Intel reaches breakthrough in transistor design

[NFS4]

http://biz.yahoo.com/rf/011125/n25277907_1.html

 

[ST4RCUTTER]

Neato! 2005 looks to be the year that a lot of changes start hitting the semiconductor industry. EUV, better materials and more efficient transitors can only mean one thing...

"Good morning Dave..."

 

[RaynorWolfcastle]

Hmmm.... sounds promising

-Ice

 

[AGodspeed]

Whoa, very awesome stuff. However, 2005 is a long ways off. By then we might have single atom processors or something crazy like that.

I also found this pretty interesting:

One element of the new transistor structure is what Intel calls a ``depleted substrate transistor,'' a form of a technology called silicon on insulator. Using this technology, a transistor is built in an ultra-thin layer of silicon on top of an embedded layer of insulator.

Um, does this mean Intel might use SOI in the future? Btw, how can Intel develop a breakthrough technology based partly on SOI technology without dealing with IBM?

 

[Wingznut]

<< "Um, does this mean Intel might use SOI in the future? Btw, how can Intel develop a breakthrough technology based partly on SOI technology without dealing with IBM?" >>

I'm sure that Intel would use a type of SOI in the future, if they find it beneficial (as with any technology). SOI is one type of insulation built into the substrate. There are different methods (i.e. materials used) to achieve this, and IBM of course doesn't own patents to all of them ( just their method.)

 

[AGodspeed]

There are different methods (i.e. materials used) to achieve this, and IBM of course doesn't own patents to all of them ( just their method.)

Oh, I see.

Also, is this the determining factor in per-unit production costs? SOI is significantly more expensive than regular Si wafers (right?); what exactly makes this so? Is it the complexity of implementing SOI, or is it just that IBM has developed an SOI technology that is based on expensive materials?

 

[Wingznut]

Yes, from what I understand, SOI wafers are 5-9 times more expensive. It's not so much the materials used, but it is simply that it's a new technology and that the process isn't as refined as "bulk" Si.

 

[ST4RCUTTER]

You guys are hitting around the correct answer. The added cost of SOI wafers is mainly due to the difficulties in the fabrication. Namely the implantation/deposition of the oxide layer. Achieving a low-defect film over the Si wafer has been the most difficult step in the process. SIMOX (Separation by Implantation of Oxygen) is beginning to bring SOI wafer costs down to that of bulk, but it will always be more expensive. The benefits could be substantial however. Some preliminary results are garnering 20-35% faster transistor switching times. Where low power switching is the desired goal, 40-50% less power than traditional CMOS has been observed. To make matters more interesting, SOI has its own "quirky" characteristics that require circuit testing. You might remember seeing something a few months back about AMD partnering with another firm for simulation/testing/design purposes...this is why. It leads one to believe that you can't switch to SOI easily until the circuits of the CPU have been tested using an SOI substrate.

Some good articles on SOI:


Link 1

Link 2

 

[Bluga]

i'm waiting for the DNA computer

 

[Adul]

bump

 

[AGodspeed]

You guys are hitting around the correct answer. The added cost of SOI wafers is mainly due to the difficulties in the fabrication. Namely the implantation/deposition of the oxide layer. Achieving a low-defect film over the Si wafer has been the most difficult step in the process. SIMOX (Separation by Implantation of Oxygen) is beginning to bring SOI wafer costs down to that of bulk, but it will always be more expensive. The benefits could be substantial however. Some preliminary results are garnering 20-35% faster transistor switching times. Where low power switching is the desired goal, 40-50% less power than traditional CMOS has been observed. To make matters more interesting, SOI has its own "quirky" characteristics that require circuit testing. You might remember seeing something a few months back about AMD partnering with another firm for simulation/testing/design purposes...this is why. It leads one to believe that you can't switch to SOI easily until the circuits of the CPU have been tested using an SOI substrate.

Whoa, thanks for the explanation.

Thanks for the links too.

 

[pm]

Nice summary, St4rcutter.

I posted some comments on the differences between partially depleted SOI (IBM, AMD) and fully depleted SOI (what this article is about) as well as a link to a more detailed article at the EETimes in this thread in the HT forum.

 

[ST4RCUTTER]

Thanks. That's another good article from EE Times. I find the sudden change in Intel's position over SOI amusing but not surprising. The science behind the new process is sound, and it's only in its early stages. We have only just begun to see the interesting new developments in this area. Both AMD and Intel will be using SOI and high K dielectrics in the near future. Most other semi makers will do so as well if they plan on staying competitive. I'm excited.

 

[Wingznut]

What "sudden change" is that? It's not as if Intel said that they will never use SOI. They just have said that it's not worth it, at this point. And judging by the success of their .13µ process, who's to argue?

 

[AGodspeed]

What "sudden change" is that? It's not as if Intel said that they will never use SOI.

Oh come on. Are you going to tell me that Intel using a form of SOI technology in the future was foreseeable? Everyone knows that Intel has opposed any type of SOI implementation since IBM revealed it way back when.

Don't you think it's kind of weird that Intel was so vocally opposed to using current SOI technology, yet they've put in god knows how much money into research on SOI? I think a "sudden change" is not all that accurate, but what is accurate is that Intel is looking quite confused by toting non-SOI technologies now but then announcing a breakthrough based on technology they have been opposed to so long.

And yes, I know the form of SOI Intel is researching is supposed to be very different from the SOI we know today. We'll see how IBM responds to this next week, and in the next few months.

 

[ST4RCUTTER]

What "sudden change" is that? It's not as if Intel said that they will never use SOI. They just have said that it's not worth it, at this point. And judging by the success of their .13µ process, who's to argue?

Sheesh Wingz I never stated that Intel said, "they will never use SOI". I just find it amusing that in the past few months Intel claimed that, "SOI was not necessary" and the statements that are coming out now are that SOI will be an important part of their process.

What I find even more amusing is the statement that Intel will, "skip traditional SOI wafers and move to a next-generation technology called thin SOI or fully-depleted SOI", considering that Intel may plan on buying their fully-depleted SOI wafers from Soitec, the company that is supplying AMD...you see where I'm going with this? Soitec is currently the largest manufacturer of SOI wafers, and arguably the most advanced. Intel will probably buy their wafers from more than one supplier, but you can bet that Soitec will be a major supplier to both Intel and AMD.



edit: Soitec is a French company that trades exclusively on the DAX right? Hmmm...maybe I need to pick up some international securities.

 

[Wingznut]

"Forseeable"?... I'm not sure that I (nor you) could predict what technologies will be used in semiconductors in 3+ years. So, (as a PC enthusiast) would it surprise me? Not in the least.

Sure, Intel has said that SOI is not necessary for .13µ, at this time (more specifically that the benefits of SOI don't outweigh the drawbacks). Are they incorrect? As far as I know, they've never said that they will never use SOI, nor that the technology wouldn't become useful in one way or another.


"Don't you think it's kind of weird that Intel was so vocally opposed to using current SOI technology, yet they've put in god knows how much money into research on SOI?"

Not "weird' at all. If they didn't experiment with SOI, how would they know that it's not beneficial to them, at this time? Intel researches every possible technology. I'm sure there are plenty of technologies that they've invested R&D into, and then determined that they aren't beneficial at this time.

"Intel is looking quite confused by toting non-SOI technologies now but then announcing a breakthrough based on technology they have been opposed to so long."

Again, you are taking this out of context. Intel has stated that the benefits of SOI don't outweigh the drawbacks at this time. But why would you assume that means they would think that similar/better technologies could never be used? I don't see how them saying that "Technology A isn't beneficial now, but Technology A2 looks promising in the future" looks confusing or amusing.

 

[John Hancock]

i'm waiting for the DNA computer


look in mirror