hcl64 (mario.smarq@gmail.com) on 4/30/12 wrote:
---------------------------
>David Kanter (dkanter@realworldtech.com) on 4/30/12 wrote:
>---------------------------
>>First of all, I'd like to ask you to refrain from turning our forums into a giant
>>bill board for EETimes. If you want to build link farms, do it somewhere else.
>>
>
>Sorry Mr. Kanter, i know i've been your guest and i must >respect the host, no matter how different the opinions can >be.

>Well... it has been an habit to provide info from disparate sources, sources can
>be wrong or less correct of course, marketing is pervasive, >so let see who is more
>correct or more wrong... what players are more inclined to >spin or to facts...

>So at least i think those links can yet provide an >excellent service to the inquiring
>mind critically seeking the truth.

That's fine and I applaud finding independent information. If you look at your post...it contains 14 links. I absolutely respect the effort that you took in putting it together.

However, there is a fine line between providing supporting evidence and arguments and inundating our forum with outgoing links.

I'd also point out that many of your links don't exactly support your point.

For example, STMicro has committed to using FD-SOI for their products going forward. However, they are not shipping any chips manufactured using FD-SOI. As I mentioned, Intel is shipping millions of IVB using FinFETs.

I'd also encourage you to avoid taking anything from SOITEC at face value. As an organization, they do some very interesting science and engineering work. However, they are very aggressive about marketing. PD-SOI was supposed to the best thing ever, and pretty much nobody used it, so they don't exactly have a reputation for producing technology that is widely adopted.

Those studies that purport to show that FD-SOI is less expensive were bought and paid for by SOITEC and are unlikely to be fully accurate. Some of the assumptions seem rather unrealistic, particularly because the results are so dramatically different from empirical observations.

I don't believe for a minute that FD-SOI is less expensive than FinFETs, otherwise TSMC would have chosen that option.

Also, many of the claims about FD-SOI assume that customers will use body biasing techniques to further enhance power/performance. Body bias adds cost to chips, because it requires additional voltage rails to apply the bias and it also adds complexity to the chip by distributing more power rails throughout the design.

I know that TI and Samsung are using body bias, and STMicro clearly intends to do so. However, high performance products generally don't. Body bias isn't used by Intel, AMD, IBM, Oracle or Nvidia for CPUs/GPUs. I don't believe it is used by Qualcomm either.

If you look at the facts, it seems pretty clear that for companies with high volume such as TSMC and Intel, FinFETs are a much better choice. There is a higher investment in NRE to get them right, but the variable cost is lower.

For companies like IBM and STMicro, where the volumes aren't quite as high as TSMC or Intel, FD-SOI may be a very logical choice. But even IBM is looking at FinFETs...

David