Intel® Data Center GPU Flex Series

Push the limits of innovation with the industry’s most open GPU.

See product brief

Intel® Data Center GPU Flex Series

Highlights

Flexibility

Specifications

Resources

Support

See product brief

VDI, media, and AI on one platform

Intel Data Center GPU Flex series is the industry's most open GPU solution, delivering powerful media processing, virtualization, and visual innovation on shared Xe-HPG hardware for VDI, media delivery, and AI inference.

Up to

55%

better visual inference performance

Compared to NVIDIA A10.¹ With up to four Xe media engines helping drive up to 256 TOPS for AI, you'll see what you've been missing.

Up to

50%

lower total cost of ownership (TCO)

Compared to the NVIDIA A16.² There are no royalty or license fees, and reliability, availability, and serviceability (RAS) are built in.

Up to

4x

more 4K media transcode streams

Vs. NVIDIA L4.² The built-in AV1 encode in each GPU reduces bandwidth requirements and enables a much higher stream density.

Up to

50%

higher throughput at 50% power usage

Vs. NVIDIA A10³ enabled by low power usage with open source oneAPI optimizations and code portability

Two people walking through a data center aisle, with one holding a laptop

Enable boundless workspaces for VDI

The Intel® Data Center GPU Flex 140 delivers impeccable quality, flexibility, and productivity at scale. This includes up to 62 virtual functions based on hardware-enabled single-root input/output virtualization (SR-IOV) as well as no licensing fees.

Watch the demo

Xᵉ architecture

Deploy a foundation of unmatched flexibility

Intel Data Center GPU cards, shown against a dark blue background

Intel® Data Center GPU Flex 140

Unlock the full potential of VDI efficiency and media processing precision with up to four media engines and 16 Xe cores in a 75W, half-height PCIe package. Enable advanced workloads with unparalleled ease, flexibility, and scalability.

See product specs

Intel® Data Center GPU Flex 170

Step into a new era of unprecedented AI analytics speed and accuracy with up to 32 Xe cores in a 150W, full-height PCIe form factor. The Intel® Data Center GPU Flex 170 is tailored for adaptive AI workloads with up to 500 TOPS.

See product specs

Intel® Data Center GPU Flex 170V

For VDI applications, experience great compute performance with lower complexity in a 150W, full-height PCIe form factor with 32 Xe cores. The Intel Data Center GPU Flex 170V is built to be a lower TCO option while providing the performance and features needed to enable world-class VDI capabilities.

See product specs

Browse Intel Data Center GPU Flex series

Filters Compare


Intel® Data Center GPU Flex 140	16	16	12 GB	336 GB/s
Intel® Data Center GPU Flex 170	32	32	16 GB	576 GB/s
Intel® Data Center GPU Flex 170V	32	32	16 GB	576 GB/s

Resources

Technician installing a server component into a data center rack

Get support for the Intel Data Center GPU Flex series

Get product support

Product and Performance Information

¹ Up to 4x more 4K media transcode streams (FFmpeg): testing by Intel as of October 9, 2023. Intel configuration: System Vendor: Intel Coyote Pass, CPU: Intel® Xeon® Platinum 6336Y CPU @ 2.40GHz, stepping 6), two sockets, 24 cores, two threads (96 CPUs), memory: GB 32-ch 3200 MT/sec, ucode: 0xd000280, BIOS: SE5C620.86B.01.01.0003.2104260124, OS: Ubuntu 22.04 LTS (Jammy Jellyfish), Kernel: 5.15.0-84-generic, GPU: Intel® Data Center GPU Flex Series 140 x 1—B1 PRQ RLG1, driver: Production_682.22_20230912, GPU firmware: ATS_M75_128_B0_ES_029_23WW08_02_GS2054_PC9815_OP1060_Headless_P2SB_DS_ECC_EN (ECC Off), HT enabled, Turbo on, FFmpeg n6.0, data collected on PVT card < 55C temperature. NVIDIA configuration: System vendor: GIGABYTE G492-H80-00, CPU: Intel® Xeon® Gold 6338 CPU @ 2.00GHz, stepping 6), two sockets, 32 cores, two threads (128 CPUs), Memory: 256 GB 32-ch 3200 MT/sec, ucode: 0xd000389, BIOS: F25, OS: Ubuntu 22.04 LTS (Jammy Jellyfish), Kernel: 5.15.0-67-generic, GPU: NVIDIA L4 (UUID: GPU-143f747e-97c2-c500-2122-a88ab89fffe5) SN: 1325222012830, Driver: docker: nvidia/cuda:12.2.0-devel-ubuntu22.04, 535.104.05 CUDA 12.2, GPU firmware: Vbios 95.04.29.00.06 (ECC Off), HT enabled, Turbo on, FFmepg version 5.1.git(g9ef20920ab), NVIDIA-smi 535.104.05, Data collected on NVIDIA L4 card < 55C temperature. See https://www.intel.com/content/www/us/en/secure/content-details/820189/competitive-proof-points-nvidia.html?DocID=820189 for additional details.

² Up to 55 percent better visual inference for object detection: testing by Intel as of September 16, 2022. Intel configuration: 2S Intel® Xeon® 4309Y processor, 256 GB DDR4-2933, 1x Intel® Data Center GPU Flex 170, Ubuntu 20.04 Kernel 5.10, Agama 419, OpenVINO™ 2022.2.0 ‒ 90 1080p AVC streams at 25 FPS with YOLO v5. NVIDIA configuration: 2S Intel® Xeon® 6336Y processor, 128 GB DDR4-3200, 1x NVIDIA A10 GPU, Ubuntu 20.04, Kernel 5.15, NVIDIA Driver 515.65.01, CUDA 11.7 Update 1, DeepStream 6.1.1, TensorRT 8.4.1.5 ‒ 58 1080p AVC streams at 25 FPS with YOLO v5. See https://www.intel.com/content/www/us/en/secure/content-details/820189/competitive-proof-points-nvidia.html?DocID=820189 for additional details.

³ Up to 50 percent lower TCO for VDI: pricing as of October 11, 2023. Intel® Data Center GPU Flex Series pricing is recommended customer pricing from Intel pricing group. Actual system-level cost will vary based on server configurations. Actual pricing may vary. Intel® Data Center GPU Flex Series 170 has 16 GB vRAM and supports 16 users. NVIDIA cost includes GPU plus NVIDIA vGPU vPC CCU per-user licensing fees for five years—1 GB per user profile. NVIDIA A16 has 64 GB and supports 64 users. Pricing sources: cdw.com/product/pny-nvidia-a16-64gb-gddr6-graphic-card/7091857 and nvidia.com/content/dam/en-zz/Solutions/design-visualization/solutions/resources/documents1/Virtual-GPU-Packaging-and-Licensing-Guide.pdf.