12GB · AI Score 2.6/100 · first-party measured on 12 AI workloads
2.6 AI Score ✓ Measured
Every number on this page is first-party: NVIDIA GeForce RTX 3060 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 delivers about 65.36 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 12GB. For image generation, SDXL runs at 1.47 it/s, while FLUX.1-dev won't fit at BF16 (needs ~26GB). 8 of the 12 workloads won't fit on 12GB at the tested precision, Qwen3 32B, Llama 3.3 70B, Z-Image Turbo, FLUX.1-dev and others. We publish those as hard gates rather than quietly dropping to a smaller quant.
The 12GB on a budget card is the whole reason to look at this, it loads things the 8GB cards can't. It also ran 6% over its official power rating, which tells you Ampere ratings were optimistic. Slow, but honest about what fits. 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 4B | 102.04 tok/s | 2.8 GB peak128 W63°C0.8 tok/WQ4_K_M | ✓ Measured |
| Llama 3.1 8B | 65.36 tok/s | 4.8 GB peak146 W67°C0.45 tok/WQ4_K_M | ✓ Measured |
| Qwen2.5-Coder 14B | 35.52 tok/s | 8.5 GB peak147 W71°C0.24 tok/WQ4_K_M | ✓ Measured |
| 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 XL | 2.94 images/min | 11.7 GB peak173 W73°C20.4 s/img | ✓ Measured |
| Z-Image Turbo | ✕ Won't fit needs ~13 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 |
|---|---|---|---|
| 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 (GA106) |
| CUDA cores | 3,584 |
| VRAM | 12GB GDDR6 |
| Memory bus | 192-bit |
| Memory bandwidth | 360 GB/s |
| Boost clock | 1,777 MHz |
| TDP | 170 W |
| Process | 8nm Samsung |
| Interface | PCIe 4.0 x16 |
| Release date | 2021-02-25 |
| Launch MSRP | $329 |
NVIDIA GeForce RTX 3060 scores 2.6/100, #73 of 102. It ran 4 of 12; 8 exceeded its 12GB. Every figure here is our own measurement.
100% = this card, AI & Machine Learning headline metric (AI Score). #34 of 61 desktop cards in this vertical.
| GPU | Relative | % | AI Score |
|---|---|---|---|
| NVIDIA GeForce RTX 4070 Ti | 115% | 3 | |
| Intel Arc A770 Limited Edition | 112% | 2.9 | |
| NVIDIA TITAN V | 112% | 2.9 | |
| GeForce GTX 1080 Ti | 100% | 2.6 | |
| NVIDIA GeForce RTX 3060 | 100% | 2.6 | |
| NVIDIA TITAN Xp | 100% | 2.6 | |
| AMD Radeon RX 7700 XT | 96% | 2.5 | |
| NVIDIA TITAN X (Pascal) | 96% | 2.5 | |
| AMD Radeon RX 6700 | 88% | 2.3 |
Same card, other workloads: NVIDIA GeForce RTX 3060 Gaming benchmarks
← All AI & Machine Learning GPU rankings
| Transistors | 12,000 million |
| Die size | 276 mm² |
| Fabricated by | Samsung |
| Transistor density | 43.5 million per mm² |
Denser than 38% of the 76 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 | 28.2 min | 69.16 Wh | measured |
Can't run: 60-second AI short film (needs Qwen3 32B), 60-second AI short film, narrated (needs Qwen3 32B), 10 short social clips (needs Qwen3 32B), 40-product photo shoot (needs FLUX.1 Kontext dev), 6-panel comic page (needs Qwen3 32B), 20 long-form articles (needs Llama 3.3 70B), Character sheet, 12 poses (needs FLUX.1 dev), 100-photo restoration batch (needs FLUX.1 Kontext dev), 24-frame storyboard (needs Qwen3 32B), 100-photo restore and enlarge (needs FLUX.1 Kontext dev).
This card is $329 to buy. The cheapest listed rate on Vast.ai is $0.034/hour, but that is the floor: we budget $0.041/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 8,064 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 | $30 | 11.0 years |
| 8 hours a day, working on it | 2,920 | $119 | 2.8 years |
| 24/7, always-on agent | 8,760 | $357 | 11.0 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.