PDA

View Full Version : The Ninety Nanometer Speed-bump


MacBytes
Jul 9, 2004, 01:07 AM
Category: Opinion/Interviews
Link: The Ninety Nanometer Speed-bump (http://www.macbytes.com/link.php?sid=20040709020704)
Posted on MacBytes.com (http://www.macbytes.com)

Approved by Mudbug

OhEsTen
Jul 9, 2004, 01:57 AM
LOL.

Just kidding. Good article, allthough I wasn't "freaked" when 3gz didn't come out. I personally am confident in the Apple/IBM relationship.

IBM has the know-how and the "intrest" in the Apple business that moto just didn't have. After 5 years with the G4 (even though I feel that it was a good chip for the time - just not "kept up"), it seemed that moto just lost interest. It felt like they just wanted to make cell phone chips and thats it.

IBM on the otherhand is not new to the high-end chip market. I think we have a lot of good years ahead of us.

crees!
Jul 9, 2004, 02:06 AM
Same here. I came across niggardly and paused. Read it again. Then again. Then proceeded on. 10 GHz, oh yea - now we're cookin'!

isgoed
Jul 9, 2004, 02:28 AM
10 GHZ!, these are the kinds of articles i love.

BakedBeans
Jul 9, 2004, 03:13 AM
10GHZ...HOLY FLASHBACKS....(thanks to whom i borrowed that from)
how can ibm tell that there going to be at 10ghz in 3 1/2 years....

Knox
Jul 9, 2004, 04:08 AM
I did wonder why that site looked familar, then i remembered. That was the one that was copying MR articles and forum posts a while back...

crees!
Jul 9, 2004, 05:01 AM
10GHZ...HOLY FLASHBACKS....(thanks to whom i borrowed that from)
how can ibm tell that there going to be at 10ghz in 3 1/2 years.... Mathematics

yamabushi
Jul 9, 2004, 05:44 AM
A couple years ago I had a discussion with an electrical engineer in Japan about moving to ever shrinking chip die processes. We generally agreed that the industry would hit a wall at some point. We thought at the time that continuing to shrink would require increasing technological challenges that would eventually make it much more economical to use other methods such as multiple cores to increase performance at some point. We only disagreed on the timing of the difficulties. He thought that chip makers would be safe until they reached close to 50nm but I thought they were due for some trouble much sooner around 100nm.

He later thought that the move to 90nm would go without a hitch. I however thought that things were moving too fast when you consider the percentage based decrease in surface area and that a more conservative shrink was wiser at first, such as to 110nm. He assured me that the engineers at Intel and IBM are very smart and could handle it. I took his word for it since he knew more about it than I.

I only mention this because I think that we may continue to see delays with each shrink due to technological challenges. I doubt the move to a 65nm die process will go smoothly. :(

wrldwzrd89
Jul 9, 2004, 06:28 AM
A couple years ago I had a discussion with an electrical engineer in Japan about moving to ever shrinking chip die processes. We generally agreed that the industry would hit a wall at some point. We thought at the time that continuing to shrink would require increasing technological challenges that would eventually make it much more economical to use other methods such as multiple cores to increase performance at some point. We only disagreed on the timing of the difficulties. He thought that chip makers would be safe until they reached close to 50nm but I thought they were due for some trouble much sooner around 100nm.

He later thought that the move to 90nm would go without a hitch. I however thought that things were moving too fast when you consider the percentage based decrease in surface area and that a more conservative shrink was wiser at first, such as to 110nm. He assured me that the engineers at Intel and IBM are very smart and could handle it. I took his word for it since he knew more about it than I.

I only mention this because I think that we may continue to see delays with each shrink due to technological challenges. I doubt the move to a 65nm die process will go smoothly. :(
I strongly suspect you'll be right when the move to 65nm comes. The industry wouldn't go to 110nm first - they've been in a sort of pattern as far as die shrinks are concerned. I suspect it's gone something like this in the past: ..., 720nm, 520nm, 360nm, 260nm, 180nm, 130nm. We're at 90nm now. I see no reason why this pattern wouldn't continue: 65nm, 45nm, 32nm, 22nm, 16nm, 11nm, 8nm, 5nm, 4nm, ...

MacsRgr8
Jul 9, 2004, 06:59 AM
G3 max clock of 450 Mhz? That was the max clock for the Blue & White.
iBook G3 went up to 900 MHz.

wrldwzrd89
Jul 9, 2004, 07:13 AM
G3 max clock of 450 Mhz? That was the max clock for the Blue & White.
iBook G3 went up to 900 MHz.
450 MHz was the PLANNED maximum clock for the G3. Due to the situation with the iBook, the ACTUAL max clock was twice that (900 MHz). Nobody was expecting the iBooks to stay G3 for so long - that only happened because of Motorola's G4 processor debacle at 500 MHz, plus all the other trouble they had with the G4.

themadchemist
Jul 9, 2004, 08:34 AM
well, it's encouraging, if this guy is right...certainly, we're not as bad off now as we were with moto.

g4cubed
Jul 9, 2004, 12:32 PM
Same here. I came across niggardly and paused. Read it again. Then again. Then proceeded on. 10 GHz, oh yea - now we're cookin'!
I stopped a had to look up the meaning :rolleyes:
I've heard it used before but wasn't sure. And for anyone else,here:

nig·gard·ly
adj.
1.) Grudging and petty in giving or spending.
2.) Meanly small; scanty or meager: left the waiter a niggardly tip.

I can only hope that we hit speeds like 10GHz in 3-4 years :D

BakedBeans
Jul 9, 2004, 05:33 PM
Mathematics

is that the type of maths that is idealistic...like sj's maths of this time next year we will be at 3ghz ect ect.... you get my meaning though...
.5ghz jump this year...next year 3ghz jump...hmmmm ;)

yamabushi
Jul 10, 2004, 03:56 AM
I strongly suspect you'll be right when the move to 65nm comes. The industry wouldn't go to 110nm first - they've been in a sort of pattern as far as die shrinks are concerned. I suspect it's gone something like this in the past: ..., 720nm, 520nm, 360nm, 260nm, 180nm, 130nm. We're at 90nm now. I see no reason why this pattern wouldn't continue: 65nm, 45nm, 32nm, 22nm, 16nm, 11nm, 8nm, 5nm, 4nm, ...
Actually, I believe at least one fab did move to 110nm. However, I think only less complicated ICs are produced there - no cpus.

I wouldn't exactly say the industry has completed the transition to 90nm yet. Yields are still low and the vast majority of chips are still built at 130nm and 180nm. 110nm is about as small as you can go before you need exotic materials and designs to counteract the leakage problem. Anyways, this is sort of 20/20 hindsight and not a suggestion for current action. Once you get down to about 45nm some really wierd effects start to appear that I am hardly expert enough to explain well.

Little Endian
Jul 10, 2004, 09:06 AM
[QUOTE=g4cubed]I stopped a had to look up the meaning :rolleyes:
I've heard it used before but wasn't sure. And for anyone else,here:

nig·gard·ly
adj.
1.) Grudging and petty in giving or spending.
2.) Meanly small; scanty or meager: left the waiter a niggardly tip.

I can only hope that we hit speeds like 10GHz in 3-4 years :D[/QUOTE

I too looked up the meaning, I wonder if the word originated in a deragotory sense as the dictionary does not provide an origin for the world Niggardly. Either way an interesting choice of word to use in the Article as I have not seen the word used in a long time.

Anyhow I don't think we will move beyond 65nm until the end of this decade. 90nm still needs alot of work and IBM and intel are really the only ones shipping 90nm in limited quantities and only as high end offerings AMD has still yet to release a 90nm core and probably won't have one shipping till mid 2005 next year at this rate and again only in limited quantities. I also doubt AMD will be able to Hit 3GHz for as long as another 9-12 months. We probably will not see a 65nm core shipped regardless of manufacturer until mid 2006 at the very earliest maybe even 2007 and even then only in very limited quantities at high prices. Dual Core and MP setups will be the way to go for the rest of this Decade as I highly doubt we will see 10Ghz or faster by 2010. That is why everyone is jumping on the Dual Core bandwagon and are pushing MP while we wait for Dual Core chips. Apple has been flaunting MP for years. Even AMD has gotten bigger on MP as with the immense popularity of the Opteron line which PC users are embracing at much higher rates than in the days of the Athlon MP.

yamabushi
Jul 10, 2004, 11:32 AM
Quad 2.5Ghz dual core G5 next tuesday! :p

Actually I think even Little Endian's timeline may be overly optimistic. :(

King Cobra
Jul 10, 2004, 06:35 PM
Unlike Motorola's lethargic climb up the megahertz ladder, IBM confidentally expects to hit 10 GHz with the PowerPC 990 in mid-2008.
Watch...IBM will have more problems with the 90nm chip and we'll only be at 6GHz by WWDC 2008. Actually, a dual 6.0GHz by then doesn't sound too bad, since now that IBM has had its first successful experience with liquid cooling in a 90nm machine and can learn from there, I feel that Intel is going to struggle a little more than IBM will.

Kagetenshi
Jul 10, 2004, 09:12 PM
I too looked up the meaning, I wonder if the word originated in a deragotory sense as the dictionary does not provide an origin for the world Niggardly. Either way an interesting choice of word to use in the Article as I have not seen the word used in a long time.

It's a valid word, and does not have negro ("black") as a root word. It was catapulted to semifame when some complete idiot made a (successful?) lawsuit for discrimination against someone who used the word.

People, we have words. They sometimes sound similar to other words. Get over it.

~J

idkew
Jul 11, 2004, 02:30 PM
It's a valid word, and does not have negro ("black") as a root word. It was catapulted to semifame when some complete idiot made a (successful?) lawsuit for discrimination against someone who used the word.

People, we have words. They sometimes sound similar to other words. Get over it.

~J

i am in the process of suing the keebler elves. i feel threatened and discriminated against by their use of the word cracker.

relimw
Jul 11, 2004, 09:53 PM
I stopped a had to look up the meaning :rolleyes:
I've heard it used before but wasn't sure. And for anyone else,here:

nig·gard·ly
adj.
1.) Grudging and petty in giving or spending.
2.) Meanly small; scanty or meager: left the waiter a niggardly tip.

I can only hope that we hit speeds like 10GHz in 3-4 years :D

Bah! Must we all think in politically correct terms everytime we see something?

Besides, >6GHz isn't excessively hard to achieve, the embedded world has just started fully moving into that area (mainly in things like high speed, high volume telco switches).

Give them some time, and you'll have that dual core, dual processor system running at 10GHz ;)

g4cubed
Jul 12, 2004, 11:16 AM
Bah! Must we all think in politically correct terms everytime we see something?

Besides, >6GHz isn't excessively hard to achieve, the embedded world has just started fully moving into that area (mainly in things like high speed, high volume telco switches).

Give them some time, and you'll have that dual core, dual processor system running at 10GHz ;)

I wasn't thinking in a politically correct view.

I was thinking "what the hell does this word mean". So I looked in a dictionary,
it's something you do when you're unsure of what a word means.
A lot of words sound derogatory but aren't, such as this.

Which, and I'm only guessing by your comment, is your way of thinking.

relimw
Jul 13, 2004, 06:02 PM
I wasn't thinking in a politically correct view.

I was thinking "what the hell does this word mean". So I looked in a dictionary,
it's something you do when you're unsure of what a word means.
A lot of words sound derogatory but aren't, such as this.

Which, and I'm only guessing by your comment, is your way of thinking.

You get that way when the NAACP is constantly telling you that all Southerns are racists...Ah well, I just hate being politically correct in any case, why else would I enjoy using a computer that has a <25% market share in any market? ;)
(Ok, there are lots of other reasons I use a mac...;)

Mav451
Jul 13, 2004, 06:32 PM
...
Anyhow I don't think we will move beyond 65nm until the end of this decade. 90nm still needs alot of work and IBM and intel are really the only ones shipping 90nm in limited quantities and only as high end offerings AMD has still yet to release a 90nm core and probably won't have one shipping till mid 2005 next year at this rate and again only in limited quantities. I also doubt AMD will be able to Hit 3GHz for as long as another 9-12 months. We probably will not see a 65nm core shipped regardless of manufacturer until mid 2006 at the very earliest maybe even 2007 and even then only in very limited quantities at high prices. Dual Core and MP setups will be the way to go for the rest of this Decade as I highly doubt we will see 10Ghz or faster by 2010. That is why everyone is jumping on the Dual Core bandwagon and are pushing MP while we wait for Dual Core chips. Apple has been flaunting MP for years. Even AMD has gotten bigger on MP as with the immense popularity of the Opteron line which PC users are embracing at much higher rates than in the days of the Athlon MP.

While AMD has not made any official releases, a few sites such as VR-Zone and Xtreme Systems have already gotten engineering samples of the 90nm part dubbed the "Winchester".

Commentary by overclockers.com can be found here:
http://www.overclockers.com/tips00615/

http://www.vr-zone.com/?i=982&s=1
@ VR Zone, they found that the numerous modifications make the 90nm faster than the current 130nm ones (based on Socket 939). These are basically the recent 3500+ and 3800+ (2.2 and 2.4Ghz respectively).