ForceWare 61.34 tested

DaveB,

In the Foundary face off drawing that you showed, the comparison is between two competitors, right? We already know that TSMC has low-k process, but does UMC also have a low-k process? The main question I have: Is the drawing comparing data on TSMC's low-k process vs UMC's low-k process, or just TSMC's low-k process vs UMC's standard process? Also, is there any comparable data from IBM? Thanks.

 

UMC is also adopting Applied Materials Black Diamond process for Low K. There appears to be a bit of a war between the two foundries over who has the better implimentation.

 

Here is a story about IBM’s woes and the difficulties with the move to .9 micron.

LINK

 

Well, it doesn't negate, because you don't know which one (FSG, diamond) goes up to 8 layers

 

Having the ability to go to more layers is an advantage not a disadvantage. Just because you have the option to go to 8 layers doesn't mean you must.

-FUDie

 

Noone said it was a disadvantage, just that it impacts yields and costs.

 

Look at the quote beneath - it clearly says that TSMC's 130nm process is capable of 8 layers with low-k options, ergo both go up to 8 layers.

 

The point of the comparison chart (in conjunction with the text from TSMC) was to highlight the fact that there are no layer differences from TSMC's low-k and standard FSG options, both are up to 8 layers. As for your questions regarding the comparison to UMC's process capabilities on comparison - the necessary information is in the chart to answer the question (hint: yes this is a comparing to UMC's Low-k process). Not seen anything similar for IBM.

 

Sure, did i say something on that? :?

 

Well, it's in general term, talking about all the capabilities. You could be right, but not necesseraly. It copes well with the "up to"

 

Its seems fairly clear.

However, the point is pertinent to my initial comments since we are talking about "up to 8 layers", we have no knowledge on how many layers ATI are actually using for R420. Given that there is only a 10% surface area difference between 420 and NV40 that increases the possability of different number of layers being used.

 

Whoa there, that's a lot of interesting stuff hidden in a forceware thread!

edited: off the long unneccessary quote...

 

Yeah, because you have made up your mind

Now, any idea of the number of layers provided by IBM?

 

ED...just what's your problem? Do you honestly believe what you have just implied? If so, then why are you even posting? Believe me whan I say you are absolutley no voice of reason....... I personally find your implication insulting......


Hmmm... just who's word would anyone with a bit of intellegence trust.... DB or Radar........

 

Nice sig there, martrox.
________
Ferrari F300 History

 

That's some general papers on the capabilities of the fab. I said he could be right, but i don't think it's that clear, or my faith in marketing figures just went up. Oh, and then should i take Nv PR for pure truth?

As for insulting, i don't know i didn't mean to, so i apologize if it did sound insulting. Moreover, perhaps i should say you just insulted me, and i didn't think that DB's quote was really very nice of him.

 

Huge differences in the target audiences - fabs are not marketting to general consumers, but other companies that have specific fabbing requirements and I'm not sure it would come across as very good if the breifs say one thing and the librarys say another. Plus, the table (independantly generated) says the same.

 

It only affects yield and costs if your design uses these layers.

Am I conversing with a wall or what?

-FUDie

 

You are very correct Dave.

Companies do resent being misled, having design libraries changed at the last minute and the guilty party refusing to admit their culpability.

Some companies resent that sort of thing so much that they take their business elsewhere.

 

Say wha.....? Only you, Radar..........Sheesh!