NVIDIA / Enthusiast tier
NVIDIA GeForce RTX 4090
NVIDIA's consumer GPU with 24GB of GDDR6X memory, used for local AI training and inference.
- Tdp
- 450W
- HDCP
- 2.3
- Vram
- 24GB GDDR6X
- Sloty
- 3 sloty (61 mm)
The decision
Where it wins
- Supports 4K and 8K resolutions with high refresh rates.
- Includes advanced AI and ray tracing capabilities.
- Multi-monitor support up to four displays.
Where it doesn't
- Requires a powerful power supply unit (PSU) to handle its 450W TDP.
- Large form factor (304 mm length) may not fit in all cases.
- High price point of $1599.
Ideal forEnthusiasts and professionals requiring high-end graphics performance.Users working with AI and machine learning applications.Gamers seeking the best visual fidelity and performance.
Key Features
- ✦24GB GDDR6X memory
- ✦4K and 8K resolution support
- ✦Advanced AI and ray tracing capabilities
- ✦Support for up to four monitors
- ✦High power consumption (450W TDP)
02 / Technical record
Specifications
Structured product data, grouped by system.
Core hardware05
| AD102 GPU Notes VDPAU | Feature Set K |
|---|---|
| Clock Speeds Base Clock | 2235 MHz |
| Graphics Features Shader Model | 6.9 |
| GPU Engine Specs / Base Clock (GHz) | 2.23 |
| Technology Support / NVIDIA Architecture | Ada Lovelace |
Memory & compute10
| Vram | 24GB GDDR6X |
|---|---|
| Memory Bandwidth | 1.01 TB/s |
| Memory Memory Bus | 384 bit |
| Memory Memory Size | 24 GB |
| Memory Memory Type | GDDR6X |
| Render Config L1 Cache | 128 KB (per SM) |
| Clock Speeds Memory Clock | 1313 MHz |
| AD102 GPU Notes Frame Generation | DLSS 4.5, FSR 3.1.6, XeSS 3 |
| Memory Specs / Memory Interface Width | 384-bit |
| Memory Specs / Standard Memory Config | 24 GB GDDR6X |
AI & acceleration20
| Cuda Cores | 16384 |
|---|---|
| NVIDIA DLSS | DLSS 3.5 Super Resolution DLAA Ray Reconstruction Frame Generation |
| Ray Tracing | Ano |
| Boost Clock (GHz) | 2,52 |
| NVIDIA GPU Boost ™ | Ano |
| Graphics Features CUDA | 8.9 |
| JáDra NVIDIA CUDA ™ | 16 384 |
| Render Config RT Cores | 128 |
| Clock Speeds Boost Clock | 2520 MHz |
| MožNosti Technologie CUDA | 8.9 |
| Render Config Tensor Cores | 512 |
| AD102 GPU Notes Tensor Cores | 4th Gen |
| NVIDIA NVLink ™ (SLI-Ready) | Ne |
| GPU Engine Specs / Shader Cores | Ada Lovelace |
| AD102 GPU Notes Ray Tracing Cores | 3rd Gen |
| GPU Engine Specs / NVIDIA CUDA Cores | 16384 |
| GPU Engine Specs / Ray Tracing Cores | 3rd Generation |
| Technology Support / CUDA Capability | 8.9 |
| Technology Support / NVIDIA GPU Boost | Yes |
| Technology Support / NVIDIA NVLink (SLI-Ready) | No |
Display & media14
| KřěDení AV1 | Ano |
|---|---|
| DekčKováNí AV1 | Ano |
| NVIDIA G-SYNC ™ | Ano |
| AD102 GPU Notes NVDEC | 5th Gen |
| AD102 GPU Notes NVENC | 8th Gen |
| NVIDIA Decoder (NVDEC) | 5. generace |
| NVIDIA Encoder (NVENC) | 2x 8. generace |
| Display Support / Multi Monitor | up to 4 |
| Technology Support / AV1 Encode | Yes |
| AD102 GPU Notes Super Resolution | DLSS 4.5, FSR 3.1.5, XeSS 3 |
| Technology Support / NVIDIA G-SYNC | Yes |
| Display Support / Standard Display Connectors | HDMI (2), 3x DisplayPort (3) |
| Thermal And Power Specs / Video Playback Power (W) | 25 |
| Display Support / Maximum Resolution & Refresh Rate | 4K at 240Hz or 8K at 60Hz with DSC, HDR |
Power & thermal09
| Tdp | 450W |
|---|---|
| Board Design TDP | 450 W |
| Board Design Suggested PSU | 850 W |
| Thermal And Power Specs / Idle Power (W) | 19 |
| Thermal And Power Specs / Average Gaming Power (W) | 315 |
| Thermal And Power Specs / Total Graphics Power (W) | 450 |
| Thermal And Power Specs / Required System Power (W) | 850 |
| Thermal And Power Specs / Maximum GPU Temperature (In C) | 90 |
| Thermal And Power Specs / Supplementary Power Connectors | 3x PCIe 8-pin cables (adapter in box) OR 450 W or greater PCIe Gen 5 cable |
Physical & connectivity11
| Sloty | 3 sloty (61 mm) |
|---|---|
| Board Design Width | 61 mm |
| Board Design Height | 137 mm |
| Board Design Length | 304 mm |
| Board Design Weight | 2181 g |
| Board Design Slot Width | Triple-slot |
| Card Dimensions / Slots | 3-Slot (61mm) |
| Card Dimensions / Width | 137 mm |
| PCI Express 4. Generace | Ano |
| Card Dimensions / Length | 304 mm |
| Technology Support / PCI Express Gen 4 | Yes |
Platform & software09
| Graphics Features OpenGL | 4.6 |
|---|---|
| Graphics Features Vulkan | 1.4 |
| Vulkan RT API, OpenGL 4.6 | Ano |
| Technology Support / Photo Mode | Yes |
| Technology Support / NVIDIA FreeStyle | Yes |
| Technology Support / Game Ready Drivers | Yes |
| Technology Support / NVIDIA Studio Drivers | Yes |
| Technology Support / NVIDIA GeForce Experience | Yes |
| Technology Support / Microsoft DirectX 12 Ultimate | Yes |
Other specifications47
| HDCP | 2.3 |
|---|---|
| DěLka | 304 mm |
| VR Ready | Ano |
| ťířKa | 137 mm |
| Typ PaměTi | GDDR6X |
| NVIDIA Reflex | Ano |
| Resizable BAR | Ano |
| VíCe Monitorů | až 4 (4) |
| NVIDIA Broadcast | Ano |
| NVIDIA Freestyle | Ano |
| NVIDIA Omniverse | Ano |
| Velikost PaměTi | 24 GB |
| NVIDIA Highlights | Ano |
| NVIDIA ShadowPlay | Ano |
| Render Config ROPs | 176 |
| Architektura NVIDIA | Ada Lovelace |
| OvladačY Game Ready | Ano |
| Matrix Performance FP8 | 330.3 TFLOPS |
| Render Config SM Count | 128 |
| Matrix Performance BF16 | 165.2 TFLOPS |
| Matrix Performance INT4 | 1,321.2 TOPS |
| Matrix Performance INT8 | 660.6 TOPS |
| Matrix Performance TF32 | 82.58 TFLOPS |
| OvladačY NVIDIA Studio | Ano |
| Graphics Features OpenCL | 3.0 |
| ZáKladní Frekvence (GHz) | 2,23 |
| Render Config Shading Units | 16384 |
| Theoretical Performance FP16 | 82.58 TFLOPS (1:1) |
| Theoretical Performance FP32 | 82.58 TFLOPS |
| Theoretical Performance FP64 | 1.290 TFLOPS (1:64) |
| Numeric Format Support Matrix | INT1, INT2, INT4, INT8, MXFP4, NVFP4, MXFP6, FP8, MXFP8, FP16, BF16, FP32, TF32, FP64 |
| Numeric Format Support Vector | INT4, INT8, INT16, INT32, FP16, BF16, FP32, FP64 |
| Pomocné NapáJecí Konektory | 3x 8pinový kabel PCIe (adaptěr v krabici) NEBO kabel PCIe Gen 5 (450 W nebo více) |
| Standardní Konfigurace PaměTi | 24 GB GDDR6X |
| ŠířKa PaměžOvéHo Rozhraní | 384 bitť |
| Konektory Pro Standardní Monitor | HDMI (2), 3x DisplayPort (3) |
| Microsoft DirectX ™ 12 Ultimate | Ano |
| NVIDIA ™ GeForce Experience ™ | Ano |
| Theoretical Performance Pixel Rate | 443.5 GPixel/s |
| PožAdovaný VěYn SystéMu (W) (8) | 850 |
| Theoretical Performance Texture Rate | 1,290.2 GTexel/s |
| ČLenáRný VěYn PřI Hraní (W) (7) | 315 |
| Celková SpotřEba Grafické Karty (W) | 450 |
| MaximáLní DigitáLní RozliťEní (1) | 4K při 240 Hz nebo 8K při 60 Hz s DSC, HDR |
| VěYnek PřI NečNosti SystéMu (W) (5) | 19 |
| MaximáLní Teplota Grafické Karty (č) | 90 |
| VěYnek PřI PřEhňOváNí Videa (W) (6) | 25 |
Evidence / qualification
Read the record
Specs not independently verifiednvidia.com
- Source on record
- nvidia.com
The NVIDIA GeForce RTX 4090 is a powerful graphics card designed for enthusiasts and professionals requiring top-tier performance in gaming, AI, and creative work. It supports 4K and 8K resolutions, multi-monitor setups, and advanced features like DLSS 3.5 and Ray Tracing. However, its large form factor and high power consumption (450W) make it less suitable for compact builds or systems with limited power supply capacity. The card's extensive feature set and high price point cater to users who demand the best in graphical fidelity and computational power.
The CUDA Juggernaut
The RTX 4090 remains the holy grail for local AI practitioners who need raw CUDA compute. While Macs win on total memory capacity, NVIDIA wins on speed, software support, and ecosystem maturity.
Why 24GB matters
For fine-tuning models using LoRA or running Stable Diffusion XL, 24GB of VRAM is the sweet spot. Anything less requires heavy quantization, which degrades model output quality.
Frequently Asked Questions
Do I need a new power supply?↓
Most likely. A high quality 850W or 1000W ATX 3.0 power supply is highly recommended.
Can I fit two in a standard PC?↓
It is extremely difficult due to the 3-slot to 4-slot thickness of most models. Blower-style cards or custom water cooling are required for multi-GPU setups.
Who is NVIDIA GeForce RTX 4090 for?↓
NVIDIA GeForce RTX 4090 is a fit for enthusiasts and professionals requiring high-end graphics performance., Users working with AI and machine learning applications. and Gamers seeking the best visual fidelity and performance..
What are the drawbacks of NVIDIA GeForce RTX 4090?↓
The trade-offs we record are: Requires a powerful power supply unit (PSU) to handle its 450W TDP., Large form factor (304 mm length) may not fit in all cases. and High price point of $1599..
What does NVIDIA GeForce RTX 4090 do well?↓
Supports 4K and 8K resolutions with high refresh rates., Includes advanced AI and ray tracing capabilities. and Multi-monitor support up to four displays..
How much does NVIDIA GeForce RTX 4090 cost?↓
$1,599, as recorded in this index. Prices move; check the retailer for the current figure.
What is the Fathom Layer design score for NVIDIA GeForce RTX 4090?↓
NVIDIA GeForce RTX 4090 scores 9.5 out of 10. The score is assigned by a human against published criteria and placement is never sold — see the methodology page for how it is built.
Alternatives to consider
See all alternatives →A point of difference
Memory on record
NVIDIA GeForce RTX 409024GB GDDR6X
Nvidia GeForce RTX 5070 Ti (16GB)16GB GDDR7
- Another option Nvidia GeForce RTX 5070 Ti (16GB)
16GB VRAM is the practical floor for comfortable local inference
16GB GDDR7Design score 9.3/10 - Another option Nvidia GeForce RTX 5090 (32GB)
32GB GDDR7 — runs 70B-class models quantised on a single card
32GB GDDR7Design score 9.7/10 - Another option Nvidia GeForce RTX 5080 (16GB)
Solid generational gain over the 4080
16GB GDDR7Design score 9.1/10
Where to buy
Purchase links may earn Fathom Layer a commission, at no extra cost to you. As an Amazon Associate we earn from qualifying purchases.
Continue the inquiry
Further reading
Featured badge
Building this product? Add the badge to your site to show it’s in the index.
<a href="https://fathomlayer.com/compute/local-ai-workstations/nvidia-rtx-4090-fe" target="_blank" rel="noopener noreferrer"><img src="https://fathomlayer.com/fathom-badge.svg" alt="Featured on Fathom Layer" width="250" height="54" /></a>