Wi-Sci
A wireless computing infrastructure solution that transparently delivers remote desktop-class GPU compute capability to terminal devices over a wireless network, so that the remote GPU is recognized as local hardware at the system level.
Timeline (last 90 days)
Wi-Sci is currently an early engineering prototype; the unit Chuck received arrived in a suitcase stuffed with bubble wrap.
Wi-Sci installs a "fake driver" that makes Windows and game engines recognize the remote GPU as a local PCIe device, so it shows as an RTX 5090 in Task Manager and games run without issue.
Through four optimizations—expert-aligned chunking, hot expert residency, compute splitting and scheduler preloading—Wi-Sci raises effective SSD bandwidth from 1 GB per second to about 11.5 GB per second, enabling DeepSeek 4.1 Flash to reach a generation speed of about 8-15 tokens/second on the Wi-Sci box.
The Wi-Sci scheduler can cut the switch time from an image model back to a chat model from about 17-18 seconds without the scheduler to about 4 seconds, and the reverse switch from about 18 seconds to about 8.6 seconds.
The Wi-Sci product comes in two versions, RTX 5090 and RTX 5060 Ti, and is now open for pre-order, with shipping expected in eight weeks.
Wi-Sci's proprietary protocol includes four mechanisms: instruction batching compression, incremental data transmission, predictive acknowledgment and compressed video-stream return.
In Chuck's demo, a Mac connected to the Wi-Sci box over Wi-Fi, used Figma to generate images, Love Art to generate video and DaVinci Resolve to render video, with the GPU scheduled across multiple tasks but not truly in parallel.
Wi-Sci uses a proprietary protocol to compress the bandwidth requirement for wireless GPU communication from 8 gigabits per second to about 400 megabits, enabling the laptop to run Cyberpunk 2077 at 92 frames per second with Wi-Fi usage steady at around 400 megabits.
The Wi-Sci scheme claims to offer a 410TB memory pool, but the author stressed "not in the way you think," implying it is not single-machine VRAM.
Traditional remote streaming (such as Steam Remote Play) captures frames and input at the application layer, whereas Wi-Sci attempts to virtualize PCIe devices at the system layer so that all GPU calls are transparently redirected to the remote end.
4 more timeline events
All Facts (14)
Wi-Sci is currently an early engineering prototype; the unit Chuck received arrived in a suitcase stuffed with bubble wrap.
50%UnverifiedWi-Sci uses a proprietary protocol to compress the bandwidth requirement for wireless GPU communication from 8 gigabits per second to about 400 megabits, enabling the laptop to run Cyberpunk 2077 at 92 frames per second with Wi-Fi usage steady at around 400 megabits.
50%UnverifiedWi-Sci installs a "fake driver" that makes Windows and game engines recognize the remote GPU as a local PCIe device, so it shows as an RTX 5090 in Task Manager and games run without issue.
50%UnverifiedThrough four optimizations—expert-aligned chunking, hot expert residency, compute splitting and scheduler preloading—Wi-Sci raises effective SSD bandwidth from 1 GB per second to about 11.5 GB per second, enabling DeepSeek 4.1 Flash to reach a generation speed of about 8-15 tokens/second on the Wi-Sci box.
50%UnverifiedThe Wi-Sci scheduler can cut the switch time from an image model back to a chat model from about 17-18 seconds without the scheduler to about 4 seconds, and the reverse switch from about 18 seconds to about 8.6 seconds.
50%UnverifiedWi-Sci's proprietary protocol includes four mechanisms: instruction batching compression, incremental data transmission, predictive acknowledgment and compressed video-stream return.
50%UnverifiedAt this stage the wireless GPU scheme is more a feasibility validation than a deliverable product.
50%UnverifiedThe demo involves only a single client; resource isolation and QoS when multiple clients share the same remote GPU remain unsolved.
50%UnverifiedWi-Sci's approach is to make the remote GPU masquerade as local hardware, with applications still running on the local CPU and only GPU compute offloaded to the remote end, in contrast to traditional thin clients (such as VDI) that run the entire desktop session remotely.
50%UnverifiedThe wireless GPU scheme is named Wi-Sci (Wireless Computing Infrastructure).
50%UnverifiedThe Wi-Sci scheme claims to offer a 410TB memory pool, but the author stressed "not in the way you think," implying it is not single-machine VRAM.
50%UnverifiedTraditional remote streaming (such as Steam Remote Play) captures frames and input at the application layer, whereas Wi-Sci attempts to virtualize PCIe devices at the system layer so that all GPU calls are transparently redirected to the remote end.
50%UnverifiedThe Wi-Sci product comes in two versions, RTX 5090 and RTX 5060 Ti, and is now open for pre-order, with shipping expected in eight weeks.
50%UnverifiedIn Chuck's demo, a Mac connected to the Wi-Sci box over Wi-Fi, used Figma to generate images, Love Art to generate video and DaVinci Resolve to render video, with the GPU scheduled across multiple tasks but not truly in parallel.
50%