Wednesday, November 26, 2008

A theory about "old" Part I : Hardware

I know a lot of people that are always running behind the latest piece of hardware and trying to play the latest game that just got out. Sincerely, I don't share those positions.

Let's start with Hardware. The latest hardware is always overpriced (it has to do with it being new and powerful, but also because prices tend to fall when something gets mass produced). When they buy it, they obtain the maximum possible performance, but they don't get the best power/price relation. The price of the latest video cards can cover the purchase of 3 great video cards, not so powerful as this one, but with enough power for most things.

Other computer pieces don't have a 3 to 1 relation, but a 2 to 1. That means that the price I pay for the greatest computer I can get right now would buy me two computers that each are capable of outrunning 95% of the other computers out there, or even 4 or 5 capable computers. I think that nobody here misses the point that 4 or 5 capable computers combined have more computing power than one extremely powerful one (of course, that's not counting the new Nvidia Tesla card).

There's of course a hidden cost in buying the latest technology: support and standards. Let's take for example a piece of hardware that got out a pair of years ago, the AGEIA PhysX, a pci board designed to crunch numbers related to physic models. This was planned as a sort of coprocessor to relieve the CPU and the video card from doing a lot of physics calculations, needed for example for the fast implementation of the rag doll concept

It was a concept similar to the first 3d acceleration cards (so similar that the technology is now inside video cards): freeing the processor from the most heavy and common 3d tasks. It wasn't an expensive card, but it lacked game support, with only one game implementing it's API by the time of launch.

A small number of other games announced projected support, but it has to be noted that the card wasn't really needed to play any of those games, being capable of emulating most of it by software. This was because standard PCs don't had that card, and as the game developers don't want to release something for just a few, they had to program it for machines without the expansion board.

It would have been nice to get one of that AGEIAs, the nerd inside me agrees, but the lack of support and the lack of standards around it, made it a risky business to buy one of those at that time. The same is happening right now with the Nvidia Tesla, although the niche market of science labs and supercomputer users might provide a nice launching area.

This is happening again and again in the PC industry. Whenever you buy something new and not common, you risk being trapped with something that might loose support quickly. This is why it's a good idea to buy proven technology. It has already been incorporated into some standard, it has already been evaluated, and they are definetely cheaper.

One little final word about going for something just because it's better. BETAMAX and VHS were fighting for the video standard. BETAMAX was superior, but VHS was cheaper. VHS finally became the standard, so betting on something just because it's better isn't always safe, there are always other things to consider.

No comments: