Wici One Puts an RTX 5090 and 4TB SSD in a Wireless eGPU Box Running Over Wi-Fi 7
Saturday, September 26, 2026By Indie Kings | September 26, 2026
Updated September 26, 2026: Wici has launched the Wici One, a wireless external GPU enclosure that ships an entire compute platform rather than just a card. The RTX 5090 version pairs the GPU with an Intel Core Ultra 7 255H and 4TB of NVMe storage, and it sends work to the box over Wi-Fi 7 instead of a cable.
Verification note: Wici's performance and pricing claims below are vendor statements reported by TechPowerUp. Wici makes no latency claims, and the Wi-Fi 7 link tops out at a fraction of the standard's theoretical maximum. We have not tested any of it.
Image: Wici One. Credit: TechPowerUp / Wici.
What the Wici One actually is
The framing that makes the Wici One interesting is that it is not primarily an external graphics card. It is a small computer that happens to contain one.
Most eGPU enclosures are a PCIe slot in a box with a Thunderbolt or OCuLink cable to a host. The Wici One replaces the cable with Wi-Fi 7, and then adds the parts an enclosure normally leaves out. Per TechPowerUp, it contains a CPU and storage as well as the GPU.
There is a notable absence in the published specifications. No RAM specs or quantities are listed, and the report notes the platform presumably relies entirely on VRAM, which for an RTX 5090 means 32GB standing in for system memory. That is an unusual architecture, and it is the single biggest open question about the product.
The bandwidth problem, stated plainly
Wireless eGPU has an obvious ceiling and it is worth putting the numbers next to each other rather than skipping past them.
| Measure | Value |
|---|---|
| Wi-Fi 7 theoretical maximum | Around 46.1 Gbps |
| Wici One radio | 4x4, running at 320 MHz with 4096-QAM |
| Wici One practical ceiling | 11.5 Gbps |
| Link type | Wireless, no cable connection |
So the radio delivers roughly a quarter of what the standard can theoretically do. TechPowerUp adds the caveat that those speeds are likely only achievable over very short distances with very little interference, which is the normal condition for Wi-Fi performance claims and should be assumed rather than tested.
For context on what that means in a wired comparison, PCIe 5.0 x16 is a different order of magnitude again, and even our coverage of Thunderbolt 5 eGPU docks sits on a far higher throughput floor than 11.5 Gbps. Wireless is not competing with cable on bandwidth and is not trying to.
The gaming numbers, and why they are not the pitch
Wici does publish gaming figures, and they are worth recording accurately before explaining why they are the least interesting part.
- Cyberpunk 2077 reaching 80 FPS at 4K Ultra with DLSS 4 on the RTX 5060 Ti version.
- Black Myth: Wukong hitting 60 FPS at the 4K Cinematic preset.
TechPowerUp is explicit that Wici makes no latency claims, and states plainly that gaming, being a latency-sensitive workload, is not the product's main selling point. That is a more credible position than a wireless eGPU marketing itself on framerate, because latency rather than throughput is what actually breaks remote rendering, and no amount of bandwidth fixes a variable round trip.
It is also worth naming what is missing. Both figures are vendor claims with no latency figure, no resolution scaling explanation, and no indication of whether DLSS 4 is doing the work or the link is. Anyone buying this for gaming is being asked to accept a frame rate without the input latency number that would make it meaningful.
The real use case is AI and rendering
The stated focus is local AI workloads and GPU-accelerated 3D rendering, and the model is straightforward: send the entire job to the Wici One's GPU over the network, complete the workload inside the enclosure, then return the result to the host machine.
That model is a much better fit for wireless than gaming is, for one reason. Inference and render jobs are throughput-bound and largely latency-tolerant compared with a game loop, and they are also jobs where the input is a prompt or a scene file rather than a stream of per-frame state. A job that takes minutes does not care that the result arrives 20 milliseconds later than it would over cable.
This is also where the internal storage stops being an oddity and becomes the point. A box with 4TB of NVMe next to the GPU means a render or model job can be sent once, run entirely locally, and the host only ever receives the finished output. That is a coherent machine for someone who wants a dedicated inference or render box that is not physically attached to the workstation driving it.
Configurations and pricing
Two configurations, and they are not close to each other in price or in parts.
| Specification | RTX 5060 Ti version | RTX 5090 version |
|---|---|---|
| GPU | GeForce RTX 5060 Ti | GeForce RTX 5090 |
| CPU | Intel Core 3 100U | Intel Core Ultra 7 255H |
| Storage | 1TB NVMe | 4TB NVMe |
| List price | $2,599 MSRP | Not announced |
| Early sign-up price | $1,999 | Not announced |
| Availability | Launching soon | Unspecified |
Per TechPowerUp, the RTX 5090 version has no listed price and the RTX 5060 Ti version is $2,599 MSRP with a $1,999 early sign-up price. The 5090 configuration at an unannounced price is the more interesting one and the less purchasable one.
Whether this is a product or a demonstration
Being fair about this, the honest read is that the Wici One is currently closer to a demonstration of an architecture than a product you would build a workflow on.
- No latency figures at all, from a product whose main use case is not latency-sensitive but whose secondary one certainly is.
- No system memory listed, with VRAM apparently doing double duty. That is workable for AI workloads and questionable for general compute.
- Gaming performance is vendor-claimed and the company does not even claim that as its focus.
- The 5090 version has no price or date, and the 5060 Ti version costs $2,599 before the discount, which is a lot for a wireless link that caps at 11.5 Gbps.
None of that means it will not matter. A wireless box that runs its own compute and returns only results is a genuinely different object from an eGPU, and if the latency turns out to be acceptable for the workloads Wici is actually targeting, the cable stops being the reason these things live next to the workstation. It just is not the product the framerate figures imply.
FAQ
What is the Wici One?
A wireless external GPU enclosure that uses Wi-Fi 7 instead of a cable, and contains a CPU, storage, and an NVIDIA GPU. It is closer to a small remote compute node than a traditional eGPU.
What GPU options are available?
GeForce RTX 5060 Ti and GeForce RTX 5090, per TechPowerUp. The RTX 5090 version does not yet have a listed price.
How fast is the wireless link?
The 4x4 radio at 320 MHz with 4096-QAM tops out at 11.5 Gbps, against a Wi-Fi 7 theoretical maximum of around 46.1 Gbps, and those speeds are likely only reachable over short distances with little interference.
What performance does Wici claim for gaming?
Cyberpunk 2077 at 80 FPS in 4K Ultra with DLSS 4 on the RTX 5060 Ti version, and Black Myth: Wukong at 60 FPS in 4K Cinematic. Wici makes no latency claims, and TechPowerUp notes gaming is not the product's main selling point.
What is the Wici One actually for?
Local AI workloads and GPU-accelerated 3D rendering. The job is sent to the enclosure over the network, completed there, and the result returned to the host, with the internal NVMe storage holding data locally alongside the GPU.
How much does it cost?
The RTX 5060 Ti version is $2,599 MSRP with a $1,999 early sign-up price. The RTX 5090 version has no announced price, and ships with an Intel Core Ultra 7 255H and 4TB NVMe against the 5060 Ti's Core 3 100U and 1TB.
Bottom Line
The Wici One is more interesting as an architecture than as a purchase. It is a complete compute node with a GPU, a CPU, and up to 4TB of local storage, reached over Wi-Fi 7 rather than a cable, which makes it the first eGPU-shaped product where the box is the computer. That is the right design for AI and render workloads, which are throughput-bound and tolerant of a slower result, and the wrong design for anything with a frame loop. The numbers deserve scepticism on both sides. There are no latency figures at all, there is no system memory specification with VRAM apparently doing double duty, and the two gaming results are vendor claims from a company that does not claim gaming as its focus. At $2,599 for the 5060 Ti version, with a 11.5 Gbps ceiling on a link the standard can theoretically do four times faster, this is not a cable replacement. It is a demonstration that a remote GPU node can be commercially packaged, and worth watching for exactly that reason.
Source: TechPowerUp, Wici Wireless eGPU Promises RTX 5060 Ti and 5090 Performance Over Wi-Fi 7, by Cpt.Jank, September 25, 2026
Related: Thunderbolt 5 eGPU dock with built-in storage | RTX 5090 at 575 watts, and the rest of the 50 series specs | GPU outlook and the RAM crisis