Summary
WiCi has introduced WiCi One, a desktop GPU system designed to share graphics and AI compute with nearby devices over Wi-Fi 7. The company lists RTX 5060 Ti and RTX 5090 versions and says shipping is anticipated to start in Q4.
WiCi has introduced WiCi One, a desktop GPU system intended to make graphics and AI compute available to nearby phones, laptops and other devices over Wi-Fi. The company says applications can use it as a virtual GPU, call it through an API or work with its software development kit (SDK), while the main device keeps the app’s interface and user input local.
The system is designed for home or office use rather than as a cloud service. WiCi says GPU-heavy work is sent over Wi-Fi to the nearby box, which contains the graphics processor and cooling hardware. Its pitch is that a user can share one powerful GPU across several devices instead of buying a separate high-end computer for each workload.
How WiCi One is designed to work
WiCi lists two configurations: one with an NVIDIA RTX 5060 Ti with 16GB of GDDR7 memory, and another with an RTX 5090 with 32GB of GDDR7. The box also includes a Wi-Fi 7 radio with 4×4 MIMO and 320MHz bandwidth. Wi-Fi 7 is the wireless networking standard the system uses to connect the GPU box with devices on the local network.
The company describes three ways for software to use the hardware. In Driver mode, WiCi One presents itself as a virtual GPU for supported applications, which WiCi says can run without modification. SDK mode provides tools for developers, including Python, JavaScript and Swift support. API mode offers an OpenAI-compatible endpoint, allowing compatible software clients to send requests to the device.
WiCi says the device handles GPU-bound work while an application’s interaction and interface stay on the user’s computer. That is different from streaming a game’s video from a remote computer: the company describes WiCi One as a nearby compute resource connected over Wi-Fi, with no cloud service in between.
Configurations, performance claims and timing
The two versions also differ in their other listed hardware. The RTX 5060 Ti edition has an Intel Core 3 100U processor and 1TB of NVMe storage; the RTX 5090 edition has an Intel Core Ultra 7 255H and 4TB of NVMe storage. WiCi says its TurboStream software can use NVMe storage for model-weight paging, a way of moving model data between storage and memory when it does not all fit in GPU memory.
WiCi publishes gaming results for both configurations. It reports 80 frames per second for Cyberpunk 2077 at 4K Ultra with DLSS 4 on the 5060 Ti edition, and 110 fps on the 5090 edition. For Black Myth: Wukong at 4K Cinematic with DLSS 4, the company reports 60 fps and 90 fps respectively. WiCi says these results came from unmodified games running over Wi-Fi 7 on a production WiCi One. They are the manufacturer’s reported figures.
The product page advertises a starting price of $1,999 for early sign-ups, compared with a stated regular starting price of $2,599. WiCi says pricing for the RTX 5090 edition will be announced closer to launch. It describes shipping as anticipated to begin in Q4, without specifying a year in the supplied product information.