You can never have too many cores. Each AI can run in it's own core. Each pixel can be a part of a giant ray tracer, where you dedicate one core per pixel, and it has the refresh time of the video display to generate it's next dot. Until you get to 4 trillion cores, an experienced developer can find a way to use them all... You have one core managing a group of cores, and each of those managers managed by another, etc... Being able to have thousands of cores all doing a little bit is the only way for exponential improvements in what can be done.
Programmers just have to think differently how the implement things.
@crashman: The number of transistors has increased exponentially since 1965, doubling every two years on average, and is expected to continue for at least another 10-20 years. In fact, as programmer learn to take advantage of massive numbers of cores, some of the extra circuitry for optimizing a single thread will be removed to allow for more cores. For example, a single thread may run 20% slower, but can get 64 cores instead of 16...
Haven't you heard of Moore's law? It is far more difficult to make a single thread go faster now, so that is why the numbers of cores is increasing recently.
One thing I think we will see more of, is more memory will be built into the chips. Drop the "cache" on the chip and actually run everything inside the cpu. If additional external memory is used, let it be swapped into the chip in a more intelligent manner instead of cached which is just a kludge anyways.
Existing User Log In
New User Registration
Register for a free account to gain full access to the VGChartz Network and join our thriving community.
