Why Prescott?

[Crashman]

OK, the Northwood is at 3.4GHz. Tejas is next year. Prescott was given a bunch of extra stages so it could clock higher...wait!

Northwood is at 3.4GHz. Don't give me any BS about putting an old design on a smaller process, we all know fixing those issues is still FAR less expensive than developing a new core.

Once 90nm is mostly debugged (by the end of the year), it will allow processors to clock 50% higher than 130nm. That would have taken the elderly Northwood core to 5GHz.

Wait, wasn't 5GHz the limit for Prescott? What's going on here?

Intel could have moved Northwood to 90nm and fixed the bugs just as easily as fixing the manufacturing problems with the Prescott design. So what's the point?

Intel COULD have kept the Northwood until Tejas. This shouldn't be hard to see. They WILL have a pinless Northwood P4EE for the new socket. So it's not a socket issue either. So why the heck did they even bother developing the Prescott?

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>

 

[P4Man]

>Once 90nm is mostly debugged (by the end of the year), it
>will allow processors to clock 50% higher than 130nm. That
>would have taken the elderly Northwood core to 5GHz.

That is what you assume, doesnt make it true though. A die shrink really doesnt guarantee a 50% clock increase, other things could be holding the core back. That being said, I think everyone agrees its likely a 90nm northwood, especially with a 1Mb cache would have been a better prescott than prescott.

>Intel could have moved Northwood to 90nm and fixed the bugs
>just as easily as fixing the manufacturing problems with
>the Prescott design. So what's the point?

Point is, Prescott was obviously meant to be more than what we have seen so far. More than 40 million transistors that are unaccounted for. They have been put in there for a reason. My guess is 64 bit extentions (not compatible with AMD), Lagrande and maybe dynamic multithreading. Since MS will not support another 64 bit extention, its quite likely Intel is changing the design to make them compatible, and the DMT might not have been working yet, just like HT on initial P4's.

So, while Prescott as it is, may not offer anything over a 90nm Nortwhood, think about it as a developement platform, and over time, the features mentioned above will be fixed while the design allows for clockspeed increases that are beyond those of NW. This is very much a typical intel strategy.

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

 

[endyen]

It was a long time between 3.2 and 3.4. The die shrink didn't go well. Maybe a straight dieshrink yielded no more than 3.4. Who needs conspiracy theory when you have reality.

 

[juin]

Even what you just say dont make 40 million transistor.They must have some thing big there.My guess is prescoot will not be debug.They will just take the hit and prepare tejas release.

Just to show dad

 

[CaptainNemo]

I'm baffled by Prescott too, especially with the 3.4 Ghz Northwood now available at competitive prices (i.e. about the same as an A64 3400).

Intel users will just have to wait and see (and upgrade their Northwoods if need be).

If a tree falls on coop, but noone is there to hear it - do less people rejoice?

 

[Mephistopheles]

I agree with P4Man on this one: I think Prescott was supposed to be a killer chip. And I really mean... a killer chip. It's got 2.5 times the logic(al transistors) that Northwood has.

Most likely, then, Scotty is a crippled beast. It's huge, big, a conundrum in itself... But as of now, it's huge, big and broken.

It would have been extremely stupid, however, if all things that are in place could not be activated at all at some point (after debugging). In the best case scenario, Prescott's hidden secrets will be activated gradually within the next several months... In the worst case scenario, Prescott's hidden secrets will only be activated by the time the next major core revision takes place (i.e., Tejas... and 2005)

So, while Prescott as it is, may not offer anything over a 90nm Nortwhood, think about it as a developement platform, and over time, the features mentioned above will be fixed while the design allows for clockspeed increases that are beyond those of NW. This is very much a typical intel strategy.

These are my thoughts too. While this still doesn't mean that prescott is interesting from the buyer's perspective, it represents a long-term strategy. I think it would be impressive to see prescott doing all it's supposed to do with its 125 million transistors and clocked at 4Ghz; it won't make us long for the times of northwoods at all.

Meanwhile, prescott seems to be a skunkworks experiment in development... Someone ought to get Intel's act together in once piece!

 

[redface]

Let's hope everything gets fixed.

A fine day!

 

[Mephistopheles]

Exactly.

<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>

 

[P4Man]

> I think Prescott was supposed to be a killer chip. And I
>really mean... a killer chip.

I kinda doubt it. Intel *never* hyped Precott, not even years ago, while they used to hype upcoming "killer chips" like the original Pentium, the Pentium II Klamath, the original P4, etc. Prescott IMHO has always been more an evolutionary design than a revolutionary one. So don't be too quick to agree with me

As for fixing whatever is wrong with it; it seems the priorities are obvious:
1) power consumption
2) power consumption
3) power consumption
4) power consumption
5) enabling 64 bit

1-4 might be fixed or at least improved this year, but anything else will have to wait until Tejas IMHO. 12 months just isnt enough for major redesigns, and if it where only a minor glitch, they would have solved it by now. Prescott has been under development for about 5 years, if they didnt get it working in that time, I doubt they will in the next 6 months,let alone it would be worth the trouble.

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

 

[Whisper]

Intel could have moved Northwood to 90nm and fixed the bugs just as easily as fixing the manufacturing problems with the Prescott design. So what's the point?

Not entirely true. A Northwood on 90 nm technology would not be able to clock much higher. One of the problems is that the core just won't allow stable clock distribution at these high frequencies. So a new core was needed anyway, with extra pipeline stages, which in turn forced them to introduce other improvements. They now also have a new program to optimize the position of transitors and their interconnections: <A HREF="http://www.geek.com/images/procspecs/p4/p4-13die.jpg" target="_new">Northwood</A> vs. <A HREF="http://images.anandtech.com/reviews/cpu/intel/pentium4/prescott/die.jpg" target="_new">Prescott</A>. This means that every new spin will bring an improved layout. Also, with the new socket they will have even more freedom to position transistors optimally. So Prescott has a lot more future. This is also a big step forward towards Tejas (which <i>will</i> run cool) and 65 nm technology. Only not very succesful as an intermediate step...

 

[addiarmadar]

Prescott is just a stupid marketing scamn to get you to waste more money for the same thing. Your northwood chips are just as good.

Barton 2500+ @ 2200mhz (10x220 vcore @ 1.8)
Asus A7N8X Dlx 440 FSB
1gb Geil GD pc3500 Dual Channel (2-3-3-6)
Segata 80gb SATA 8.5ms seek
ATI Radeon 9800 Pro(420/720)

 

[juin]

The new clock distribution could have been done on a improve NW

Just to show dad

 

[Whisper]

The new clock distribution could have been done on a improve NW

Prescott <b>is</b> an improved Northwood.

 

[CaptainNemo]

Yes, but it is kinda stillborn atm.

I don't want to get involved in any silly fanboyism, but there is very little reason for buying one until it is sorted.

If a tree falls on coop, but noone is there to hear it - do less people rejoice?

 

[juin]

No it layout have nothing to do with NW.

Just to show dad

 

[Whisper]

I don't want to get involved in any silly fanboyism, but there is very little reason for buying one until it is sorted.

I fully agree actually. I just wanted to make clear that Prescott and its direct successors have a lot more future than Athlon 64's right now. It's just too early to entirely label it as crap, just like Wilamette turned out to become the very succesful Northwood chip, and Athlon 64 didn't directly look very promising at first either. I can only agree that, for the moment, Athlon 64's get you great value at their current prices.

 

[Whisper]

No it layout have nothing to do with NW.

That's what I've been telling. The layout has been reworked by the optimizer. But it's undenyable that they used an architecture quite close to Northwood as a starting point. So I think it's not wrong to call it an improved Northwood.

 

[P4Man]

> fully agree actually. I just wanted to make clear that
>Prescott and its direct successors have a lot more future
>than Athlon 64's right now.

That is so ironic. If intel releases a crap chip, it has "more future" than better, existing, shipping AMD chips.. a way of looking at the facts I guess :/ One thing is for sure, it will *need* a better future if intel wants to continue competing in the high end.

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

 

[Mephistopheles]

Just had to say that... I agree. Don't label Prescott as crap right now, even if 2.8E, 3.0E and 3.2E don't exactly shine at all.

<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>

 

[P4Man]

>Also, with the new socket they will have even more freedom
>to position transistors optimally. So Prescott has a lot
>more future.

Wow.. bright future there. I'm sure better positioned transistors will solve all the issue.. not..

>This is also a big step forward towards Tejas

I would call it a big step backwards from Northwood for now. If that is the direction things wil go towards Tejas, intel is in a lot trouble.

>(which will run cool)

Got a link ? Or is it just a pipedream ?

>and 65 nm technology

If 90nm SS->65nm SS (without SOI) is anything like 130 bulk ->90nm SS.. well.. some hot prospects there. Being an optimist is good, but assuming intel will do everything right from now on, and AMD will falter may not give you the most accurate predictions...

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

 

[P4Man]

So they are crap. May or may not get better in the future, but for now the shipping products are crap, period.

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

 

[Zoron]

If intel releases a crap chip, it has "more future" than better, existing, shipping AMD chips.. a way of looking at the facts I guess :/

Hasn't this been happening since the days of the Pentium? AMD released better chips... yet Intel was the one making money...

<font color=red> If you design software that is fool-proof, only a fool will want to use it. </font color=red>

 

[Spitfire_x86]

Making money is a completely different thing. Most of the PC buyers are n00b and they'll keep Intel market leader and profitable

----------------
<b><A HREF="http://geocities.com/spitfire_x86" target="_new">My Website</A></b>

<b><A HREF="http://geocities.com/spitfire_x86/myrig.html" target="_new">My Rig & 3DMark score</A></b>

 

[P4Man]

In the Pentium days, AMD really didnt have a better processor (you mean the 486 ? or that 5x86 hybrid ?). But they did make money (oh irony).

In the early Athlon days, they had a slightly better processor (than P3), and if memory serves, they made some money. Not much, since basically no big oem backed the chip.

Today, they have (IMHO) a superior platform, much more than they had in the Athlon vs P3 days, they are getting massive oem and ISV backing (at least, for AMD standards), especially in the server/ws market, and they are making money (*). Unlike in the Athlon or Pentium days, however, I don't see intel being able to strike back in short order. Intel fans hope intel will reduce power consumption, hope Tejas will be everything prescott is not, but fact is, in spite of a new core, and a new process, they are barely competitive with AMDs "old" kit. This will give AMD at least a year, and an important one at that. For 2005 and beyond, all bets are off, but if Intel doesnt implement SOI or a similar technology, I see them even getting further behind, instead of regaining the lead in the desktop market. If they don't implement a cpu to cpu interconnect, I think they will have a very hard time competing in the servermarket, and if they don't implement a 64 bit extention, they'll have trouble in both. IOW, intel basically has to clone K8, or focus on the mobile market, and let AMD win the benchmarks and gain marketshare everywhere else.

Anyway, i could be dead wrong, but last time I thought all this, I bought AMD shares, which have increased by 5x since then. I would not be surprised to see them double again over the next year.

(*) don't get me wrong, I don't want to make it sound like earning money is a frequent thing for AMD though

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

 

[Crashman]

I'm not so upset about a minor drop in per clock performance, after all, my P4 2.6C at 3.2GHz is still 2.5x as fast as my PIII 1000EB was. But the fact is, the Prescott isn't meant to clock that high, 5GHz at most. It just seems that Intel put way too much into it when it can't even double the clock speed of the Northwood.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>