Does anyone think PC simulators will ever look this good ?

[Golden_Eagle_FM]

I am ready to bet that in ten years from know real time interactive raytrayced animation of the quality of what we have on those short films will be available at less than 1000 US$ (today value) for general purpose desktop machines.

This is if CPU design evolution goes as planned that is the massive multicore CPU's of the Larabelle type that Intel has already demonstrated with a raytraced version of a very popular FPS game. We are talking of CPU's with 80 cores and plus.
We are now at the stage were we have quadcore CPU's with a max frequency around 3GHz available on the market and in general use. I do not believe the frequency will increase much, but a breaktghrough in technology may allow some major steps. Shrinking the size of CPU components in combination of high voltages for high frequencies, and high heat dissipation problems is already starting to hit some fundamental physical barriers that generate many unwanted effects as for ex. electron migration. We see that it is a few years know that with INTEL and AMD have stalled between 2.5 and 3.2 GHz. Maybe we get to 4GHz but it is not really of much use. On the other hand the fundamental limits to core multiplication in the CPU are much further away.

If we apply Moore's law in conservative manner that is a doubling every 24 months then in 2009 we have 8 cores in 2011 16 cores, in 2013 32 cores, in 2015 64 cores and in 2017 128 cores. The advantage of this approach is that you can lower the frequency, and run say at 1.5 GHZ max and you can get full use of the shrinkage of components as you do not have any more the heat issue. So all the real estate can go to low heat dissipating and simpler architectured cores and so also with even more shrinking without major problems (but for sure there is also an atomic phisical limit here).
The final result is promising. Better 80 cores at 1 GHz then 4 at 4 Ghz. Roughly speaking and to have it simple you get 80 Ghz of power instead of 16.

So it is better to go the lower frequency+low heat dissipation+simpler architecture and higher multiple-core path then the other way round.

If these multi-multicore CPU's do get on the market then there will be enough processing power to get all the software rendering algorithms executed in realtime (60FPS).

The problem is that the whole polygon based culture and tools in game design will have to be readapted to optimally generate raytraced images.
But also the whole software generation community will have to be geared to use massively parallel hardware. This last point may be the most difficult issue.

So to finish I would say that we will have in ten years the hardware capability on the table for sure. I will not bet even a dime that we will be able to use that capability. That will be the major roadblock.

Gold