3D chip to allow for 50 GHZ CPUs?

[MUYA]

http://forums.vr-zone.com/showthread.php?t=212665

http://www.theregister.co.uk/2007/12/08/unisantis_and_ime_3d_transistor/


50GHZ FTW!!!! :D

But seriously, wonder how long before they can get this to production level?

[CaptNKILL]

I wonder how the Crysis physics demo would run on a 50Ghz CPU. :p

I found this pretty interesting though:
The design work is headed by CTO of Unisantis Fujio Masuoka - a man credited with the invention of flash memory.

That tells me that this might actually work. :)

[rhink]

Other articles I read indicated 50 ghz transistor speed, not cpu clock speed. If that's accurate, we're looking at a healthy clock speed increase, but not to 50 ghz.

Also, how much does it decrease power dissipation? Can these things operate at high clock speed with low leakage current? Found vague references to "produce less heat", but no mention of how much. That's really the limiting factor in clock speeds atm- building chips that don't vaporize at high clock speeds/power consumption.

[sharvin]

mmmmm physalicious.

[DRen72]

I think we could all benefit more from a 50GHz GPU instead.

[ephmrl]

I think we could all benefit more from a 50GHz GPU instead.

At the moment, the transistor they're working on will get you a 20+ Ghz solid state drive rather than a cpu/gpu.

http://arstechnica.com/news.media/figure2.gif

This pic implies a much more complex process than what's used today (particularly the vertical triple layer insulator), and doesn't really hint how you would apply the design to cmos technologies (or how it would eliminate the need for cmos). So, maybe 5 to 10 years till they have something figured out and working for alu's that is manufacturable?

[Q]

I think this is a Pentium 4 derivative.

[SH64]

But what about teh c3ll ?!!

[Monolyth]

But what about teh c3ll ?!!

it just got pwned.

[DRen72]

At the moment, the transistor they're working on will get you a 20+ Ghz solid state drive rather than a cpu/gpu.

http://arstechnica.com/news.media/figure2.gif

This pic implies a much more complex process than what's used today (particularly the vertical triple layer insulator), and doesn't really hint how you would apply the design to cmos technologies (or how it would eliminate the need for cmos). So, maybe 5 to 10 years till they have something figured out and working for alu's that is manufacturable? Thats pretty cool. William Shockley would be proud.

[thor1182]

seems like it would get used the same way as current transistors, just do the same job quicker? So all you would do is change out the transistor type in your circuits, and you are good to go... except the part where it looks like that would be a lot harder to make at a very small level.

[Koko56]

it just got pwned.
:lol2:

[GitDat]

So now we can expect raytracing in real-time? :afro:

[LORD-eX-Bu]

yes in about 50 years :lol: