7 replies to this topic

[maestro949]

Saw this at digg. IBM and Georgia Tech get a silcon based chip to run at 500GHz. Time to start working on those whole cell and full organism simulations.

Chip breaks speed record in deep freeze

Digg

[rjws]

Thats cool read alot of other articles near it seems like chips will keep getting faster until atleast 2025 where the 1.5 nanometer wall is at. But then again they have ideas for that too.

[maestro949]

I'm sure by then we'll have thought of a few more too. I don't think CPU cycles will be the bottleneck 20 years from now. It'll be the complexity of the algorithms and calculations needed to simultaneously simulate all of the laws of physics such that they replicate atomic particle interaction.

[spins]

More impressive for me is the fact that at room temperature the chip runs at 350GHz! I bet it needs an equally impressive power supply though!

[JohnDoe1234]

Yeah that is pretty impressive.... I wish they would have metioned a bit more about its design... eg. size, power requirements, setup... but oh well

[olaf.larsson]

I'm sure by then we'll have thought of a few more too. I don't think CPU cycles will be the bottleneck 20 years from now. It'll be the complexity of the algorithms and calculations needed to simultaneously simulate all of the laws of physics such that they replicate atomic particle interaction.

Every single particle has not to be simulated. If you simulate a bouncing ball, do you then calculate possitions e.t.c. for every particle in the ball? Ofcourse not. The simulation will be could be pretty good and accurate anyway.

[caston]

You need to assume that things at a lower level always behave the same. Whatever makes up quarks and leptons could have variable interactions.

Edited by caston, 12 August 2006 - 01:32 PM.

[maestro949]

I don't think every particle has to be simulated but their effects at a higher level do need to be accounted for with a good degree of accuracy. In most biomolecular simulations accuracy is often lost as the scale of the simulation is increased. There are a lot of factors for this but mostly because the cost of calculating some of the molecular and subatomic interactions (e.g. electrostatics) are just too computationally expensive. Shifting from Newtonian physics to quantum mechanics, better algorithms and calculations combined with more computing horsepower will be key to better large scale simulations.

In vivo protein folding within the ribosome seems like it will be one of the more challenging yet rewarding simulations. Ribosome Simulation

Edited by maestro949, 12 August 2006 - 01:12 PM.