Software for Intel® Data Center GPU Flex Series
Flexible, Robust, and the Industry's Most Open GPU Solution for the Intelligent Visual Cloud
The Intel® Data Center GPU Flex Series supports an open, standards-based software stack optimized for density and quality. With critical server capabilities for high reliability, availability, and scalability, the GPU supports media streaming, AI visual inference, and cloud gaming. Later capabilities will include support for virtual desktop infrastructure workloads.
Intel Data Center GPU Flex Series reduces the need for data centers to use disparate solutions to manage heterogeneous or proprietary environments.
Media Delivery
The Intel Data Center GPU Flex Series is the industry's first data center GPU with a hardware AV1 encoder. This GPU provides stunning video performance, optimized by the Intel® oneAPI Video Processing Library.†
- 5x media transcode and decode throughput at half the power of the competition
- Eight simultaneous 4Kp60 streams, or more than 30 simultaneous 1080p60 streams per PCIe* card
Choose Your Operating System
Before You Begin
Get your system ready to run media workloads with Intel Data Center GPU Flex Series. Get the latest Linux* driver and check that it is installed correctly.
Get Started with Transcoding
To ensure your media stack is running properly, build a media-delivery Docker* image and run an AVC to HEVC FFmpeg* transcoding sample within a container.
Explore Media Delivery Further
This GitHub* project has a collection of samples that demonstrate best practices to achieve optimal video quality and performance on Intel GPUs for content delivery networks. Watch the demo for recommended command lines and quality and performance-measuring tools.
Before You Begin
Get your system ready to run media workloads with Intel Data Center GPU Flex Series. Get the latest Linux* driver and check that it is installed correctly.
Get Started with Transcoding
To ensure your media stack is running properly, build a media-delivery Docker* image and run an AVC to HEVC FFmpeg* transcoding sample within a container.
Explore Media Delivery Further
This GitHub* project has a collection of samples that demonstrate best practices to achieve optimal video quality and performance on Intel GPUs for content delivery networks. Watch the demo for recommended command lines and quality and performance-measuring tools.
Before You Begin
Get your system ready to run media workloads with Intel Data Center GPU Flex Series. Get the latest Linux* driver and check that it is installed correctly.
Explore Media Delivery Further
This GitHub* project has a collection of samples that demonstrate best practices to achieve optimal video quality and performance on Intel GPUs for content delivery networks. Watch the demo for recommended command lines and quality and performance-measuring tools.
Before You Begin
Get your system ready to run media workloads with Intel Data Center GPU Flex Series. Get the latest Linux* driver and check that it is installed correctly.
Getting Started with Transcoding
To ensure your media stack is running properly, build a media-delivery Docker* image and run an AVC to HEVC FFmpeg* transcoding sample within a container.
Explore Media Delivery Further
This GitHub* project has a collection of samples that demonstrate best practices to achieve optimal video quality and performance on Intel GPUs for content delivery networks. Watch the demo for recommended command lines and quality and performance-measuring tools.
Android* Cloud Gaming
The Intel Data Center GPU Flex Series provides a standards-based, high throughput, low total cost of ownership foundation for cloud gaming that helps providers build efficient, effective infrastructure. The platform brings together high gameplay quality with a high density of game instances per server.
- Over 90 games validated
- Intel®-optimized Android* cloud gaming reference stack
- Intel® GPU and CPU compatible code stacks using oneAPI
- Up to 68 select 720p30 game streams on Intel Data Center GPU Flex Series 170
- Up to 46 select 720p30 game streams on Intel Data Center GPU Flex Series 140
Software for this solution is available for service providers that host Android cloud games (NDA access is required).
Sign In to Your Premier Account and Apply for Software Access
Windows* Cloud Gaming
Intel Data Center GPU Flex Series is a high-quality, high-density accelerator for Windows* remote gaming. Intel provides a cloud gaming reference stack for Windows to demonstrate how to unleash the Intel Data Center GPU Flex Series for remote gaming, including:
- How to enumerate and retarget multiple game titles to run concurrently on each of multiple Intel Data Center Flex Series adapters1
- Two types of virtual displays, including a built-in driver and new Microsoft* indirect display drivers
- How-to guides for bare metal, Single-Root I/O Virtualization (SR-IOV), and Passthrough virtualization
- Ultra-low delay, transport, and content adaptive encode of AVC, HEVC, and AV1.
- High-efficiency streaming using the Open WebRTC Toolkit
- Optimized Microsoft DirectX* 9, 11, and 12 API render target and desktop capture
1 Select titles, where not limited by vendor anti-cheat systems
Choose Your Environment
Software for this solution is available from your system original equipment manufacturer (OEM) or through service providers that host Windows cloud games (NDA access is required). Reference the following steps after you access the Software Packages and Collaterals solution. You can also get the software through Intel by signing up for a Premier Account.
Sign In to Your Premier Account and Apply for Software Access
Before You Begin
Ensure that the correct drivers are installed for your hardware. These drivers support Windows Server 2019 and 2022.
Install the Intel Data GPU Flex Series driver for Windows from the Software Packages and Collateral directory.
Get Software for Your Environment
Get the cloud gaming reference stack for Windows from the Software Packages and Collaterals solution.
This setup supports Ubuntu* 22.04 as the host and Windows Server 2022 as a guest operating system.
Software for this solution is available from your system original equipment manufacturer (OEM) or through service providers that host Windows cloud games (NDA access is required). Reference the following steps after you access the Software Packages and Collaterals solution. You can also get the software through Intel by signing up for a Premier Account.
Sign In to Your Premier Account and Apply for Software Access
Before You Begin
Download drivers for SR-IOV or a Passthrough setup:
- Both drivers are required for SR-IOV.
- Only a guest driver is needed for a Passthrough setup.
SR-IOV:
Host Operating System Driver: Ubuntu 22.04 Installation Guide for Ubuntu 22.04
Guest Operating System Driver:
Windows Server 2022 (from your original equipment manufacturer [OEM] or through the previous Premier Account link)
Passthrough Driver:
Windows Server 2022 (from your OEM or through the previous Premier Account link)
Set Up Your Environment
Download the guide on how to set up a virtualization environment on the Intel Data Center GPU Flex Series.
Get Software for Your Environment
Get the Windows Cloud Gaming Stack.
After accessing the Windows Cloud Gaming Software Packages and Collaterals section, check for instructions inside the Windows cloud gaming software package for SR-IOV or Passthrough environments.
AI Inference
The Intel Data Center GPU Flex Series is ideal for AI inference workloads, including smart city, library indexing and compliance, and AI-guided video enhancement workloads. The Intel® AI Analytics Toolkit supports the most common AI frameworks, TensorFlow* and PyTorch*, and the optimization of AI algorithms from other frameworks through the Intel® Distribution of OpenVINO™ toolkit.
- Supports the most popular AI frameworks
- Scales code from Intel® Xeon® processors to Intel Data Center GPU Flex Series
- Delivers AI optimized kernels using the Intel® oneAPI Deep Neural Network Library (oneDNN), Intel AI Analytics Toolkit
How to Deploy AI Models
This setup supports Ubuntu* 22.04.
Before You Begin
Get your system ready to run media workloads with Intel Data Center GPU Flex Series. Get the latest Linux* driver and validate that it is installed correctly.
Intel Model Zoo
Get optimized workloads for the Intel Data Center GPU Flex Series.
AI Visual Inference
Intel Data Center GPU Flex Series is ideal for visual inference and media analytics workloads that include intelligent traffic management, smart buildings and factories, and frictionless retail applications. The platform brings together highly optimized media decoding with AI inferencing. The software stack from Intel for visual inference includes open source FFmpeg* and GStreamer multimedia frameworks, and the Intel® Distribution of OpenVINO™ toolkit inference engine for creating complex visual inference pipelines for the cloud or at the edge. This sample implementation covers:
- Designing, creating, building, and running visual inference and analytics pipelines via command line, C++ and Python* APIs.
- Benchmarking and optimizing performance of visual inference and analytics pipelines with a catalog of ready-to-use media analytics pipelines.
- Distributing a workload across a GPU and host CPU for achieving higher stream density and lower end-to-end latency.
How to Get Started with Visual Inference
This setup supports Ubuntu* 22.04.
Before You Begin
Get your system ready to run visual inference workloads with Intel Data Center GPU Flex Series. Get the latest Linux* driver and check that it is installed correctly.
Get Started with Visual Inference Object Detection
To ensure your visual inference software stack is running properly, download and set up a media analytics Docker* image. Then run an object detection sample within a container.
Explore Visual Inference
This GitHub* project has a collection of tools and samples to develop visual inference workloads for Intel® Data Center GPUs. Using the OpenVINO toolkit, develop confidently with APIs that are cross compatible to CPUs. Review the samples and developer documentation for recommended command lines and performance-measuring tools. Combine your inference workload with FFmpeg media framework and Intel® oneAPI Video Processing Library (oneVPL) media acceleration library to design end-to-end media analytics workflow.
Virtual Desktop Infrastructure (VDI)
The Intel Data Center GPU Flex Series supports multiple virtual GPU (vGPU) configuration and scheduling options for VDI solutions to meet different customer workload and quality of service (QoS) requirements. Intel provides virtual GPU software licenses to customers for free to help lower their total cost of ownership in VDI deployments.
- Hardware-based Single-Root I/O Virtualization (SR-IOV)
- Flexible vGPU management such as support for multiple configurations and scheduling options
- Strong ecosystem support that including VMware ESXi*, VMware Horizon*, Citrix* Desktop as a Service (DaaS), and more
Virtualization
The Intel Data Center GPU Flex Series supports virtualization through Passthrough and Single Root I/O Virtualization (SR-IOV) approaches.
- Passthrough is based on Intel® Virtualization Technology (Intel® VT) for Directed I/O (Intel® VT-d). This approach allows direct assignment of an entire GPU to a single user, passing native driver capabilities through the hypervisor without any limitations.
- Interface is an extension of the PCIe specification. SR-IOV allows a device, such as a GPU card or network adapter, to separate access to its resources among various PCIe hardware functions. This "fractionalizing" of the GPU allows separate workloads or users to share the GPU resources simultaneously.
The Virtualization Guide for Intel Data Center GPU Flex Series includes:
- The fundamentals of virtualization at the platform and system level
- Various alternatives for graphics device virtualization setup on the host server
- Implementation of graphics device virtualization for Passthrough and SR-IOV
Digital Content Creation
The Intel Data Center GPU Flex Series accelerates a variety of ray tracing, simulation, and image-enhancement workloads using up to 32 dedicated hardware ray tracing units.
- Compute-intensive ray traced rendering can be scaled across data center GPUs for speed and efficiency.
- Professional graphics in scientific visualization, synthetic data generation, digital twins, engineering, gaming, and motion pictures surpasses limitations with photorealistic computer-generated imagery.
Deliver high-fidelity, high-performance rendering and denoising that enables ray tracing acceleration hardware using the following libraries:
These libraries are part of the Intel® Rendering Toolkit.
Software Stack for Intel Data Center GPU Flex Series
Using key open source technologies with Intel oneAPI cross-architecture tools, developers can:
- Innovate faster
- Reduce development time, complexity, and cost
- Overcome the constraints of proprietary environments that limit code portability and adoption of new architectures
Test Intel GPUs in Intel® Developer Cloud
Get free access to GPUs. Developers can build, test and optimize media, AI and visual inference, content creation, cloud gaming, virtualization, and VDI applications. Run small to large-scale workloads that demonstrate performance and efficiency.
Intel® XPU Manager (Intel® XPUM)
Intel® XPU Manager (Intel® XPUM) is a free and open source solution built on top of the Intel® oneAPI Level Zero interface for monitoring and managing Intel® Data Center GPUs. Intel XPUM is a complete solution for GPU administration, location and topology, telemetry and health monitoring, diagnostics, firmware updating, setting GPU policies, and configuration. Use Intel XPUM as a stand-alone solution through its command-line interface for local GPU management, and through RESTful APIs for remote GPU management. It also includes a local library for third-party solution integration, such as workload and cluster managers, job schedulers, and monitoring solutions.
Intel Data Center GPU Flex Series
Intel Data Center GPU Flex Series is flexible, robust, and the most open GPU solution for the intelligent visual cloud in the industry.
Product and Performance Information
† For workloads and configurations, see the Intel Data Center GPU Flex Series Performance Index. Results may vary.