I believe that the current hardware technology has yet to be fully exploited by lower level coding. As covered here, memory utilization has yet to be connected most effectively with processing.

Some questions for areas of progress:

1) Binary - Where can improvement be made to help the average game developer?

2) Virtual transistors - Why can't coding emulate the digital transistor better than it does?

3) 3D light processors (3 dimensions instead of the 2 dimensional chip and using light in the processing in conjunction with electricity, such as micro-LEDs)

4) What is the probability that several immerging different technologies compete for many years and each will be promoted by competing companies?

5) Related to the previous question: Are we going to see more niche deployment for small and middle size developers and cross-platform left mostly to the biggest game developers?

6) Will performance leap enable more inefficient coding to be competitive?

7) Could performance leap stay beyond the cost range of most consumers and developers?

2) Virtual transistors - Why can't coding emulate the digital transistor better than it does?

What does this mean? Is this "transistor" you speak of in a digital configuration, or are you talking about analog technology? In any case, you're far better off emulating it in software, due to spacial constraints and acquisition complexity in hardware, not to mention drivers supporting it.

3) 3D light processors (3 dimensions instead of the 2 dimensional chip and using light in the processing in conjunction with electricity, such as micro-LEDs)

Chips are already 3-dimensional, so I also don't understand what you mean here?

Using LEDs to process light is probably the worst possible idea, because 1) it's not consistent over time due to degradation, 2) it's analog technology, which introduces all sorts of problems, such as data acquisition and control 3) it'll take up far more space than a digital processing unit, 4) it won't be flexible, and 5) it'll be slow because LEDs require multiple microseconds before they "stabilise" their emission of light.

4) What is the probability that several immerging different technologies compete for many years and each will be promoted by competing companies?

Ever heard of AMD, nVidia, Intel, etc....?

6) Will performance leap enable more inefficient coding to be competitive?

That's an interesting question. Comparing code on embedded systems (where resources are scarce) to code for front-end applications yields some key differences. Embedded code is far more compact, filled with all sorts of hacks and weird optimisations, because the code only targets a single instruction set so you can get away with it. Front-end code is always more structured and favours clean and comprehensible code over speed.

So with that in mind, increased performance certainly relaxes constraints, but I don't think it directly means that programmers become sloppy.

6) Will performance leap enable more inefficient coding to be competitive?

It's going to take new ideas.

As far as software goes, I have seen a lot of great innovations bought out by larger companies. Then the tech gets sat on.

As far as hardware goes, this gamedev stuff has made me more interested in robotics programming, which would be the best merger of software and hardware.

But in all honesty, I know where the tech industry is heading, and I don't like it one bit.

I'd be better off with my own farm, a windmill, and a well to get my own water from.

p.s. I wouldn't doubt it if there are people who are paid to find new ideas and then give the creators an offer they can't refuse. $$$

I'd personally rather see a unified memory format before another leap in processing power. It's already tough enough to keep current processors fully loaded with current memory hierarchies. A non-volatile (or selectively volatile) memory medium that had the speed of on-die cache and the cost of metallic disk storage would all by itself represent a huge leap forward in overall system performance as well as tremendously simplifying the entire system architecture.

Ok, let's try thinking about it from a more direct viewpoint.

Think of your current main PC. What would make it better for you now?

The first thing I thought of was the hard disk. Most applications are using the hard disk all the time and seek / read times are very slow compared to cpu clock rates.

So if hard disks were replaced with hi-tech solid state devices, that would make a big improvement to me.

Would a huge increase in processing power be nice? Hell yes, but is it essential?

No. You see programmers are good at making the best of what they have, they can get a pint out of a half pint glass. Most of us don't run out of CPU power when we do the normal things we do. We don't run out of ram. We might run out of GPU power if we really push it, but then you just buy a better graphics card.

GPU's will get better and better. The graphics in our games will get better and better. We'll probably get new spins in our daily life, voice commands that actually work, text to speech that sounds like a human, things like that, but we won't get a massive change in computing until a radical new technology arrives.

Call of Duty 27 will be visibly accurate to the molecular level, but will have the same crap network code as it has now. Skyrim 17 will look good enough to make you orgasm, but will still be full of glitches that mean you have to save ever three minutes. FIFA 2056 will look so realistic you can almost smell the spittle dropping from the players mouths, but you'll still spend most of your life sat on the bench in career mode.

Direct brain injection like in the matrix is one that would make a massive change.

What other technologies do you think would be a major seed change?

What are the limitation on CPU cache, it's a form of memory right?

So why can't we bump up the available cache to replace system memory if it's system memory bandwidth that's a bottleneck, I assume there are optimal sizes for cache before they too become bottlenecks, or maybe it's down to economics..?

So why can't we bump up the available cache to replace system memory if it's system memory bandwidth that's a bottleneck, I assume there are optimal sizes for cache before they too become bottlenecks, or maybe it's down to economics..?

Because cache is stupidly expensive in dollars per gigabyte (a sizeable fraction of what you pay for your CPU goes towards those few megabytes of cache), and furthermore SRAM - the technology used for caches - does not scale well with cache size, meaning the larger it is, the slower it gets, so CPU's have a bunch of tricks to mitigate that, including prefetching, using multiple cache levels, and so on.

I still dream on a fastest main memory (like at least 10e3 times faster at least).

But yea, something like hUMA would be fine (but please nothing related to OpenCL stuffs... yes, actually HSA is still a joke and a pain in the axx)

A friend of mine is working on a $300 supercomputer

He's basically working on a reworking of the old transputer, but with each node having 2 meg of cache and a super fast communication network.

He wants the basic version to have 16 nodes running at 2 gig hertz producing 3 teraflops

Well the raspberry pi now exists, so he has a chance