GeekRadar / The comparison desk
COMPARE.
Tested systems8 × AMD Instinct MI350X 288GB HBM3e
Tested system
- Processor
- AMD EPYC 9755
- Graphics
- AMD Instinct MI350X 288GB HBM3e
- Accelerator count
- 8
- Nodes
- 1
Recorded results
Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.
MLPerf Inference / gpt-oss-120bOffline · v6.1 · Tokens/s
8 × AMD Instinct MI350X 288GB HBM3e
97,019.70* Tokens/s
- Accuracy target / result
- exact_match: 83.639
- Weight precision
- fp4
- Division
- closed
- System type
- datacenter
- Availability
- available
- Cores per processor
- 128
- Processors per node
- 2
- Nodes
- 1
- Total accelerators
- 8
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Operating system
- Ubuntu 24.04.4 LTS
MLPerf Inference / gpt-oss-120bServer · v6.1 · Tokens/s
8 × AMD Instinct MI350X 288GB HBM3e
87,549.50* Tokens/s
- Accuracy target / result
- exact_match: 83.316
- Weight precision
- fp4
- Division
- closed
- System type
- datacenter
- Availability
- available
- Cores per processor
- 128
- Processors per node
- 2
- Nodes
- 1
- Total accelerators
- 8
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Operating system
- Ubuntu 24.04.4 LTS
MLPerf Inference / llama2-70b-99.9Offline · v6.1 · Tokens/s
8 × AMD Instinct MI350X 288GB HBM3e
84,016.40* Tokens/s
- Accuracy target / result
- ROUGE1: 44.4551 ROUGE2: 22.0787 ROUGEL: 28.8207 TOKENS_PER_SAMPLE: 283.0
- Weight precision
- fp4
- Division
- closed
- System type
- datacenter
- Availability
- available
- Cores per processor
- 128
- Processors per node
- 2
- Nodes
- 1
- Total accelerators
- 8
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Operating system
- Ubuntu 24.04.4 LTS
MLPerf Inference / llama2-70b-99.9Server · v6.1 · Tokens/s
8 × AMD Instinct MI350X 288GB HBM3e
85,004.80* Tokens/s
- Accuracy target / result
- ROUGE1: 44.5423 ROUGE2: 22.1706 ROUGEL: 28.9236 TOKENS_PER_SAMPLE: 284.5
- Weight precision
- fp4
- Division
- closed
- System type
- datacenter
- Availability
- available
- Cores per processor
- 128
- Processors per node
- 2
- Nodes
- 1
- Total accelerators
- 8
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Operating system
- Ubuntu 24.04.4 LTS
MLPerf Inference / llama3.1-8bOffline · v6.1 · Tokens/s
8 × AMD Instinct MI350X 288GB HBM3e
129,943* Tokens/s
- Accuracy target / result
- ROUGE1: 38.5694 ROUGE2: 16.0367 ROUGEL: 24.4456 ROUGELSUM: 35.7078 GEN_LEN: 8150329
- Weight precision
- fp8
- Division
- closed
- System type
- datacenter
- Availability
- available
- Cores per processor
- 128
- Processors per node
- 2
- Nodes
- 1
- Total accelerators
- 8
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Operating system
- Ubuntu 24.04.4 LTS
MLPerf Inference / llama3.1-8bServer · v6.1 · Tokens/s
8 × AMD Instinct MI350X 288GB HBM3e
121,706* Tokens/s
- Accuracy target / result
- ROUGE1: 38.5208 ROUGE2: 16.0104 ROUGEL: 24.424 ROUGELSUM: 35.666 GEN_LEN: 8168845
- Weight precision
- fp8
- Division
- closed
- System type
- datacenter
- Availability
- available
- Cores per processor
- 128
- Processors per node
- 2
- Nodes
- 1
- Total accelerators
- 8
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Operating system
- Ubuntu 24.04.4 LTS
Processing & platform
- Processor
- AMD EPYC 9755
- Graphics
- AMD Instinct MI350X 288GB HBM3e
Connectivity & software
- Operating system
- Ubuntu 24.04.4 LTS
System configuration
- Reported host core count
- 128
- Processors per node
- 2
- Accelerator count
- 8
- Nodes
- 1
- Software stack
- Pytorch 2.10, ROCm 7.2.2
- Submission division
- closed
0 selected Compare 0
* Provisional data