TSMC confirms 40nm yield issues resolved

[Jos]

Following AMD's recent comments on the improved availability of Radeon HD 5000 series graphics cards, TSMC's Senior VP of Operations, Mark Liu, today confirmed that yield rates on their 40nm manufacturing process are finally about the same level as the mature 65nm process. The executive didn't go too deep into details but said the chamber matching problems that had hampered supply for the last quarter have now been fixed.

Indeed the yield rate increase was reflected by the sudden widespread availability of AMD's 40nm cards around the turn of the year. The timing also works quite well for Nvidia as it prepares the release of its first billion transistor 40nm GPU towards the end of this quarter, the Fermi-based GF100.

In addition to the good news above, TSMC announced it has opened a new factory building named Phase 5, which is part of the company's Fab 12 located at the Hsinchu Science Park in Taiwan. Once Phase 5 is operational it will be used for volume production of 28nm products in the third quarter of 2010.

Permalink to story.

https://www.techspot.com/news/37645-tsmc-confirms-40nm-yield-issues-resolved.html

 

[KG363]

28nm? Why do they keep taking out beloved nanometers. I want more Voltage and more nanometers!

 

[Technochicken]

^ Wait, what?

I believe 28 nanometers is getting uncomfortably close to the limit of transistor size. At that rate transistors will only have 5-10 years of shrinking before the limit is hit an a new technology must be used to be able to efficiently increase processing speeds.

 

[red1776]

28nm? Why do they keep taking out beloved nanometers. I want more Voltage and more nanometers!

:haha:

 

[eafshar]

red1776 said:

28nm? Why do they keep taking out beloved nanometers. I want more Voltage and more nanometers!

:haha:

we will see carbon nanotubes red1776 when that happens.

 

[eafshar]

haha i meant to respond to technochicken. ooobs

 

[9Nails]

What spins my head is that the machines which make those chips have to be isolated from Planet Earth because "ambient ground motion" is such a problem that it will obliterate the fine pattern that they're trying to etch in scales that small.

 

[14 of a hotdog]

technochicken said:
^ Wait, what?

I believe 28 nanometers is getting uncomfortably close to the limit of transistor size. At that rate transistors will only have 5-10 years of shrinking before the limit is hit an a new technology must be used to be able to efficiently increase processing speeds.

I think molecular computing would work. I think molecular processors still behave like traditional processors (unlike quantum ones) but they can use a single molecule in place of each transistor. Think of how many more molecules you could fit in that space than you could transistors. A while back I also looked into diamond processors and it seemed like one article said they could dope the synthetic diamond to behave like the types of silicon in microprocessors but if you keep oxygen away from diamond it wont melt until it hits 3500f or something like that and it also conducts heat several times better then copper! Imagine a remake of a current processor that could run at 100ghz I had even read that a company claimed that it could reach speeds above 250ghz with pure synthetic diamond. I know about the megahertz myth but I think having the clock 50 to 100 times higher must count for something.

 

[Guest]

Would it be naive to think that once it gets below the 10nm mark, quantum computing will not be too far over the horizin?

 

[HaMsTeYr]

If it makes my games run faster, hot digiddy it don't matter how small it gets... though irony comes in such a way such that the past graphics cards were all single slots and now all the "smaller" die size graphics cards have massive heatsinks lol....

 

[slh28]

I think I read somewhere recently (sorry, not very specific) that a transistor was made from a single atom. Did anybody else remember this?

 

[compdata]

slh28 said:
I think I read somewhere recently (sorry, not very specific) that a transistor was made from a single atom. Did anybody else remember this?

Not yet, but it is close. Not familiar with the exact structure of the silicon/metal chanells on the chip but for reference:

1 silicon atom has a Van der Waal radius of approximately 200 pm, which gives a diameter of ~400pm or .4nm. So ~70 silicon atoms could fit in 28nm.

 

[compdata]

9Nails said:
What spins my head is that the machines which make those chips have to be isolated from Planet Earth because "ambient ground motion" is such a problem that it will obliterate the fine pattern that they're trying to etch in scales that small.

I actually hadn't thought about that for chip lithography, but it makes sense. This is similar to some of the issues being worked on with the James Webb Space Telescope in that they had to deal with fluctuation in earths gravity field distorting their thin mirror segments (one of the reasons they want to park it at the L2 Lagrange point).