Specifications
Product
Intel Data Center GPU Flex 170
Architecture
Intel Xe-HPG (Xe Matrix Extensions / XMX)
Memory
16 GB GDDR6, 256-bit
Memory Bandwidth
576 GB/s
TBP
150 W
Host Interface
PCIe 4.0 x8
Form Factor
Full-height, dual-slot, passive (server-cooled)
Card Dimensions
254 mm × 111 mm × 19 mm
Media Engine
Dual media engines, up to 8 streams of 4K60 transcode; AV1 / HEVC / AVC
Display Outputs
Up to 4 × DisplayPort (virtual / VDI)
AI Acceleration
Xe Matrix Extensions (XMX), INT8 / FP16 inference
Graphics APIs
DirectX 12, Vulkan, OpenGL, OpenCL
Compute APIs
oneAPI, SYCL, OpenVINO
Target Workloads
Cloud gaming, VDI, media transcode, edge AI inference
Availability
In stock — quote for volume
Overview
The Intel Data Center GPU Flex 170 is a single-slot-class server GPU aimed at workloads that are not pure dense matrix math: media transcode, virtual desktop and cloud gaming, and medium-scale AI inference at the edge. It pairs a Xe-HPG graphics engine with 16 GB of GDDR6 on a 256-bit bus and a 150 W board power envelope.
What makes the Flex 170 interesting for infrastructure buyers is density per watt in the media pipeline. The dual media engines support AV1 encode as well as HEVC and AVC, so a single card can retire a rack of older transcode nodes. The Xe Matrix Extensions (XMX) add INT8 and FP16 inference acceleration for OpenVINO-based vision pipelines running on the same hardware.
QS Compute supplies Flex 170 boards together with the Intel Arc Pro and NVIDIA L40S / A-series accelerators so integrators can match the right card to the workload instead of over-buying training-class GPUs for media and inference tasks.
Key Benefits
16 GB GDDR6 at 576 GB/s gives headroom for multiple concurrent transcode or inference streams. AV1 encode cuts bitrate at equal quality for streaming and VDI. XMX INT8/FP16 accelerates OpenVINO inference on the same card. 150 W passive fits standard server airflow and power budgets.
Applications
Cloud gaming and game-streaming nodes; virtual desktop infrastructure (VDI); broadcast and OTT media transcode with AV1; edge AI inference for vision and analytics; and dense media servers replacing discrete transcoder farms.
Request a Quote — INTEL DATA CENTER GPU FLEX 170 — 16 GB GDDR6 MEDIA & AI ACCELERATOR
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →