GPUDesktopReleased Jan 30, 2025

NVIDIA GeForce RTX 5090 D

512-bit · 21,760 cores

The GeForce RTX 5090 D is the China-market version of the GeForce RTX 5090, released on January 30, 2025 to comply with US export controls. It keeps the full CUDA, Tensor and RT core counts and 32 GB of GDDR7 memory of the global model, but its AI performance is limited to 2,375 AI TOPS. In August 2025 NVIDIA added the RTX 5090 D v2, which has the same core configuration and price but 24 GB of memory on a 384-bit bus.

General

Manufacturer
NVIDIA
Model
GeForce RTX 5090 D
Segment
Desktop
GPU chip
GB202
Architecture
Blackwell
Release date
Jan 30, 2025
Price
CN¥16,499Search on Amazon

Memory

Memory size
32 GB
Memory type
GDDR7
Bus width
512-bit
Bandwidth
1,792 GB/s
Memory clock
3,500 MHz

Clocks & silicon

GPU clock
2,017 MHz
Boost clock
2,407 MHz
Shader units
21,760 CUDA Cores
Tensor / AI cores
680
RT cores
170
Transistors
92.2 billion
Technology (nm)
4 nm
TSMC 4N
TDP / TBP
575 W

Interface & APIs

Bus interface
PCIe 5.0 x16
SLI / multi-GPU
None (no SLI / NVLink)
DirectX
12 Ultimate (12_2)
OpenGL
4.6
Vulkan
1.4
Other APIs
CUDA 12.0, OpenCL 3.0

AI support

Precisions
FP8FP16BF16INT8INT4TF32FP64
AI accelerator
5th Gen Tensor Cores (680) + 4th Gen RT Cores (170)
Peak AI TOPS
2,375 TOPS
Notes
Fifth-generation Tensor Cores add FP4 alongside FP8, BF16, FP16, TF32, INT8 and INT4. NVIDIA rates the RTX 5090 D at 2,375 AI TOPS (FP4 with sparsity), about 29% below the 3,352 AI TOPS of the global RTX 5090, to meet US export-control limits. FP64 runs at 1/64 of the FP32 rate.

Related NVIDIA GPUs

All NVIDIA GPUs →

Desktop gpus from NVIDIA announced around the same time as the NVIDIA GeForce RTX 5090 D. Add any of them to your compare list to see the specs side by side.