Brand: Lenovo | Category: GPUs
SKU: 4X67A89325 | Part #: 4X67A89325 | MPN: 4X67A89325
Contact for Pricing — Request a Quote
The Lenovo ThinkSystem NVIDIA H100 NVL 94GB PCIe Gen5 Passive GPU (part number 4X67A89325) is a data centre accelerator for Lenovo ThinkSystem servers. It carries 94 GB of HBM3 memory with 3.9 TB/s of memory bandwidth and has ECC support.
The GPU provides 16,896 FP32 CUDA cores, with peak FP64 performance of 34 TFLOPS, peak FP64 Tensor Core performance of 67 TFLOPS and INT8 integer performance of 3,958 TOPS*. Interconnect bandwidth is 600 GB/sec over NVLink and 128 GB/sec over PCIe Gen5. Maximum power consumption is 400 W, and the card uses a passive air-cooled thermal solution, so it relies on airflow from the host server chassis.
This GPU is intended for server environments running compute-intensive workloads such as AI training and inference, high-performance computing and large-scale data analytics, where high memory capacity and bandwidth are needed.
Send the part number through our RFQ form and our team will reply within 48 hours.
| Brand | Lenovo |
| Category | GPUs |
| SKU | 4X67A89325 |
| Part Number | 4X67A89325 |
| Condition | New |
| Manufacturer Part Number | 4X67A89325 |
| GPU Memory | 94 GB HBM3 |
| Memory Bandwidth | 3.9 TB/s |
| FP32 CUDA Cores | 16,896 FP32 CUDA Cores |
| Peak FP64 performance | 34 TFLOPS |
| Peak FP64 Tensor Core performance | 67 TFLOPS |
| INT8 Integer Performance | 3,958 TOPS* |
| ECC | Yes |
| Interconnect Bandwidth | NVLink: 600 GB/sec PCIe Gen5: 128 GB/sec |
| Max Power Consumption | 400 W |
| Thermal Solution | Passive air-cooled |
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.
Yes — Omnixon supplies 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.
