Let’s turn to the problem of constant growth of processor power. As you see the very first problems connected with growth of processor power are the questions of system memory and cache performance. Intel solves the problem with cache very simple – corporation has operating 20 mb chip SRAM which is piled with processor on one crystal and because of this it has huge bandwidth – near terabyte per second and even more. As for the system bus, it is much more interesting.
It’s not a secret that the possibilities of solid-state circuits and conductors exhaust their physical potentialities. That’s why if processor power can be upgraded with the help of more complicated set of regulations, multinuclear technology and other extensive-intensive methods without changing radically the production technology, this won’t do with conductors. You can resort to various tricks, but is not possible to increase the ability of the material to transmit electric signal from A to B with the help of regulations set.
However we should give due to Intel Corporation – in their last version of processors and chipsets the work with memory is optimized extremely religiously. There is no sense to enumerate all used technologies and tricks of engineers and programmers, which let to use the bus efficiently – an end-user doesn’t need this information. But the consequence of this know-how became an attempt of system bus over-clocking, which shows that the interaction between processor and memory increases slightly. At least, the performance acceleration falls short of expenses on over-clocking. Scientist and engineers began to develop new methods of data communications – from experiments with superconductors in liquid nitrogen to absolutely futuristic attempts to create biocomputer. But Intel seems to be the first company developed the technology ready for use right now – the question is only in implementation and tweaking. At the heart of the technology there is denial from electrical conductors in favour of optical, and even much simpler – laser. The advantages of optical data communication are known for long time. Good example of this is fiber-optic network interfaces which are faster than any Ethernet. But up to now the main problem of implementation of optic at the heart of the computer with miniaturization of transmitter-receiver devices.
During IDF Intel presented proudly their new piping hot creation – chip of hybrid silicon laser, comparable in the matter of size with modern chips. Of course, laser power on this chip is not great, but we don’t need to transmit information on more than several inches on the motherboard. By the way such downsizing gives possibilities to accelerate optical data communication in comparison with modern fiber-optic networks. On the other hand up to now the main problem of implementation of the fiber optic technologies was their transcendental price resulted from common complexity of production process. But invention of laser head in the format of chip and mass production let the price of one sample reduce to acceptable level, but Intel hasn’t announced concrete prices yet.
Intel seems the primary direction of developing their processors in multinuclear technology. This is provided not only by the announcement of company representatives, not only by the fact of presentation of tetra nuclear processor Core 2 Extreme – any news release presentation and schemes enclose neatly drawing processor with 16 and more cores. And here the problem of data communication between cores become the stumbling block which can be “removed†with the help of new technology of laser. As you see, in the word of IT the revolution is about to happen in many directions. DirectX10 will absolutely change the architecture of video adapter as well as the “relations†between video and processor; usual hard disks live their last days – Samsung sells subnotebooks with hard disks on flash memory (the price is up to $1000 per 30 Gb). On a waiting list – system and other channels of communication.