16GB · AI Score 3.7/100 · first-party measured on 12 AI workloads
3.7 AI Score Includes estimates
Every number on this page is first-party: GeForce RTX 5060 Ti 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 5060 Ti delivers about 84.2 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 16GB. For image generation, SDXL runs at 2.61 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.
This one surprised me. 1.49 tokens/watt, ran at 90% of its power rating, and because it's the 16GB SKU it loads models that 12GB cards flat-out refuse, Z-Image and LTX video both ran. For the money, it might be the quietest smart buy in the AI lineup right now. Quick note on the setup: all my AI benchmarking was done on rented cloud GPUs, I used all three of Vast.ai, RunPod and Modal depending on which had the card, and they all have their pros and cons. Same pinned harness on every run, and everything here got double-checked before it went up.
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Qwen3 0.6B | 419.09 tok/s | 52 W59°CQ4_K_M | ✓ Measured |
| Llama 3.2 1B | 400.21 tok/s | 65 W65°CQ4_K_M | ✓ Measured |
| MiniCPM5 2B | 200.58 tok/s | 55 W49°CQ4_K_M | ✓ Measured |
| LFM2.5 2.6B | 188.2 tok/s | 60 W49°CQ4_K_M | ✓ Measured |
| Granite 4.1 3B | 130.68 tok/s | 63 W49°CQ4_K_M | ✓ Measured |
| Nemotron 3 Nano 4B | 128.01 tok/s | 61 W47°CQ4_K_M | ✓ Measured |
| Qwen3 4B | 134.96 tok/s | 2.8 GB peak91 W52°C1.49 tok/WQ4_K_M | ✓ Measured |
| DeepSeek Coder 7B Instruct v1.5 | 87.21 tok/s | 74 W51°CQ4_K_M | ✓ Measured |
| Qwen2.5-7B | 87.24 tok/s | 118 W70°CQ4_K_M | ✓ Measured |
| Qwen2.5-Coder 7B | 87.19 tok/s | 116 W64°CQ4_K_M | ✓ Measured |
| Llama 3 8B | 78.51 tok/s | 79 W52°CQ4_K_M | ✓ Measured |
| Llama 3.1 8B | 84.2 tok/s | 4.8 GB peak124 W57°C0.68 tok/WQ4_K_M | ✓ Measured |
| Qwen3 8B | 76.63 tok/s | 78 W53°CQ4_K_M | ✓ Measured |
| Nemotron Nano 9B v2 | 58.98 tok/s | 74 W51°CQ4_K_M | ✓ Measured |
| Ornith 1.5 9B | 68.96 tok/s | 75 W49°CQ4_K_M | ✓ Measured |
| Gemma 4 12B | 48.66 tok/s | 79 W53°CQ4_K_M | ✓ Measured |
| Qwen2.5-Coder 14B | 45.06 tok/s | 8.5 GB peak124 W63°C0.36 tok/WQ4_K_M | ✓ Measured |
| Qwen3 14B | 41.93 tok/s | 85 W53°CQ4_K_M | ✓ Measured |
| gpt-oss-20b | 143.33 tok/s | 67 W59°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 | 25.14 images/min | 84 W53°C | ✓ Measured |
| LCM DreamShaper v7 | 106.19 images/min | 115 W59°C | ✓ Measured |
| SDXL Turbo | 260.84 images/min | 48 W48°C | ✓ Measured |
| SSD-1B | 4.56 images/min | 93 W61°C | ✓ Measured |
| Z-Image Turbo | 1.56 images/min | ✓ Measured 3 hosts ±9% | |
| Sana 1.6B | 12.64 images/min | 103 W60°C | ✓ Measured |
| Stable Diffusion XL | 5.22 images/min | 14.3 GB peak156 W70°C11.5 s/img | ✓ Measured |
| DreamShaper XL Lightning | 28.35 images/min | 88 W52°C | ✓ Measured |
| DreamShaper XL Turbo | 18.31 images/min | 155 W66°C | ✓ Measured |
| Playground v2.5 | 2.93 images/min | 90 W63°C | ✓ Measured |
| PixArt-Sigma XL | 6.16 images/min | 78 W59°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 | 0.67 clips/min | 101 W65°C | ✓ Measured |
| LTX-Video (image to video) | 1.08 clips/min | 105 W68°C | ✓ Measured 2 hosts ±4% · CPU offload |
| Wan 2.2 TI2V-5B (image to video) | 0.42 clips/min | 122 W73°C | ✓ Measured 2 hosts ±2% · CPU offload |
| Stable Video Diffusion XT | 0.42 clips/min | 154 W76°C | ✓ Measured 2 hosts ±1% · CPU offload |
| CogVideoX-5B I2V | 0.11 clips/min | 131 W72°C | ✓ Measured 2 hosts ±1% · CPU offload |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Wan 2.1 1.3B | 0.2 frames/s | 121 W69°C241.7 s/clip | ✓ Measured 2 hosts ±3% · CPU offload |
| CogVideoX-2B | 0.16 frames/s | 135 W72°C309.3 s/clip | ✓ Measured 2 hosts ±1% · CPU offload |
| LTX-Video (distilled) | 1.71 frames/s | 106 W59°C49.4 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 |
| CUDA cores | 4,608 |
| VRAM | 16GB GDDR7 |
| Memory bus | 128-bit |
| Memory bandwidth | 448 GB/s |
| Boost clock | 2,572 MHz |
| TDP | 180 W |
| Process | 3nm |
| Interface | PCIe 5.0 x16 |
| Release date | 2025-04-16 |
| Launch MSRP | $429 |
GeForce RTX 5060 Ti scores 3.7/100, #50 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). #16 of 61 desktop cards in this vertical.
| GPU | Relative | % | AI Score |
|---|---|---|---|
| GeForce RTX 5070 Ti | 127% | 4.7 | |
| NVIDIA RTX 4000 (Ada Generation) | 119% | 4.4 | |
| NVIDIA GeForce RTX 4070 Ti Super | 116% | 4.3 | |
| AMD Radeon RX 7900 XTX | 100% | 3.7 | |
| GeForce RTX 5060 Ti | 100% | 3.7 | |
| NVIDIA GeForce RTX 3080 Ti | 95% | 3.5 | |
| NVIDIA GeForce RTX 4060 Ti 16GB | 86% | 3.2 | |
| NVIDIA GeForce RTX 4070 Ti | 86% | 3.2 | |
| GeForce RTX 5070 | 86% | 3.2 |
Same card, other workloads: GeForce RTX 5060 Ti Gaming benchmarks
← All AI & Machine Learning GPU rankings
| Transistors | 21,900 million |
| Die size | 181 mm² |
| Process node | 4 nm |
| Fabricated by | TSMC |
| Transistor density | 121 million per mm² |
Denser than 93% 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 | 22.2 min | 45.98 Wh | measured |
| Animate a batch of images | 47.3 min | 96.38 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 $429 to buy. The cheapest listed rate on Vast.ai is $0.121/hour, but that is the floor: we budget $0.145/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 2,955 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 | $106 | 4.0 years |
| 8 hours a day, working on it | 2,920 | $424 | 1.0 years |
| 24/7, always-on agent | 8,760 | $1,272 | 4.0 months |
At steady usage this card pays for itself inside a normal ownership window. 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.