12GB · AI Score 3.2/100 · anchored estimate vs 51 measured cards
3.2 AI Score Includes estimates
We have not run NVIDIA GeForce RTX 4070 Ti on our bench. These figures are anchored estimates, interpolated per workload against the 51 GPUs we did measure. On Llama 3.1 8B (Q4_K_M) NVIDIA GeForce RTX 4070 Ti should deliver about 103.9 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 12GB. For image generation, SDXL should run near 4.21 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.
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| MiniCPM5 2B | 237.51 tok/s | 103 W60°CQ4_K_M | ✓ Measured |
| LFM2.5 2.6B | 220.82 tok/s | 108 W62°CQ4_K_M | ✓ Measured |
| Granite 4.1 3B | 170.68 tok/s | 120 W63°CQ4_K_M | ✓ Measured |
| Nemotron 3 Nano 4B | 143.04 tok/s | 117 W61°CQ4_K_M | ✓ Measured |
| Qwen3-4B | 151.72 tok/s | 124 W61°CQ4_K_M | ✓ Measured |
| DeepSeek Coder 7B Instruct v1.5 | 103.87 tok/s | 139 W66°CQ4_K_M | ✓ Measured |
| Llama 3 8B | 92.13 tok/s | 144 W65°CQ4_K_M | ✓ Measured |
| Llama-3.1-8B | 92.09 tok/s | 139 W61°CQ4_K_M | ✓ Measured |
| Qwen3 8B | 89.83 tok/s | 145 W67°CQ4_K_M | ✓ Measured |
| Nemotron Nano 9B v2 | 65.66 tok/s | 140 W64°CQ4_K_M | ✓ Measured |
| Ornith 1.5 9B | 79.04 tok/s | 140 W71°CQ4_K_M | ✓ Measured |
| Gemma 4 12B | 58.07 tok/s | 142 W64°CQ4_K_M | ✓ Measured |
| Qwen2.5-Coder-14B | 50.73 tok/s | 158 W65°CQ4_K_M | ✓ Measured |
| Qwen3 14B | 51.24 tok/s | 158 W70°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 | VRAM-gated at this precision | Est. |
| Llama 3.3 70B | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| Stable Diffusion 1.5 | 44.44 images/min | 208 W60°C | ✓ Measured |
| Sana 1.6B | 21.64 images/min | 229 W61°C | ✓ Measured |
| Stable Diffusion XL | 8.65 images/min | 240 W66°C | ✓ Measured |
| Playground v2.5 | 5.29 images/min | 239 W64°C | ✓ Measured |
| FLUX.1 dev | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Z-Image Turbo | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| FLUX.1 Kontext dev | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Qwen-Image-Edit | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Workload | Result | Telemetry | Data |
|---|---|---|---|
| LTX-Video (distilled) | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Wan 2.2 5B (720p) | ✕ Won't fit | VRAM-gated at this precision | Est. |
| Architecture | Ada Lovelace (AD104) |
| CUDA cores | 7,680 |
| VRAM | 12GB GDDR6X |
| Memory bus | 192-bit |
| Memory bandwidth | 504 GB/s |
| Boost clock | 2,610 MHz |
| TDP | 285 W |
| Process | 4nm |
| Interface | PCIe 4.0 x16 |
| Release date | 2023-01-05 |
| Launch MSRP | $799 |
NVIDIA GeForce RTX 4070 Ti scores 3.2/100, #54 of 102. It ran 4 of 12; 8 exceeded its 12GB. Figures are anchored estimates, not measurements, we flag that on every row.
100% = this card, AI & Machine Learning headline metric (AI Score). #19 of 61 desktop cards in this vertical.
| GPU | Relative | % | AI Score |
|---|---|---|---|
| AMD Radeon RX 7900 XTX | 116% | 3.7 | |
| GeForce RTX 5060 Ti | 116% | 3.7 | |
| NVIDIA GeForce RTX 3080 Ti | 109% | 3.5 | |
| NVIDIA GeForce RTX 4060 Ti 16GB | 100% | 3.2 | |
| NVIDIA GeForce RTX 4070 Ti | 100% | 3.2 | |
| GeForce RTX 5070 | 100% | 3.2 | |
| NVIDIA GeForce RTX 4070 Super | 97% | 3.1 | |
| NVIDIA GeForce RTX 4070 | 97% | 3.1 | |
| NVIDIA TITAN V | 91% | 2.9 |
Same card, other workloads: NVIDIA GeForce RTX 4070 Ti Gaming benchmarks
← All AI & Machine Learning GPU rankings
| Transistors | 35,800 million |
| Die size | 294.5 mm² |
| Process node | 4 nm |
| Fabricated by | TSMC |
| Transistor density | 121.6 million per mm² |
Denser than 94% 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 | 19.8 min | 51.88 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 $799 to buy. The cheapest listed rate on RunPod is $0.190/hour, but that is the floor: we budget $0.228/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 3,504 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 | $166 | 4.8 years |
| 8 hours a day, working on it | 2,920 | $666 | 1.2 years |
| 24/7, always-on agent | 8,760 | $1,997 | 4.8 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.