Designer puts 96 cores on single chip

“Chip designer ClearSpeed has put 96 computing cores onto a single semiconductor, as the race to secure a niche in the emerging market for co-processors heats up.

Code-named Avebury, the company’s upcoming chip contains 96 separate internal units tuned for doing particular types of math problems. Inserted into Intel- or Advanced Micro Device-based servers, Avebury will handle the repetitive, computational grunt work involved in preparing a study on how a single protein will react with thousands of others or a financial analysis that charts how slight changes in a stock portfolio could affect a person’s financial position over the course of several years. “

[Original Article]

Now, this article comments on the market for “local high-performance” computing. I think that this is a narrow view of the possibilities. Let me tell you about a concept I came up with some 20 years ago and then form your own opinions.

The time was the late 80’s and the personal computer was as hot an item as any billionaire could ask for. The state of the art in the IBM PC was the 80286 and access to the nearly 1 meg of memory it allowed. It was also the time of the birth of the hologram.

The concept of CPU design I came up with was what I called ‘Holographic processing’.

When you start to make processors faster, you run into a huge wall of diminishing returns. To make the chip faster, as opposed to merely uping the internal clock speed, you need to shorten the path length between components- see the shrinking size of conductors inside today’s CPU’s if you want confirmation of that. Now on to the fun stuff!

Imagine, if you will, a CPU built using the currently available technology that lets you design objects as three-dimensional constructs, rather than 2D designs. Everyone knows that the sphere represents the ultimate in short-path objects. That means that, in a sphere, you have the maximum number of shortest distances between any two internal points. The story above shows that stacking of processors is possible, so just take the whole thing one more step by using a holographic laser to fuse pathways in a spherical matrix.

Then kick it up another notch by imagining that instead of stacking 96 identical processors, you stack… well… maybe 8 or 10. Then you start stacking on the GPU (graphical processing unit). Then Wi-Fi. Then… RAM? Think about that for a while. If you’re a true geek, the mere idea of all the major compnents of a PC in one ‘chip’. I think I’d rather call it The MonoBlock. Anyway, system speeds become faster by an order of ten (i.e. 30 GHz instead of 3 GHz). Intercomponent speeds become HUGE with very little latency, because there’s next to no distance between the CPU’s and RAM.

Well, there you have my dreams of a better computing world. I’ve been waiting a long time for them to come true. Maybe I’ll live to see it happen…