NVIDIA DGX Spark
GPUBlackwell (GB10) 架构;128GB 显存,带宽 273 GB/s,TDP 140W,CUDA 核心 未知。发布日期:2025-01-07。来源:LCZ 硬件知识库,缺项经 NVIDIA 官方资料补充。
85
运行记录
22
覆盖模型
6
覆盖框架
L0 自报 × 9 · L2 跨框架验证 × 76
证据分布
典型性能表现
同硬件上各模型 × 框架 × 量化的已上架实测均值。
- vLLM · IQ1_S Decode 306.37 · Prefill 1243.41 ·
- Ollama · IQ2_XS Decode 258.67 · Prefill 1172.95 ·
- llama.cpp · Q6_K Decode 110.08 · Prefill 496.59 ·
- Strata · Q3_K_M Decode 209.28 · Prefill 1036.76 ·
- Ollama · IQ3_S Decode 119.95 · Prefill 607.06 ·
- Strata · Q4_K_S Decode 169.69 · Prefill 928.54 ·
- vLLM · IQ2_XS Decode 310.29 · Prefill 1000.44 ·
- Ollama · IQ2_XS Decode 348.98 · Prefill 1689.57 ·
- Ollama · Q3_K_M Decode 387.92 · Prefill 1612.07 ·
- vLLM · IQ3_S Decode 178.22 · Prefill 972.1 ·
- vLLM · Q5_K_M Decode 122.43 · Prefill 399.31 ·
- Strata · Q3_K_S Decode 245.76 · Prefill 803.01 ·
- vLLM · Q3_K_M Decode 167.2 · Prefill 861.79 ·
- Strata · Q5_K_S Decode 197.21 · Prefill 856.13 ·
- llama.cpp · Q3_K_M Decode 206.5 · Prefill 1076.24 ·
- Strata · Q3_K_S Decode 338.04 · Prefill 1544.38 ·
- Strata · Q5_K_M Decode 221.01 · Prefill 864.66 ·
- vLLM · Q2_K Decode 374.76 · Prefill 1672.78 ·
- vLLM · Q5_K_M Decode 240.21 · Prefill 1012.49 ·
- Strata · IQ3_XS Decode 315.95 · Prefill 1231.57 ·
- Strata · Q4_K_S Decode 375.7 · Prefill 1600.4 ·
- vLLM · IQ1_S Decode 291.5 · Prefill 1360.04 ·
- vLLM · Q5_K_M Decode 151.71 · Prefill 832.95 ·
- Strata · Q5_K_M Decode 220.61 · Prefill 780.86 ·
- Ollama · Q3_K_L Decode 131.46 · Prefill 705.73 ·
- Ollama · Q4_K_S Decode 128.78 · Prefill 450.1 ·
- Strata · IQ2_M Decode 261.64 · Prefill 916.28 ·
- Strata · IQ2_XS Decode 259.54 · Prefill 1190.55 ·
- llama.cpp · IQ2_XS Decode 344.66 · Prefill 1710.46 ·
- Strata · IQ2_XS Decode 295.77 · Prefill 1437.09 ·
- Strata · Q5_K_M Decode 294.11 · Prefill 1222.22 ·
- vLLM · Q3_K_M Decode 258.63 · Prefill 1322.45 ·
- vLLM · Q5_K_M Decode 183.75 · Prefill 675.77 ·
- Ollama · IQ2_M Decode 292.67 · Prefill 1302.54 ·
- vLLM · IQ1_S Decode 353.22 · Prefill 1765.39 ·
- vLLM · Q5_K_S Decode 364.71 · Prefill 1715.93 ·
- Ollama · IQ2_M Decode 338.92 · Prefill 1553.22 ·
- vLLM · IQ2_M Decode 380.91 · Prefill 2066.12 ·
- vLLM · IQ3_XS Decode 360.01 · Prefill 1436.84 ·
- Strata · Q3_K_L Decode 389.31 · Prefill 1363.67 ·
- Strata · Q4_K_M Decode 326.71 · Prefill 1358.52 ·
- llama.cpp · Q2_K Decode 258.21 · Prefill 920.61 ·
- vLLM · FP16 Decode 107.3 · Prefill 444.7 ·
- vLLM · Q2_K Decode 266.26 · Prefill 1108.22 ·
- vLLM · Q3_K_L Decode 219.42 · Prefill 756.56 ·
- llama.cpp · IQ2_M Decode 181.94 · Prefill 684.54 ·
- vLLM · IQ2_M Decode 210.69 · Prefill 1114.84 ·
- vLLM · Q8_0 Decode 101.73 · Prefill 523.44 ·
- Ollama · IQ3_XS Decode 391.28 · Prefill 1950.82 ·
- Ollama · IQ4_XS Decode 338.64 · Prefill 1719.5 ·
Fit 判定
按显存与内存带宽做物理估算:选择模型,查看各量化能否装下及预计速度
估算公式:tok/s =(带宽 ÷ 模型占用)× 0.55,借鉴 llmfit;实测以运行记录为准
核心规格
- 厂商
- NVIDIA
- 架构
- Blackwell (GB10)
- 显存
- 128 GB
- 显存带宽
- 273 GB/s
- TDP 功耗
- 140 W
- 发布日期
- 2025-01-07
运行记录
共 50 组配置
| 模型 | 框架 | 量化 | Decode | 生成耗时 | 样本数 |
|---|---|---|---|---|---|
| Qwen3.5-4B | vLLM | IQ1_S | 306.4 tok/s | — | 1 |
| Qwen3.5-4B | Ollama | IQ2_XS | 258.7 tok/s | — | 1 |
| Qwen3.5-9B | llama.cpp | Q6_K | 110.1 tok/s | — | 1 |
| Qwen3.5-9B | Strata | Q3_K_M | 209.3 tok/s | — | 1 |
| Qwen3.5-27B | Ollama | IQ3_S | 120 tok/s | — | 1 |
| Qwen3.5-27B | Strata | Q4_K_S | 169.7 tok/s | — | 1 |
| Qwen3.5-0.8B | vLLM | IQ2_XS | 310.3 tok/s | — | 1 |
| Qwen3.5-0.8B | Ollama | IQ2_XS | 349 tok/s | — | 1 |
| Qwen3.5-0.8B | Ollama | Q3_K_M | 387.9 tok/s | — | 1 |
| Qwen3.8-9B | vLLM | IQ3_S | 178.2 tok/s | — | 1 |
| Qwen3.8-9B | vLLM | Q5_K_M | 122.4 tok/s | — | 1 |
| Qwen3.8-9B | Strata | Q3_K_S | 245.8 tok/s | — | 1 |
| Qwen3-8B | vLLM | Q3_K_M | 167.2 tok/s | — | 1 |
| Qwen3-8B | Strata | Q5_K_S | 197.2 tok/s | — | 1 |
| Qwen3-4B | llama.cpp | Q3_K_M | 206.5 tok/s | — | 1 |
| Qwen3-4B | Strata | Q3_K_S | 338 tok/s | — | 1 |
| Qwen3-4B | Strata | Q5_K_M | 221 tok/s | — | 2 |
| Qwen3-1.7B | vLLM | Q2_K | 374.8 tok/s | — | 1 |
| Qwen3-1.7B | vLLM | Q5_K_M | 240.2 tok/s | — | 1 |
| Qwen3-1.7B | Strata | IQ3_XS | 315.9 tok/s | — | 1 |
| Qwen3-1.7B | Strata | Q4_K_S | 375.7 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | vLLM | IQ1_S | 291.5 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | vLLM | Q5_K_M | 151.7 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | Strata | Q5_K_M | 220.6 tok/s | — | 1 |
| Qwen3-Coder-Next | Ollama | Q3_K_L | 131.5 tok/s | — | 1 |
| Qwen3-Coder-Next | Ollama | Q4_K_S | 128.8 tok/s | — | 1 |
| Qwen3-Coder-Next | Strata | IQ2_M | 261.6 tok/s | — | 1 |
| Qwen3-Coder-Next | Strata | IQ2_XS | 259.5 tok/s | — | 1 |
| Qwen2.5-0.5B-Instruct | llama.cpp | IQ2_XS | 344.7 tok/s | — | 1 |
| Qwen2.5-0.5B-Instruct | Strata | IQ2_XS | 295.8 tok/s | — | 1 |
| Qwen2.5-0.5B-Instruct | Strata | Q5_K_M | 294.1 tok/s | — | 1 |
| Qwen2.5-Omni-3B | vLLM | Q3_K_M | 258.6 tok/s | — | 1 |
| Qwen2.5-Omni-3B | vLLM | Q5_K_M | 183.8 tok/s | — | 1 |
| Qwen2.5-Omni-3B | Ollama | IQ2_M | 292.7 tok/s | — | 1 |
| Qwen3-Embedding-0.6B | vLLM | IQ1_S | 353.2 tok/s | — | 1 |
| Qwen3-Embedding-0.6B | vLLM | Q5_K_S | 364.7 tok/s | — | 1 |
| Qwen3-Embedding-0.6B | Ollama | IQ2_M | 338.9 tok/s | — | 1 |
| gemma-4-E2B-it | vLLM | IQ2_M | 380.9 tok/s | — | 1 |
| gemma-4-E2B-it | vLLM | IQ3_XS | 360 tok/s | — | 1 |
| gemma-4-E2B-it | Strata | Q3_K_L | 389.3 tok/s | — | 1 |
| gemma-4-E2B-it | Strata | Q4_K_M | 326.7 tok/s | — | 1 |
| gemma-4-E4B-it | llama.cpp | Q2_K | 258.2 tok/s | — | 1 |
| gemma-4-E4B-it | vLLM | FP16 | 107.3 tok/s | — | 1 |
| Qwen3.5-4B | vLLM | Q2_K | 266.3 tok/s | — | 1 |
| Qwen3.5-4B | vLLM | Q3_K_L | 219.4 tok/s | — | 1 |
| Qwen3.5-9B | llama.cpp | IQ2_M | 181.9 tok/s | — | 1 |
| Qwen3.5-9B | vLLM | IQ2_M | 210.7 tok/s | — | 1 |
| Qwen3.5-9B | vLLM | Q8_0 | 101.7 tok/s | — | 1 |
| Qwen3.5-0.8B | Ollama | IQ3_XS | 391.3 tok/s | — | 1 |
| Qwen3.5-0.8B | Ollama | IQ4_XS | 338.6 tok/s | — | 1 |