AMD / ACCELERATOR VARIANT

MI355X.

AMD Instinct MI355X 288GB HBM3e · 288 GB · HBM3e

Logical ID accelerator:amd-mi355x-hbm3e-288-gb
Content version cc7e26283971c092279c11e82400ca9ccfe97e76d5694095b60b1f03db1007a8

Submitted systems

AMD · 8xMI355X_2xEPYC_9575F

1 nodes × 8 accelerators = 8 total

Topology: XGMI
Software: PyTorch 2.9.1+git8907517, ROCm 7.0.0

Pinned system specification

gpt-oss-120b · Offline: 95,004 tokens/s · source

gpt-oss-120b · Server: 82,136.1 tokens/s · source

Dell · 8xMI355X_2xEPYC_9655

1 nodes × 8 accelerators = 8 total

Topology: XGMI
Software: vLLM 0.14.0rc3.dev30+gb026cf14e.d20260212.rocm700, Pytorch 2.9.1+git8907517, ROCm 7.0.0.70000-38~22.04

Pinned system specification

gpt-oss-120b · Offline: 91,395.5 tokens/s · source

gpt-oss-120b · Server: 78,603.5 tokens/s · source

GigaComputing · 8xMI355X_2xEPYC-9575F

1 nodes × 8 accelerators = 8 total

Topology: XGMI
Software: PyTorch 2.9.1+git8907517, ROCm 7.0.0

Pinned system specification

gpt-oss-120b · Offline: 92,332.2 tokens/s · source

gpt-oss-120b · Server: 80,089.6 tokens/s · source

HPE · HPE_ProLiant_XD685_MI355X_288GBx8

1 nodes × 8 accelerators = 8 total

Topology: AMD Infinity Fabric
Software: vLLM

Pinned system specification

gpt-oss-120b · Offline: 91,846.6 tokens/s · source

gpt-oss-120b · Server: 75,354.9 tokens/s · source

ORACLE · 8xMI355X_2xEPYC_9575F

1 nodes × 8 accelerators = 8 total

Topology: AMD Infinity Fabric
Software: N/A

Pinned system specification

gpt-oss-120b · Offline: 91,706.3 tokens/s · source

gpt-oss-120b · Server: 78,186.3 tokens/s · source