Radeon PRO W7500 (8GB) 提供 8 GB 显存/内存与 176 GB/s 带宽,支持通过 ROCm 与 Vulkan 加速进行本地推理。
总显存
8 GB
可用显存上限
8 GB
显存带宽
176 GB/s
最大推理模型级别
8B 模型 (Q4)
热设计功耗
70W
随上下文长度增长的显存需求评估。当曲线超过参考线时将导致显存不足(OOM)。
量化精度:
主流开源大模型在 Radeon PRO W7500 (8GB) 上的推理速度、显存兼容性与生成吞吐表现。
排序:
上下文长度:
KV 缓存:
显卡数量:
| 排名 | 模型 | 参数量 | 量化精度 | 预估 TPS | TTFT | 可运行 |
|---|---|---|---|---|---|---|
#51 | Inkling 2 Small | 2B (17B active) | INT4 Q8 FP16 | 15 tok/s 8 tok/s 4 tok/s | ~8.5s ~8.6s ~8.7s | |
#129 | Qwen3.5-9B | 9B | INT4 | 26 tok/s | ~4.7s | |
#145 | Qwen3.5-4B | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
#153 | K2 Horizon 7B | 7B | INT4 | 32 tok/s | ~3.7s | |
#157 | NVIDIA Nemotron 3 Nano 30B-A3B | 3.5B (30B active) | INT4 Q8 | 8 tok/s 4 tok/s | ~15.0s ~16.4s | |
#164 | Granite 4.2 8B | 8B | INT4 | 29 tok/s | ~4.2s | |
#167 | Gemma 4 E4B | 8B | INT4 | 29 tok/s | ~4.2s | |
#192 | Gemma 3 12B | 12B | INT4 | 18 tok/s | ~7.4s | |
#196 | Ministral 3 8B | 8B | INT4 | 18 tok/s | ~5.0s | |
#197 | OLMo 3 7B Think | 7B | INT4 | 21 tok/s | ~4.1s | |
#198 | Qwen3.5-2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
#199 | MiniCPM5-2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
#205 | Gemma 4 E2B | 5.1B | INT4 Q8 | 42 tok/s 24 tok/s | ~2.7s ~3.0s | |
#209 | DeepSeek-R1 8B | 8B | INT4 | 30 tok/s | ~4.2s | |
#211 | Qwen3-4B | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
#212 | Ministral 3 3B | 3B | INT4 Q8 | 40 tok/s 30 tok/s | ~1.7s ~1.7s | |
#214 | Qwen3.5-0.8B | 0.8B | INT4 Q8 FP16 | 144 tok/s 107 tok/s 68 tok/s | ~473 ms ~475 ms ~481 ms | |
#215 | LFM3 1.7B | 1.7B | INT4 Q8 FP16 | 93 tok/s 63 tok/s 37 tok/s | ~991 ms ~996 ms ~1.0s | |
#217 | Qwen3-8B | 8B | INT4 | 29 tok/s | ~4.2s | |
#218 | Gemma 3 4B | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
#219 | Qwen3-1.7B | 1.7B | INT4 Q8 FP16 | 93 tok/s 63 tok/s 37 tok/s | ~991 ms ~996 ms ~1.0s | |
#220 | Gemma 2 9B | 9B | INT4 | 26 tok/s | ~4.7s | |
#221 | Gemma 3 270M | 0.27B | INT4 Q8 FP16 | 183 tok/s 161 tok/s 125 tok/s | ~160 ms ~161 ms ~162 ms | |
#222 | DeepSeek-R1 7B | 7B | INT4 Q8 | 33 tok/s 18 tok/s | ~3.7s ~4.4s | |
#223 | Qwen3-0.6B | 0.6B | INT4 Q8 FP16 | 166 tok/s 128 tok/s 84 tok/s | ~351 ms ~352 ms ~356 ms | |
#225 | Llama 3.1 8B | 8B | INT4 | 29 tok/s | ~4.2s | |
#228 | Llama 3 8B | 8B | INT4 | 29 tok/s | ~4.2s | |
#229 | Phi-4-Mini | 3.8B | INT4 Q8 | 53 tok/s 33 tok/s | ~2.1s ~2.1s | |
#231 | Qwen2.5-7B | 7B | INT4 | 32 tok/s | ~3.7s | |
#232 | DeepSeek-R1 1.5B | 1.5B | INT4 Q8 FP16 | 101 tok/s 69 tok/s 41 tok/s | ~877 ms ~881 ms ~891 ms | |
#233 | LFM3 VL 2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
#234 | Ministral-8B-2410 | 8B | INT4 | 29 tok/s | ~4.2s | |
#235 | Llama 3.2 3B | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
#237 | Phi-3-mini | 3.8B | INT4 Q8 | 53 tok/s 33 tok/s | ~2.1s ~2.1s | |
#238 | Gemma 3n E2B IT | 6B (2B active) | INT4 | 38 tok/s | ~1.4s | |
#239 | Mistral-7B-Instruct-v0.1 | 7.3B | INT4 | 31 tok/s | ~3.9s | |
#240 | Llama 3.2 1B | 1B | INT4 Q8 FP16 | 127 tok/s 92 tok/s 57 tok/s | ~600 ms ~603 ms ~610 ms | |
#241 | Gemma 2 2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
#243 | OLMo 3 7B Instruct | 7B | INT4 | 21 tok/s | ~4.1s | |
#245 | Qwen2-7B | 7B | INT4 | 32 tok/s | ~3.7s | |
#246 | Phi-3-small | 7B | INT4 | 32 tok/s | ~3.7s | |
#247 | Gemma 3 1B | 1B | INT4 Q8 FP16 | 127 tok/s 92 tok/s 57 tok/s | ~600 ms ~603 ms ~610 ms | |
#248 | Mistral-7B-Instruct-v0.2 | 7.3B | INT4 | 31 tok/s | ~3.9s | |
#250 | Gemma 1 7B | 7B | INT4 | 21 tok/s | ~4.1s | |
#251 | Gemma 1 2B | 2B | INT4 Q8 FP16 | 90 tok/s 58 tok/s 33 tok/s | ~1.2s ~1.2s ~1.2s | |
#252 | ChatGLM3-6B | 6B | INT4 | 24 tok/s | ~3.5s | |
#253 | ChatGLM2-6B | 6B | INT4 | 24 tok/s | ~3.5s | |
#254 | ChatGLM-6B | 6B | INT4 | 24 tok/s | ~3.5s | |
未上榜模型(按推理速度排序) | ||||||
- | SmolVLM2 256M Video Instruct | 0.26B | INT4 Q8 FP16 | 219 tok/s 187 tok/s 141 tok/s | ~154 ms ~155 ms ~156 ms | |
- | DINOv3 ViT-L/16 | 0.3B | INT4 Q8 FP16 | 211 tok/s 177 tok/s 131 tok/s | ~177 ms ~177 ms ~179 ms | |
- | Qwen2-0.5B | 0.5B | INT4 Q8 FP16 | 179 tok/s 141 tok/s 96 tok/s | ~291 ms ~293 ms ~296 ms | |
- | Qwen2.5-0.5B | 0.5B | INT4 Q8 FP16 | 179 tok/s 141 tok/s 96 tok/s | ~291 ms ~293 ms ~296 ms | |
- | ERNIE-4.5-0.3B | 0.3B | INT4 Q8 FP16 | 174 tok/s 152 tok/s 117 tok/s | ~178 ms ~178 ms ~180 ms | |
- | ERNIE-4.5-0.3B-Base | 0.3B | INT4 Q8 FP16 | 174 tok/s 152 tok/s 117 tok/s | ~178 ms ~178 ms ~180 ms | |
- | Liquid D1-Omni 600M | 0.6B | INT4 Q8 FP16 | 166 tok/s 128 tok/s 84 tok/s | ~351 ms ~352 ms ~356 ms | |
- | PixArt-α XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 165 tok/s 127 tok/s 84 tok/s | ~356 ms ~358 ms ~362 ms | |
- | PixArt-Σ XL 2 1024-MS | 0.61B | INT4 Q8 FP16 | 165 tok/s 127 tok/s 84 tok/s | ~356 ms ~358 ms ~362 ms | |
- | Nemotron 3.5 Parse | 0.63B | INT4 Q8 FP16 | 162 tok/s 124 tok/s 82 tok/s | ~370 ms ~372 ms ~376 ms | |
- | SD 2.1 Turbo | 0.86B | INT4 Q8 FP16 | 139 tok/s 102 tok/s 64 tok/s | ~510 ms ~512 ms ~518 ms | |
- | Stable Diffusion 1.4 | 0.86B | INT4 Q8 FP16 | 139 tok/s 102 tok/s 64 tok/s | ~510 ms ~512 ms ~518 ms | |
- | Stable Diffusion 2.1 | 0.86B | INT4 Q8 FP16 | 139 tok/s 102 tok/s 64 tok/s | ~510 ms ~512 ms ~518 ms | |
- | Falcon-1B | 1B | INT4 Q8 FP16 | 137 tok/s 97 tok/s 59 tok/s | ~600 ms ~603 ms ~609 ms | |
- | Falcon3-1B | 1B | INT4 Q8 FP16 | 127 tok/s 92 tok/s 57 tok/s | ~600 ms ~603 ms ~610 ms | |
- | MiniCPM-V 4.7 1B | 1B | INT4 Q8 FP16 | 127 tok/s 92 tok/s 57 tok/s | ~600 ms ~603 ms ~610 ms | |
- | Kandinsky 2.2 Decoder | 1.22B | INT4 Q8 FP16 | 114 tok/s 80 tok/s 49 tok/s | ~723 ms ~726 ms ~734 ms | |
- | Wan 2.1 T2V 1.3B | 1.3B | INT4 Q8 FP16 | 110 tok/s 77 tok/s 46 tok/s | ~764 ms ~768 ms ~777 ms | |
- | Qwen2-1.5B | 1.5B | INT4 Q8 FP16 | 101 tok/s 69 tok/s 41 tok/s | ~877 ms ~881 ms ~891 ms | |
- | Qwen2.5-1.5B | 1.5B | INT4 Q8 FP16 | 101 tok/s 69 tok/s 41 tok/s | ~877 ms ~881 ms ~891 ms | |
- | SVD Image-to-Video XT | 1.5B | INT4 Q8 FP16 | 101 tok/s 69 tok/s 41 tok/s | ~877 ms ~881 ms ~891 ms | |
- | SVD Image-to-Video XT 1.1 | 1.5B | INT4 Q8 FP16 | 101 tok/s 69 tok/s 41 tok/s | ~877 ms ~881 ms ~891 ms | |
- | LFM3 1.7B Base | 1.7B | INT4 Q8 FP16 | 93 tok/s 63 tok/s 37 tok/s | ~991 ms ~996 ms ~1.0s | |
- | Spark X2.5 1.7B | 1.7B | INT4 Q8 FP16 | 93 tok/s 63 tok/s 37 tok/s | ~991 ms ~996 ms ~1.0s | |
- | Hy-MT2 1.8B | 1.8B | INT4 Q8 FP16 | 90 tok/s 60 tok/s 35 tok/s | ~1.0s ~1.0s ~1.1s | |
- | Moondream 3.1 | 9B (2B active) | INT4 | 84 tok/s | ~1.3s | |
- | CogVideoX 2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
- | LFM3 Audio 2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
- | LFM3 Omni 2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
- | Liquid D1 2B | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
- | LTX-Video 0.9.1 | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
- | Stable Diffusion 3 Medium | 2B | INT4 Q8 FP16 | 83 tok/s 56 tok/s 32 tok/s | ~1.2s ~1.2s ~1.2s | |
- | Kandinsky 3.0 | 2.1B | INT4 Q8 FP16 | 81 tok/s 54 tok/s 31 tok/s | ~1.2s ~1.2s ~1.2s | |
- | LTX-2 | 2.2B | INT4 Q8 FP16 | 79 tok/s 52 tok/s 29 tok/s | ~1.3s ~1.3s ~1.3s | |
- | SDXL 1.0 Refiner | 2.31B | INT4 Q8 FP16 | 76 tok/s 50 tok/s 28 tok/s | ~1.3s ~1.3s ~1.3s | |
- | Stable Diffusion 3.5 Medium | 2.5B | INT4 Q8 FP16 | 72 tok/s 47 tok/s 26 tok/s | ~1.4s ~1.4s ~1.4s | |
- | SDXL 1.0 Base | 2.56B | INT4 Q8 FP16 | 71 tok/s 46 tok/s 24 tok/s | ~1.4s ~1.4s ~1.6s | |
- | SDXL Turbo | 2.56B | INT4 Q8 FP16 | 71 tok/s 46 tok/s 24 tok/s | ~1.4s ~1.4s ~1.6s | |
- | Alpamayo2 Edge | 2.65B | INT4 Q8 FP16 | 69 tok/s 45 tok/s 24 tok/s | ~1.5s ~1.5s ~1.6s | |
- | Falcon-3B | 3B | INT4 Q8 FP16 | 68 tok/s 42 tok/s 22 tok/s | ~1.7s ~1.7s ~1.8s | |
- | CroissantLLM Base | 1.3B | INT4 Q8 FP16 | 66 tok/s 54 tok/s 37 tok/s | ~770 ms ~773 ms ~782 ms | |
- | Phi-1 | 1.3B | INT4 Q8 FP16 | 66 tok/s 54 tok/s 37 tok/s | ~770 ms ~773 ms ~782 ms | |
- | Phi-1.5 | 1.3B | INT4 Q8 FP16 | 66 tok/s 54 tok/s 37 tok/s | ~770 ms ~773 ms ~782 ms | |
- | DeepSeek-R1 3B | 3B | INT4 Q8 FP16 | 65 tok/s 41 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Falcon3-3B | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | LFM2-VL 3B | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Liquid D1 3B | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Liquid D1-VL 3B | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Liquid LFM2-VL 3B D1 | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Qwen2.5-3B | 3B | INT4 Q8 FP16 | 64 tok/s 40 tok/s 21 tok/s | ~1.7s ~1.7s ~1.8s | |
- | Mellum 2.1 12B A2.5B Thinking | 12B (2.5B active) | INT4 | 58 tok/s | ~1.9s | |
- | FLUX.2 Klein 4B | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
- | FrogNano 4B | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
- | Mage-VL Cowork | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
- | Spark X2.5 4B | 4B | INT4 Q8 | 51 tok/s 32 tok/s | ~2.2s ~2.2s | |
- | EmbeddingGemma 2 | 2B | INT4 Q8 FP16 | 49 tok/s 39 tok/s 24 tok/s | ~1.2s ~1.2s ~1.3s | |
- | CogVideoX 1.5 5B | 5B | INT4 Q8 | 43 tok/s 25 tok/s | ~2.7s ~2.9s | |
- | CogVideoX 5B | 5B | INT4 Q8 | 43 tok/s 25 tok/s | ~2.7s ~2.9s | |
- | Wan 2.2 TI2V 5B | 5B | INT4 Q8 | 43 tok/s 25 tok/s | ~2.7s ~2.9s | |
- | Phi-2 | 2.7B | INT4 Q8 FP16 | 42 tok/s 32 tok/s 18 tok/s | ~1.5s ~1.5s ~1.8s | |
- | MaLLaM-3B | 3B | INT4 Q8 | 40 tok/s 30 tok/s | ~1.7s ~1.7s | |
- | SmolLM3 3B | 3B | INT4 Q8 | 40 tok/s 30 tok/s | ~1.7s ~1.7s | |
- | Falcon-7B | 7B | INT4 Q8 | 34 tok/s 18 tok/s | ~3.7s ~4.4s | |
- | AuraFlow v0.2 | 6.8B | INT4 Q8 | 33 tok/s 18 tok/s | ~3.6s ~4.3s | |
- | AuraFlow v0.3 | 6.8B | INT4 Q8 | 33 tok/s 18 tok/s | ~3.6s ~4.3s | |
- | Falcon3-7B | 7B | INT4 | 32 tok/s | ~3.7s | |
- | FastFrames-L | 7B | INT4 | 32 tok/s | ~3.7s | |
- | Hy-MT2 7B | 7B | INT4 | 32 tok/s | ~3.7s | |
- | Mistral-7B-v0.1 | 7.3B | INT4 | 31 tok/s | ~3.9s | |
- | Youtu Parsing Omni | 8B | INT4 | 30 tok/s | ~4.2s | |
- | Stable Diffusion 3.5 Large | 8.1B | INT4 | 29 tok/s | ~4.3s | |
- | Stable Diffusion 3.5 Large Turbo | 8.1B | INT4 | 29 tok/s | ~4.3s | |
- | FLUX.2 Klein 9B | 9B | INT4 | 26 tok/s | ~4.7s | |
- | LensVLM 9B | 9B | INT4 | 26 tok/s | ~4.7s | |
- | MiMo V2.6 Distill Qwen 9B | 9B | INT4 | 26 tok/s | ~4.7s | |
- | ChatGLM3-6B-32K | 6B | INT4 | 24 tok/s | ~3.5s | |
- | Yi-6B | 6B | INT4 | 24 tok/s | ~3.5s | |
- | Falcon2-11B | 11B | INT4 | 22 tok/s | ~6.3s | |
- | Falcon3-10B | 10B | INT4 | 22 tok/s | ~5.7s | |
- | Mochi 1 Preview | 10B | INT4 | 22 tok/s | ~5.7s | |
- | Hunyuan Lite | 7B | INT4 | 21 tok/s | ~4.1s | |
- | OLMo 3 7B Base | 7B | INT4 | 21 tok/s | ~4.1s | |
- | SEA-LION-7B | 7.1B | INT4 | 21 tok/s | ~4.1s | |
- | SEA-LION-7B-Instruct | 7.1B | INT4 | 21 tok/s | ~4.1s | |
- | FLUX.1 Canny Dev | 11.9B | INT4 | 18 tok/s | ~7.3s | |
- | FLUX.1 Depth Dev | 11.9B | INT4 | 18 tok/s | ~7.3s | |
- | FLUX.1 Dev | 11.9B | INT4 | 18 tok/s | ~7.3s | |
- | FLUX.1 Fill Dev | 11.9B | INT4 | 18 tok/s | ~7.3s | |
- | FLUX.1 Schnell | 11.9B | INT4 | 18 tok/s | ~7.3s | |
- | Sahabat-AI-Llama3-8B-Instruct | 8B | INT4 | 18 tok/s | ~5.0s | |
- | Typhoon-2-8B | 8B | INT4 | 18 tok/s | ~5.0s | |
- | HunyuanVideo 1.5 720P T2V | 13B | INT4 | 17 tok/s | ~8.0s | |
不同量化精度下可运行的最大模型参数级别。
上下文长度:
4位量化
8B 模型
8位精度
3B 模型
16位全精度
< 1B 模型
本地模型训练与适配器微调支持能力。
微调上下文:
QLoRA
8B 模型
LoRA
3B 模型
全参数微调
< 1B 模型
Radeon PRO W7500 (8GB) 上的最佳量化精度与运行规格建议。
最佳推理区间
推荐以 FP16/Q8 精度运行 1B 至 3B 轻量级模型,或在 4-bit 量化下运行 7B/8B 模型。
显存带宽与推理速率
具备 176 GB/s 聚合带宽,单批次推理处于显存带宽受限模式。8B Q4 模型预估生成速度约 39 tok/s,70B Q4 模型预估生成速度约 5 tok/s。