Intel's Plans for 478 & T Socket processors

[MikeC]

This summary of an X-bit Labs news piece was just posted in the front page news. I'm confused. Are you confused? Intel seems to be confused, too.

[Ralf Hutter]

Intel seems to have been confused ever since the initial reports of 100+W Prescotts were announced late last year. A delay in Prescott's release, mediocre to poor benchmarks, high temps and the cancellation of Tejas followed. Obviously they're still scrambling.

[HammerSandwich]

I guess the concerns about high-end Prescotts sucking too much current for 875 mobos have been allayed.

[RaNDoMMAI]

what is intel trying to do with these new processors?

Are they going to be more bang for the buck or something?

~RaNDoM

[shathal]

More money (for Intel) is what they are after - obviously .

Slightly off-topic, I know, but I have seen Intel's "new" Model #'s creeping up (I.e. A Pentium 4 550, 540, and whatever). I had a rant when AMD went to the 1800+ (and so on) ratings, in a marketing move to not look too bad in comparison with P4's MHz lead.

As we all know, The consumer has been educated for years that "More MHz is better" - therefor, Intel better than AMD, on grounds of more MHz.

However, there was SOME understanding to the 1800+ (and so on) ratings, in that they gave an idea how the CPU performed.

What on earth the intel decision-makers have been smoking when they came up with the new rating-system is beyond me. It's aggravatingly counter-intuitive and bloody annoying.

[vortex222]

i think intel is still doing well, they do offer high end product (at high prices), and low end product (at high prices). they also have many very feature rich chipsets to support those prossessors.

But it would look like they dont know what to do for the long term. must meen they have somthing big coming up? (even higher prices)

all i have to say is, as there prossessors get faster, they get much hotter. its becoming rediculess. even the new A64's tend to be warmer then i would like them. even tho they are not that bad i know.

Intel would have been better off just doing a simple dieshrink of northwood. then we would have a decent proformer at much nicer thermal proformance. but intell never does that. they only put more crap on the die to make the dieshrink inefective in making the chip cooler.

at least AMD will do simple dieshrinks of existing product. aka theroughbred, and the upcoming .09 A64's.

presscott is the embodiment of Bloatware in the hardware form.

[HammerSandwich]

Intel would have been better off just doing a simple dieshrink of northwood. then we would have a decent proformer at much nicer thermal proformance.

Such a chip would still have power problems thanks to the current leakage that everyone's seeing at .09.

[smilingcrow]

Just when it looked as if AMD were running away with the award for the most Byzantine CPU product range, Intel stick their nose out with this and reach for the finishing post.
It gives the impression that they don’t know what to do in the short term, so are revamping product lines to keep their staff busy and give them something new to release.
At this rate even Northwood on 90 nm is a possibility.

But simultaneously releasing Sempron on SoA & S754 was a masterstroke by AMD that is hard to beat. And pricing the Sempron S754 1.8 GHz 256k cache, so close in price to the A64 S754 2.0 GHz 512k cache, seems just plain dumb. Makes me wonder whether they really want to sell any of those chips at the moment? Sempron S754 looks like a good future budget chip though, when compared to socket A. And they can turn on 64 bit support relatively quickly I imagine, if Celeron makes that leap. Great, then there’ll be even more model numbers to choose from.

[vortex222]

.09 leakage is acualy not as bad as people think it is. given the fact that Northwood had 55 million transistors, and that Prescott had 130 million.

a simple die shrink of northwood would probibly put its power consumption back to the days of the tualatin p3 or even the coppermine.

Ahh to have cool running chips like that again. the centrino based chips did not have the same problems shrinking to 09 that the p4 did.

the extra hot nature of the prescott is do to the overabondunce of garbage they put on the core.

[shathal]

But without "adding extra garbage", Marketing would have precious little to sell.

And as in most companies (much to my dismay), it's Marketing who get the say, for the better part.

[ceraf]

the extra hot nature of the prescott is do to the overabondunce of garbage they put on the core.

Might be true to some degree, but maybe it's because of the way they designed the shrunk core, using "Strained Silicon". They strain the atoms to create more space between them to let electrons travel faster. I have no idea why they need more power, but Intel has their plans..

[aidanjm2004]

This summary of an X-bit Labs news piece was just posted in the front page news. I'm confused. Are you confused? Intel seems to be confused, too.

I guess it's a good thing that a few Northwoods will be available on Socket T. If you wanna try out that platform's new technology (hi-def audio, PCI-E, DDR2, etc.) you can do it without Prescott. (Although most reviews I've read indicate no dramatic performance improvements are to be gained just yet from the shift to DDR2 and PCI-E.)

[smilingcrow]

I guess it's a good thing that a few Northwoods will be available on Socket T. If you wanna try out that platform's new technology (hi-def audio, PCI-E, DDR2, etc.) you can do it without Prescott. (Although most reviews I've read indicate no dramatic performance improvements are to be gained just yet from the shift to DDR2 and PCI-E.)

It’s typical for the first iteration of a new Intel platform to not offer much in terms of a performance gain. But they’ve out done themselves in the number of new features that they’ve added to the platform. The next revision of chipsets with 1066 FSB and maybe higher frequency DDR-2 support, sounds more compelling. Although you will probably have to use a non-Intel chipset for the later?

But it’s good to see Northwood on S775, although I’m curious to know what Intel’s reasons are for offering it up?

The industry seems to have hit an inflection point, whereby a relatively linear process has started to go non-linear. All bets are off and it’ll be interesting to see how things develop from here. It must be pretty intense inside Intel right now.

[aidanjm2004]

But it’s good to see Northwood on S775

The pin retraction mechanism for socket T is supposedly quite delicate (it breaks easily). If you break a pin or two, the motherboard is useless. Apparently motherboards will be guaranteed for a limited number of insertions (20 insertions is what I've read). This situation seems kind of pathetic.

[smilingcrow]

The pin retraction mechanism for socket T is supposedly quite delicate (it breaks easily). If you break a pin or two, the motherboard is useless. Apparently motherboards will be guaranteed for a limited number of insertions (20 insertions is what I've read). This situation seems kind of pathetic.

The Inquirer has a very informative series of articles on S775, which also addresses the issue that you raise. http://www.theinquirer.net/?article=17756

It seems a little premature to be jumping to conclusions on this matter, as not that many S775 systems have been shipped yet. The average user isn’t likely to find 20 insertions a limitation either. Hardware review sites are more likely to find this sort of thing a problem.

Who knows whether Intel will release figures on whether return rates have increased since they introduced the new socket? I imagine that independent motherboard manufacturers will leak their displeasure, if it becomes an issue. I’ll wait for the dust to settle before jumping to conclusions.

For the record, I’m currently planning to build an Athlon 64 S754 system, so I have no allegiance to S775 or Intel.

[aidanjm2004]

It seems a little premature to be jumping to conclusions on this matter

Where's the fun in all this, if we're not jumping to conclusions?

But seriously, thanks for posting those links on the new socket design, it was interesting to see the reasons why the changes in socket design were necessary. I was intrigued by the author's speculation (in the last article) that a dual core Pentium M based chip will fit into the new socket format.

[Straker]

Who knows whether Intel will release figures on whether return rates have increased since they introduced the new socket? I imagine that independent motherboard manufacturers will leak their displeasure, if it becomes an issue. I’ll wait for the dust to settle before jumping to conclusions.

*increased*? a cynical explanation for the socket change is that something like 30-40% of Intel's returns right now are due to bent CPU pins.

granted, even if this was the entire reason why the new socket is like this, it's still reasonable since motherboards are usually cheaper than CPUs. on the other hand no manufacturer should be taking returns like that in the first place, unless the socket/clip etc is truly defective.

i do really like the way the socket T CPU retention clip looks though.

[vortex222]

It seems a little premature to be jumping to conclusions on this matter

Where's the fun in all this, if we're not jumping to conclusions?

But seriously, thanks for posting those links on the new socket design, it was interesting to see the reasons why the changes in socket design were necessary. I was intrigued by the author's speculation (in the last article) that a dual core Pentium M based chip will fit into the new socket format.

I would love to see Pentium M on a platform that lets it shine. i always maintained that one of the big bottlenecks of the P3 was its GTL protocall. I thought that at one time intel was supposed to increase the p3's FSB to 200 mhz but that never panned, then we got the pentium M, but have yet to see a really kickass platform for it.

if it had the ability to plug into a 925X platform, then we would have a real monster on our hands. and somthing that wont heat my house in the winter.