Do you think Intel is having trouble with the 90nm process?

[AgaBoogaBoo]

I get the feeling that Intel is having troubles with the 90nm process because if you remember Dothan, the next Pentium M processor and the Prescott CPU's are to be both 90nm. Prescott is said to be creating large amounts of heat, and Dothan according to a new post on the front page of Anandtech.com is having trouble. I think the two are connected in some similar way, linked to the 90nm production, kind of like the problem when switching to a .13 micron process for Nvidia.

Of course, this is all speculation because on the other side, Intel may just need the cpu fan to spin faster and copper heatsinks to cool Prescott and Dothan may be temporarily replaced with faster versions of the current Pentium M processor.

 

[Nebor]

Prescott is said to be dissipating large amounts of heat

That sounds like a good thing to me. I think we'd all like our processor to dissipate a large amount of heat. Then we wouldn't need such beefy heatsinks. Anyways, I went over the intel production facilities, and said I'd like to see how production of the prescott is coming along, and sit in on a few tech meetings. THey said, "Who are you?" I pulled out my stock certificates and said, "I'm one of the owners!" The security guards pulled out lightsabers and chased me away...

 

[OverVolt]

Originally posted by: Nebor

Prescott is said to be dissipating large amounts of heat

That sounds like a good thing to me. I think we'd all like our processor to dissipate a large amount of heat. Then we wouldn't need such beefy heatsinks. Anyways, I went over the intel production facilities, and said I'd like to see how production of the prescott is coming along, and sit in on a few tech meetings. THey said, "Who are you?" I pulled out my stock certificates and said, "I'm one of the owners!" The security guards pulled out lightsabers and chased me away...

lol... That makes nooo sense but yeh anyway.

I have no idea if Intels havin trouble with the 90nm i wouldn't know.

 

[AgaBoogaBoo]

Originally posted by: OverVolt

Originally posted by: Nebor

Prescott is said to be dissipating large amounts of heat

That sounds like a good thing to me. I think we'd all like our processor to dissipate a large amount of heat. Then we wouldn't need such beefy heatsinks. Anyways, I went over the intel production facilities, and said I'd like to see how production of the prescott is coming along, and sit in on a few tech meetings. THey said, "Who are you?" I pulled out my stock certificates and said, "I'm one of the owners!" The security guards pulled out lightsabers and chased me away...

lol... That makes nooo sense but yeh anyway.

I have no idea if Intels havin trouble with the 90nm i wouldn't know.

Sorry, I meant the other thing. I meant that it generated a lot of heat... hehe, I don't know how dissipated slipped in there

 

[0roo0roo]

intels not having trouble until amd puts out processors that truely rival it again.

 

[Luagsch]

well mybe the time is comong when we have a general chance to phase-coolers and watercooling. remember the time when we only had passive heatsinks on our proc's and everybody had wide-opened eyes because the need of active cooling seemed so out-of-the-world?

 

[Mingon]

Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

 

[dullard]

This is all my opinion, I'm not a processor designer so this may all be BS:

All processors have current leaks - where current is flowing through places where it shouldn't. This results in wasted power, which simply turns into heat. These leaks take time to find and correct which is why newer revisions of processors often overclock a lot better. As the process size gets smaller, the insulating regions get smaller as well - and I assume that means there is more points where current can leak. Thus it gets harder and harder to make a process shrink. Not only that, but the process shrink to 0.09 microns is a bit larger than the rest:

• 0.35 to 0.25 is a 29% drop.
  0.25 to 0.18 is a 28% drop.
  0.18 to 0.13 is a 28% drop.
  0.13 to 0.09 is a 31% drop.

That means they are trying to drop it 10% faster than they used to at a time where drops are harder and harder. No wonder companies are struggling. 0.09 micron is just too optomistic, and 0.10 micron may have been a better choice (like ATI went to 0.15 micron instead of 0.13 micron).

With Dothan and the 0.09 micron Prescotts both running significantly hotter than originally predicted, it is probably due to these process leaks. So yes Intel is having troubles with its 0.09 micron process. I assume all the other companies will struggle as well.

 

[Lonyo]

It's gonna be hell when they try and get right down to 0.45 then.

IBM are starting work on 0.65 at the moment I believe, for future use of course, and IIRC, memory companies are now getting down to 0.1 technology and putting it in to practice, although I think some have gone to 0.11.

Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

SMP isn't all that great for use unless we have optimised applications.
An Unreal game with one proc used for AI, and others for physics/whatever else would be good, but as it is, most games can't take advantage of 2 processors, so unless software developers do something different, it's not going to be the best way.
There's something like a 15% improvement through 2 procesors (logical, not HT) overall (from some random other thread here).

I would say that a Intel had a good plan with their Banias (Centrino/Pentium-M) architecture, making something as efficiant as possible, and the future may be to work on that sort of approach, since it gives good performance while using low power and producing less heat than the raw MHz approach they seem to have at the moment in their desktop lineup.

A move to better architecture is being made, because there are rumours of Banias variants being prepared for use in servers to reduce power consumption and heat.
Imagine a room full of new Xeon type things based on Prescott putting out 100w, then think what even 150 Banias type chips putting out maybe 40w would be like in comparison. Much easier to cool and much less power consumption.

But it would seem they are having 0.9 problems as both their 0.9 micron chips have been delayed.

 

[Chadder007]

Originally posted by: Mingon
Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

:werd: I think its time for them to just quit bumping the clock and make the CPU more efficient like the Pentium M when using its cycles.

 

[AgaBoogaBoo]

Originally posted by: dullard
This is all my opinion, I'm not a processor designer so this may all be BS:

All processors have current leaks - where current is flowing through places where it shouldn't. This results in wasted power, which simply turns into heat. These leaks take time to find and correct which is why newer revisions of processors often overclock a lot better. As the process size gets smaller, the insulating regions get smaller as well - and I assume that means there is more points where current can leak. Thus it gets harder and harder to make a process shrink. Not only that, but the process shrink to 0.09 microns is a bit larger than the rest:

• 0.35 to 0.25 is a 29% drop.
  0.25 to 0.18 is a 28% drop.
  0.18 to 0.13 is a 28% drop.
  0.13 to 0.09 is a 31% drop.

That means they are trying to drop it 10% faster than they used to at a time where drops are harder and harder. No wonder companies are struggling. 0.09 micron is just too optomistic, and 0.10 micron may have been a better choice (like ATI went to 0.15 micron instead of 0.13 micron).

With Dothan and the 0.09 micron Prescotts both running significantly hotter than originally predicted, it is probably due to these process leaks. So yes Intel is having troubles with its 0.09 micron process. I assume all the other companies will struggle as well.

Isn't that what SOI is supposed to do? (Silicon On Insulator?)

 

[AgaBoogaBoo]

Originally posted by: Chadder007

Originally posted by: Mingon
Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

:werd: I think its time for them to just quit bumping the clock and make the CPU more efficient like the Pentium M when using its cycles.

Yeah, but they'd then have to start marketing differently and that would cost a ton because people right now only know about speed in terms of Mhz. If you give the average user a choice from a Celeron 2.6Ghz and a Pentium 4 2.4C, they will likely pick the Celeron as long as the system looks good in terms of case, monitor, etc.

 

[grant2]

Originally posted by: AgaBooga
If you give the average user a choice from a Celeron 2.6Ghz and a Pentium 4 2.4C, they will likely pick the Celeron as long as the system looks good in terms of case, monitor, etc.

Do you know this from actual investigation, or are you just guessing?

 

[JediJeb]

Ummm SOI is an AMD product isn't it, yell at me if I'm wrong.

 

[Barnaby W. Füi]

Originally posted by: Mingon
Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

No way, I have dual 1.3 durons and they definitely aren't as quick as a >2ghz p4. *However*, if you like to do cpu-intensive things a lot, it's nice to have smp, since your machine doesn't go down the drain during that period of time. Most "normal" people don't do many cpu-intensive things at once, so smp isn't all that great for them. Now, scsi hard drives are something that would boost performance for EVERYONE, yet basically no one uses them. $200 for a FAST 15k fujitsu drive, and < $50 for a plain jane scsi controller, that's not *too* much cash, and it'll completely rape any ide drive, including the raptors that everyone is drooling over (oh, and it'll last forever, which isn't something you can reliably say about ide drives).

 

[OverVolt]

Originally posted by: BingBongWongFooey

Originally posted by: Mingon
Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

No way, I have dual 1.3 durons and they definitely aren't as quick as a >2ghz p4. *However*, if you like to do cpu-intensive things a lot, it's nice to have smp, since your machine doesn't go down the drain during that period of time. Most "normal" people don't do many cpu-intensive things at once, so smp isn't all that great for them. Now, scsi hard drives are something that would boost performance for EVERYONE, yet basically no one uses them. $200 for a FAST 15k fujitsu drive, and < $50 for a plain jane scsi controller, that's not *too* much cash, and it'll completely rape any ide drive, including the raptors that everyone is drooling over (oh, and it'll last forever, which isn't something you can reliably say about ide drives).

What is up with ppl bashing my Raptors !

And we can always use more CPU power!

 

[Barnaby W. Füi]

Originally posted by: OverVolt
What is up with ppl bashing my Raptors !

I'm not bashing them, they are good for their intended purpose, but they are not equivelant to a 15k scsi drive.

 

[Jeff7181]

Originally posted by: JediJeb
Ummm SOI is an AMD product isn't it, yell at me if I'm wrong.

YOU'RE WRONG!!!

=) Now that that's out of the way... SOI is the process AMD and IBM are currently using... Intel hasn't switched to that yet, there using a technique called strained silicon, I believe they stated they plan on going to SOI (in combination with strained silicon technologies) when they switch to .065 micron.

*EDIT* Now that I think about it... if Intel really is having trouble with .09 micron they may have to switch to SOI sooner. I wonder if strained silicon isn't as beneficial as Intel had predicted.

 

[Jeff7181]

Originally posted by: BingBongWongFooey

Originally posted by: OverVolt
What is up with ppl bashing my Raptors !

I'm not bashing them, they are good for their intended purpose, but they are not equivelant to a 15k scsi drive.

And a Mustang isn't as fast as a Corvette either... talk about stating the obvious...

 

[devers]

Originally posted by: Nebor


Anyways, I went over the intel production facilities, and said I'd like to see how production of the prescott is coming along, and sit in on a few tech meetings. THey said, "Who are you?" I pulled out my stock certificates and said, "I'm one of the owners!" The security guards pulled out lightsabers and chased me away...

lol! I would love to see somebody actually try that.

 

[magomago]

Originally posted by: AgaBooga

Originally posted by: Chadder007

Originally posted by: Mingon
Perhaps its time for lower clocked smp machines, say 2 x Pentium M's at 1.0ghz should be plenty quick

:werd: I think its time for them to just quit bumping the clock and make the CPU more efficient like the Pentium M when using its cycles.

Yeah, but they'd then have to start marketing differently and that would cost a ton because people right now only know about speed in terms of Mhz. If you give the average user a choice from a Celeron 2.6Ghz and a Pentium 4 2.4C, they will likely pick the Celeron as long as the system looks good in terms of case, monitor, etc.

My Uncle did that - chose a 1.7Ghz Celeron over a 1600+ ; My dad would have done that had I not led him "to the light"

Hell even when I thought I knew a lot about computers had I been given the choice between even a 2.8GHz Celeron and a 2.4C I would have taken the 2.8Celeron provided it was cheaper and looked nice and had ram (I probably wouldn't inquire about mobo, speed of ram or anything. My friend's dad has a 1.5Ghz Pentium4 and the thing is coupled with pc133 ram and the father was gloating how he upgraded it by 256 Megabytes of ram so it would reduce "disc cache" - he would've gone well to stick some faster ram instead of choking that poor Willamette ).
I would probably think I'm paying the extra money for the "Pentum" name.

And I'm sure a lot of people think they know a lot about computers when in reality their knowledge really isn't that much (And I'm sure we'll all been there ) so they can eaily make mistakes based on what they perceive to be the truth.

 

[AgaBoogaBoo]

Originally posted by: Jeff7181

Originally posted by: JediJeb
Ummm SOI is an AMD product isn't it, yell at me if I'm wrong.

YOU'RE WRONG!!!

=) Now that that's out of the way... SOI is the process AMD and IBM are currently using... Intel hasn't switched to that yet, there using a technique called strained silicon, I believe they stated they plan on going to SOI (in combination with strained silicon technologies) when they switch to .065 micron.

*EDIT* Now that I think about it... if Intel really is having trouble with .09 micron they may have to switch to SOI sooner. I wonder if strained silicon isn't as beneficial as Intel had predicted.

I think you may be right because I remember them planning to use SOI after the .90 micron process in the .65 micron. I think Intel may have to switch to using SOI with the .90nm process. I wonder why the Prescott CPU needs so much more power than before. Anyone have an idea?

 

[AgaBoogaBoo]

Originally posted by: grant2

Originally posted by: AgaBooga
If you give the average user a choice from a Celeron 2.6Ghz and a Pentium 4 2.4C, they will likely pick the Celeron as long as the system looks good in terms of case, monitor, etc.

Do you know this from actual investigation, or are you just guessing?

A little bit of both because some people think they're paying just for the brand name. In a way like when you buy shoes, the $60 Nike's or the $20 Wal*Mart generic shoes. A person who doesn't care what others think and just wants the cheapest shoes because its going to get dirty, will choose the Wal*mart shoes. I know thats not an accurate comparison, but its what I can think of right now.

I th ink I made the speeds too close, it should have been like a P4 2.2Ghz and a Celeron 2.7 or 2.8Ghz cpu. The average person will choose the Celeron CPU. That I know by actual investigation because out of the many people I know, who think they know much about computers, will not have the slightest clue what the difference between a Celeon and P4 is other than that the P4 has been reaching higher speeds sooner than the Celeron.

 

[AgaBoogaBoo]

Originally posted by: Jeff7181

Originally posted by: BingBongWongFooey

Originally posted by: OverVolt
What is up with ppl bashing my Raptors !

I'm not bashing them, they are good for their intended purpose, but they are not equivelant to a 15k scsi drive.

And a Mustang isn't as fast as a Corvette either... talk about stating the obvious...

Yes, but, when people buy $2000 or $3000 systems, the company can probably add a SCSI 15,000 RPM drive even if it costs $100 more total... I mean, if a person is spending $3000 on a system, why not top it off with a really fast HD?