PDA

View Full Version : Let's discuss Low-K, IBM, and the future...


Pages : [1] 2

ragejg
09-25-03, 10:41 AM
From EE Times Sept 22 issue... there's 3 parts to this, so there's three posts... dang ChrisRay, you interrupted the Part 1/2/3 flow!! :p

http://www.nvnews.net/vbulletin/attachment.php?s=&postid=202373

ChrisRay
09-25-03, 10:47 AM
Thats interesting. Think you could summarize it for me :) I'm curious if this coulda affected Nvidias recent announcement too?

ragejg
09-25-03, 10:51 AM
part 2:

http://www.nvnews.net/vbulletin/attachment.php?s=&attachmentid=3274

ragejg
09-25-03, 10:53 AM
part 3, cut to fit...

http://www.nvnews.net/vbulletin/attachment.php?s=&attachmentid=3275

Hellbinder
09-25-03, 11:26 AM
This is how bad things have gone for Nvidia this year...

They Switched from TSMC to IBM specifically becuase it was supposed to enable them a more stable process for their upcomming .13u Chips. Like Nv40.

Now TSMC is going full bore... and IBM is hozed...

This must be the single greatest example of Reaping what you sow, or as some would say... Negative Karma i ahve ever seen.

Flavius
09-25-03, 11:45 AM
Haven't you people heard of Optical Character Recognition?! My god....

Malfunction
09-25-03, 12:01 PM
I got a question... meh, :)

What has changed in the manufacturing process (other than nanometer size) from the conceptions of the NV25, R200 up to the present time? It appears to me that in the effort to cut manufacturing cost, they have also cut the quality as well.

I don't remember the NV25 or R200 having as much manufacturing problems as the latest R3xx and NV3x are having.

While it seems that the IHV are benefitting from lower cost manufacturing and higher prices for there products, we are stuck with poor product(s) (Nvidia) and outrageous cost. (ATi and Nvidia)

Peace,

:confused:

ragejg
09-25-03, 12:14 PM
Originally posted by Flavius
Haven't you people heard of Optical Character Recognition?! My god....

... time was of the essence... I'm @ work, & I did this as fast as possible... and it is readable... the content's more important than the look IMO.

ragejg
09-25-03, 12:20 PM
I wonder how the conversion to Coral wil affect nVidia, whether it will be a boon or the twist of the knife...

sxotty
09-25-03, 01:49 PM
HellBinder the funny thing is IBM most likely did exactly what Nvidia has been doing.

IBM "Oh yeah the low-k is great if you use our fab we'll have your cards out just like you want"
Nvidia "Really? TSMC seemed to find it difficult"
IBM "Well WE aren't TSMC we have better foundries and engineers, our support is legendary"

Nvidia "Well ok it sure sounds good ....":)

Flavius
09-26-03, 07:21 PM
Strangely, Nvidia said in a press conference early this month that IBM's process basically kicked TSMC's ass and cost less too and that yields were good. Take the article with a grain of salt.

Hellbinder
09-26-03, 07:29 PM
Your serious.. :confused:

Accept Another Nvidia PR statement as the Truth..

But Take the respected "EE Times" article with a "Grain of salt".

This is not the Inq you know...

ragejg
09-26-03, 09:36 PM
nope, it's not... heck, I've been reading that mag since I was 13... it's pretty far from crap editorialization... well, there is one guy... :rolleyes: ;)

Uttar
09-27-03, 05:23 AM
The NV40 was never meant to be on Low K, or at least not meant to as soon as it switched to IBM ( no idea about the plans of when it was at TSMC )
However, it should have SOI - which is at least as good as Low-K AFAIK ( just different ). Both could be combined for it to be even better, but obviously, neither TSMC or IBM is there yet.


Uttar

dexiter
09-28-03, 05:43 AM
Originally posted by Hellbinder
This is how bad things have gone for Nvidia this year...

They Switched from TSMC to IBM specifically becuase it was supposed to enable them a more stable process for their upcomming .13u Chips. Like Nv40.

Now TSMC is going full bore... and IBM is hozed...

This must be the single greatest example of Reaping what you sow, or as some would say... Negative Karma i ahve ever seen.

Maybe the scanned article was a little hard to read. Here's a link to the EE times article:

http://www.eetimes.com/story/OEG20030922S0027

The article discusses how IBM's decision in 2000 to use SiLK in 0.13u process didn't pan out as IBM planned. So when did the problems start to show up? Summer 2002. By the time NVIDIA announced its partnership with IBM (March 2003), it was already known that IBM's having problems with SiLK and IBM had already made a decision to go with CVD. So if you think this is "the single greatest example of Reaping what you sow", I think your post may be the single greatest example of misunderstanding.

AnteP
10-01-03, 06:33 PM
9600 XT = on low k

The Baron
10-01-03, 06:34 PM
Originally posted by AnteP
9600 XT = on low k
Gawha?

Why is that, exactly? Hell, a lot of the 9600 Pros without any sort of exotic cooling can hit 500Mhz no problem...

Skuzzy
10-01-03, 07:13 PM
It has been ATI's pattern to introduce new processes on simpler parts. Makes sense. That methodology seems to be working for them very well.

EMunEeE
10-01-03, 07:14 PM
Originally posted by The Baron
Gawha?

Why is that, exactly? Hell, a lot of the 9600 Pros without any sort of exotic cooling can hit 500Mhz no problem...

Yea i know. It'll be interesting to see how high the 9600XT OCs

reever2
10-01-03, 07:21 PM
Originally posted by The Baron
Gawha?

Why is that, exactly? Hell, a lot of the 9600 Pros without any sort of exotic cooling can hit 500Mhz no problem...

Power requirements, despite the high performance and clocks it will require no extra power connectors, and it still helps OC'ing

StealthHawk
10-01-03, 09:08 PM
Originally posted by The Baron
Gawha?

Why is that, exactly? Hell, a lot of the 9600 Pros without any sort of exotic cooling can hit 500Mhz no problem...

No need for the Molex power connector on the r9600XT :)

The Baron
10-01-03, 09:12 PM
Originally posted by StealthHawk
No need for the Molex power connector on the r9600XT :)
No need on the 9600 Pro either... which is why I'm confused. I'd think they could easily have 500/400 clocks with the RV350 (might need some better cooling, but that'd be it). I guess it really was, "Hey, dudes, let's try low-K on the RV360 so we can get ready for R420."

SlyBoots
10-01-03, 10:05 PM
Originally posted by The Baron
I guess it really was, "Hey, dudes, let's try low-K on the RV360 so we can get ready for R420."

3 on 1 w/12 pipes? HB's rumor?

The Baron
10-01-03, 10:15 PM
No, I meant they're just testing the process, making sure they get the kinks worked out of everything before they use it for their huge complex killer chip.

Sazar
10-01-03, 10:40 PM
Originally posted by The Baron
No, I meant they're just testing the process, making sure they get the kinks worked out of everything before they use it for their huge complex killer chip.

yah pretty much :)

ibm may have gone about their low-k process the wrong way... there are different ways to tackle the same issue... but ibm may have gone the wrong route...

I do want to see if there is going to be any changes whatsoever in the design of the 9600xt's layout (ie components on the pcb) due to the nature of low-k dielectric materials...

:)

better to use it with a small hsf than a large one I reckon to start off with... see how it handles things...