PDA

View Full Version : Intel Cancels Top-Speed 4Ghz Pentium 4 Chip


MacBytes
Oct 14, 2004, 08:17 PM
Category: Microsoft
Link: Intel Cancels Top-Speed 4Ghz Pentium 4 Chip (http://www.macbytes.com/link.php?sid=20041014211742)
Posted on MacBytes.com (http://www.macbytes.com)

Approved by Mudbug

paulypants
Oct 14, 2004, 09:21 PM
aaaaawwwwww... :(

;)

Dr. Dastardly
Oct 14, 2004, 09:37 PM
I'm just glad these companies are FINALLY figuring out its not just about clock speed.

nagromme
Oct 14, 2004, 10:59 PM
They're working on dual-core instead, aiming for next year. Not unlike a certain Power-series chipmaker...

Timelessblur
Oct 15, 2004, 01:10 AM
I might like to point out them mac theroical speed of a cpu is something like 5-6 ghz and the close you get to that wall the hard it is going to be to go faster

isgoed
Oct 15, 2004, 02:36 AM
I might like to point out them mac theroical speed of a cpu is something like 5-6 ghz and the close you get to that wall the hard it is going to be to go faster

The theoretical speed of a transostor lies at about 200Ghz. Of a cpu that will not be as high, but certainly higher than 6Ghz.

I don't understand what the problem is with 4Ghz. Overclockers were able to achieve 6Ghz (windows stable). It seems strange that those 200mhz or %5 increase pose a threshold.

Finally: why is this placed under topic "microsoft"?

Analog Kid
Oct 15, 2004, 02:43 AM
I might like to point out them mac theroical speed of a cpu is something like 5-6 ghz and the close you get to that wall the hard it is going to be to go faster
Funny, they said that theoretically it was impossible to break a 4 minute mile...

So what's the race going to be now? How many cores you have on a chip? Core count x clock rate?

This is going to force an interesting change in the software world... I'm amazed by how many applications are single threaded. I guess Intel's Hyperthreading has eased the industry into the transition, and giving a years warning will help.

Are games multi-threaded yet?

Analog Kid
Oct 15, 2004, 02:48 AM
I don't understand what the problem is with 4Ghz. Overclockers were able to achieve 6Ghz (windows stable). It seems strange that those 200mhz or %5 increase pose a threshold.
A lot of the overclocking relies on LN2 cooling technology and the like, not something you can really ship.

I'm sure they didn't cancel it because they can't do it at all, but probably because they can't get the yields where they want them. There's a big difference between one hack getting one chip to operate in a controlled environment and outfitting an entire company with high reliability systems on their desks.

Of course Intel's MTBF doesn't really need to exceed Windows mean time between infections...

Finally: why is this placed under topic "microsoft"?

Should probably be called "Wintel" instead of Microsoft... I've noticed the Intel stories get lumped in with Windows as a platform. I think they mean Microsoft as in "the other guys".

thatwendigo
Oct 15, 2004, 03:23 AM
A lot of the overclocking relies on LN2 cooling technology and the like, not something you can really ship.

It's also interesting to note that, while they did reach a supposed 6.0ghz mark with the machine, they had to scale back to 5.6ghz to be able to run anything. The feat was accomplished with a submersion bath of N2, which is so ridiculously unlikely in the home market that it doesn't even bear discussing.

I'm sure they didn't cancel it because they can't do it at all, but probably because they can't get the yields where they want them. There's a big difference between one hack getting one chip to operate in a controlled environment and outfitting an entire company with high reliability systems on their desks.

There's also the small matter of needing liquid nitrogen so that the chip doesn't crack from overheating, and the fact that even 3.2ghz parts were running up over the 100 watt power draw mark.

In fact, I think I'll look up the latest releases...

Intel Pentium 4 3.6ghz runs at 115 watts (http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7J9&ProcFam=483&PkgType=ALL&SysBusSpd=ALL&CorSpd=ALL)

That makes the 970 look pretty cool, after all. :D

vollspacken
Oct 15, 2004, 04:47 AM
BAAAAAAAAAAM!!!

medic please, someone has hit the wall...

looks like the Pentium M is the way to go for the x86-world (and a nice chip that is...)

vSpacken

AmigoMac
Oct 15, 2004, 05:46 AM
It will be a long time before people "forget" the GHz myth, Dealers and shops probably won't help a lot in a short time, since a very famous question when people want a new PC is about processor speed ...

Maybe it will be until longhorn when the real publicity starts with system requierements:

Intel "Write the name here" processor.
XY GB RAM
blblabla

by now I don't see a radical change in that respect, but I could be wrong, AMD will help with that

AMD (Intel "Name here")+ processor... :rolleyes:

Dont Hurt Me
Oct 15, 2004, 05:54 AM
It will be a long time before people "forget" the GHz myth, Dealers and shops probably won't help a lot in a short time, since a very famous question when people want a new PC is about processor speed ...

Maybe it will be until longhorn when the real publicity starts with system requierements:

Intel "Write the name here" processor.
XY GB RAM
blblabla

by now I don't see a radical change in that respect, but I could be wrong, AMD will help with that

AMD (Intel "Name here")+ processor... :rolleyes:So true, i was wondering what is Amd going to do now? so i have a 3500+ so that means its......nevermind.

Sol
Oct 15, 2004, 10:08 AM
Intel has hit the x86 wall. When the dual-core CPUs are ready Intel would prefer that Longhorn was available. Otherwise, Windows and all its applications need a major update to see a noticable effect from two cores.

CaptainHaddock
Oct 16, 2004, 06:39 PM
The theoretical speed of a transostor [sic] lies at about 200Ghz. Of a cpu that will not be as high, but certainly higher than 6Ghz.

The problem is not the switching speed of a transistor, but the speed of an electron—a true physical limit. As a poster on Slashdot (http://slashdot.org/comments.pl?sid=125718&cid=10530013) pointed out, an electron can only travel .75cm—less than the diameter of a chip die—per cycle at 4 GHz. So to go any faster, you have to make sure that no possible operation has circuits that long. The faster you go, the more constrained your circuit designs get. And then there's the distribution of the clock cycle signal (which must reach all parts of the die) to worry about.

Heck, even the Cray X-1 supercomputer (http://www.cray.com/products/x1/specifications.html) only runs at 800 MHz!

That's why Intel is falling behind. They've been too concerned about clock speed, a marketing gimmick, while AMD and IBM have been focusing on getting more out of each clock cycle.

The only way to radically increase clock speed in the future will be to adopt optical circuitry. In the meantime, it'll be nice to see chip-makers concentrate on improving other features.