HPE ProLiant Compute XD685 (Instinct MI355X)
Overview
Performance Metrics
| Precision | Peak (TFLOPS) | Per GPU (TFLOPS) | TFLOPS/W |
|---|---|---|---|
| FP64 | 629 | 79 | — |
| FP32 | 1,258 | 157 | — |
| FP16 | 20,133 | 2,517 | — |
| BF16 | 20,133 | 2,517 | — |
| FP8 | 40,266 | 5,033 | — |
| FP4 | 80,530 | 10,066 | — |
| INT8 | 40,266 | 5,033 | — |
| FP6 | 80,530 | 10,066 | — |
All figures are dense. Vendors commonly headline the sparse number, which is twice the dense one.
Configuration
- Accelerator
- Instinct MI355X OAM
- Count
- 8
- Memory per GPU
- 288 GB
- Bandwidth per GPU
- 8.0 TB/s
Power & Efficiency
- Aggregate Power
- —
- TDP per GPU
- 1400 W
- FP8 Efficiency
- —
- Aggregate Bandwidth
- 64.0 TB/s
System Details
Interconnect
4th Gen AMD Infinity Fabric across the eight-OAM OCP baseboard; 4x FHHL PCIe Gen5 x16 slots for fabric, NIC or DPU
Notes
The eight-way XD685 in its AMD Instinct MI355X configuration, liquid cooled, with accelerator kit S6J64A. This is an all-AMD machine, pairing eight MI355X OAM modules with two fifth-generation AMD EPYC processors. No chassis part number is recorded: HPE names the accelerator kit but never states which chassis it requires, and the two candidates differ. Compute comes from AMD's eight-GPU Instinct platform, where the sparse figures cover only the four datatypes AMD publishes them for, being FP8, FP16, BF16 and INT8; AMD gives no structured-sparsity figure for MXFP4 or MXFP6. HPE publishes no system power draw for the XD685, only twelve 3,000 W Titanium supplies, so nothing is summed.