Intel plots 4MB L2 64-bit x86 desktop CPU

[eden]

<A HREF="http://www.theregister.com/content/51/36314.html" target="_new">http://www.theregister.com/content/51/36314.html</A>

Looks interesting.
I don't think they'll need to toy much around with the Pentium M on Desktop, since already with power saving features, it performs so well. They will just need clock speed.


--
<i>Ede</i>
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>More updates and added sites as over <font color=red>62</font color=red> no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol:

 

[tlaughrey]

I guess Intel's answer to everything now is to just add more cache to make up for the deficiencies in the architecture.

<i>Money talks. Mine always likes to say "goodbye." </i>

 

[eden]

I guess Intel's answer to everything now is to just add more cache to make up for the deficiencies in the architecture.

Ironically, if you modified something in this phrase, it'd just mean the same:
"I guess Intel's answer to everything now is to just add more ca<b>sh</b> to make up for the deficiencies in the architecture."



--
<i>Ede</i>
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>More updates and added sites as over <font color=red>62</font color=red> no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol: <P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 03/17/04 01:29 PM.</EM></FONT></P>

 

[Mephistopheles]

Maybe with P4, but P-M is good architecture. So it's not really deficient...

<i><font color=red>You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete</font color=red> - Buckminster Fuller </i>

 

[tlaughrey]

Maybe with P4, but P-M is good architecture. So it's not really deficient...

That's true. I like the Pentium M a lot better than the P4, although I'm not sure how much more speed they can get out of it.

<i>Money talks. Mine always likes to say "goodbye." </i>

 

[tlaughrey]

I guess Intel's answer to everything now is to just add more ca<b>sh</b> to make up for the deficiencies in the architecture.

Yep, that's equally true ... good one, lol.

<i>Money talks. Mine always likes to say "goodbye." </i>

 

[Xeon]

I guess Intel's answer to everything now is to just add more cache to make up for the deficiencies in the architecture.

Last I checked the P-M's had higher IPC than the K-7 and K-8 line. You be sure to show me when this changed.

Xeon

<font color=orange>Scratch Here To Reveal Prize</font color=orange>

 

[tlaughrey]

You're right. For some reason I was thinking P4 instead of P-M, but this article does say the new processor will be based on the P-M. I like the P-M a lot better than the P4 and hope that Intel continues to go in that direction.

My comment was directed toward the P4 EE and the new Xeons with their increased cache sizes.

<i>Money talks. Mine always likes to say "goodbye." </i>

 

[Xeon]

Oh honest mistake, I figured thats what you meant.

Xeon

<font color=orange>Scratch Here To Reveal Prize</font color=orange>

 

[eden]

I wish they would show more on the Pentium M architecture. While I personally LOVE the NetBurst architecture (sad it turned out so weak per clock), the PM remains a mystery to most, and the whole "P6-P7" fusion with 15 stages approximately, is all weird really. With its powerful core and a great power management functionality, who knows, it may define the future desktop CPU norms in a sense that companies will realize they were wasting power in some CPU areas when unneeded, and that power can easily be brought down. As well, it might scale alot for 15 stages, definitely past the Athlon 64 if the latter has a huge IPC, runs with 10 stages integer and scales excellently.

I always thought there was something deficient with the NetBurst though. 20 stages should offer MUCH more than this.

--
<i>Ede</i>
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>More updates and added sites as over <font color=red>62</font color=red> no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol:

 

[coolsquirtle]

The P4 Architecture is from years ago.... of course its not that great compare to current CPUs. HT is an awesome concept IMO. HT has served me well 😛

they need to get P5 out .......fast.........

RIP Block Heater....HELLO P4~~~~~
120% nVidia Fanboy
never tried to go crazy when it comes to o/cing.
THGC's resident Asian and Graphics Forum's resident nVboy 😀

 

[imgod2u]

I guess Intel's answer to everything now is to just add more cache to make up for the deficiencies in the architecture.

It's weird, a few months ago, the architectural features weren't considered a deficiency when it was the top end processor.
My point is, this "deficiency" as you put it, isn't something experienced only by the P4, but by all architectures. It's an inherent problem when you have your processor clocked at 2+ GHz and memory running at 200MHz. The Athlon had this problem, the Pentium-M has this problem, every desktop/server microprocessor out there has this problem. Would you consider this one of the "deficiencies" of the Athlon/K7 as well?
You are right that Intel's attempts at resolving this problem so far has been just to throw more cache at it (although arguably, SMT is another method to hide latency). And I do prefer the K8's approach better (integrate a memory controller and reduce latency dramatically).
However, each method has its drawbacks. For the K8, it's flexibility. Although in a market dominated by one type of DRAM, this isn't much of a problem, if you wanted to use, say, DDR-2 in your K8, you won't simply have to buy a new motherboard, you'll need a new processor. Not that this isn't the case with other chips, but it's not a technical requirement with say, the P4 or the PPC 970.

Personally, I like the PPC970's approach to memory latency the best. A low-latency, point to point, flexible, high frequency serial bus.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

 

[Mr_Nuke]

I guess Intel's answer to everything now is to just add more ca<b>sh</b> to make up for the deficiencies in the architecture.

Buying an Intel CPU is like <b>gambling</b>.

 

[tlaughrey]

You make some excellent points in reference to my tongue-in-cheek comment. Every CPU design is a compromise since you can't have it all. I also prefer the AMD approach over Intel's MHz-at-all-costs approach. But Intel seems to be redirecting itself with the Pentium M and its potential offspring, so that's encouraging.

<i>Money talks. Mine always likes to say "goodbye." </i>

 

[eden]

I think one of the problems that may have affected our view lately on Intel's approach, is that the AMD processors are proving themselves, even with an "inferior" architecture that the IPC approach seems to work much better, especially since they can still clock it very high, and yet not overheat or need a new process. The P4 is about done at 3.4GHZ and its IPC is simply not doing too well, and it runs at a hefty 85W or so on average, and trying to raise the IPC will simply may it even more power hungry. While the A64s get excellent performance, can be OCed very high, (2.7GHZ)and still maintain excellent temperatures and power usage.

So I am sometimes wondering if the balance approach or the MHZ are nothing but myths and IPC is the way to go, if you followed AMD's approaches with their core.

Ahh, how nice it would be to have a K8 on 20 stages. A tad less IPC, but it would FLY!

--
<i>Ede</i>
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>More updates and added sites as over <font color=red>62</font color=red> no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol:

 

[P4Man]

A few random thoughts.. K8's performance at lower clockspeeds is not only because of its core architecture (being a braniac), I think many people are underestimating how much the ODMC brings. Integrate one (or better: two) on a P4, and you might see some very impressive performance.

>The P4 is about done at 3.4GHZ

I think not. Prescott seems thermally limited, but if (big if) intel can resolve that, it might have quite a bit of headroom. And if not prescott, my guess is Northwood still has a bit of life left in it, even at 130nm. A theoretical 90nm northwood would likely not be as thermally limited as prescott, and should be good for 4 GHz at least. I think 3.6/3.8 is not even out of reach on 130nm.

>Ahh, how nice it would be to have a K8 on 20 stages. A tad
>less IPC, but it would FLY!

What makes you think that ? Nah, I'd rather have some dual core chips, wether it be K8 or (64 bit) dothan based. New dual cored PA risc and sparc chips seem to give some serious performance boosts at relatively low transistor and power consumption costs. Here is my guess: around this time next year, Athlon FX (and opteron) will be dual core, the A64 will remain single core. Now there would be a reason for me to consider shelling out $750 instead of $250 for a cpu.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[IIB]

That's true. I like the Pentium M a lot better than the P4, although I'm not sure how much more speed they can get out of it.

Made in Israel Baby 😉

and "Merom" is also a Hebrew name, so I guess it will be made by Intel Israel aswell

This post is best viewed with common sense enabled