8GB · AI Score 2.0/100 · first-party measured on 12 AI workloads
2 AI Score Includes estimates
Every number on this page is first-party: NVIDIA GeForce RTX 3060 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) NVIDIA GeForce RTX 3060 Ti delivers about 76.93 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 8GB. For image generation, SDXL runs at 1.1 it/s, while FLUX.1-dev won't fit at BF16 (needs ~26GB). 9 of the 12 workloads won't fit on 8GB at the tested precision, Qwen2.5-Coder 14B, Qwen3 32B, Llama 3.3 70B, Z-Image Turbo and others. We publish those as hard gates rather than quietly dropping to a smaller quant.
The 3060 Ti is basically a lesson in what 8GB means in 2026, 9 of my 12 workloads wouldn't even load. What it does run, it runs hot: SDXL pushed it to 81°C at 99% of its 200W rating, and that was also my noisiest run on this card. If you're buying for AI, the 8GB is the whole story. 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 |
|---|---|---|---|
| MiniCPM5 2B | 205.86 tok/s | 110 W55°CQ4_K_M | ✓ Measured |
| LFM2.5 2.6B | 198.79 tok/s | 117 W54°CQ4_K_M | ✓ Measured |
| Granite 4.1 3B | 146.23 tok/s | 129 W56°CQ4_K_M | ✓ Measured |
| Nemotron 3 Nano 4B | 130.09 tok/s | 125 W56°CQ4_K_M | ✓ Measured |
| Qwen3 4B | 118.21 tok/s | 2.9 GB peak130 W63°C0.91 tok/WQ4_K_M | ✓ Measured |
| DeepSeek Coder 7B Instruct v1.5 | 96.35 tok/s | 132 W57°CQ4_K_M | ✓ Measured |
| Llama 3 8B | 85.46 tok/s | 135 W57°CQ4_K_M | ✓ Measured |
| Llama 3.1 8B | 76.93 tok/s | 4.8 GB peak145 W67°C0.53 tok/WQ4_K_M | ✓ Measured |
| Qwen3 8B | 82.96 tok/s | 137 W57°CQ4_K_M | ✓ Measured |
| Nemotron Nano 9B v2 | 61.7 tok/s | 140 W58°CQ4_K_M | ✓ Measured |
| Ornith 1.5 9B | 73.84 tok/s | 137 W57°CQ4_K_M | ✓ Measured |
| Gemma 4 12B | 52.67 tok/s | 139 W58°CQ4_K_M | ✓ Measured |
| Qwen2.5-Coder 14B | ✕ Won't fit needs ~11.5 GB | VRAM-gated at this precision | ✓ Measured |
| Qwen3 14B | ✕ Won't fit needs ~11 GB | VRAM-gated at this precision | Est. |
| 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 | 20.42 images/min | 188 W68°C | ✓ Measured |
| SD Turbo | 327.8 images/min | 82 W53°C | ✓ Measured |
| LCM DreamShaper v7 | 80.24 images/min | 160 W63°C | ✓ Measured |
| SDXL Turbo | 201.93 images/min | 80 W56°C | ✓ Measured |
| Sana 1.6B | ✕ Won't fit needs ~11 GB | VRAM-gated at this precision | Est. |
| Stable Diffusion XL | 2.2 images/min | 6.5 GB peak151 W81°C27.3 s/img | ✓ Measured CPU offload |
| Z-Image Turbo | ✕ Won't fit needs ~13 GB | VRAM-gated at this precision | ✓ Measured |
| PixArt-Sigma XL | ✕ Won't fit needs ~14 GB | VRAM-gated at this precision | Est. |
| 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 |
|---|---|---|---|
| LTX-Video (distilled) | ✕ Won't fit needs ~14 GB | VRAM-gated at this precision | ✓ Measured |
| Wan 2.2 5B (720p) | ✕ Won't fit needs ~18 GB | VRAM-gated at this precision | ✓ Measured |
| Architecture | Ampere (GA104) |
| CUDA cores | 4,864 |
| VRAM | 8GB GDDR6 |
| Memory bus | 256-bit |
| Memory bandwidth | 448 GB/s |
| Boost clock | 1,665 MHz |
| TDP | 200 W |
| Process | 8nm |
| Interface | PCIe 4.0 x16 |
| Release date | 2020-12-02 |
| Launch MSRP | $399 |
NVIDIA GeForce RTX 3060 Ti scores 2.0/100, #86 of 102. It ran 3 of 12; 9 exceeded its 8GB. Every figure here is our own measurement.
100% = this card, AI & Machine Learning headline metric (AI Score). #47 of 61 desktop cards in this vertical.
| GPU | Relative | % | AI Score |
|---|---|---|---|
| Intel Arc B580 | 105% | 2.1 | |
| NVIDIA GeForce RTX 2070 SUPER | 100% | 2 | |
| NVIDIA GeForce RTX 2070 | 100% | 2 | |
| NVIDIA GeForce RTX 2080 Founders Edition | 100% | 2 | |
| NVIDIA GeForce RTX 3060 Ti | 100% | 2 | |
| Intel Arc A770 Limited Edition | 100% | 2 | |
| NVIDIA GeForce RTX 2060 Super | 95% | 1.9 | |
| GeForce RTX 4060 | 95% | 1.9 | |
| NVIDIA GeForce RTX 5050 | 95% | 1.9 |
Same card, other workloads: NVIDIA GeForce RTX 3060 Ti Gaming benchmarks
← All AI & Machine Learning GPU rankings
| Transistors | 17,400 million |
| Die size | 392.5 mm² |
| Process node | 8 nm |
| Fabricated by | Samsung |
| Transistor density | 44.3 million per mm² |
Denser than 80% 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.
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), Full codebase review (needs Qwen2.5-Coder 14B).
This card is $399 to buy. The cheapest listed rate on Vast.ai is $0.075/hour, but that is the floor: we budget $0.090/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 4,433 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 | $66 | 6.1 years |
| 8 hours a day, working on it | 2,920 | $263 | 1.5 years |
| 24/7, always-on agent | 8,760 | $788 | 6.1 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.