Gigabyte G482-Z55 8-GPU Server with Intel Flex 170

The Gigabyte G482-Z55 is a 4U rackmount server platform engineered to support up to eight Intel Data Center GPU Flex 170 accelerators, delivering a high-density GPU compute architecture purpose-built for enterprise data centers. The Intel Data Center GPU Flex 170 is based on Intel's Xe-HPC-derived Xe HPG media architecture, integrating 32 Xe-cores and 4 GB of GDDR6 memory per GPU unit, with a TDP of 75W — a thermally efficient design that enables dense multi-GPU deployments without specialized liquid cooling infrastructure. The G482-Z55 chassis accommodates dual AMD EPYC 7003-series processors alongside its GPU payload, providing a balanced CPU-to-accelerator ratio suited to orchestrated workloads across virtualized and bare-metal environments.

The Intel Flex 170 GPU is specifically optimized for media transcoding, cloud gaming, and inferencing workloads at scale, supporting AV1, HEVC, AVC, and VP9 hardware encode and decode engines. Each Flex 170 card supports PCIe 4.0 x16 connectivity and is driver-compatible with Intel's oneAPI software stack, enabling deployment on Linux-based hypervisors including VMware vSphere and KVM. The G482-Z55 platform supports SR-IOV (Single Root I/O Virtualization), allowing individual GPU resources to be partitioned and allocated across multiple virtual machines, a critical capability for media streaming service providers and cloud desktop infrastructure operators.

Designed for regional enterprise markets including the UAE, GCC, EMEA, and APAC, the G482-Z55 with Intel Flex 170 GPUs represents a scalable, standards-compliant inference and media platform that integrates cleanly into existing data center power and cooling envelopes. The combination of eight low-TDP Flex 170 GPUs in a single 4U chassis delivers a dense GPU-to-rack-unit ratio while remaining within typical 30A branch circuit budgets, making it a practical choice for operators expanding AI inference and video-as-a-service capabilities without infrastructure overhauls.

Ideal for

• High-density video transcoding for streaming platforms requiring simultaneous multi-format encode and decode across AV1, HEVC, and AVC codecs at scale
• Cloud gaming infrastructure where SR-IOV GPU partitioning enables multiple concurrent player sessions to be served from a single physical server node
• AI and deep learning inference serving for classification, object detection, and recommendation models using Intel OpenVINO and oneAPI runtime environments
• Virtual desktop infrastructure (VDI) deployments requiring hardware-accelerated graphics delivered to thin clients across enterprise campuses or remote workforces
• Media processing pipelines in broadcast and OTT environments that demand real-time transcode, thumbnail generation, and quality-of-experience analytics
• Edge data center AI inference consolidation where low per-GPU TDP and compact 4U form factor reduce facility power and space requirements

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.