New Infineon tech: what do you make of this?

[rubank]

http://www.infineon.com/news/press/208_133e.htm

This is is not eDRAM, but it´s not far off either as I understand it.
Whith this tech at least memory bus limitations doesn´t seem a problem.

So soon we will have true single-chip solutions

 

[Nappe1]

Interesting...

I need to look for this a bit deeper.

 

[rubank]

N1,
I thought you would...

 

[ChrisK]

Could...

An interesting technology, but could heat dissipation be a hindrance for the use of this technology in high performance products?

 

[Nappe1]

Re: Could...

An interesting technology, but could heat dissipation be a hindrance for the use of this technology in high performance products?

that is my concern too. But how on earth they claim it able to reach 200 Ghz? heat dissipation should be a big problem a way before that.

Rubank, as knowledge BB doesn't have much to do with Inifineon anymore (mainly just waiting new revisions of Axe getting out.) and I suspect that development of this was one reason for shutdown of eDRAM line up. BB has now a new fab partner, but it's up to partner make the official announcement. (announcing co-operation with Bitboys isn't really good idea right now.)

 

[MfA]

The 200 GHz thing is probably about the possible bandwith of the chip to chip connection when compared to a pin on a standard package, Im assuming the PR fucked it up.

Also they seem to mix up two things ... the press release is mostly about a dual chip setup where the bottom chip has slightly larger edges where standard wirebond or flipchip pads can be put (assuming the top chip is thin enough to still allow flipchip assembly). Some of their promises seem to assume the bottom chip will be used rather as a substrate for multiple top chips and maybe even a sort of PCB though.

I have always said this was a good idea ... it would be great to use area interconnects to flip chip bond a 3D core to an underlying DRAM chip. Just as fast as eDRAM but no need to get the headaches of putting it in with the logic (you gain the ability to use seperate processes, seperate testing etc). Unfortunately you would probably need new tools to be able to effectively use area interconnects

 

[RussSchultz]

I'm not so sure the 200Ghz is a typo. Inline impedance of the lead frame cuts the maximum frequency, whereas direct wires wouldn't.

Of course, that's a highly optimistic number, (sort of like when fabs announce they've made a transistor that can run at 200GHz. Sure, one can, but when you stack them together...)

As for the technology. Sounds cool, but price will ultimately dictate whether it succeeds or not.

 

[MfA]

Im assuming the comparison to the internal clock of a desktop processor was a fuck up, not the number itself.

 

[PC-Engine]

Didn't AMD display a transistor that ran at 3.3 THz?