NVIDIA GeForce RTX 4070 Ti Super, AI & Machine Learning Benchmarks & Specs

16GB · AI Score 4.3/100 · first-party measured on 12 AI workloads

4.3 AI Score Includes estimates

Every number on this page is first-party: NVIDIA GeForce RTX 4070 Ti Super 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 4070 Ti Super delivers about 119.63 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 16GB. For image generation, SDXL runs at 4.45 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.

Bench notes: from the person who ran it

The 16GB actually changes this card's story: only 6 of 12 workloads gated instead of 8, and it never got near its 285W rating (I peaked it at 250W). The Z-Image and LTX video runs were my noisiest on this card, which makes sense, those sit right at the VRAM edge. 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.

AI & Machine Learning benchmark results

Text Generation tok/s 19

MiniCPM5 2B295.94

LFM2.5 2.6B290.14

Granite 4.1 3B212.25

Qwen3 4B187.75

Nemotron 3 Nano 4B186.32

DeepSeek Coder 7B Instruct v1.5136.19

Llama 3 8B120.81

Llama 3.1 8B119.63

Qwen3 8B117.39

Ornith 1.5 9B104.04

Nemotron Nano 9B v286.2

Gemma 4 12B74.94

Workload	Result	Telemetry	Data
MiniCPM5 2B	295.94 tok/s	110 W66°CQ4_K_M	✓ Measured
LFM2.5 2.6B	290.14 tok/s	128 W69°CQ4_K_M	✓ Measured
Granite 4.1 3B	212.25 tok/s	121 W58°CQ4_K_M	✓ Measured
Nemotron 3 Nano 4B	186.32 tok/s	131 W66°CQ4_K_M	✓ Measured
Qwen3 4B	187.75 tok/s	2.7 GB peak159 W59°C1.18 tok/WQ4_K_M	✓ Measured
DeepSeek Coder 7B Instruct v1.5	136.19 tok/s	152 W62°CQ4_K_M	✓ Measured
Llama 3 8B	120.81 tok/s	158 W63°CQ4_K_M	✓ Measured
Llama 3.1 8B	119.63 tok/s	4.6 GB peak191 W61°C0.63 tok/WQ4_K_M	✓ Measured
Qwen3 8B	117.39 tok/s	162 W63°CQ4_K_M	✓ Measured
Nemotron Nano 9B v2	86.2 tok/s	164 W62°CQ4_K_M	✓ Measured
Ornith 1.5 9B	104.04 tok/s	159 W66°CQ4_K_M	✓ Measured
Gemma 4 12B	74.94 tok/s	166 W64°CQ4_K_M	✓ Measured
Qwen2.5-Coder 14B	65.89 tok/s	8.6 GB peak201 W65°C0.33 tok/WQ4_K_M	✓ Measured
Qwen3 14B	67.09 tok/s	182 W67°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

Image Generation images/min 9

SDXL Turbo532.27

Stable Diffusion 1.546.09

DreamShaper XL Turbo30.49

Sana 1.6B22.59

PixArt-Sigma XL12.6

Stable Diffusion XL8.9

Playground v2.55.57

Z-Image Turbo2.3

Workload	Result	Telemetry	Data
Stable Diffusion 1.5	46.09 images/min	188 W66°C	✓ Measured
SDXL Turbo	532.27 images/min	65 W51°C	✓ Measured
Z-Image Turbo	2.3 images/min		✓ Measured 3 hosts ±3%
Sana 1.6B	22.59 images/min	211 W67°C	✓ Measured
Stable Diffusion XL	8.9 images/min	14.6 GB peak247 W69°C6.8 s/img	✓ Measured
DreamShaper XL Turbo	30.49 images/min	217 W63°C	✓ Measured
Playground v2.5	5.57 images/min	221 W73°C	✓ Measured
PixArt-Sigma XL	12.6 images/min	213 W69°C	✓ Measured
FLUX.1 dev	✕ Won't fit needs ~26 GB	VRAM-gated at this precision	✓ Measured

Image Editing images/min 2

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

Video Generation frames/s 2

Workload	Result	Telemetry	Data
LTX-Video (distilled)	2.77 frames/s	119 W39°C38.3 s/clip	✓ Measured 3 hosts ±7% · CPU offload
Wan 2.2 5B (720p)	✕ Won't fit needs ~18 GB	VRAM-gated at this precision	✓ Measured

How we measured this. Every result comes from our own pinned, reproducible AI suite, 12 workloads: the Qwen3-4B to Llama-70B LLM ladder (llama.cpp, Q4_K_M), SDXL / Z-Image / FLUX-dev generation, FLUX-Kontext / Qwen-Edit editing, and LTX / Wan video, run first-party on rented hardware with under 0.5% run-to-run variance. Peak VRAM, power draw, temperature and tokens-per-watt are captured per workload. “Won’t fit” rows are real data: where a model exceeds the card’s VRAM at the tested precision we record a hard gate rather than silently dropping to a smaller quant. Measured 2026-07-11 · harness 2.0.0-standalone.

NVIDIA GeForce RTX 4070 Ti Super specifications

Architecture	Ada Lovelace (AD103)
CUDA cores	8,448
VRAM	16GB GDDR6X
Memory bus	256-bit
Memory bandwidth	672 GB/s
Boost clock	2,610 MHz
TDP	285 W
Process	4nm
Interface	PCIe 4.0 x16
Release date	2024-01-24
Launch MSRP	$799

Verdict, capable, but 16GB sets the ceiling

NVIDIA GeForce RTX 4070 Ti Super scores 4.3/100, #47 of 102. It ran 6 of 12; 6 exceeded its 16GB. Every figure here is our own measurement.

Relative performance: where the NVIDIA GeForce RTX 4070 Ti Super lands

100% = this card, AI & Machine Learning headline metric (AI Score). #14 of 61 desktop cards in this vertical.

GPU	%	AI Score
NVIDIA GeForce RTX 4080	112%	4.8
GeForce RTX 4080 Super	109%	4.7
GeForce RTX 5070 Ti	109%	4.7
NVIDIA RTX 4000 (Ada Generation)	102%	4.4
NVIDIA GeForce RTX 4070 Ti Super	100%	4.3
AMD Radeon RX 7900 XTX	86%	3.7
GeForce RTX 5060 Ti	86%	3.7
NVIDIA GeForce RTX 3080 Ti	81%	3.5
NVIDIA GeForce RTX 4060 Ti 16GB	74%	3.2

One tier up: NVIDIA RTX 4000 (Ada Generation) (+2%) · compare head-to-head
One tier down: AMD Radeon RX 7900 XTX (-14%)
Best value alternative ($749): GeForce RTX 5070 Ti (+9%) · compare head-to-head

Same card, other workloads: NVIDIA GeForce RTX 4070 Ti Super Gaming benchmarks

← All AI & Machine Learning GPU rankings

The silicon

Transistors	45,900 million
Die size	378.6 mm²
Process node	4 nm
Fabricated by	TSMC
Transistor density	121.2 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.

What this card can build

Whole-job timings, composed from our measured per-model results on this card.

Workflow	Time	Energy	Basis
Full codebase review	15.2 min	50.72 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).

Rent or buy?

This card is $799 to buy. The cheapest listed rate on Vast.ai is $0.252/hour, but that is the floor: we budget $0.302/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 2,642 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	$221	3.6 years
8 hours a day, working on it	2,920	$883	10.9 months
24/7, always-on agent	8,760	$2,649	3.6 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.

Rental price

$0.252/hr+28.6% since 2026-09-30low $0.149 · high $0.252

Cheapest of the RunPod and Vast on-demand rates we see, sampled daily. Spot and interruptible pricing runs lower.