WiCi One Puts an RTX GPU on Wi-Fi 7, With Big Claims Attached

WiCi One is making a bold technical promise: put a desktop Nvidia GPU in a box, connect that box over Wi-Fi 7, and let a laptop or tablet use the graphics hardware as though it were local. I like the idea. I am not ready to trust the performance numbers yet.

This is not being pitched as ordinary game streaming. WiCi says an application continues to run on the client machine while its proprietary software sends GPU work to the nearby box. That distinction matters. A remote-play service renders a whole game elsewhere and streams compressed video back. WiCi One instead wants its driver and runtime to make a remote GPU available to existing software.

There is a complete PC inside the box

The entry configuration pairs a GeForce RTX 5060 Ti with 16GB of GDDR7, an Intel Core 3 100U processor, and a 1TB PCIe 4.0 NVMe SSD. The higher-end model steps up to an RTX 5090 with 32GB of GDDR7, a Core Ultra 7 255H, and a 4TB PCIe 5.0 SSD.

Those CPUs and SSDs are a clue that this is more than a wireless PCIe adapter. WiCi One has its own runtime, scheduler, storage, and networking stack. The listed radio is Wi-Fi 7 with 4x4 MIMO, a 320MHz channel, and multi-link operation. Driver mode is listed for Windows, macOS, and Linux. Developers can also reach the hardware through Python, JavaScript, and Swift SDKs, plus an OpenAI-compatible API for local models.

WiCi names caching and GPU trace replay as parts of its system. That makes sense. If frequently used assets and commands can remain beside the GPU, the product does not need to drag every byte across the wireless link for every frame. The exact transport is proprietary, though, so we do not yet know how often that optimization works or what happens when a workload changes constantly.

Wi-Fi 7 does not erase the bandwidth problem

A desktop GPU normally talks to the CPU over PCIe. A PCIe 4.0 x16 link offers roughly 31.5GB/s in each direction before protocol overhead. Even a fast Wi-Fi 7 connection works on a very different scale, and real throughput falls when distance, walls, neighboring networks, or a busy access point enter the picture.

Raw bandwidth is only part of it. Games also care about latency and consistency. An extra delay that is harmless during a video export can show up as an uneven frame in a shooter. Wireless jitter is especially awkward because average frame rate can look good while one-percent-low performance feels rough.

That does not make the design impossible. A Blender render or a local language-model request can tolerate some transfer delay if the large job stays on the WiCi box. A game that repeatedly moves assets and frame data is a harder case. I would expect results to vary much more by workload than they do with a conventional GPU installed in a desktop.

The published benchmarks come from WiCi

WiCi reports that its RTX 5060 Ti model reaches about 80fps in Cyberpunk 2077 and 60fps in Black Myth: Wukong at 4K Ultra with DLSS 4. The RTX 5090 version is listed at 110fps and 90fps in those same games. A demonstration also showed Cyberpunk 2077 running above 90fps, although the exact graphics settings were not visible.

These are useful targets, not independent test results. I want to see the complete client specification, render settings, DLSS mode, frame generation setting, network topology, distance from the box, average latency, and one-percent lows. I also want side-by-side results with the same GPU installed locally. Without that information, a large frame-rate number says very little about how responsive the game felt.

The same caution applies to the product's participation in Nvidia's Inception program. That program can give a company useful technical and business support, but it is not a third-party certification of WiCi's latency or benchmark claims.

The price puts it in workstation territory

The RTX 5060 Ti version is advertised at $1,999 for early signups, with a stated regular price of $2,599. Pricing for the RTX 5090 model has not been announced. Shipping is currently anticipated for the fourth quarter of 2026, so this remains a pre-release product rather than hardware anyone has independently reviewed at retail.

At $1,999, the practical comparison is not just another eGPU enclosure. It is a capable gaming desktop or workstation. A desktop will be less portable, but it also provides a direct PCIe connection, mature drivers, predictable latency, and easier upgrades. A wired Thunderbolt 5 or OCuLink eGPU is less tidy than Wi-Fi, yet the cable removes a major source of variability.

Where the idea could actually fit

I can see a useful niche in a small studio with thin laptops, especially when long renders or local AI jobs can be queued on shared hardware. It could also help someone who wants one expensive GPU available to several low-power devices without maintaining a full tower at every desk.

Sharing brings another open question: contention. If one person starts a 4K export while another launches a game, the scheduler has to divide GPU time and network capacity without making both sessions miserable. WiCi says multiple devices can connect, but published peak numbers do not explain behavior under simultaneous load.

Before spending anything, I would wait for independent tests covering latency, frame-time graphs, interference, range, thermals, noise, software compatibility, and recovery after a dropped connection. If WiCi One handles those basics, a wireless GPU could be genuinely useful. Until then, it is an interesting architecture attached to unverified numbers.

Sources: TechSpot's September 27 report and the WiCi One product page.

Share: