Speed Up Your Computer!

Method identified to double computer processing speeds (msn.com)

Somebody figured out a way to actually speed up your computer or cellphone, by simply multiplexing everything better. Of course, the new MX2 motherboards will make his work trivial, using ballistic electrons that travel at almost the speed of light, but his work can be applied to even your existing cellphone. Currently, we have a wild variety of chips all slapped onto the same motherboard technology we've been using for fifty years. This work is relevant to AI researchers, providing yet another way to cut the system demands in half. Currently, most LLM AI require 56gb vram, and at least four graphics cards and, at the rate they're going, it will soon be cheap enough to make your own AI at home. At least four graphics cards and 56gb vram will cost as much as a used car, and cutting that cost in half is a huge priority for many, and my own work is related to making it even more efficient and productive.

AMD is in the best position to leverage this work, and it means we will have a new ALL-IN-WONDER chip in another five years, that requires half the damn parts! The mathematicians are making such fast progress, it could be ten years before anyone figures out how to make the most efficient design. Already, they've found ways to hardwire things like ray tracing, to cut the system demands in half, and the issue is how to combine all these shortcuts. The faster they keep finding more new shortcuts, published in the public domain, the faster their own super-chips and super computers become obsolete, and their operators as well.