RX 9070 GRE LLM benchmarks
29 results · 41 measurements · 4 sources
Device specifications
| Variant | Memory | Memory type | Bus | Bandwidth | FP16 matrix | INT8 matrix |
|---|---|---|---|---|---|---|
| RX 9070 GRE ↗ | 12 GB | GDDR6 | 192-bit | 432 GB/s | 137 TFLOPS | 274 TOPS |
Manufacturer specifications. Shared system RAM is not dedicated VRAM. Unreported values remain —; measured LLM results retain their reported configuration.
Selected reported tests
24 of 29 resultsSource-reported results, selected across reports by documentation quality. Match configurations before comparing values.
| View Reportllama-bench | DeepSeek-R1-Distill-Llama-8B | Token speed | — | — | 49.33 tok/sGeneration · mean · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
|---|---|---|---|---|---|---|
| View ReportGeneric LLM | Llama-2-13b-chat | Procyon | — | — | 25.58 tok/sGeneration · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | MLPerf Client v1.0 (selected models unreported) | Token speed1.0 · MLPerf Client | — | — | 86 tok/sGeneration · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Llama-3.1-8B-Instruct | Procyon | 1,594 pointsBenchmark score · Per request | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | DeepSeek-R1-Distill-Llama-8B | Token speed | — | 2,955.16 tok/sPrefill · mean · per request | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Llama-3.1-8B-Instruct | Procyon | — | — | 65.51 tok/sGeneration · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | MLPerf Client v1.0 (selected models unreported) | Latency1.0 · MLPerf Client | 520 msTime to first token · Per request | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | gpt-oss-20b | Token speed | — | — | 131.07 tok/sGeneration · mean · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Mistral-7B (Procyon) (checkpoint unspecified) | Procyon | — | — | 76.04 tok/sGeneration · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | gpt-oss-20b | Token speed | — | 943.71 tok/sPrefill · mean · per request | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Phi-3.5-mini-instruct | Procyon | — | — | 105.95 tok/sGeneration · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | granite-3.0-3b-a800m-instruct | Token speed | — | — | 254.58 tok/sGeneration · mean · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | granite-3.0-3b-a800m-instruct | Token speed | — | 9,515.59 tok/sPrefill · mean · per request | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3.1-Tulu-3-8B | Token speed | — | — | 49.34 tok/sGeneration · mean · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3.1-Tulu-3-8B | Token speed | — | 2,961.98 tok/sPrefill · mean · per request | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Mistral-7B-Instruct-v0.3 | Token speed | — | — | 52.09 tok/sGeneration · mean · per request | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Mistral-7B-Instruct-v0.3 | Token speed | — | 2,964.4 tok/sPrefill · mean · per request | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3.1-Tulu-3-8B | Power & energy2K input · 0 output | 36 WPower draw · Per GPU | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3.1-Tulu-3-8B | Power & energy2K input · 0 output | 205.31 WPower draw · Per GPU | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3.1-Tulu-3-8B | Power & energy2K input · 0 output | 220 WPower draw · Per GPU | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3.1-Tulu-3-8B | Other LLM measurements2K input · 0 output | 14.43 tokens/s/Wtokens per second per watt · Per request | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Mistral-7B-Instruct-v0.3 | Power & energy0 input · 128 output | 27 WPower draw · Per GPU | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Mistral-7B-Instruct-v0.3 | Power & energy0 input · 128 output | 215.18 WPower draw · Per GPU | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Mistral-7B-Instruct-v0.3 | Power & energy0 input · 128 output | 220 WPower draw · Per GPU | — | — | 1 × RX 9070 GRESingle GPU · 12 GB VRAM/GPUConfiguration
|
Models and software in this collection
Model families: DeepSeek R1, GPT-OSS, Granite 3.0, Llama 2, Llama 3.1, Mistral, Phi 3.5
Engines: ONNX Runtime, llama.cpp
Methodology and evidence standards · Download the research dataset