4GHz seems to be an elusive target. When do you think we'll see it.

[Hulk]

Since the failure of Netburst to scale as predicted due to thermal limitations we have seen a regression in clock speeds to just below 3GHz.

1GHz was a big milestone but then Intel blew through the 2GHz and 3GHz marks with little fanfare. But hitting 4GHz seems to be much more difficult.

AFAIK there is nothing on the roadmaps showing 4GHz parts right?

When do you think we'll get there? And when we do hit 4GHz do you think it will be right the end of one architecture like with the PIII where it just barely hit it or at the beginning of a new architecture like the P4 when it hit 2GHz and continued to sail on?

 

[allies]

Originally posted by: Hulk
Since the failure of Netburst to scale as predicted due to thermal limitations we have seen a regression in clock speeds to just below 3GHz.

1GHz was a big milestone but then Intel blew through the 2GHz and 3GHz marks with little fanfare. But hitting 4GHz seems to be much more difficult.

AFAIK there is nothing on the roadmaps showing 4GHz parts right?

When do you think we'll get there? And when we do hit 4GHz do you think it will be right the end of one architecture like with the PIII where it just barely hit it or at the beginning of a new architecture like the P4 when it hit 2GHz and continued to sail on?

I'd say single core variants of 45nm intel chips.

 

[Zim]

9/15/2007 at 4:32pm EST

 

[coldpower27]

Originally posted by: Hulk
Since the failure of Netburst to scale as predicted due to thermal limitations we have seen a regression in clock speeds to just below 3GHz.

1GHz was a big milestone but then Intel blew through the 2GHz and 3GHz marks with little fanfare. But hitting 4GHz seems to be much more difficult.

AFAIK there is nothing on the roadmaps showing 4GHz parts right?

When do you think we'll get there? And when we do hit 4GHz do you think it will be right the end of one architecture like with the PIII where it just barely hit it or at the beginning of a new architecture like the P4 when it hit 2GHz and continued to sail on?

The Penryn family will get pretty close, and if not then Nehalem.

 

[Cheex]

Nehalem gets my vote.

 

[caboob]

I could care less about clockspeed. I just want to see an improvement of performance with each passing chip generation.

 

[AnotherGuy]

Yeah it seems like lately AMD and Intel get better performance improvements from the new generations than from clockspeed increases... and these generations are starting to move faster than in the past.... like Intel already has the map layed out within 3 years we have core2duo, penryn, Nehalem... amd the same almost with Am2, barcelona and then 45 nm transition... all supossed to happen within a short period... in three years i think we will still be sitting at low 4ghz clockspeed... and very hard to surpass that barrier...

 

[lopri]

Originally posted by: caboob
I could care less about clockspeed. I just want to see an improvement of performance with each passing chip generation.

QTF

4GHz isn't hard with Netburst. Intel already has a SKU with 3.80GHz. What's another 200MHz?

For example:

4GHz?

Celeron 331 does 4GHz comfortably @stock voltage. (1.312V) I gave a bit of boost for longer stress run with tight memory timing. If 4GHz was a pure goal, it was achieved a long time ago with Celerons. All you need to do is to cut a huge chunk of CPU cache, which in turn reduces heat dissipiation and lower the stress on each transistors. But who wants a chip @4GHz that performs like ****?

 

[Hulk]

Originally posted by: caboob
I could care less about clockspeed. I just want to see an improvement of performance with each passing chip generation.

Really? So I guess you buy the slowest clocked version of the most efficient chip out there and DON'T overclock it either?

 

[Cogman]

Originally posted by: Hulk
Really? So I guess you buy the slowest clocked version of the most efficient chip out there and DON'T overclock it either?

The fact that clockspeed is part of preformance is not what he was disputing, he we disputing the push for high clock speeds. We know that Overclocking provides preformance increases, but if I had a 1 GHZ cpu that preformed just as well as a 5 GHZ oc part that you have, guess which part I would choose? Thats right, the 1 GHZ CPU. I would probibly try and overclock it as well to get speeds faster then your 5 GHZ CPU.

If we never have a company release a 4 GHZ CPU I would not be all that depressed (heck, it would not phase me at all)

 

[Hulk]

I know. I was just being ornery.

My point is that we all know that both IPC and frequency are important in determing the overall performance of a CPU.

IPC is great but you've gotta have clockspeed as well. If C2D topped out at 2GHz I don't think it would be in many desktops.

And we all know that a 3.8GHz P4 isn't a great performer.

I'm wondering if we're going to remain at sub 4GHz frequencies and IPC and core count will increase to improve performance or it we will see another run in frequency like we saw a few years ago.

 

[Cogman]

Thats the way it looks right now. 4 GHZ just has not been as important as Multi Cores and IPC improvements to AMD/Intel. Sub-4GHZ is adiqute, and probably will be for the next 10 years.

 

[orion23]

It's all about cooling, quiet, power use andperformance.

I think that once a NEW type of cooling is created, or when water / phase becomes user friendly, then we will move onto faster cpu's, quads or not.

@ this point just like in the past, it is the cooling method that needs to be solved first!

 

[Hulk]

Process technology improvements and shrinks will probably allow for faster frequencies without such high current and therefore heat output.

 

[Cheex]

Intel should just change their HSF design to a heatpipe tower style and then release "E6300's" at 3GHz and "X6800's" at 4GHz...There you have it...

 

[magreen]

Originally posted by: Cogman
Thats the way it looks right now. 4 GHZ just has not been as important as Multi Cores and IPC improvements to AMD/Intel. Sub-4GHZ is adiqute, and probably will be for the next 10 years.

Yes to IPC, no to Multi Cores. Hardware makers are probably pushing Multi Cores specifically because they're running into limitations producing faster single cores.