Wink Pings

Can a Surface with 128GB of unified memory make it through Tuesday at GitHub Universe?

Microsoft sent Chubby♨️ a Surface Laptop Ultra with 128GB of unified memory. He had just gotten home from NVIDIA GTC and planned to take it to GitHub Universe on Monday as his main work laptop, and to test local AI with Hermes and qwen 3.8 27b. The comments section immediately started arguing over whether $3,000, Windows, and local AI are worth it.

Chubby♨️ received a Surface Laptop Ultra from Microsoft, with 128GB of unified memory. He had just gotten home from NVIDIA GTC, and the first test plan he gave after unboxing was very practical: take it to GitHub Universe on Monday and use it as his main work laptop.

The point of this machine isn't the unboxing. 128GB of unified memory means the CPU and GPU share the same memory pool, making capacity more flexible for running large models locally. For AI developers, it offers the possibility of running an IDE, a browser, and local models at the same time on a Windows laptop. For ordinary users, 128GB is more like an expensive label; what really affects the experience is battery life, cooling, compatibility, and price.

Chubby♨️'s test checklist is not light: local AI, real work, and an irresponsible number of browser tabs. He asks directly whether it can replace a MacBook Pro, or whether he'll have to borrow someone else's Mac on Tuesday. Someone in the comments picked up the joke, saying the MacBook has already packed its bags.

As for models, Chubby♨️ replied that he has already downloaded Hermes and qwen 3.8 27b, and will test larger models later. This detail is more interesting than benchmarks. It's not surprising that a 27B model can run locally; the hard part is running it continuously in a main workflow without turning the machine into a hand warmer. Larger models will further stress memory capacity, bandwidth, and cooling.

The comments section quickly split into camps. Some calculated the price, saying this is a $3,000 PC; some asked how much Microsoft paid; some simply said Windows put them off; others shared their Surface experience, saying their Surface Windows 2 had fewer problems than two newer Macs. Another commenter mentioned that Matthew Berman recently made a video arguing that local AI is currently a bad deal, so they were more interested in seeing whether this machine can run hybrid AI or fine-tuning.

These reactions say more about the issue than the unboxing video. Microsoft is pushing local AI PCs, while Apple's unified memory has already claimed a place in developers' minds. The Surface Laptop Ultra needs to prove it can be a main work laptop, not just run models. Monday's GitHub Universe is a suitable exam room: meetings, code, browsers, and local inference all at once—the fans and compatibility will speak for themselves.

Chubby♨️ promised an honest review. Someone asked how much Microsoft paid, and that doubt won't disappear because of a promise. Fortunately, the test plan is specific enough, and the Tuesday borrow-a-Mac flag has been planted. Results are more useful than positions.

Source: Chubby♨️'s X post

发布时间: 2026-10-25 12:31