NVIDIA GeForce RTX 3080 Ti, AI & Machine Learning Benchmarks & Specs

12GB · AI Score 3.5/100 · first-party measured on 12 AI workloads

3.5 AI Score Includes estimates

Every number on this page is first-party: NVIDIA GeForce RTX 3080 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 3080 Ti delivers about 144.11 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 12GB. 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.

Bench notes: from the person who ran it

This was my hottest run of the whole fleet, 86°C at a flat 350W on the coder model. It'll do the work, but plan cooling around sustained load, not gaming bursts. 12GB gates 8 of 12 workloads. 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

LFM2.5 2.6B320.54

MiniCPM5 2B304.63

Granite 4.1 3B228.53

Nemotron 3 Nano 4B222.47

Qwen3 4B206.45

DeepSeek Coder 7B Instruct v1.5164.84

Llama 3 8B146.52

Llama 3.1 8B144.11

Qwen3 8B140.71

Ornith 1.5 9B125.18

Nemotron Nano 9B v2107.7

Gemma 4 12B88.43

Workload	Result	Telemetry	Data
MiniCPM5 2B	304.63 tok/s	180 W44°CQ4_K_M	✓ Measured
LFM2.5 2.6B	320.54 tok/s	171 W44°CQ4_K_M	✓ Measured
Granite 4.1 3B	228.53 tok/s	210 W44°CQ4_K_M	✓ Measured
Nemotron 3 Nano 4B	222.47 tok/s	208 W45°CQ4_K_M	✓ Measured
Qwen3 4B	206.45 tok/s	2.7 GB peak195 W66°C1.06 tok/WQ4_K_M	✓ Measured
DeepSeek Coder 7B Instruct v1.5	164.84 tok/s	244 W44°CQ4_K_M	✓ Measured
Llama 3 8B	146.52 tok/s	240 W45°CQ4_K_M	✓ Measured
Llama 3.1 8B	144.11 tok/s	4.9 GB peak275 W73°C0.52 tok/WQ4_K_M	✓ Measured
Qwen3 8B	140.71 tok/s	236 W45°CQ4_K_M	✓ Measured
Nemotron Nano 9B v2	107.7 tok/s	245 W45°CQ4_K_M	✓ Measured
Ornith 1.5 9B	125.18 tok/s	230 W45°CQ4_K_M	✓ Measured
Gemma 4 12B	88.43 tok/s	244 W45°CQ4_K_M	✓ Measured
Qwen2.5-Coder 14B	78.31 tok/s	8.7 GB peak305 W86°C0.26 tok/WQ4_K_M	✓ Measured
Qwen3 14B	82.77 tok/s	257 W45°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 8

SDXL Turbo440.89

Stable Diffusion 1.540.7

DreamShaper XL Turbo28.09

Sana 1.6B20.44

Stable Diffusion XL8.03

Playground v2.54.9

Workload	Result	Telemetry	Data
Stable Diffusion 1.5	40.7 images/min	334 W45°C	✓ Measured
SDXL Turbo	440.89 images/min	150 W63°C	✓ Measured
Sana 1.6B	20.44 images/min	347 W47°C	✓ Measured
Stable Diffusion XL	8.03 images/min	346 W51°C	✓ Measured
DreamShaper XL Turbo	28.09 images/min	392 W70°C	✓ Measured
Playground v2.5	4.9 images/min	347 W53°C	✓ 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

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)	✕ 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

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 3080 Ti specifications

Architecture	Ampere (GA102)
CUDA cores	10,240
VRAM	12GB GDDR6X
Memory bus	384-bit
Memory bandwidth	912.4 GB/s
Boost clock	1,665 MHz
TDP	350 W
Process	8nm
Interface	PCIe 4.0 x16
Release date	2021-06-03
Launch MSRP	$1,199

Verdict, capable, but 12GB sets the ceiling

NVIDIA GeForce RTX 3080 Ti scores 3.5/100, #51 of 102. It ran 3 of 12; 8 exceeded its 12GB. Every figure here is our own measurement.

Relative performance: where the NVIDIA GeForce RTX 3080 Ti lands

100% = this card, AI & Machine Learning headline metric (AI Score). #17 of 61 desktop cards in this vertical.

GPU	%	AI Score
NVIDIA RTX 4000 (Ada Generation)	126%	4.4
NVIDIA GeForce RTX 4070 Ti Super	123%	4.3
AMD Radeon RX 7900 XTX	106%	3.7
GeForce RTX 5060 Ti	106%	3.7
NVIDIA GeForce RTX 3080 Ti	100%	3.5
NVIDIA GeForce RTX 4060 Ti 16GB	91%	3.2
NVIDIA GeForce RTX 4070 Ti	91%	3.2
GeForce RTX 5070	91%	3.2
NVIDIA GeForce RTX 4070 Super	89%	3.1

One tier up: GeForce RTX 5060 Ti (+6%) · compare head-to-head
One tier down: NVIDIA GeForce RTX 4060 Ti 16GB (-9%) · compare head-to-head
Best value alternative ($429): GeForce RTX 5060 Ti (+6%) · compare head-to-head

Same card, other workloads: NVIDIA GeForce RTX 3080 Ti Gaming benchmarks

← All AI & Machine Learning GPU rankings

The silicon

Transistors	28,300 million
Die size	628.4 mm²
Process node	8 nm
Fabricated by	Samsung
Transistor density	45 million per mm²

Denser than 81% 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	12.8 min	64.96 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 $1,199 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 8,258 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	11.3 years
8 hours a day, working on it	2,920	$424	2.8 years
24/7, always-on agent	8,760	$1,272	11.3 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.

Rental price

$0.121/hr-32.8% since 2026-08-14low $0.108 · high $0.180

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