When it comes to pushing your laptop’s performance to the next level, few upgrades make a bigger difference than connecting an external GPU. GIGABYTE’s new AI BOX, powered by Thunderbolt™ 5, lets you unlock desktop-class performance for both gamers and AI creators alike.
But what makes AI BOX truly special is that it’s not only designed for lightweight ultrabooks — it also works alongside gaming laptops with built-in discrete GPUs (dGPUs). You can split the workload, running different tasks on each GPU to maximize your productivity and fun.
For the vast majority of games and applications (including some software that usually runs alongside others, like OBS Studio and Discord), Windows provides a straightforward method for GPU assignment.
Right-click on your desktop and select "Display Settings."
Navigate to "Graphics."
You’ll see a list of applications. If the application you want to configure isn’t listed, click "Add desktop app" and browse to its executable file.
Once added, click to expand the options for a certain app, and navigate the “GPU preference” drop-down menu.
Choose your preferred GPU: Windows will display the GPU name, and have several preset options like power saving, high performance.
If you are assigning the apps to the AI BOX, they should be displayed as “NVIDIA GeForce RTX 5090” or “NVIDIA GeForce RTX 5060 Ti”, without a suffix like “Laptop GPU”.
However, some AI applications require a more direct approach that bypasses Windows’ graphics settings. Let’s look at two popular examples: LM Studio and ComfyUI.
LM Studio is a popular entry point for experiencing local AI models. Unlike typical applications, it bypasses Windows’ graphics settings and handles GPU selection internally.
Open LM Studio.
Click on the gear icon (App Settings) in the lower‑right corner.
In the pop-up, select Mission Control → Hardware.
You’ll see a list of available GPUs: your laptop’s dGPU and the AI BOX eGPU.
You can choose a specific GPU, utilize multiple GPUs simultaneously, and adjust settings for offloading data to DRAM.
ComfyUI, a powerful visual workflow tool for AI image generation, requires a bit more hands-on configuration.
Locate the installation root folder of ComfyUI on your machine.
Inside, navigate to /user/default/comfy.settings.json.
Open the file in Notepad (or any plain‑text editor).
Find the last closing brace } at the end of the file and paste the following before it:
,
"Comfy.Server.LaunchArgs": {
"cuda-device": "1"
},
"Comfy.Server.ServerConfigValues": {
"cuda-device": 1
}
Save the file and restart ComfyUI.
The "cuda-device" number tells ComfyUI which GPU to use (0 = first GPU, 1 = second, etc.). Because the AI BOX is a hot-pluggable device and Windows resource allocation can be dynamic, the numbering is not fixed. We recommend testing a few times to see if your image generation is using the intended GPU.
Whether you’re a gamer, a streamer, or an AI hobbyist, the AORUS AI BOX gives your laptop a flexible way to scale.
Ultrabook users can enjoy a massive performance upgrade for both gaming and AI workloads.
Gaming laptop users can enjoy dual-GPU multitasking — playing games while simultaneously running AI tasks, rendering, or streaming without performance drops.
Ultra‑thin ultrabook with no discrete GPU: Plug in the AI BOX for all heavy graphics and AI workloads.
Content creator streaming live for gaming: Use the AI BOX for gaming, as it delivers significantly better performance than any laptop GPU; Assign the laptop’s dGPU to rendering tasks like OBS Studio.
AI image generation and ComfyUI on the AI BOX; adjust via manual GPU configuration for optimal performance.while gaming: Use the AI BOX for gaming, as it delivers significantly better performance than any laptop GPU; Offload ComfyUI or Stable Diffusion tasks to the laptop’s dGPU.
AI workflow + video editing: Assign LM Studio or ComfyUI to the AI BOX for faster response; Keep video editing software like Premiere Pro on the dGPU for lower latency with the CPU.
Running multiple local AI tools (e.g LM Studio + ComfyUI + OpenAI Whisper): Distribute workloads: Run LM Studio and Whisper on the laptop GPU, and ComfyUI on the AI BOX.
Ultra‑thin ultrabook with no discrete GPU: Plug in the AI BOX for all heavy graphics and AI workloads.
Content creator streaming live for gaming: Use AI BOX for rendering tasks like OBS Studio; Keeps the game on the laptop dGPU for less latency.
AI image generation while gaming: Use the AI BOX for ComfyUI or other AI generation workloads; Keep the game on the laptop dGPU for less latency.
AI workflow + video editing: Keep LM Studio or ComfyUI on the dGPU, as the shorter workload duration reduces the chance of throttling; Assign video editing software like Premiere Pro to the AI BOX.
Running multiple local AI tools (e.g LM Studio + ComfyUI + OpenAI Whisper): Assign one AI app per GPU to maintain stability and avoid VRAM contention.