Data Center GPU Flex 140 is ideal for VDI deployments targeting knowledge workers persona - office productivity, video, and browser workloads whose graphics and compute performance requirements are low.
High Performance Virtual Desktop
The Intel® Data Center GPU Flex 170 is best suited for VDI deployment targeting power users persona – 3D and media content creation workloads whose graphics and compute performance requirements are high.
Software and Tools
Discover and get started with the open software ecosystem that is powering the industry’s most flexible GPU for the intelligent visual cloud.
Product Details
Built on Intel’s Xe Architecture, the Intel Flex Series GPU has up to 32 Xe-cores and ray tracing units, up to 4 Xe Media Engines, AI acceleration with Intel® Xe Matrix Extensions (Intel® XMX) and support for hardware-based SR-IOV virtualization. Leveraging Intel® oneAPI Video Processing Library (oneVPL) and Intel® Deep Link Hyper Encode, the Flex Series 140 accelerator with its two GPUs can meet the industry’s one-second delay requirement while providing 8K 60 real-time transcode. This capability is available for AV1 and HEVC HDR format.
Solution Providers
Extending the benefits of its standards-based open architecture and optimized software stack, the Intel Flex Series GPU draws on a broad ecosystem of service providers, independent software vendors (ISVs), original equipment manufacturers (OEMs), and others.
Systems featuring Flex Series GPUs are now available and coming soon from providers, including Cisco, Dell Technologies, HPE, H3C, IEIT, Lenovo, Nettrix, PowerLeader, Supermicro and xFusion. Solutions supported by the Flex Series GPU are also ramping and now include media delivery, cloud gaming workloads for both Windows and Android, and AI visual inference. VDI workload support will follow later this year.
Disclaimers
† Reflects capabilities of Intel® Data Center GPU Flex Series that will be available when product is fully mature.
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.
Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure.
Your costs and results may vary.
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.
Intel technologies may require enabled hardware, software, or service activation.
Cisco Systems, “VNI Complete Forecast Highlights.” https://www.cisco.com/c/dam/m/en_us/solutions/service-provider/vni-forecast-highlights/pdf/Global_2021_Forecast_Highlights.pdf.
Altman Solon. Referenced data is from Slides 132, 135 of their February 17, 2022 study for Intel.