4 × R9700 LLM benchmarks
102 results · 158 measurements · 9 sources
Device specifications
| Variant | Memory | Memory type | Bus | Bandwidth | FP16 matrix | INT8 matrix |
|---|---|---|---|---|---|---|
| R9700 ↗ | 32 GB | GDDR6 | 256-bit | 640 GB/s | 191 TFLOPS | 383 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 102 resultsSource-reported results, selected across reports by documentation quality. Match configurations before comparing values.
| View ReportLocalMaxxing | DeepSeek-V4-Flash | Token speed | — | 389.19 tok/sPrefill · median · per request | 54.27 tok/sGeneration · median · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
|---|---|---|---|---|---|---|
| View ReportGeneric LLM | DeepSeek-R1-Distill-Qwen-14B | Token speed | 41.5 tok/sThroughput · per request | — | — | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportBetterBench | Qwen3.8-Flash-Next | BetterBench | — | 3,207.8 tok/sPrefill · median · per request | — | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-27B | Token speed1 · halo-arena-v1 | 34.08 tok/sThroughput · aggregate | — | 35.58 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-35B-A3B | Token speed1 · halo-arena-v1 | 34.25 tok/sThroughput · aggregate | — | 39.85 tok/sGeneration · per request | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-35B-A3B | Token speed1 · halo-arena-v1 | 96.85 tok/sThroughput · aggregate | — | 103.47 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.8-Flash-Next | Token speed | — | — | 219 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | 17,636 tok/sPrefill · peak · unknown | — | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | DeepSeek-R1-Distill-Qwen-32B | Token speed | 40 output tokens/sToken throughput · Aggregate | — | — | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.6-35B-A3B | Token speed | 6,001.4 tok/sThroughput · median · aggregate | — | — | 4 × R9700Independent replicas · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | DeepSeek-R1-Distill-Qwen-14B | Token speed | — | — | 42 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportBetterBench | Qwen3.8-Flash-Next | BetterBench | — | 3,508.6 tok/sPrefill · median · per request | — | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-27B | Token speed1 · halo-arena-v1 | 115.22 tok/sThroughput · aggregate | — | 32.42 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-35B-A3B | Token speed1 · halo-arena-v1 | 53.68 tok/sThroughput · aggregate | — | 19.9 tok/sGeneration · per request | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-35B-A3B | Token speed1 · halo-arena-v1 | 289.86 tok/sThroughput · aggregate | — | 82.6 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportBetterBench | Qwen3.8-Flash-Next | Token speed | 172 tok/sThroughput · aggregate | — | — | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | — | 36.6 tok/sGeneration · peak · unknown | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | DeepSeek-R1-Distill-Qwen-32B | Token speed | 189 output tokens/sToken throughput · Aggregate | — | — | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.6-35B-A3B | Token speed | 5,993.3 tok/sThroughput · median · aggregate | — | — | 4 × R9700Independent replicas · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | gemma-4-26B-A4B-it | Token speed | 24.2 tok/sThroughput · per request | — | — | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportBetterBench | Qwen3.8-Flash-Next | BetterBench | — | 3,570.9 tok/sPrefill · median · per request | — | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-27B | Token speed1 · halo-arena-v1 | 14.43 tok/sThroughput · aggregate | — | 0.91 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-35B-A3B | Token speed1 · halo-arena-v1 | 3.11 tok/sThroughput · aggregate | — | 7.03 tok/sGeneration · per request | 4 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen3.6-35B-A3B | Token speed1 · halo-arena-v1 | 134.34 tok/sThroughput · aggregate | — | 29.48 tok/sGeneration · per request | 4 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
Models and software in this collection
Model families: DeepSeek R1, DeepSeek V4, GPT-OSS, Gemma 4, Llama 3.1, Qwen3.5, Qwen3.6, Qwen3.8
Engines: hipfire, llama.cpp, vLLM, vLLM + r9700-stack, vLLM Radiance
Methodology and evidence standards · Download the research dataset