Intel Announces 90nm Prescott P4s

[Cyan]

Users willing to upgrade their socket 478 systems are left out in the rain due to the uncertain compatibility situation and last but not least Prescott is badly overclockable, which made us skipping the overclocking tests. So there is only one argument left that makes Prescott really attractive: It is a new technology and it is not more expensive than Northwood.

Do you believe buying a Prescott processor is a future-proof investment? You really shouldn't, because that is what Intel wants us to believe, as the real generation and architecture change to socket 775, DDRII memory and PCI Express are few months away yet. within a few months range.



http://www20.tomshardware.com/cpu/20040201/index.html

[zybch]

I can't believe that the pipeline length has been increased. This means even higher latencies if the branch predictor makes a mistake, even with hyper threading.
Why won't they reduce the length but add a couple more pipes?!

I guess it will help increase the benchmark numbers with audio/video encoding so that Intel can say that have the fastest CPU in that field.

[Someone Stupid]

31 pipelines? That isn't a good sign at all. Those add-ons for hyperthreading are **** well going to be needed.

[iceblue]

I'll be waiting to see how it performs under the Grantsdale Chipset. :cool:

[causticVapor]

Folks, there's no point to this chip other than more cores/wafer. Intel has done modifications to the core so that it at least has BALLPARK performance figures when compared with the northwood. Intel was forced to go to 31 pipeline stages to at least get these things to Northwood speeds and to tighten up the branch predictor and add more cache to lessen the deleterious effects of an (even further) stretched pipeline.

The mere fact that these cores run on average 10C hotter than northwoods and consume more power is evidence enough to show how much trouble the 90nm process shrink has been for intel. That and the long delay to get yields to reasonable levels.

Prescott shows slight improvement over Northwood in some areas, but is slightly slower in most benchmarks. Intel is lucky to have a processor with 33% more pipeline stages and still keep the performance of the core reasonable.

There have been rumors that prescott has Yamhill (64-bit extensions) and is a dual-core. The former theory has many experts with conclusive evidence backing it up. The latter simply doesn't logically make sense - why would the die size be so small if it were a dual core? And why would intel still POWER one core even if it is not doing anything? Either that or prescott is a real PI$$ performer and two cores are actually doing the work to bring performance up to northwood levels. But I doubt it.

All in all - best to do what Ed at overclockers.com says... sleep till 2005. :)

I'm sleeping. Are you? )-|

[Someone Stupid]

Yes I am.

[zybch]

I'd love to know when Intel will give up on their 'Megahertz at any cost' mantra.

As the new prescot shows, to get the MHz increased they needed a longer pipeline.

To keep the pipeline fed they needed to double the cache.

To do both the above things, they have to increase the heat by around 10C from the Northwood range of chips.

And even then, it still runs like a legless dog.


The problem is the whole 'more MHz = better performance' thing that they still stick with, even though an identically clocked Athlon (or even duron) will beat its P4 competitor in (probably) all benchmarks.

What will we see in a years time? The new Intel 5MHz CPU, with an 80 stage pipeline, 16Mb of cache, running at 50C over ambient and still no faster than the current Northwood 3.2GHz parts?

[causticVapor]

If they get these leakage problems fixed, it could do a whole world of good.

"What will we see in a years time? The new Intel 5MHz CPU, with an 80 stage pipeline, 16Mb of cache, running at 50C over ambient and still no faster than the current Northwood 3.2GHz parts?"

If not, then basically what is said above holds true. :mad:

[Beeblequix]

I'm planning on building a modest P4 system real soon.

So what should a person like me do? Do I go with my 2.8C P4 setup or opt for the Prescott?

ß

[iceblue]

Only opt for a Prescott if you are the serious overclocker.. otherwise, at stock speeds it won't be much faster than a Northwood.

[Someone Stupid]

See www.overclockers.com

The Prescott should be avoided like the plague. A 3.6C Northwoord (overclocked of course) they estimate would be running close the the Prescott's upper end limits of 4GHZ. You also would need extreme cooling to push a prescott harder. Watercooling only pretty much, and those guys can do more with a Northwood than a Prescott most likely anyhow. So the Northwood C core easily over the Prescott. More work per cycle (never though I'd say that about Intel). A 2.8C is faster than a 2.8 Prescott from some benches I've seen.

[RamonGTP]

Originally posted by iceblue
Only opt for a Prescott if you are the serious overclocker.. otherwise, at stock speeds it won't be much faster than a Northwood.

According to TOM's, the prescott is a very poor overclocker.

[iceblue]

Who to trust.. Anandtech or TOM's? ;)
Prescott's enhancements actually give it a steeper increase in performance per increase in clock. Not only can Prescott be clocked higher than Northwood, but as its clock speed is increased, it will start to outperform similarly clocked Northwood CPUs.

[Someone Stupid]

Why would a chip do more work per cycle faster is what I'm wondering from Andandtech's "review." It runs slow at slow speeds but once you've ramped the speeds up to past we are now, it gets faster. Trust me. Sounds like a used car salesman.

[Vampiel]

Originally posted by iceblue
Who to trust.. Anandtech or TOM's? ;)

ICEBLUE, you misunderstand there quote. They were not talking about overclocking. They meant that the prescott "Not only can Prescott be clocked higher than Northwood" can be clocked higher, not overclocked higher.

It looks like most of you have the prescott wrong. The extra pipelines slow it down at lower clock speeds, but as the mhz increases the 31 pipelines helps the speed. Thats what gives it "but as its clock speed is increased, it will start to outperform similarly clocked Northwood CPUs". So in other words the higher the prescott is clocked the more it will outperform a northwood, BUT it is not very overclockable from the stock speed. They are also going to increase the bus speed, add ddr2 and pci express, so if you care going to buy an intel system I would wait till later this year when the prescott clock speeds are higher and they add the new bus,pci,ddr2 suport.

[zybch]

A similar thing happened with the P4. Tye first few speed grades were slower than the previous P-III chips, but as the MHz increased we saw a speed increase that equalled more % than just the MHz increase.
Even so, Intel must have known that it would receive a battering (or even deep frying) about the initial speeds which begs the question, why release a 3.2 part when a 3.6 is probably only a month or 3 away that would let the prescot start showing its potential?
Its a rush job to try to keep customers from going AMD64.

[Someone Stupid]

They want to fab as few different cores as possible. If they can introduce the Prescott core and push away the Northwood core (which will take time, but the sooner they start, the sooner they can end the line) they save money. They also can tout that the current P4's do all these new things (thanks to added instructions). It won't be a very overclockable chip. It was meant to become a bridge btw northwood and tejas over about 2 years, now that is looking like 1 year or so. It is a temporary solution to a permanent problem for Intel. If they are to sell by megahertz, they have to keep ramping up speeds and Prescott is how. Who cares if it does less work per cycle if your in marketing for Intel? With DDR2 that should be the true test of this chip on socket T. Maybe it needs the extra bandwidth DDR2 provides to run optimally, but I doubt it making that much of a difference. Maybe enough to compensate for the longer pipeline to the Northwood, but still a pain in the **** to overclock thanks to small die size and high power consumption.

Intel isn't worried about the AMD 64, AMD isn't capitilizing off it but instead trying to play with the big boy on the playground and it isn't working. They aren't getting out what is needed soon enough to support it, not to mention they also don't have desktop boards for them, still short DDR2, and are they behaving in a way to really push away any large OEM buyers (which is where the money is, not us). AMD doesn't know how to market either, they have to market to get anywhere and they really aren't doing that except to people who already know about computers, so it is wasted ad money that could be used to educate others so the megahertz myth is shattered to Joe Sixpack. Intel has little to worry about AMD as they have a 64 bit processor and have hinted that they are working on another one for desktops (so Itanium would stay for servers, and Itanium sales are up not down for the last reported quarter). AMD's management really needs to get it's act together (well it has been needing to do that for years :rolleyes: ).

[Vampiel]

Originally posted by Someone Stupid
AMD's management really needs to get it's act together (well it has been needing to do that for years :rolleyes: ).

I completely disagree with this entire post. First of all you need to get your facts straight. (from anantech)

"Prescott's enhancements actually give it a steeper increase in performance per increase in clock."

You said it doesnt get more from each clock.

2nd of all you said all this stuff about AMD trying to play with the big boy and its not working, and how there managment needs to get there act together. No matter what you may think about intel not sweating AMD64 (yea right why do you think they came out with P4E? there sweating there balls off about amd64) the bottom line is money.

If you want to know how much AMD sales are up over the past year read any article on it. Tell me what it says, 88% SALES INCREASE, wow management you need to get your act together cause your not doing to well against intel (even though intels server market is shrinking not gaining)

read some facts before you post plz

[iceblue]

If the Prescott really can't overclock well as you say, explain how one review got their 3.2 to 4GHz+ speeds with little effort.

[Vampiel]

I only know from what I read, I havent actually tested one myself.
THG says its not really overclockable

http://www.tomshardware.com/cpu/20040201/prescott-30.html

Although I did just read the whole anantech article and it does appear that it says its easier to OC than a Northwood. So im a bit confused here. Either way they both end up with the same general conclusion....

"If you find yourself using Microsoft Office for most of your tasks and if you’re a gamer the decision is clear: the Athlon 64 is for you. The Pentium 4 continues to hold advantages in content creation applications, 3D rendering and media encoding; if we just described how you use your computer then the Pentium 4 is for you"
http://www.anandtech.com/cpu/showdoc.html?i=1956&p=25

[iceblue]

Surprisingly there are still people out there that take THG seriously..
and to add to that, in their latest article they claim to have a GeForce FX 5900 Pro :rolleyes:

[Vampiel]

I dont know why you dont take THG seriously. There awards are a big deal for man. otherwise they wouldnt send them to THG for testing.

What about a GF FX 5900 ultra? You didnt know about those, you can allready get them.

http://www.axiontech.com/prdt.php?src=PW&item=45911

THG is very reliable and not biased like alot of review web sites. Im surprised there are ppl that dont take THG seriously.

[zybch]

Every review site is biased to one degree or another.
THG seems a little biased towards AMD, but they have been much more even handed lately.
The sad truth is that you can prove any result you want by the choice of benchmarks that you perform. There still isn't any one benchmark package that is scrupulously fair handed.

Just take a look at the recent kerfuffle with Sysmark 2002 and how almost all the tests that favoured AMD were removed and replaced with tests that just about all favoured the long pipeline of Intel's P4 chip. Not quite so strange when you consider that Intel pays bapco lots of $$ whereas AMD has ony started doing to since.

[iceblue]

What about a GF FX 5900 ultra? You didnt know about those, you can allready get them.

http://www.axiontech.com/prdt.php?src=PW&item=45911

Seriously now, did you really think I didn't know there was a 5900 Ultra? Why do you think I specifically pointed out that TGH had a 5900 "Pro". Just to let you know, there is no 5900 Pro.

[Vampiel]

Originally posted by iceblue
Surprisingly there are still people out there that take THG seriously..
and to add to that, in their latest article they claim to have a GeForce FX 5900 Pro :rolleyes:

No were on THG does it refer to a 5900 pro. Post a link so I can see it b/c I cant find it anywere.

[iceblue]

Originally posted by Vampiel
No were on THG does it refer to a 5900 pro. Post a link so I can see it b/c I cant find it anywere.
It's in the link you posted. )-|

[Vampiel]

OK I see what you are looking at. That is a misprint that I can verify from the sentence of "There are several things that we changed since our last CPU article. The most important change is the upgraded graphics card. Instead of the GeForce FX 5900 Pro".
Obviously someone wasnt thinking when they typed that cause when you go back to "our last CPU article" They are clearly using a "MSI FX5900U-VTD256
GPU: NVIDIA GeForce FX 5900 Ultra"

Good eye though, but do a little research and you can find that someone printed the wrong word "Pro" instead of "Ultra" maybe you should be there editor.:D

[causticVapor]

Originally posted by Vampiel
I dont know why you dont take THG seriously. There awards are a big deal for man. otherwise they wouldnt send them to THG for testing.

What about a GF FX 5900 ultra? You didnt know about those, you can allready get them.

http://www.axiontech.com/prdt.php?src=PW&item=45911

THG is very reliable and not biased like alot of review web sites. Im surprised there are ppl that dont take THG seriously.

ROFLMAO

[Vampiel]

okok no tester is completly unbiased.:rolleyes:

but I would like to know why (if any) of you dont take THG seriously. I read it all the time, and it always seems pretty accurate. Ive never read any SUN articles in it with BATMAN BABIES.

[porsch1909]

intel have recognised that they need an affordable high performance CPU to combat the AMD 64. i refer to myself as the intel guardian angel. i always stick up for intel. in my opinion intel dont want to release their 64 bit CPU yet because windows doesnt support it. so they will wait till the window 64 bit edition is out. in the mean time they probably thought the extreme edition could manage in the front line of battle and the EE would win the battle if it wasnt to expensive, so the affordable prescott is brought out in the meantime. its all marketing strategy, they dont care about how good a product is just as long as it sells and on the marketing front intel are kicking ****. they have 80% market share AMD and 20%. the EE is affordable in the sense that its cheaper than the AMD64. this is all opinon anybody agree and please AMD lover dont bite my head off.