header image
[ # ] Dream Machine (take 2)
January 18th, 2011 under Computers, Gadgtes, Hardware, rengolin, Technology, Thoughts

More than three years ago I wrote about the desktop I really wanted… Now it’s time to review that and make some new speculations…

Back Then

The key issues I raised back then were wireless technology, box size, noise, temperature and the interface.

Wireless power hasn’t progressed as much as I’d like, but all the rest (including wireless graphic cards) are already at full steam. So, apart from power, you don’t need any cables. Also, batteries are getting a bit better (not as fast as I’d like, too), so there is another stop-gap for wireless power.

Box size has reduced dramatically since 2007. All the tablets are almost full computers and with Intel and ARM battling for the mid-size form-factor, we’ll see radical improvements with lower power consumption, smaller sizes, much cooler CPUs and consequently, no noisy fans. Another thing that is bound to reduce temperature and noise is the speed in which solid-state drives are catching up with magnetic ones.

But with regard to the interface, I have to admit I was a bit too retro. Who needs 3D glasses, or pointer hats to drive the cursor on the screen? Why does anyone needs a cursor in the first place? Well, that comes to my second dream machine.

Form Factor

I love keyboards. Writing for (int i=0; i<10; i++) { a[i] = i*M_PI; } is way easier than try to dictate that and hope it gets the brackets, increments and semi-colons correctly. Even if the dictation software was super-smart, I still would feel silly dictating that. Unless I can think and the computer creates the code for me the way I want, there no better interface than the keyboard.

Having a full-size keyboard also allows you to spare some space for the rest of the machine. Transparent CPUs, GPUs and storage are still not available (nor I think will be in the next three years), so putting it into the monitor is a no-go. Flat keyboards (like the Mac ones) are a bit odd and bad for ergonomics, so a simple ergonomic keyboard with the basic hardware inside would do. No mouse, of course, nor any other device except the keyboard.

A flat transparent screen, of some organic LED or electronic paper, with the camera built-in in the centre of the screen, just behind it. So, on VoIP conversations, you look straight into the eyes of the interlocutor. Also, transparent speakers are part of the screen, half-right and half-left are screen + speakers, with transparent wiring as well. All of that, wireless of course. It should be extra-light, so just a single arm to hold the monitor, not attached to the keyboard. You should be able to control the transparency of the screen, to change between VoIP and video modes.


CPUs and GPUs are so 10's. The best way to go forward is to have multi-purpose chips, that can turn themselves (or their parts) on and off at will, that can execute serial or vector code (or both) when required. So, a 16/32 core machine, with heavily pipelined CPU/GPUs, on multiple buses (not necessarily all active at the same time, or for the same communication purpose), could deal with on-demand gaming, video streaming, real-time ray-tracing and multi-threaded compilation without wasting too much power.

On a direct comparison, any of those CPU/GPU dies would have a fraction of the performance of a traditional mono-block chip, but their inherent parallelism and if the OS/drivers are written based on that assumption, a lot of power can be extracted from them. Also, with so many chips, you can selectively use only as much as you need for each task for specific applications. So, a game would use more GPUs than CPUs, probably with one or two CPUs to handle interface and sound. When programming, one or two CPUs can handle the IDE, while the other can compile your code in background. As all of this is on-demand, even during the game you could have a variable number of chips working as GPUs, depending on the depth of the world it's rendering.

Memory and disk are getting cheaper by the second. I wouldn't be surprised if in three years 128GB of memory and 10TB of solid-state disk are the new minima. All that, fitting nicely alongside the CPU/GPU bus, avoiding too many hops (NB+PCI+SATA+etc) to get the data in and out would also speed up the storage/retrieval of information. You can probably do a 1s boot up from scratch without the necessity of sleeping any more, just pure hibernate.

Network, again, wireless of course. It's already a reality for a while, but I don't expect it to increase considerably in the next 3 years. I assume broadband would increase a few percent, 4G will fail to deliver what it promises when the number of active clients reach a few hundred and the TV spectrum requires more bureaucracy than the world can handle. The cloud will have to wait a bit more to get where hard drives are today.


A few designs have revolutionized interfaces in the last three years. I consider the pointer-less interface (decent touch screen, camera-ware) and the brain interface as the two most important ones. Touch-screens are interesting, but they are cumbersome as your limbs get in the way of the screen you're trying to interact with. The Wii-mote was a pioneer, but the MS Kinect broke the barrier of usability. It's still in its early stages, but as such, it's a great revolution and because of the unnatural openness of Microsoft about it, I expect it to baffle even the most open minded ones.

On the other hand, brain interfaces only began this year to be usable (and not that much so), the combination of a Kinect, with a camera that reads your eyes and the brain interface to control interactions with the items on the screen should be enough to work efficiently and effectively.

People already follow the mouse with their eyes, it's easy to teach people to make the pointer follow their eyes. But to remove uncertainties and get rid once and for all of the annoying cursor, you need a 3D camera to take into account your position relative to the screen, the position of other people (that could also interact with the screen on a multi-look interface) and think together to achieve goals. That has applications from games to XP programming.

Voice control could also be used for more natural commands such as "shut-up" or "play some jazz, will ya?". Nothing too complex, as that's another field that is crawling for decades and hasn't have a decent sprint since it started...


The cost of such a machine wouldn't be too high, as the components are cheaper than today's complex motherboard designs, with multiple interconnection standards, different manufacturing processes and tests (very expensive!). The parts themselves would maybe be a bit expensive, but in such volumes (and standardised production) the cost would be greatly reduced.

To the environment, not so much. If mankind continues with the ridiculous necessity of changing their computers every year, a computer like that would fill up the landfills. The integration of the parts is so dense (eg monitor+cameras+speakers in one package) that would be impossible to recycle that cheaper than sending it to the sun to burn (a not so bad alternative).

But in life, we have to choose what's really important. A nice computer that puts you in a chair for the majority of your life is more important that some pandas and bumble bees, right?

Creative Commons License
We Support



National Autistic Society

Royal Society for the Prevention of Cruelty to Animals


End Software Patents

See Also

The information in this weblog is provided “AS IS” with no warranties, and confers no rights.

This weblog does not represent the thoughts, intentions, plans or strategies of our employers. It is solely our opinion.

Feel free to challenge and disagree, and do not take any of it personally. It is not intended to harm or offend.

We will easily back down on our strong opinions by presentation of facts and proofs, not beliefs or myths. Be sensible.

Recent Posts