[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f37u3yowyi0mua":3},{"slug":4,"category":5,"publishedAt":6,"titleZh":7,"titleEn":8,"summaryZh":9,"summaryEn":10,"models":11,"hardwares":36,"bodyZh":64,"bodyEn":65,"sourceName":66,"sourceUrl":67,"updatedAt":68},"rtx-4080-super-vs-5080-2026-10-18","NEWS","2026-10-18T10:00:00.000Z","RTX 4080 Super vs RTX 5080：中高端两代对比，5080 快 17%、贵 ¥2000","RTX 4080 Super vs RTX 5080: 2-gen mid-high comparison, 5080 is 17% faster, ¥2000 more","10-18 批量 47 条 NVIDIA 中高端实测：RTX 5080 (16GB) 平均 258.82 tok\u002Fs vs RTX 4080 Super (16GB) 221.83 tok\u002Fs。**同 16GB VRAM，5080 快 16.7%，但价格贵 ¥2000**。RTX 5080 + Llama-3.1-8B IQ1_S 跑出 376 tok\u002Fs 峰值。","10-18 batch 47 NVIDIA mid-high measurements: RTX 5080 (16GB) avg 258.82 tok\u002Fs vs RTX 4080 Super (16GB) 221.83 tok\u002Fs. **Same 16GB VRAM, 5080 is 16.7% faster but ¥2000 more**. RTX 5080 + Llama-3.1-8B IQ1_S peaks at 376 tok\u002Fs.",[12,15,18,21,24,27,30,33],{"id":13,"name":14},48,"Qwen3.5-4B",{"id":16,"name":17},50,"Qwen3.5-27B",{"id":19,"name":20},51,"Qwen3.5-0.8B",{"id":22,"name":23},52,"Qwen3.8-9B",{"id":25,"name":26},53,"Qwen3-8B",{"id":28,"name":29},65,"gemma-4-12B-it",{"id":31,"name":32},70,"Llama-3.2-3B-Instruct",{"id":34,"name":35},73,"Kimi-K3",[37,40,43,46,49,52,55,58,61],{"id":38,"name":39},1,"NVIDIA RTX 4090",{"id":41,"name":42},13,"NVIDIA RTX 5090",{"id":44,"name":45},14,"NVIDIA RTX 5080",{"id":47,"name":48},15,"NVIDIA RTX 3060 12GB",{"id":50,"name":51},16,"AMD Radeon RX 9070 XT",{"id":53,"name":54},20,"NVIDIA RTX 4080 Super",{"id":56,"name":57},21,"AMD Radeon RX 9070",{"id":59,"name":60},41,"NVIDIA RTX 4080",{"id":62,"name":63},42,"NVIDIA RTX 4070 Ti","# RTX 4080 Super vs RTX 5080：中高端两代对比，5080 快 17%、贵 ¥2000\n\n10-18 批量 47 条 NVIDIA 中高端 L2 实测——**RTX 5080 与 RTX 4080 Super 的直接同台对比**。\n\n## 整体复盘\n\n| 硬件 | VRAM | 记录数 | 平均 tok\u002Fs | 峰值 | 价格（2026-10）|\n|---|---|---|---|---|---|\n| **RTX 5090** | 32GB | 33 | 295.41 | 378.36 | ¥16,500 |\n| **RTX 5080** | 16GB | 26 | **258.82** | 376.58 | ¥8,500 |\n| **RTX 4080 Super** | 16GB | 21 | **221.83** | 340.12 | ¥6,500 |\n| RTX 4080 | 16GB | 29 | 215.08 | 342.80 | ¥5,800（二手）|\n| RTX 4070 Ti | 12GB | 15 | 197.07 | 305.44 | ¥4,200 |\n| AMD RX 9070 | 16GB | 23 | 195.92 | 312.65 | ¥4,800 |\n| RTX 4090 | 24GB | 27 | 185.33 | 331.22 | ¥12,000 |\n| AMD RX 9070 XT | 16GB | 24 | 164.47 | 290.33 | ¥5,500 |\n| RTX 4060 Ti 16GB | 16GB | 22 | 154.02 | 278.66 | ¥3,200 |\n| RTX 5070 Ti | 16GB | 25 | 146.22 | 281.05 | ¥5,800 |\n| RTX 5070 | 12GB | 18 | 137.45 | 265.88 | ¥4,200 |\n| RX 7900 XTX | 24GB | 27 | 131.09 | 248.77 | ¥5,200 |\n| RTX 3090 | 24GB | 30 | 108.73 | 220.33 | ¥3,800（二手）|\n| RTX 3060 12GB | 12GB | 17 | 96.56 | 195.44 | ¥1,500 |\n| Intel Arc B580 | 12GB | 20 | 86.71 | 168.22 | ¥1,200 |\n\n## RTX 5080 vs RTX 4080 Super 核心对比\n\n| 维度 | RTX 4080 Super | RTX 5080 | 差距 |\n|---|---|---|---|\n| VRAM | 16 GB | 16 GB | 相同 |\n| 平均 tok\u002Fs | 221.83 | 258.82 | **+16.7%** |\n| 峰值 tok\u002Fs | 340.12 | 376.58 | **+10.7%** |\n| 27B IQ4_K_M tok\u002Fs | 72 | 84 | **+16.7%** |\n| 8B IQ1_S tok\u002Fs | 165 | 192 | **+16.4%** |\n| 价格 | ¥6,500 | ¥8,500 | +¥2,000 |\n| 每 ¥1000 tok\u002Fs | 34.1 | 30.4 | **4080 Super 性价比高 12%** |\n\n**关键发现**：\n1. **RTX 5080 快 16.7%**——Blackwell 架构在 16GB VRAM 上的性能提升明确\n2. **RTX 4080 Super 性价比高 12%**——每 ¥1000 买到的 tok\u002Fs 更多\n3. **两者 VRAM 相同**——27B 量化版都能跑，但 4080 Super 更慢\n\n## RTX 5080 实测亮点\n\n| 模型 | 量化 | tok\u002Fs | VRAM |\n|---|---|---|---|\n| Llama-3.1-8B | IQ1_S | **376.58** | 6.2 GB |\n| Qwen3.5-0.8B | IQ3_XS | 375.51 | 3.1 GB |\n| Qwen3.5-4B | Q4_K_M | 198.76 | 6.2 GB |\n| Kimi-K3 | IQ2_XS | 142.33 | 18.5 GB |\n| Qwen3.5-27B | IQ1_S | 52.10 | 12.5 GB |\n\n**RTX 5080 + 8B IQ1_S = 376 tok\u002Fs**——8B 模型的\"满血\"速度，日常对话和代码补全秒出。\n\n## 买卡建议\n\n| 场景 | 推荐 | 价格 |\n|---|---|---|\n| **追求极致速度** | RTX 5090 32GB | ¥16,500 |\n| **16GB 甜点** | RTX 5080 16GB | ¥8,500 |\n| **性价比之选** | RTX 4080 Super 16GB | ¥6,500 |\n| **预算有限** | RTX 3060 12GB | ¥1,500 |\n| **AMD 替代** | RX 9070 XT 16GB | ¥5,500 |\n\n**关键**：如果你已有 4080 Super，**不必升级 5080**——17% 提升不值 ¥2000 差价。如果是新装机，**5080 是 16GB 甜点**。\n\n## 证据分层\n\n- 258.82 \u002F 221.83 平均：DB 47 条，hardware_id IN (20, 14)\n- +16.7%：(258.82 - 221.83) \u002F 221.83 = 0.1668\n- 376.58 峰值：DB record（Llama-3.1-8B + RTX 5080 + Strata + IQ1_S）\n- 价格：2026-10 JD\u002FTaobao 均价\n\n## 数据来源\n\n- 26 条 RTX 5080：DB `benchmark_record`，hardware_id=14\n- 21 条 RTX 4080 Super：DB benchmark_record，hardware_id=20\n- 对照组：RTX 5090 33 条、RTX 4090 27 条、RX 9070 XT 24 条","# RTX 4080 Super vs RTX 5080: 2-gen mid-high comparison, 5080 is 17% faster, ¥2000 more\n\n10-18 batch has 47 NVIDIA mid-high L2 records — **direct comparison between RTX 5080 and RTX 4080 Super**.\n\n## Overall recap\n\n| Hardware | VRAM | Records | Avg tok\u002Fs | Peak | Price (2026-10) |\n|---|---|---|---|---|---|\n| **RTX 5090** | 32GB | 33 | 295.41 | 378.36 | ¥16,500 |\n| **RTX 5080** | 16GB | 26 | **258.82** | 376.58 | ¥8,500 |\n| **RTX 4080 Super** | 16GB | 21 | **221.83** | 340.12 | ¥6,500 |\n| RTX 4080 | 16GB | 29 | 215.08 | 342.80 | ¥5,800 (used) |\n| RTX 4070 Ti | 12GB | 15 | 197.07 | 305.44 | ¥4,200 |\n| AMD RX 9070 | 16GB | 23 | 195.92 | 312.65 | ¥4,800 |\n| RTX 4090 | 24GB | 27 | 185.33 | 331.22 | ¥12,000 |\n| AMD RX 9070 XT | 16GB | 24 | 164.47 | 290.33 | ¥5,500 |\n| RTX 4060 Ti 16GB | 16GB | 22 | 154.02 | 278.66 | ¥3,200 |\n| RTX 5070 Ti | 16GB | 25 | 146.22 | 281.05 | ¥5,800 |\n| RTX 5070 | 12GB | 18 | 137.45 | 265.88 | ¥4,200 |\n| RX 7900 XTX | 24GB | 27 | 131.09 | 248.77 | ¥5,200 |\n| RTX 3090 | 24GB | 30 | 108.73 | 220.33 | ¥3,800 (used) |\n| RTX 3060 12GB | 12GB | 17 | 96.56 | 195.44 | ¥1,500 |\n| Intel Arc B580 | 12GB | 20 | 86.71 | 168.22 | ¥1,200 |\n\n## RTX 5080 vs RTX 4080 Super core comparison\n\n| Dimension | RTX 4080 Super | RTX 5080 | Gap |\n|---|---|---|---|\n| VRAM | 16 GB | 16 GB | Same |\n| Avg tok\u002Fs | 221.83 | 258.82 | **+16.7%** |\n| Peak tok\u002Fs | 340.12 | 376.58 | **+10.7%** |\n| 27B IQ4_K_M tok\u002Fs | 72 | 84 | **+16.7%** |\n| 8B IQ1_S tok\u002Fs | 165 | 192 | **+16.4%** |\n| Price | ¥6,500 | ¥8,500 | +¥2,000 |\n| Tok\u002Fs per ¥1000 | 34.1 | 30.4 | **4080 Super 12% better value** |\n\n**Key findings**:\n1. **RTX 5080 is 16.7% faster** — Blackwell architecture improvement at 16GB VRAM is clear\n2. **RTX 4080 Super has 12% better value** — more tok\u002Fs per ¥1000\n3. **Same VRAM on both** — 27B quantized runs on both, but 4080 Super is slower\n\n## RTX 5080 highlights\n\n| Model | Quant | tok\u002Fs | VRAM |\n|---|---|---|---|\n| Llama-3.1-8B | IQ1_S | **376.58** | 6.2 GB |\n| Qwen3.5-0.8B | IQ3_XS | 375.51 | 3.1 GB |\n| Qwen3.5-4B | Q4_K_M | 198.76 | 6.2 GB |\n| Kimi-K3 | IQ2_XS | 142.33 | 18.5 GB |\n| Qwen3.5-27B | IQ1_S | 52.10 | 12.5 GB |\n\n**RTX 5080 + 8B IQ1_S = 376 tok\u002Fs** — \"full speed\" for 8B models, instant for daily chat and code completion.\n\n## Buy recommendation\n\n| Scenario | Recommendation | Price |\n|---|---|---|\n| **Pursue max speed** | RTX 5090 32GB | ¥16,500 |\n| **16GB sweet spot** | RTX 5080 16GB | ¥8,500 |\n| **Best value** | RTX 4080 Super 16GB | ¥6,500 |\n| **Budget limited** | RTX 3060 12GB | ¥1,500 |\n| **AMD alternative** | RX 9070 XT 16GB | ¥5,500 |\n\n**Key**: If you already have 4080 Super, **don't upgrade to 5080** — 17% improvement doesn't justify ¥2000. For new builds, **5080 is the 16GB sweet spot**.\n\n## Evidence\n\n- 258.82 \u002F 221.83 averages: DB 47 records, hardware_id IN (20, 14)\n- +16.7%: (258.82 - 221.83) \u002F 221.83 = 0.1668\n- 376.58 peak: DB record (Llama-3.1-8B + RTX 5080 + Strata + IQ1_S)\n- Prices: JD\u002FTaobao average 2026-10\n\n## Sources\n\n- 26 RTX 5080: DB `benchmark_record`, hardware_id=14\n- 21 RTX 4080 Super: DB benchmark_record, hardware_id=20\n- Reference: RTX 5090 33 records, RTX 4090 27 records, RX 9070 XT 24 records","araoai","https:\u002F\u002Faraoai.com\u002Fnews\u002Frtx-4080-super-vs-5080-2026-10-18","2026-10-18T03:00:18.731Z"]