16GB · AI Score 4.9/100 · first-party measured on 12 AI workloads
5.2 AI Score Includes estimates
Every number on this page is first-party: GeForce RTX 5080 was run on our pinned 12-workload AI suite on 2026-07-11, with under 0.5% run-to-run variance. On Llama 3.1 8B (Q4_K_M) GeForce RTX 5080 delivers about 150.71 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 16GB. For image generation, SDXL runs at 4.43 it/s, while FLUX.1-dev won't fit at BF16 (needs ~26GB). 6 of the 12 workloads won't fit on 16GB at the tested precision, Qwen3 32B, Llama 3.3 70B, FLUX.1-dev, FLUX.1 Kontext and others. We publish those as hard gates rather than quietly dropping to a smaller quant.
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Qwen3 0.6B | 692.51 tok/s | 57 W54°CQ4_K_M | ✓ Measured |
| Llama 3.2 1B | 726.57 tok/s | 88 W63°CQ4_K_M | ✓ Measured |
| MiniCPM5 2B | 367.12 tok/s | 126 W45°CQ4_K_M | ✓ Measured |
| LFM2.5 2.6B | 377.14 tok/s | 132 W47°CQ4_K_M | ✓ Measured |
| Granite 4.1 3B | 253.62 tok/s | 121 W49°CQ4_K_M | ✓ Measured |
| Nemotron 3 Nano 4B | 255.18 tok/s | 138 W46°CQ4_K_M | ✓ Measured |
| Qwen3 4B | 215.89 tok/s | 2.8 GB peak86 W53°C2.51 tok/WQ4_K_M | ✓ Measured |
| DeepSeek Coder 7B Instruct v1.5 | 180.65 tok/s | 182 W51°CQ4_K_M | ✓ Measured |
| Qwen2.5-7B | 173.49 tok/s | 186 W67°CQ4_K_M | ✓ Measured |
| Qwen2.5-Coder 7B | 173.53 tok/s | 190 W68°CQ4_K_M | ✓ Measured |
| Llama 3 8B | 158.84 tok/s | 184 W52°CQ4_K_M | ✓ Measured |
| Llama 3.1 8B | 150.71 tok/s | 4.7 GB peak144 W56°C1.05 tok/WQ4_K_M | ✓ Measured |
| Qwen3 8B | 153.89 tok/s | 178 W51°CQ4_K_M | ✓ Measured |
| Nemotron Nano 9B v2 | 119.94 tok/s | 181 W50°CQ4_K_M | ✓ Measured |
| Ornith 1.5 9B | 138.12 tok/s | 180 W49°CQ4_K_M | ✓ Measured |
| Gemma 4 12B | 96.61 tok/s | 193 W49°CQ4_K_M | ✓ Measured |
| Qwen2.5-Coder 14B | 81.97 tok/s | 8.7 GB peak146 W59°C0.56 tok/WQ4_K_M | ✓ Measured |
| Qwen3 14B | 87.02 tok/s | 209 W54°CQ4_K_M | ✓ Measured |
| gpt-oss-20b | 265.35 tok/s | 115 W64°CQ4_K_M | ✓ Measured |
| Gemma 4 26B A4B | ✕ Won't fit needs ~20 GB | VRAM-gated at this precision | Est. |
| Qwen3 30B A3B | ✕ Won't fit needs ~20 GB | VRAM-gated at this precision | Est. |
| Gemma 4 31B | ✕ Won't fit needs ~22 GB | VRAM-gated at this precision | Est. |
| Qwen3 32B | ✕ Won't fit needs ~23 GB | VRAM-gated at this precision | ✓ Measured |
| Llama 3.3 70B | ✕ Won't fit needs ~46 GB | VRAM-gated at this precision | ✓ Measured |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Stable Diffusion 1.5 | 61.19 images/min | 276 W56°C | ✓ Measured |
| SDXL Turbo | 512.76 images/min | 83 W46°C | ✓ Measured |
| Stable Diffusion XL | 12.24 images/min | 313 W77°C | ✓ Measured |
| Z-Image Turbo | 3.56 images/min | 197 W67°C | ✓ Measured 3 hosts ±11% |
| Sana 1.6B | 30.02 images/min | 310 W56°C | ✓ Measured |
| Playground v2.5 | 7.6 images/min | 309 W60°C | ✓ Measured |
| PixArt-Sigma XL | 16.22 images/min | 276 W56°C | ✓ Measured |
| FLUX.2 klein 4B | ✕ Won't fit needs ~19 GB | VRAM-gated at this precision | ✓ Measured |
| FLUX.1 dev | ✕ Won't fit needs ~26 GB | VRAM-gated at this precision | ✓ Measured |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| FLUX.1 Kontext dev | ✕ Won't fit needs ~26 GB | VRAM-gated at this precision | ✓ Measured |
| Qwen-Image-Edit | ✕ Won't fit needs ~42 GB | VRAM-gated at this precision | ✓ Measured |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Stable Video Diffusion | 1.67 clips/min | 329 W66°C | ✓ Measured |
| LTX-Video (image to video) | 2.39 clips/min | 193 W59°C | ✓ Measured 2 hosts ±11% · CPU offload |
| Wan 2.2 TI2V-5B (image to video) | 0.96 clips/min | 239 W61°C | ✓ Measured 2 hosts ±7% · CPU offload |
| Stable Video Diffusion XT | 0.94 clips/min | 280 W60°C | ✓ Measured 2 hosts ±0% · CPU offload |
| CogVideoX-5B I2V | 0.26 clips/min | 239 W57°C | ✓ Measured 2 hosts ±2% · CPU offload |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Wan 2.1 1.3B | 0.45 frames/s | 253 W71°C106.9 s/clip | ✓ Measured 2 hosts ±1% · CPU offload |
| CogVideoX-2B | 0.36 frames/s | 281 W72°C134.9 s/clip | ✓ Measured 2 hosts ±1% · CPU offload |
| LTX-Video (distilled) | 4.33 frames/s | 217 W71°C22.4 s/clip | ✓ Measured 3 hosts ±8% · CPU offload |
| Wan 2.2 5B (720p) | ✕ Won't fit needs ~18 GB | VRAM-gated at this precision | ✓ Measured |
| Architecture | Blackwell (GB203) |
| CUDA cores | 10,752 |
| VRAM | 16GB GDDR7 |
| Memory bus | 256-bit |
| Memory bandwidth | 960 GB/s |
| Boost clock | 2,617 MHz |
| TDP | 360 W |
| Process | 5nm |
| Interface | PCIe 5.0 x16 |
| Release date | 2025-01-30 |
| Launch MSRP | $999 |
GeForce RTX 5080 scores 4.9/100, #39 of 102. It ran 6 of 12; 6 exceeded its 16GB. Every figure here is our own measurement.
100% = this card, AI & Machine Learning headline metric (AI Score). #9 of 61 desktop cards in this vertical.
| GPU | Relative | % | AI Score |
|---|---|---|---|
| NVIDIA GeForce RTX 4090 | 204% | 10.6 | |
| NVIDIA GeForce RTX 3090 Ti | 163% | 8.5 | |
| NVIDIA Titan RTX | 158% | 8.2 | |
| NVIDIA GeForce RTX 3090 | 148% | 7.7 | |
| GeForce RTX 5080 | 100% | 5.2 | |
| NVIDIA GeForce RTX 4080 | 92% | 4.8 | |
| GeForce RTX 4080 Super | 90% | 4.7 | |
| GeForce RTX 5070 Ti | 90% | 4.7 | |
| NVIDIA RTX 4000 (Ada Generation) | 85% | 4.4 |
Same card, other workloads: GeForce RTX 5080 Gaming benchmarks
← All AI & Machine Learning GPU rankings
| Transistors | 45,600 million |
| Die size | 378 mm² |
| Process node | 4 nm |
| Fabricated by | TSMC |
| Transistor density | 120.6 million per mm² |
Denser than 92% of the 746 cards we have silicon data for. Density is the clearest measure of what a process node bought: a card that gained it without growing the die got its speed from the fab rather than the architecture.
Silicon figures from Wikipedia (CC BY-SA 4.0). Benchmarks on this page are our own. Compare every chip.
Whole-job timings, composed from our measured per-model results on this card.
| Workflow | Time | Energy | Basis |
|---|---|---|---|
| Full codebase review | 12.2 min | 29.75 Wh | measured |
| Animate a batch of images | 20.9 min | 83.14 Wh | measured |
Can't run: Product photo shoot (needs FLUX.1 Kontext dev), Photo restoration batch (needs FLUX.1 Kontext dev), Restore and enlarge photos (needs FLUX.1 Kontext dev), Character sheet, 12 poses (needs FLUX.1 dev), Short social clips (needs Qwen3 32B), 60-second AI short film (needs Qwen3 32B), 24-frame storyboard (needs Qwen3 32B), 6-panel comic page (needs Qwen3 32B), Long-form article batch (needs Llama 3.3 70B).
This card is $999 to buy. The cheapest listed rate on Vast.ai is $0.223/hour, but that is the floor: we budget $0.268/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 3,733 GPU-hours. Below it you are paying for idle silicon.
| How you would use it | GPU-hours a year | Rental cost a year | Time to break even |
|---|---|---|---|
| 2 hours a day, hobby | 730 | $195 | 5.1 years |
| 8 hours a day, working on it | 2,920 | $781 | 1.3 years |
| 24/7, always-on agent | 8,760 | $2,344 | 5.1 months |
At hobby usage this card is very unlikely to pay for itself before it is superseded. Rent it. Rental figures include a 20% premium over the cheapest listed rate. Ignores electricity, resale and the fact that a rented card can be a newer one tomorrow.
Cheapest of the RunPod and Vast on-demand rates we see, sampled daily. Spot and interruptible pricing runs lower.