Anyone for a possible 4.5GHz G5 by 2005?

[King Cobra]

Macbidouille: English version

" The maximum frequency that these chips will reach is not fixed yet with certainty, but they count on 4,5 GHz at the beginning of 2005, when the engraved successor with 65nm takes the changing."

 

[howard]

definitely possible....two years ago the 800 mhz powermac was out...now its 2 ghz....so going from 2 ghz to 4.5 ghz seems about right

 

[Mr. Anderson]

I'll be wanting one of those, thats for sure - I just hope the heating issue with the new boxes has been fixed by then.....

I'm really liking the fact that we have a decent road map in terms of CPUs with these new IBM chips.....so much better than Moto...

D

 

[MacFan25]

This is really good. I just hope Intel won't keep up.

 

[Billicus]

It wouldn't hurt my feelings to have Apple back ahead in the processor speed rankings.

 

[jethroted]

I'll take one!

 

[shadowfax]

Originally posted by Mr. Anderson
I'll be wanting one of those, thats for sure - I just hope the heating issue with the new boxes has been fixed by then.....

I'm really liking the fact that we have a decent road map in terms of CPUs with these new IBM chips.....so much better than Moto...

D

the heating issue should be fixed. it will probably be fixed by a migration to a .065 micron process or whatever it is called... though they may change up other stuff to help.

 

[Freg3000]

G4s were (are) at 180nm, the G5 is at 130nm, then they will be made at 90nm, and now we see 65nm in the future. I have no idea how lowering that number makes a processor run faster and cooler, but it is damn exciting.

 

[macphoria]

Either Steve or someone involved in the development said they will be hitting 3Ghz mark within 6 months. Or maybe it was 12 months. Either way, going beyond 4Ghz by 2005 seems possible.

 

[G5orbust]

I would have to say that I predict:

A speed bump of 1GHz every year

a drop in micron process level every year and half

So, that means, in 2005, I would expect to see around 4 GHz and the 90 micon process (.09mm), though Big Blue may pull a fast one and migrate to the .065mm (65 micron) process, depending on what micron process the rest of the CPU market is holding at.

Ive heard Intel has a big chip announcement/update planned for 2005 or so, and I assume AMD will have a chip to counter Intel, so 2005 seems to be a big year to wait for!

 

[Cubeboy]

Sounds reasonable enough, scaling from 2 GHz (using .13 micron process) to 3 GHz (using .09 micron process) is an increase in clock rate of 50%, assuming that same increase in clock rate is true when moving from .09 micron process to .65 micron process, another 50% increase in clock rate would effectively give us 4.5 GHz.

This should come as little surprise to most of you as it's just linear increase due to smaller and smaller transistors, and thus, would be true for any processor. Personally, I'd hope for alot more from the G5 in 2005 than just a large increase in clockrates.

 

[Catfish_Man]

Well, here's how I look at it:

Current G5's go to 2GHz on .13 micron

Apple has announced that they will have 3GHz (50% boost) in 12 months or less. This pretty much has to be a .09 micron chip, imo. 50% frequency boost seems to be the rule for die shrinks (I think the .18 micron G3 reached 700MHz, then the .13 micron one hit 900, and now 1.1GHz with SOI and a bunch of tweaking. Similar things happened with the AthlonXP and P4).

Assuming similar speed transitioning to .065 micron, that would mean 4.5GHz in 2005. I don't think IBM will hit .065 micron before the second half of 2005, but I could be wrong (I hope I am). Dual 3+ GHz 9xx's should be able to compete quite nicely in '04 though.

According to the leaked roadmap at extremetech, Intel will be introducing the successor to Prescott (P5?), called Tejas (P6?) at the end of 2004. I can't see how Prescott would be able to compete at .065 micron, I think the 9xx will have it hopelessly outclassed. Tejas is an unknown factor though.

http://www.extremetech.com/article2/0,3973,900185,00.asp

 

[G5orbust]

Originally posted by Catfish_Man



According to the leaked roadmap at extremetech, Intel will be introducing the successor to Prescott (P5?), called Tejas (P6?) at the end of 2004. I can't see how Prescott would be able to compete at .065 micron, I think the 9xx will have it hopelessly outclassed. Tejas is an unknown factor though.

http://www.extremetech.com/article2/0,3973,900185,00.asp

According to what I hear, neither the "Tejas" nor "Prescott" chips will be the P5. They both will be highly clocked P4's, with the "Tejas" being on the 90 micon (.09mm) process and having a 1000MHz+ bus with all this other crap (hyperthreading, improved bus architchture, etc.)

 

[Catfish_Man]

Originally posted by Freg3000
G4s were (are) at 180nm, the G5 is at 130nm, then they will be made at 90nm, and now we see 65nm in the future. I have no idea how lowering that number makes a processor run faster and cooler, but it is damn exciting.

Sorry about the double post.

The process size number refers to the average (I think, it could be the smallest) size of the features on the chip (wires and such). Since shorter wires means less distance to travel, and less power lost due to leakage and resistance, the chip can go faster, be smaller, and run cooler. A .13 micron chip is half the size of a .18 micron one, .09 is half the size of .13, etc... (this is because it scales both dimensions of the chip down)

 

[Catfish_Man]

Originally posted by G5orbust
According to what I hear, neither the "Tejas" nor "Prescott" chips will be the P5. They both will be highly clocked P4's, with the "Tejas" being on the 90 micon (.09mm) process and having a 1000MHz+ bus with all this other crap (hyperthreading, improved bus architchture, etc.)

Prescott still has the same architecture as the P4 (Intel has announced a lot of details about it). I'm guessing they're going to market Prescott as the P5 though, since it's quite a major enhancement. Similar to the way the P Pro, P2, and P3 were variants on the same core. Tejas will not be a highly clocked P4, it apparently is a brand new core. This may have changed since the stuff I've read, though.

 

[G5orbust]

Originally posted by Freg3000
G4s were (are) at 180nm, the G5 is at 130nm, then they will be made at 90nm, and now we see 65nm in the future. I have no idea how lowering that number makes a processor run faster and cooler, but it is damn exciting.

wow, Id sure love something on the 90 nanometer process.

I think you got confused with that one. Computer chips use the micron (micrometer) for process sizes. The nano meter is much smaller than a micron.

Micron: 10-6 (10 to the negative 6th power)

Nanometer: 10-9 (10 to the negative 9th power)

 

[Cubeboy]

Originally posted by G5orbust
According to what I hear, neither the "Tejas" nor "Prescott" chips will be the P5. They both will be highly clocked P4's, with the "Tejas" being on the 90 micon (.09mm) process and having a 1000MHz+ bus with all this other crap (hyperthreading, improved bus architchture, etc.)

A processor doesn't necessarily have to be a different core to be considered, the next Pentium, look at the history of the Pentium chips, the Pentium Pro, Pentium II, and Pentium III are all based off of the Pentium Pro core, similarly, I'd expect the Pentium 4, Pentium 5 (Prescott), and Pentium 6 (Tejas) to all be based off the Pentium 4 core (Netburst architecture). Thats not to say their aren't significant improvements, if you look at the differences between a Pentium Pro and Pentium II, the changes are huge and the performance obviously reflects it.

It's probably only until the Pentium 7 (Nehalem) that we'll probably see a entirely new core. From what I hear, Nehalem will probably be introduced in mid-late 2005 so it'll probably compete with the G5 as well.

 

[Zeke]

Originally posted by G5orbust
wow, Id sure love something on the 90 nanometer process.

I think you got confused with that one. Computer chips use the micron (micrometer) for process sizes. The nano meter is much smaller than a micron.

Micron: 10-6 (10 to the negative 6th power)

Nanometer: 10-9 (10 to the negative 9th power)

Actually...90nm = 0.09 microns which is what he said. And new standards are using nm for process sizing because it's simpler than 0.09 micron (since now we're at 2 decimal points in micron).

 

[G5orbust]

Originally posted by Zeke
Actually...90nm = 0.09 microns which is what he said. And new standards are using nm for process sizing because it's simpler than 0.09 micron (since now we're at 2 decimal points in micron).

Wow, youre right. Im sorry for anyone who got confused by my mistake.

It actually is the .09 micron process, or 90 nanometer; likewise with the .13 micron process and so on.

Thanks for that, I sure messed up on that one

I apologize specifically to Freg3000, because I said he was wrong, even though I was the incorrect one.

 

[Catfish_Man]

Originally posted by Cubeboy
A processor doesn't necessarily have to be a different core to be considered, the next Pentium, look at the history of the Pentium chips, the Pentium Pro, Pentium II, and Pentium III are all based off of the Pentium Pro core, similarly, I'd expect the Pentium 4, Pentium 5 (Prescott), and Pentium 6 (Tejas) to all be based off the Pentium 4 core (Netburst architecture). Thats not to say their aren't significant improvements, if you look at the differences between a Pentium Pro and Pentium II, the changes are huge and the performance obviously reflects it.

It's probably only until the Pentium 7 (Nehalem) that we'll probably see a entirely new core. From what I hear, Nehalem will probably be introduced in mid-late 2005 so it'll probably compete with the G5 as well.

OOPS. I was getting Tejas confused with Nehalem. My bad. That makes the 9xx series chances in 2005 look much better to me, although I imagine Tejas will have some pretty cool new features.

 

[Mr. MacPhisto]

Originally posted by Catfish_Man
OOPS. I was getting Tejas confused with Nehalem. My bad. That makes the 9xx series chances in 2005 look much better to me, although I imagine Tejas will have some pretty cool new features.

I'd imagine that the IBM desktop chips will be scaled down POWER5s by 2005, reaching .065 by summer and, possibly, dual cores with hyperthreading. IBM has said they expect to be at 8GHz by 2006 with hyperthreading and dual cores. Could be very exciting.

 

[Freg3000]

Originally posted by G5orbust

I apologize specifically to Freg3000, because I said he was wrong, even though I was the incorrect one.

Don't worry about it. I am surprised I even got it right, I screw it up all the time.