Specifications
Compute
1,835 TFLOPS BF16/FP8 Matrix
Memory
128GB HBM2e, 3.67 TB/s BW
Cores
64 TPC + 8 MME (Matrix Engines)
Networking
24×200GbE RoCE RDMA NICs
Interface
PCIe Gen5 x16 (128 GB/s)
TDP
600W, passive cooling
Process
TSMC 5nm
Form Factor
Dual-slot FHFL, 10.5" length
SKU
HL-338 PCIe CEM Card
Use Case
Deployed in hyperscale AI data centers, cloud GPU clusters, enterprise AI inference platforms, and HPC environments. Ideal for large language model training, distributed inference, vector databases, and real-time AI pipelines with demanding throughput requirements.
Request Quote — Intel Gaudi 3 HL-338 PCIe Accelerator
Contact us for pricing, stock availability, and bulk orders.
Email for Quote