So I love Intel even more

[HDL_CinC_Dragon]

Yeah, they reinvented the transistor.
Instead of using Silicon Dioxide they use "hafnium" now and have manage to squeeze 20% faster switching speeds out of them and drop 30% of the power consumption off of it. Theyve also dropped the form factor from 65nm to 45nm.
20% faster + 30% less hungry - 20nm = 1000% more sexifying

"Intel says new material boosts chip performance"
"Intel launches faster, smaller, greener Hafnium 45nm chips"


I want one... I want one good.

[Admin]

Man, serious, you've been smoking too much of something this morning
(I just finished reading your last few posts)

I read about the 45nm in IEEE Spectrum (electrical engineer society magazine) a week or so ago . . . apparently they had to stop using silicon to do this . . . appears silicon is going to be outdated technology in the processor business . . .

[Rebelgium]

I'm also interested in this area, just as in alot of other technological areas, so I keep myself up to date by reading similar magazines...
btw: Is IEEE spectrum worth reading?

ontopic:
Let's hope they are willing to pump enogh money into it so it can get into production fast.
45nm is the near future, then 30nm, ... but someday there'll be a border where it's just too small.

[Admin]

btw: Is IEEE spectrum worth reading?

It rarely has robotics related stuff . . . but it has occasionally had useful/interesting articles. I get it with my IEEE membership . . .

Let's hope they are willing to pump enogh money into it so it can get into production fast.
45nm is the near future

they already make them, and they work - not only much faster, but also much more power efficient than the last breed . . . probably not long before computers are sold with them . . .

[paulstreats]

theyve used them to build faux quad core processor. I think currently though its really just 2 dual cores in one package. It should be able to make packaging of memory devices smaller, faster and more efficient aswell.

Its about time there was a new development in processors since this area seems to have ground to a halt over the past few years. I bought my 3.2 ghz laptop 4 years ago and that seems to have been a bench mark speed for a since. The processor that has been built with the new structure size has been clocked up to 3.8 ghz with no problems which is good for a device in development, if i read right the 65 nm structure has been the same since 1984 its definately due fora change

[Tsukubadaisei]

I dont love intel at all. Actually I hate it.
I think intel is a horrible company for keeping alive a outdated and inefficient computer architecture called x86.
I want my IBM/Toshiba/Sony Cell based computer. But a PPC based computer is not bad either. My lab got 4 ps3 and a cell blade and a single cell processor performance it just owns the quad core in my laptop (using the same program but with different builds). I think all that new semiconductor technology developed by intel is awesome but alone it is far from being revolutonary. The performace boost is not that much due to the outdated architecture. But the numbers can easily seduce the ignorant masses. (I understand the to change the standard from x86 to cell requires a lot of money, specially from the software developers, but for progress some sacrifices must be done).

[HDL_CinC_Dragon]

This was the first big change to the transistor in 60 years

[fr4ncium]

I got to do research at a Japanese lab over the summer into ballistic rectifiers and got to use some nanotech semiconductor fabrication equipment.  We used a Indium Gallium Arsenide heterostructure for our devices.  It would be ridiculously awesome to use the stuff in today's semiconductors, but it would be impossible (well, now at least) to automate the creation of the substrate, let alone etching the device.  Once the fabrication machinery improves to where it can etch really small, we may see a shift back to silicon.

Semiconductor physics is very interesting, but robotics is even more so

I dont love intel at all. Actually I hate it.
I think intel is a horrible company for keeping alive a outdated and inefficient computer architecture called x86.
I want my IBM/Toshiba/Sony Cell based computer. But a PPC based computer is not bad either. My lab got 4 ps3 and a cell blade and a single cell processor performance it just owns the quad core in my laptop (using the same program but with different builds). I think all that new semiconductor technology developed by intel is awesome but alone it is far from being revolutonary. The performace boost is not that much due to the outdated architecture. But the numbers can easily seduce the ignorant masses. (I understand the to change the standard from x86 to cell requires a lot of money, specially from the software developers, but for progress some sacrifices must be done).

The big money consideration is not the software developers, but the actual hardware.  This is prevalent in any category of electronics, not just CPUs.  Companies like Intel decide which technology is currently the most viable, spend billions on fabrication facilities, and then improve that technology for several years before even considering moving to a new one, because that means making entirely new facilities.  That's the price you pay for cheap computer parts - they've refined the technology so well and invested so much money in their facilities ahead of time.  You may have heard about the flash technology that is better than today's in every way (it's magnetic based so it doesn't have "cycles" and retains its memory without power, meaning we would have instant boot times).  Unfortunately, the technology rolled around after all the major flash memory companies had just finished investing a ton of money in fabrication plants for the current type of flash memory, so it will be another 4 or 5 years before they move onto that kind of technology.

[Admin]

Once the fabrication machinery improves to where it can etch really small, we may see a shift back to silicon.

Well the reason they had to stop using silicon is because the transistor wall layer had already been shrunk down to just 6 silicon atoms! Considering the wavelength of an electron is already greater than the distance of those atoms, there were serious electron leakages in the 65nm (hence power inefficiencies). So they swapped over to a new better insulating material for the 45nm, which they didn't entirely reveal. I think they said it was gallium based but I'm too lazy to check . . .

[Tsukubadaisei]

The big money consideration is not the software developers, but the actual hardware.  This is prevalent in any category of electronics, not just CPUs.  Companies like Intel decide which technology is currently the most viable, spend billions on fabrication facilities, and then improve that technology for several years before even considering moving to a new one, because that means making entirely new facilities.  That's the price you pay for cheap computer parts - they've refined the technology so well and invested so much money in their facilities ahead of time.  You may have heard about the flash technology that is better than today's in every way (it's magnetic based so it doesn't have "cycles" and retains its memory without power, meaning we would have instant boot times).  Unfortunately, the technology rolled around after all the major flash memory companies had just finished investing a ton of money in fabrication plants for the current type of flash memory, so it will be another 4 or 5 years before they move onto that kind of technology.

I dont agree with you (but I understand and actually experience what you said). Changing from 65nm to 45nm requires a lot of change in the manufacturing hardware. Specially now, because they are not only going to decrease the size of the circutry by 31% but also change the material they are working with. Such a change will require replacement of pretty much all hardware plus extra training for the staff. While changing the architecture of a chip is mainly just changing its internal circuit schematic for something different. The chip architecture problem is a software problem, no doubt about that. If right now the standard become, for example, powerpc most of OS will have to be recompiled if not rewritten, lots of programs will not work properly, not to mention that some will stop working entirely. Making all those changes and buying the new recompiled software wont be cheap at all. Keep the old standard is financially convinient.9
Notice that I am not against the new 45nm semiconductors. I think it will take at least 2 years until they(intel start mass producting and selling this stuff for cheap. I think they should take this oportunity to also change the standard computer architecture. If they do that, I will definatelly change my point of view about them. Until then IBM will still be the best.