Gigabyte has debuted a new Aorus External graphics card box featuring the flagship desktop RTX 5090. Using the Thunderbolt 5 interface, Gigabyte has brought the majority of that GPU's power to the e-GPU market, offering extreme gaming performance. It's not quite native grade, with this RTX 5090 performing around 73% as fast as the standard version. That puts it in the ballpark of a desktop RTX 4090.
Mobile gaming has always fallen behind desktop gaming in terms of raw performance. Even when desktop CPUs and GPUs are somehow fitted inside the confines of a laptop, the limited cooling and power of the form factor mean it's never quite as powerful as it could be on a desktop. External GPUs give more room and power to the card itself (and use a full-size desktop GPU to boot), but it's the interface that slows them down. Thunderbolt 5 holds a lot of promise, though, and this latest example from Gigabyte shows just how fast mobile gaming can be in the right circumstances.
The new Gigabyte Aorus RTX 5090 box is probably the most powerful external graphics you can find in 2025 (for now). Notebookcheck managed to get hold of one and put it through its paces, showing it hitting 73fps in its top gaming tests. That's far ahead of a desktop RTX 5080 and 9070 XT, and just behind a desktop RTX 4090. It was around 27% slower than a desktop 5090 in real-world gaming, though only 18% slower in overall performance.
Notebook Check's performance results are impressive.Credit: Notebook Check
This is an impressive threshold to hit, and it shows that you can get high-end, desktop-grade gaming performance with an external GPU enclosure. Sure, it costs $3,000 and requires its own dedicated (fortunately included) power supply. You'll also need a laptop or desktop PC with a full-bandwidth Thunderbolt 5 port and a compatible cable.
It's crazy expensive to play some games, but for the specific audience with deep pockets who has a high-end, lightweight laptop with long battery life and wants to play high-end games at home, it's just about the perfect solution.
In theory, it could allow drop-in AI performance upgrades for developers who want to bring their laptops to work with them.
