Finding a market

I had a break at the recent SuperComm in Chicago, where I was taking pictures for work. I talked to one of the publishers on an embedded systems magazine. Like everyone I talked to about the BitGrid idea, he had his own idea of what was missing... he said I needed a market, or as we would say at work, a Target Audience... he's right.

I have an idea which could be used to build really cheap computing engines, or really huge supercomputers... but the flexibility makes targeting an audience a problem. I think right now I remain the only target.

Jonathan Schwarts talks about Big data, small compute and the different scales of computing. I see this as a cheap way to do a very fast bit of computing on a big dataset. Piping a 5Gigabyte database from RAM through a 64 bit interface could happen in 1 second, with a somewhat aggressive 700 Mhz clock. I'm going to have to figure out what he's really talking about in terms of compute tasks to see if it's a good fit.

The fine line between Genius and Insanity

I've figured out why nobody has come up with a true bitgrid yet...

The idea is insane

Everyone worries themselves sick trying to reduce either transistor count, power consumption, or signal delays. It's just plain nuts to put a signal through 100,000 extra gates, if you really don't need to. So, nobody would ever come up with the bitgrid.

I, on the other hand, saw the benefit of having a homogeneous grid with no restrictions on routing, which lead to the bitgrid concept. I was totally ignorant of the importance of keeping the transistor counts low, reducing the power consuption, and maximizing the efficiency of layout of a chip. Thus, I was too ignorant to know better

22 years have come and gone, and I'm dipping into the well to pull the concept back out. I've got a lot of life lessons learned, and have decided to get this thing built. I started doing the research, and reducing my ignorance of the subject. The realization last night of the truely awful state of the bitgrid concept was frightening. I thought I had just wasted a bit of time, and took a fair self-esteem hit.

This morning, I now see the bigger picture, and I'm comforted by it. I don't have to worry about competition, prior art, etc. I've learned quite a bit from this experience.

Now, all I have to do is figure out if it really genius to pass a signal through 100,000 transistors instead of a wire, to get the flexibility of a bitgrid. It'll take time and effort, but I'm going to make it happen.

Custom Silicon

Thanks to Wes Felter and Doc Searls for the pointers to MOSIS and ChipX.

I've learned that I've got a lot to learn. The investment to get a design cast is going to be at least $1500 for the raw manufacturing of the smallest die size available. I haven't even begun to consider the design tools, and the time to learn them. It's a formidable task, but one that I'm sure I can do, eventually.

Since that end of the process is such a big investment in terms of time, effort, and % of my life, I'm going to focus more on the big picture, and emulate the heck out of my concepts first. Software simulation is the number one priority.

Meanwhile, on the social network side, I'm going to see options are available to do a GPL type license with a patent. I'm expecting to dump at least $10,000 and 1000 hours into this project at a bare minimum, so I want everyone to get some return on my investment. (I'm not really a money motivated person)

So, now the focus moves to a software simulation that I can share with others, work out concepts, and get some virtual benchmarks against. I'm going to assume perfect hardware for the moment, because it's easiest to write. 8)

Join Me

I'm trying to change computing as we know it, in a small way. I've got an idea for a chip design, which I intend to see cast into silicon. I need all the help I can get. I want to make samples, test them, and GPL the idea (as well doing something like a GPL Patent).

It's going to be a lot of work, but it should be a very rewarding ride.