Settings

Comparing NVIDIA GeForce RTX 3090 vs. 4090

Below you can compare the latest GPU-s across multiple benchmarks. Explore detailed graphs, performance scores, and rankings to quickly find the model that best suits your workflow and creative needs.

search

No match

compare_arrows
search

No match

speed

Verdict

The RTX 3090 is 50% slower than the RTX 4090

Averaged over 4 rendering benchmarks, the RTX 4090 leads in 4 of them. The biggest gap is in OctaneBench 2025, where the RTX 4090 scores 108% higher.

642 · RTX 3090
1,337 · RTX 4090
18,921 · RTX 3090
34,772 · RTX 4090
3,325 · RTX 3090
6,861 · RTX 4090
5,864 · RTX 3090
11,733 · RTX 4090
savings

Price and value

RTX 3090 No reliable price in United States Fewer than five shops list it new.
RTX 4090 No reliable price in United States Fewer than five shops list it new.

Buy a card or render on the farm

Your RTX 3090 or 100+ GPUs at once?

frames at min per frame on the RTX 3090
RTX 3090
20 h 50 min
RTX 4090
10 h 0 min
Drop & Render
9 min 67× faster
Sapphire 18 min €48.60
Emerald 12 min €77.76
Diamond 9 min €145.80

Estimated with our live price calculator for 250 frames. Local times scale with OctaneBench 2025 scores; farm times assume enough free nodes. From €48.60 on Sapphire.

grouped_bar_chart

Ranking by benchmarks

OctaneBench 2025

Cinebench 2024

V-Ray benchmark

Blender benchmark

list

Specifications

Model GeForce RTX 3090 GeForce RTX 4090
Manufacturer nVidia nVidia
Year September 2020 October 2022
Code Name GA102-300-A1 AD102
Memory 24576 MB 24576 MB
Core Speed 1395 MHz 2235 MHz
Memory Speed 1219 GB/s 1325 GB/s
Power (Max TDP) 350 watts 450 watts
Bandwidth 958668 MB/sec 1042432 MB/sec
Texel Rate 457560 Mtexels/sec 1144320 Mtexels/sec
Pixel Rate 156240 Mpixels/sec 393360 Mpixels/sec
Unified Shaders 10496 16384
Texture Mapping Units 328 512
Render Output Units 112 176
Bus Type GDDR6X GDDR6X
Bus Width 384-bit 384-bit
Fab Process 8 nm 4 nm
Transistors 28300 million 76300 million
Bus PCIe 4.0 x16 PCIe 4.0 x16
DirectX Version DirectX 12 DirectX 12
OpenGL Version OpenGL 4.6 OpenGL 4.6
search

Methodology

Our data above is compiled from multiple sources:

  • Our own benchmarks: We conduct extensive testing with the GPU and CPU models we use on the farm. Over the past 10 years we have tested over 1,000 GPUs and 500 CPUs.

  • Model scores: Once we have two or more benchmarks for a device we can perform some machine learning computations to estimate a realistic score in any of the other benchmarks we list. These are always marked with an asterisk (*), and while usually accurate shouldn't be taken as the complete truth.

  • Public benchmark data: We may supplement our dataset with scores from public OctaneBench, V-Ray Benchmark and Blender Benchmark websites. We cannot test every device available ourselves, so this gives us a nice alternative to get You the results you are looking for.

We'll start you off with €25 credits!

We use cookies.