GPU Wiki

GeForce RTX 3090 vs GeForce RTX 4090

Popular comparisons

Drag a GPU to change the column order. With a keyboard, focus a GPU's move handle, use the arrow keys, then press Enter to apply.

Edit GPUs

Add a third or fourth GPU

Key Differences

  • No game was measured on every one of these cards on one shared test system, so there is no speed comparison to make.
  • RTX 4090 has the most memory bandwidth: 1,010 GB/s against 936.2 GB/s.
  • RTX 3090 draws the least power: 350 W against 450 W.
Spec GeForce RTX 3090 GeForce RTX 4090
Overview
Product name NVIDIA GeForce RTX 3090 NVIDIA GeForce RTX 4090
Market price Not tracked on Market Watch $4,395 Last price seen · US · New retail MSI Gaming X Trio · Newegg US · last seen 2 weeks ago, not in stock now
Architecture Ampere Ada Lovelace
Generation / family GeForce RTX 30 Series GeForce RTX 40 Series
GPU codename GA102 AD102
Launch date 2020-09-24 2022-10-12
Market segment Desktop, Flagship in GeForce RTX 30 Series Desktop, Flagship in GeForce RTX 40 Series
Board type / entry type Reference GPU Reference GPU
GPU
Process node Samsung 8 nm TSMC 5 nm
Die size 628 mm² 609 mm²
Transistor count 28.3 billion 76.3 billion
GPU chip GA102 AD102
SM / multiprocessor count 82 128
CUDA Cores 10496 16384
Tensor Cores 328 512
RT Cores 82 128
ROPs 112 176
TMUs 328 512
Clocks
Base / Core clock 1,395 MHz 2,235 MHz
Boost clock 1,695 MHz 2,520 MHz
Memory clock 1,219 MHz 1,313 MHz
Memory
Memory size 24 GB 24 GB
Memory type GDDR6X GDDR6X
Memory bus width 384-bit 384-bit
Memory bandwidth 936.2 GB/s 1,010 GB/s
Bus / Interface
Bus interface PCIe 4.0 x16 PCIe 4.0 x16
AGP / PCI / PCIe distinction PCIe PCIe
Power
TDP / TBP / board power 350 W 450 W
Recommended PSU 750 W 850 W
Power connectors 1x 12-pin 1x PCIe Gen 5 16-pin
Slot width Triple-slot, 336 mm length, 61 mm width Triple-slot, 304 mm length, 61 mm width
Display / Media
Display outputs 1x HDMI 2.1, 3x DisplayPort 1.4a 1x HDMI 2.1, 3x DisplayPort 1.4a
Features
DirectX support DirectX 12.2 DirectX 12.2
OpenGL support OpenGL 4.6 OpenGL 4.6
Vulkan support Vulkan 1.4 Vulkan 1.4
CUDA support 8.6 8.9
Shader Model 6.8 6.8
Ray tracing support Hardware RT units present Hardware RT units present
Performance / Benchmarks
FP32 throughput 35.58 TFLOPS 82.58 TFLOPS
FP16 throughput 35.58 TFLOPS 82.58 TFLOPS
Muted cells indicate unavailable, not applicable, or not directly comparable specifications.

DLAA counts as native resolution; DLSS, FSR and XeSS upscaling never do. A ray tracing setting without a readable on or off value is read as on.

Gaming benchmarks

Every card ran the same games at the same settings on the same test system. The index sets the fastest card to 100 using the eligible test runs.

No compatible approved gaming benchmark yet

Published observations exist, but no exact compatible game run is shared by every selected GPU. No result was estimated or filled in.