Brand: Intel | Category: GPUs
SKU: HLS-GAUDI3-PCIE | Part #: HLS-GAUDI3-PCIE | MPN: HLS-GAUDI3-PCIE
Contact for Pricing — Request a Quote
The Intel Gaudi 3 AI Accelerator PCIe Card (HLS-GAUDI3-PCIE) is a full-height, full-length PCIe card engineered for high-performance AI inference and training workloads in enterprise data centers. This Intel accelerator delivers exceptional compute density with 3,456 TFLOPS peak BF16 throughput and 1,728 TFLOPS peak FP32 throughput, backed by 96 GB of HBM2e memory and 2.5 TB/s memory bandwidth. Designed for infrastructure teams deploying large-scale machine learning clusters, the Gaudi 3 integrates seamlessly into existing server architectures via PCIe Gen5 connectivity and operates within a 0–55°C temperature range.
The card draws up to 600 W of power through dual 6-pin and 16-pin server-class connectors, making it suitable for modern enterprise-grade chassis. Intel backs the HLS-GAUDI3-PCIE with a 3-year manufacturer warranty, with extended coverage available for mission-critical deployments. This accelerator brings production-ready performance to organizations seeking to optimize their AI infrastructure without major architectural overhauls. An AI infrastructure team evaluating next-generation accelerators will find the Gaudi 3's combination of memory capacity, bandwidth, and throughput well-suited to demanding inference pipelines.
Contact Omnixon Global today to request a quote and discuss your AI acceleration requirements.
| Brand | Intel |
| Category | GPUs |
| SKU | HLS-GAUDI3-PCIE |
| Part Number | HLS-GAUDI3-PCIE |
| Condition | New |
| Power Connector | 12V-2x6 / 16-pin (server-class) |
| Warranty | 3-year manufacturer warranty (extended available) |
| Accelerator Model | Gaudi 3 |
| Manufacturer Part Number | HLS-GAUDI3-PCIE |
| Interface | PCIe Gen5 |
| Form Factor | Full-height, full-length (FHFL) PCIe card |
| Peak BF16 Throughput | 3,456 TFLOPS |
| Peak FP32 Throughput | 1,728 TFLOPS |
| HBM2e Memory | 96 GB |
| Memory Bandwidth | 2.5 TB/s |
| Maximum Power Consumption | 600 W |
| Operating Temperature | 0–55°C |
Reference servers include Dell PowerEdge XE9680 / XE9712, HPE Cray XD670, Lenovo ThinkSystem SR685a / SR675 V3, Supermicro AS-A21GE / SYS-821GE, Gigabyte G593 / G894, ASUS ESC. Share your target platform in the RFQ and we will confirm chassis-to-GPU compatibility and recommended NIC pairing.
Highly model-dependent. L40S / RTX-class: typically 3-6 weeks. H100/H200/B200 in SXM form factor: 12-16 weeks for whole-platform allocations. We quote authorised-channel ETAs only — no grey-market promises.
Yes — Omnixon stocks the full NVIDIA networking lineup (Quantum-2 / Quantum-X InfiniBand, Spectrum-X Ethernet, ConnectX NICs, BlueField DPUs) so we can quote a complete training-cluster BOM, not just the GPUs.
Yes. We hold authorised channels for NVIDIA AI Enterprise software subscriptions. Add it to your RFQ and we quote node-aligned licensing along with the hardware.