Neandertal wrote:
I don't believe there is a hard line. The subatomic behavior is simply to small to be relevant.
That is why I say that they don't behave like that, and that is technically wrong, I simply mean that the effect is so small that you can simply ignore it.
Apparently there are circumstances under which macroscopically visible amounts of matter can have common quantum states.
See
Bose-Einstein Condensate
especially
superfluid Helium
Neandertal wrote:
I would still like to hear your thoughts on my original question.
Im basing this off the latest nVidia single-gpu whichs performance i know (and that is abou half a year old, most likely even older).
We have a processor with 3TFLOP, 3E12 flop.
this processor runs at about 1GHz (it surely uses higher frequencies on some internal parts but i assume they scale analogous to my known 1GHz)
.
Now we assume that we can scale this thing up to teraherz frequencies.
now we are at about 3E15 flop.
Not enough.
i now lets take UV frequencies, ~ 1E15Hz -> 3E18 flop.
Close enough for now.
The actually used material thickness of the GPU is about 100nm, or 0.1micrometres.
(while the die itself is much thicker, the electronics do not need much z-area)
This gives us about 100E6 processors in 1cm^3.
So we have 300E24 Flop in 1cm^3 for UV optronics/plasmonics that i assume have the same volume density than todays electronics.
This is only using conventional computing and technology we can currently concieve.
No quantum computing, no density (flop/(ghz*cm^3)) advances, no esoteric computation methods.
I think with LT's future setting we can assume that the H-Drive is believeable, at least from an computational needs viewpoint ^^
(This are very rough ballpark estimations, the die might be a bit larger, the electronics are most likely thinner, but i should hit the right area)
Edit:i think i can actually sqeeze another 1E3 out of this by using 0.1 nanometre graphene electronics instead of 0.1 micrometre silicon electronics, so we are at 300E27 FLOP.
Edit2: the minimum feature width archieved for graphene is around 4nm, about 1/7th of the feature width of the nvidia gpu (28nm iirc).
So we have 7^2 more areal density.
->14.7E30 flop
edit3:
the newest nVidia GPU has 5Tflop, so the overall FLOP go up by 5/3
24.5E30 FLOP
edit4:
the 5Tflop GPU has ~5.5cm² area, so all flop go down by this amount.
4.45E30flop