Request a product
Photo 1 of 34

NVIDIA GeForce RTX 4090 · Photo source

NVIDIA GeForce RTX 4090
Photo 1 of 34

NVIDIA GeForce RTX 4090 · Photo source

NVIDIA · GeForce

GeForce RTX 4090

Released October 2022

NVIDIA's flagship Ada Lovelace graphics card, with 24 GB of GDDR6X, for 4K gaming, 3D rendering and AI work.

Memory
24 GB
Shaders
16,384
Boost clock
2,520 MHz
Bandwidth
1,008 GB/s
Power
450 W

What reviewers say

Reviewers agree it was by far the fastest graphics card at launch, with a huge generational jump over the RTX 3090 and DLSS 3 frame generation that impressed them, and TechRadar praised it for creative work too. The common complaints are the very high price and power demands and the card's sheer size; PC Gamer found its lead shrinks below 4K.

From tomshardware.com, pcgamer.com and techradar.com

Specifications

Report a wrong or missing spec

Performance

Chip maker
NVIDIA
Architecture
Ada Lovelace
Shader cores
16,384
Boost clock
2,520 MHz
Architecture
AD102 GPU; 128 SMs, 512 texture units, 176 ROPs
Manufacturing process
TSMC 4N NVIDIA custom process; 76.3 billion transistors; 608.5 mm² die
Ray tracing
128 third-generation RT Cores; 191 RT TFLOPS
AI cores
512 fourth-generation Tensor Cores; 1321 AI TOPS
Upscaling and frame generation
DLSS 3.5 with Super Resolution, DLAA, Ray Reconstruction and Frame Generation
Base clock
2.23 GHz
Cache
72 MB L2

Memory

Video memory
24 GB GDDR6X
Memory bus
384-bit
Memory bandwidth
1,008 GB/s
Memory detail
GDDR6X at 21 Gbps

Connections and power

Board power
450 W
Interface
PCIe 4.0
Display outputs
3 × DisplayPort 1.4a, 1 × HDMI 2.1a; up to 4 displays; HDCP 2.3
Maximum resolution
4K at 240 Hz or 8K at 60 Hz with DSC, HDR
Power connector
1 × 450 W or greater PCIe Gen 5 cable, or 3 × PCIe 8-pin cables with the adapter in the box
Recommended power supply
850 W
Slot width
3-slot, 61 mm (Founders Edition)
Size (reference card)
304 × 137 mm, 3-slot (Founders Edition)
Video encode and decode
2 × eighth-generation NVENC, 1 × fifth-generation NVDEC; AV1 encode and decode

Pros and cons

Pros

  • 16,384 CUDA cores and 24 GB of GDDR6X on a 384-bit bus (1,008 GB/s)
  • 72 MB of L2 cache
  • Two eighth-generation NVENC encoders with AV1

Cons

  • Founders Edition is a three-slot card, 304 mm long
  • 450 W total graphics power; NVIDIA asks for an 850 W power supply
  • DisplayPort 1.4a outputs and PCIe 4.0; no DLSS Multi Frame Generation

Closest rivals

Model
MemoryShadersBoost clockBandwidthPowerReleased
NVIDIAGeForce RTX 4080
16 GB9,7282,510 MHz716.8 GB/s320 WNov 2022
NVIDIAGeForce RTX 3090
24 GB10,4961,700 MHz936 GB/s350 WSept 2020
AMDRadeon RX 7900 XTX
24 GB6,1442,500 MHz960 GB/s355 WDec 2022

Newer model: GeForce RTX 5090.

Sources

  1. nvidia.com/en-gb/geforce/graphics-cards/40-series/rtx-4090
  2. nvidia.com/en-gb/geforce/graphics-cards/compare
  3. nvidia.com/en-gb/geforce/news/rtx-40-series-graphics-cards-announcements
  4. nvidia.com/en-us/geforce/news/rtx-40-series-graphics-cards-announcements
  5. images.nvidia.com/aem-dam/Solutions/geforce/blackwell/nvidia-rtx-blackwell-gpu-…

Photo: images.nvidia.com

Spot a mistake? Tell us