← Back to GPUs
NVIDIA GeForce RTX 5090

NVIDIA · RTX 50

NVIDIA GeForce RTX 5090

$2800$1999 MSRP

The NVIDIA GeForce RTX 5090 is the most powerful consumer GPU ever made, built on the Blackwell architecture with 32GB of GDDR7 memory and 1,792 GB/s bandwidth. It is the first consumer card to break the 24GB VRAM barrier, making it capable of running 70B parameter LLMs at 8-bit quantization entirely in VRAM. For gamers, it delivers unmatched 4K performance with DLSS 4 Multi Frame Generation. For AI developers, it is the best single-GPU solution available outside of data center hardware.

Best ForRunning 70B+ LLMs locally and 4K gaming without compromise
VerdictThe ultimate GPU for AI and gaming — if you can find one at a reasonable price.
AI
10/10
Gaming
10/10

Specifications

VRAM32GB GDDR7
Memory Bandwidth1792 GB/s
CUDA Cores21,760
Boost Clock2407 MHz
TDP575W
Power Connector1x 16-pin
Length340mm
Form FactorTriple Slot
Release Year2025

AI Capabilities

Unrivaled32GB VRAM

Run 70B+ models, no compromises. The AI power user's dream.

Can run (Q4 quantized)

Llama 3.1 8BQwen 2.5 32BQwen 2.5 14BMistral 7BFLUX.1 DevStable Diffusion XLStable Diffusion 3.5 LargeHunyuanVideoCogVideoX-5BMochi 1LTX VideoStable Video DiffusionWan Video 14BCodestral 22BQwen 2.5 Coder 32BLLaVA 1.6 34BAlphaFold 2ESMFold (ESM-2 15B)ESM-2 3BscGPTRFdiffusionFine-tune Llama 8BTrain SDXL LoRATrain FLUX LoRA

Tight fit (may need CPU offload)

Llama 3.1 70B (40GB Q4)Qwen 2.5 72B (42GB Q4)DeepSeek R1 70B (40GB Q4)Fine-tune Llama 70B (40GB Q4)

Recommended system RAM for AI: 64GB+ (2x GPU VRAM for model overflow)

Performance Estimates

Estimated tokens/sec for LLM inference based on 1792 GB/s memory bandwidth — not hardware benchmarks. Methodology · What is Q4/Q8?

Llama 3.1 70B70B
Offload~1-3 tok/sVery slow
Llama 3.1 8B8B
FP16~65-80 tok/sExcellent
Qwen 2.5 72B72B
Offload~1-3 tok/sVery slow
Qwen 2.5 32B32B
Q8~34-42 tok/sFast
Qwen 2.5 14B14B
FP16~37-46 tok/sFast
Mistral 7B7B
FP16~74-92 tok/sExcellent
DeepSeek R1 70B70B
Offload~1-3 tok/sVery slow
Codestral 22B22B
Q8~50-62 tok/sFast
Qwen 2.5 Coder 32B32B
Q8~34-42 tok/sFast

Pros

  • +Best gaming performance available
  • +32GB GDDR7 for large AI models
  • +DLSS 4 Multi Frame Generation

Cons

  • -Extremely expensive
  • -Very high power draw
  • -Hard to find in stock
aigamingworkstation

Where to Buy

Best Buy→Newegg→

Will It Run?

Llama 3.1 70B70B
Offload
Llama 3.1 8B8B
FP16
Qwen 2.5 72B72B
Offload
Qwen 2.5 32B32B
Q8
Qwen 2.5 14B14B
FP16
Mistral 7B7B
FP16
DeepSeek R1 70B70B
Offload
FLUX.1 Dev12B
FP16

People Also Compared

NVIDIA GeForce RTX 5090 vs NVIDIA GeForce RTX 409024GB · $1400NVIDIA GeForce RTX 5090 vs NVIDIA GeForce RTX 508016GB · $1150NVIDIA GeForce RTX 5090 vs NVIDIA GeForce RTX 3090 Ti24GB · $1000NVIDIA GeForce RTX 5090 vs NVIDIA GeForce RTX 4080 SUPER16GB · $950NVIDIA GeForce RTX 5090 vs NVIDIA GeForce RTX 309024GB · $900

Featured in a Recommended Build

AI Workstation70B models at Q8. No compromises on VRAM or speed.View full build →
Compare with Other GPUsBuild a PC with This GPUQuantization Guide