NVIDIA GeForce RTX 5090 Laptop

The Nvidia GeForce RTX 5090 Laptop is a mobile graphics card from the Blackwell family for laptops, which was presented at CES 2025 and was commercially available in April 2025.

Architecture

The GeForce RTX 5090 Mobile utilizes the GB203 chip and employs 10,496 shaders/CUDA cores, which is the maximum of 10,752 that the chip supports. This makes the RTX 5090 Laptop more similar to the RTX 5080 desktop variant (same chip) than the Desktop RTX 5090 with 20,760 cores and higher clock rates.

As a novelty, Blackwell offers support for Multi-Frame Generation (MFG) with DLSS4, thanks to the 5th generation of Tensor cores.

Performance

As our tests show, performance gains over a 175W RTX 4090 Laptop GPU are barely noticeable in 3DMark tests. Still, when it comes to overall gaming performance, the card is 2-19% faster than the previous mobile flagship. The fastest laptop variant we tested is only 5% behind the RTX 5080 desktop. This means that the 5090L is usually capable to display full details at 4k fluently (see below). Furthermore, with the mentioned Multi-Frame Generation (MFG) you can get impressive frame rates, but keep latency in mind.

Power consumption

The GB203 chip is manufactured at TSMC in the 4N (custom 4NP 5nm) process. The maximum TGP of the card can reach 175 watts (150 watts + 25 watts Dynamic Boost). 

GeForce RTX 5000 Laptop GPU Series GeForce RTX 5090 Laptop 10496 @ 0.99 - 1.52 GHz 256 Bit @ 28000 MHz GeForce RTX 5080 Laptop 8192 @ 0.98 - 2.29 GHz 256 Bit @ 28000 MHz GeForce RTX 5070 Ti Laptop 5888 @ 0.85 - 2.22 GHz 192 Bit @ 32000 MHz GeForce RTX 5070 Laptop 4608 @ 0.91 - 2.35 GHz 128 Bit @ 32000 MHz GeForce RTX 5060 Laptop 3328 @ 0.95 - 1.46 GHz 128 Bit @ 24000 MHz GeForce RTX 5050 Laptop 2560 @ 1.02 - 2.66 GHz 128 Bit @ 14000 MHz
Codename	GN22-X11
Architecture	Blackwell
Pipelines	10496 - unified
TMUs	328
ROPs	112
Raytracing Cores	82
Tensor / AI Cores	328
Core Speed	990 - 1515 (Boost) MHz
Theoretical Performance	31.8 TFLOPS FP16, 31.8 TFLOPS FP32
AI GPU	1824 TOPS INT8
Cache	L1: 10.3 MB, L2: 64 MB
Memory Speed	28000 effective = 1750 MHz
Memory Bus Width	256 Bit
Memory Type	GDDR7
Max. Amount of Memory	24 GB
Shared Memory	no
Memory Bandwidth	896 GB/s
API	DirectX 12_2, OpenGL 4.6, OpenCL 3.0, Vulkan 1.4
Power Consumption	150 Watt (95 - 150 Watt TGP)
Man. Technology	5 nm
PCIe	5
Displays	4 Displays (max.), HDMI 2.1b, DisplayPort 2.1b
Notebook Size	large
Date of Announcement	02.01.2025
Predecessor	GeForce RTX 4090 Laptop GPU

Benchmarks