Move over rover or you'll get burned, here comes Prescott

[crees!]

Seems Prescott is out but not available yet. Check out this link from Macbidouille about Intel's new baby. They also have a link to a French site about a review of the processor. You can translate it via altavista.com

Hardmac Article (http://hardmac.com/niouzcontenu.php?date=2004-02-02#1513)

Of interest in the Macbidouille article is that it states the French article says Prescott "is the first 0.09 Microns engraved CPU". MacBD says the PPC 970 Fx is. MacBD says [Prescott] "has two cores, one of which is inactivated, and 64 bits instructions which are inactive as well. ... the loss of performance is 10% to 15% as compared to a former generation P4."

I also find interesting that MacBD reports "moreover, despite their coming to the 0.09 microns engraving, its consumption is no less than 100W! Which may the reason behind the fact the second core isn't active." Sounds like we have a hot puppy on our hands!

And finally, based on MacBD assumptions - "At a given frequency, the Prescott could give away twice as much heat as a PPC 970 Fx."

I haven't fully read and translated the French article but I'm sure it'll pick up for some good reading.

[crees!]

So it is .09 or .13? I guess the real question is did I link this to the correct page?

Pentium 4 EE (Prescott) (http://intel.com/products/desktop/processors/pentium4HTXE/index.htm?iid=ipp_browse+dsktopprocess_p4phtxe&)

Knowing me I have it all ass-backwards and they are two completely different processors which are poorly named to confuse consumers.

[bousozoku]

Prescott are 90 nanometre process chips and somewhat slower than the current P4s.

[abhishekit]

prescott has cache of 1 gb,
and clock speed would go upto 4 gig by the end of the year.
even now 3 gig prescott are available. With the price being the same, its a good deal for windows users..
edit : oops .. 1 mb :D thanks counterfeit

[Counterfit]

Um, a 1GB cache? Right, that's a typo...


ZDNet did a good comparison between the Prescott and Northwood P4's and the Athlon 64 3200+ and 3400+. Seems the Prescott is usually slower than the Northwood by a slim margin, while the Athlon is almost always faster. The new Prescott has 31 (!!!!!! :eek: !!!!!!) pipeline stages, and has an itty bitty L1 cache of 8KB, compared to the 64KB (32 instruction 32 data) of the G4 currently in the PowerBooks. I just wonder how long Intel is going to make their pipeline before we see extra wide processors from them :p

[bousozoku]

They did make room for 1 MB L2 cache. :D Hot data! Hot data! Get your hot data! (The processor is still hot, despite its 90 nm figure.

[Counterfit]

I'm searching for a link to that ZDNet comparison, they had power usage figures, with the Prescott nearing 250W!


okay, this (http://news.com.com/2100-1006_3-5151363.html?tag=nefd_lede) is as close as I can get for now. Checking some logs for the ZDNet one.

[Counterfit]

Found it! (http://reviews.zdnet.co.uk/hardware/processorsmemory/0,39024015,39145079,00.htm) :D

[crees!]

Originally posted by Counterfit
Found it! (http://reviews.zdnet.co.uk/hardware/processorsmemory/0,39024015,39145079,00.htm) :D

$1,000 more for a 10%-15% performance increase.. that doesn't seem like a bang for your buck now does it?

[Hodapp]

I'll sum up every reply this thread will get... "lol anything that isn't made by Apple or IBM sucks."

[bousozoku]

Originally posted by io_burn
I'll sum up every reply this thread will get... "lol anything that isn't made by Apple or IBM sucks."

Well, actually, it's not our replies that should worry Intel. None of the reviews, so far, are more than mildly favourable, if that.

[Flynnstone]

Check Tom's Hardware. They do a test of the new Prescotts. Very lack luster performance.
Good article.

If it was a good processor, Intel would have generated a lot more hoopla. Perhaps Intel and the Prescott designers are a little embarrassed by this one.

The pipelines are over 50% longer (20 to 31) poor move. It sounds like a process problem as well that the power is in the same ball park as the 130nm processors. IBM and the 970FX looks like it has the advantage.

[Krizoitz]

is Prescott? Honestly, it makes me think of east coast country club preppy aristocrats.

[kettle]

The only thin "Prescott" means to me is a "Fat Northern b@stad" with two Jaguar Cars at his beck and call, whilst simultaniously making it very hard for the average person to afford a motor vehicle or use a functioning form of public transport. Oh, and he's the Deputy Prime Minister of Great Britain.

So now we know that the two products have inefficiency in common.:)

[Mav451]

the biggest thing is that Tom's is a HUGE supporter for Intel, and pretty much publishes Intel-biased benchmarks (the Athlon 64 is just "ok", the FX-51 is only "ok").

that's what makes this even more significant...i assume we can expect a williamette effect to occur again (start at 800fsb, moves to 1066fsb along with new process or something, cooler more efficient)...

something must happen down the road. Intel, as much as many people hate them, have too much money, too much to simply lose out on this.

[Makosuke]

If you read the Tom's article, their appraisal is something like this:

The Prescott is a hedge against the future; Intel has a significantly different architecture coming relatively soon (2005--I think it's going to be sort of an Itanium lite, but I'm not sure), but the "old" P4 architecture has obviously totally petered out--it's only incresed a few hundred MHz in the past year and a half.

So, Intel is doing what it's always done: dig out the pipelines ever further to give more headroom for ever-increasing clock speed, and tack on a few more SSE instructions (which according to Tom's in this case don't help much).

At this point they're apparently (speculation on the part of Tom's) having trouble getting decent yields, hence the 3.2 GHz top rating for the Prescott currently, but it should give them room to go up higher before they retire the P4 architecture completely.

In the past, Intel made the mistake of giving changes like this a new name (the PII and P4), when at the same clock rate they were actually slower than the previous generation of chips due to the deep pipelines; the changes eventually paid off with high clock speeds, but looked really bad at first. This time they apparently learned and instead of calling it a P5 just stuck with P4 so as not to embarass themselves.

One more interesting note from Tom's: Thanks to the 90nm process, Intel is apparently able to produce a whole lot of Prescott chips on the cheap; hence, they're currently putting them in what's really the midrange, leaving the P4 Extreme Edition (home Xeon, really) as the high end (and an outrageously expensive one at the 3.4GHz speed, at that). One assumes the Prescott will take over the high end when they eventually manage to bump the clock speed up higher, but at this point it's really just a more affordable midrange chip at the same speeds P4s were already at (and it runs a bit slower than those for the most part).

I personally find it interesting that Intel's mantra has, for the past many years, been "Clock rate = Good". Hence almost all of their design changes have been geared toward kicking up the clock rate at the expense of everything else, mostly through ever deeper pipelines. This worked well for short bursts, but it's left them stuck around 3 GHz for a long time now, and the same thing happened way back in the PI/PII days, too (remember the big 200MHz stall?).

Now that they've actually got a fast chip (the Itanium) that uses some more interesting methods of increasing actual speed, rather than just big clockrates (they're wicked fast at only 1GHz or so), one wonders how Intel is going to try to explain that once that technology becomes the only place they have to go to increase speed versus AMD. Or will they just continue to milk the P4 architecture, regardless of actual speed, and just try to lie through marketing?

[themadchemist]

This is the time for IBM to fervently develop chips so that PPC clearly defeats the Pentium line once and for all. Certainly, there is no worry of a frequency stall on IBM's end at this point: It's just a question of fab.

Anyway, I just see this as a good time to take advantage of Intel's struggling line.

[ksz]

Originally posted by Counterfit
The new Prescott has 31 (!!!!!! :eek: !!!!!!) pipeline stages, and has an itty bitty L1 cache of 8KB, compared to the 64KB (32 instruction 32 data) of the G4 currently in the PowerBooks.
Prescott has a 16KB L1 cache and 1MB L2 cache, both of which are twice the size of Northwood. More technical details here:

http://www.extremetech.com/article2/0,3973,1478686,00.asp

[MorganX]

Originally posted by themadchemist
This is the time for IBM to fervently develop chips so that PPC clearly defeats the Pentium line once and for all. Certainly, there is no worry of a frequency stall on IBM's end at this point: It's just a question of fab.

Anyway, I just see this as a good time to take advantage of Intel's struggling line.

The problem isn't defeating the PIV line with your top processor (dual) system. It's beating the PIV line across all tiers.

A 3.2GHz Celeron in a system costing $599; what you gonna defeat that with?

Prescott is a revamp of the PIV, extra cache is to offset the longer pipelines. So now, top of the line systems can be priced for the mainstream and that architecture can go to 4GHz. Since Intel is not selling this as the P5, it's really a win.

And don't forget Prescott has either dual cores and/or 64-bit extensions. Either or both of those going at 3-4GHz, that's not chump change. Intel isn't where they are by not being ruthless. Just when their competitor's think they've delivered the knockout blow, Intel will deliver a dual-core, 64-bit, or both CPU for $200 at 3-4GHz. Sounds like a solid plan to me. And what's the 64-bit rush? None. There's hardly any software; Windows64 is still beta, you have limited Linux, and no 64-bit MacOS. There's more 64-bit software for Itanic.

So what's the rush? There isn't one.

[mactastic]

Originally posted by io_burn
I'll sum up every reply this thread will get... "lol anything that isn't made by Apple or IBM sucks."

Wow, you are so prescient. You know us 'fanboys' too well. Troll away!

[ShadowHunter]

What is it with Intel's strategy these days? They did this with the first P4s too.

The first P4 core was blazingly hot and performed LESS per clock speed then the P3; a 1.5ghz P4 was SLOWER then a 1ghz P3. Intel countered by ramping up clock speeds asap, and putting the FSB on steroids.

This can only go on so long before it comes back to bite them. It already is in some way, with the Centrino line. "Why should I buy a 1.4Ghz P-M when I can buy a 2.6 P4-M," I hear that one alllll the time.

[Hodapp]

I almost wonder if Intel isn't just selling off their 'test' chips, so to speak, before they get to the 'real batch.' Intel probably knows they'll sell thousands of mediocre Prescotts to [H]ard OCP idiots who spend every dime they make whilst living in their mother's basement on having the latest and greatest names in the computer world in their forum signature.

If I recall correctly, the last chip that came out without a major performance boosting revision was the Pentium 2. I feel sorry for people who have non-hyperthreaded 533mhz FSB P4's.

I have a feeling the 2nd batch of Prescotts will be quite the ass-kickers, whenever they come out.

[hugemullens]

rumor has it that its got 64 bit extensions and dual cores. i really belive the rumor is wrong. Its just a revamped p4 with SSE3, a longer pipeline who's sole purpose is to go faster in mhz. Tejas, the new technology expected for the xeon and pentium line is said to be 64 bit like the opetron. if you check around the internet you will see lots on tejas and what it means for the pentium and if it will wind up killing itanium since it emulates 32 bit, where tejas should run it natively. We know socket 775 is coming from intel sometime this year, along with the BTX form factor, and PCI express. Not a good time to be p4 shopping.

[MorganX]

Originally posted by hugemullens
rumor has it that its got 64 bit extensions and dual cores. i really belive the rumor is wrong. Its just a revamped p4 with SSE3, a longer pipeline who's sole purpose is to go faster in mhz. Tejas, the new technology expected for the xeon and pentium line is said to be 64 bit like the opetron. if you check around the internet you will see lots on tejas and what it means for the pentium and if it will wind up killing itanium since it emulates 32 bit, where tejas should run it natively. We know socket 775 is coming from intel sometime this year, along with the BTX form factor, and PCI express. Not a good time to be p4 shopping.

The rumors are based on photos of the Prescott silicon. It's got more than SS3 and extra cache. It may never be 'enabled' in Prescott, at least not in 478-pin Prescott's, but something more is in there. Way too many transistors as well.

[hugemullens]

I haven't seen a pic of the presscott without the heat spreader off. now i'm wondering.............

[Counterfit]

Originally posted by ksz
Prescott has a 16KB L1 cache and 1MB L2 cache, both of which are twice the size of Northwood. More technical details here:

http://www.extremetech.com/article2/0,3973,1478686,00.asp Hmm, right you are. Still tiny though. *glares at clees! for providing misleading info*

[Hodapp]

Originally posted by hugemullens
I haven't seen a pic of the presscott without the heat spreader off. now i'm wondering.............

If you find one, post it. :)

[Dippo]

Originally posted by io_burn
I'll sum up every reply this thread will get... "lol anything that isn't made by Apple or IBM sucks."

What are you talking about, AMD does make some good processors too!

[MorganX]

Originally posted by hugemullens
I haven't seen a pic of the presscott without the heat spreader off. now i'm wondering.............

[Flynnstone]

Intels using socket 478 now and proceeding to socket 775. That's a difference of 300 pins.
I'm assuming they're moving the memory controller on board.
Anyone know?