The fuss about Prescott ?

[benjamat]

The boards are full of people decrying the new prescott cpu saying it is rubbish compared with the northwood so I thought I wouls relate my experience.

I wanted to update my old P4 2.4 533 as I have a P4P800 Asus mobo and obviously needed an 800 fsb cpu.

Looking on the web ( I live in UK ) I found Watford Savastore had Both Northwood and Prescott 2.8 cpus at the same price so I thought what the hell, I know there is all this talk but at the end of the day Prescott is the new chip.

It duly arrived and I stuck it in with some trepidation, my old P4 2.4 533 was happily running a stock 30% o|c on my P4P800D.

Stuck the settings to 30% and nothing happened, would not even boot, Oh S--t I thought, Ive made a mistake here.

Went to manual O\C and set the fsb at 250 and my 2700 memory at 3200 with MAM and HT enabled. Booted perfectly into windows and ran all evening with UT 2004, Call of duty and Command and conquer, no problems at all.

I have a chip that cost the same as a northwood 2.8 but is latest technology, it runs happily at 1000 fsb with a 25%o\c which by any standards is a solid achievement and I have no doubt that I could run an even faster setup with faster memory as I am sure it is the memory stopping me getting the simple 30% on the P4P800.

 

[Zebo]

There's nothing wrong with it per se. It's just at exact same MHZ/BUS the northwood is about the same (worse in some better in others). I guess people expect more with a new chip. But they lengthend the pipeline, again, for higher clock which sells. To some it's a step back to me it makes sense and I can't fiqure out why AMD is'nt doing the same.



Good read

 

[Lonyo]

Originally posted by: Zebo
There's nothing wrong with it per se. It's just at exact same MHZ/BUS the northwood is about the same (worse in some better in others). I guess people expect more with a new chip. But they lengthend the pipeline, again, for higher clock which sells. To some it's a step back to me it makes sense and I can't fiqure out why AMD is'nt doing the same.



Good read

The AMD CPU's work in a different way, lengthening the pipeline would probably really hit performance and make it sink, unless they pretty much 80% redesigned the basics of the processor and made it more like the P4's.

Prescott's run hotter than Northwood's, but also seem to overclock better (if you can get some good RAM/mobo et) but most 2.8c's seem to hit 3.5GHz+ and usually seem cheaper and run cooler, and that's fine for most people.

 

[Zebo]

You said it redesign or pretty soon AMD will still be pushing 2700 Mhz chips while Intel is pushing 6000Mhz. I don't care what PR ratings and the reviewers say about parity no one will believe it.

 

[McArra]

Originally posted by: Zebo
You said it redesign or pretty soon AMD will still be pushing 2700 Mhz chips while Intel is pushing 6000Mhz. I don't care what PR ratings and the reviewers say about parity no one will believe it.

I like a more intelligent design better than a XXXXXghz system. Is like saying I like a Buggati EB16.4 Veyron 1001hp better than an F1 because it has more hps, when any F1 (even a minardi (800hp?)) would blow the Buggati because it weights a lot less and has a more intelligent and efficient design.

 

[LED]

The Prescott Core starts to shine over the Northwood when the speed hits above 3.4Ghz and Intel knows this, hence the release. The downside is the heat it produces which will be corrected with a release of a new stepping (D0?) and 4GHz on air will be easily achieved ao I doubt if any Northwood will be able to touch it.
Complaints that I have read is that older Mobo's Caps can not handle the Power Hugry square although Intel states it will with a BIOS update.

 

[mastertech01]

Overclocking Press Hot is the problem. Hopefully that will change soon.

 

[DAPUNISHER]

Originally posted by: Zebo
You said it redesign or pretty soon AMD will still be pushing 2700 Mhz chips while Intel is pushing 6000Mhz. I don't care what PR ratings and the reviewers say about parity no one will believe it.

AMD will have CPUs based on the 90nm SOI process in the 2nd half this year, and they did add a couple stages to the pipe so clockspeeds should get a nice bump once the process matures. I should think overall performance should continue to remain close between the two companies for some time to come.

I'm guessing you are intentionally being a bit facetious about the clockspeed discrepency, but understand your argument concerning the public just not understanding that it isn't the key indicator to performance. However, I do think you are failing to consider several factors that may substantially negate much of the need for concern over that lack of comprehension on the consumers' part. One factor is that Intel will be switching to a number system to indentify microprocessors and not promoting clockspeed as the way to judge performance if what is being reported is correct.The other factor I feel is pertinent is that Intel is now producing a line of CPUs that are incredibly powerful but don't have a high clockspeed, much like AMD. I've read that this has been a issue in getting non-corporate consumers to embrace the P-M. For instance, I watched the lady selling the new GateWay centrino tablet *it looks sweet* on HSN the other night and she continued to re-iterate that although it was a 1.4ghz CPU that it performed more like a 2.8ghz CPU *she didn't state which one * I wouldn't be at all surprised if her emphasis about the performance belying the clockspeed resulted in quite a few question about that statement from callers.

My point is that I believe clockspeed will cease to be a major handicap as it appears that Intel now has a practical reason for obscuring clockspeed in favor of a performance rating system like their numbering system will provide. Being able to label a P-M and a much higher clocked P4-M with the same series number for performance would help dispell the doubts consumers have about it's performance vs the P4-M wouldn't it? Certainly it'll take time for consumers to adapt but I'm certain they have a transitional marketing strategy *Intel's marketing division is damned good* that will help indoctrinate the general public.

 

[FelixDeCat]

Originally posted by: mastertech01
Overclocking Press Hot is the problem. Hopefully that will change soon.

No problems here - 2.8@3.5, 250fsb, 1:1 500.

 

[ntrights]

Originally posted by: LED
The Prescott Core starts to shine over the Northwood when the speed hits above 3.4Ghz and Intel knows this, hence the release. The downside is the heat it produces which will be corrected with a release of a new stepping (D0?) and 4GHz on air will be easily achieved ao I doubt if any Northwood will be able to touch it.
Complaints that I have read is that older Mobo's Caps can not handle the Power Hugry square although Intel states it will with a BIOS update.

Excactly, I've been waiting for the 3.0E D0 stepping myself. I would say that it's around 3.8 that Prescott pulls away, ~3.6 its just about equal, little faster in some benches and a little slower in others. Around 4 - 4.2 thats where prescott should be able to runn in current socket 478 with air cooling thats what im aiming for anyway with my current mobo and my zalman heatsinkfan.

 

[LouPoir]

In my experience, the Prescott overclocks as well as a Northwood, just runs hotter and does not perform as well as a Northwood. I had a 3.0e that I was able to compare to a 3.0b and that was my conclusion.

IMHO

Lou

 

[redpriest_]

Here's some 'scaling numbers' for Prescott compared to Northwood at 4.2 ghz. Er, remind me, which speed was Prescott supposed to "absolutely blow away" Northwood?

http://www.xbitlabs.com/articles/other/display/vapochill_11.html
Xbit labs review of 4.2 ghz processors - northwood/gallatin/prescott

 

[Soulkeeper]

yeah my prescott is kinda weak when compared to a P4c oc'd or not
at 3.5ghz my prescott benchmarks like a 3.2 p4C
at 3.2ghz it performs like a 2.8

it seems that heat is the only think keep my oc from going over 3.4ghz stabily
idle temps are regularly over 60c on my system with some of the best air cooling available
i will need water cooling to hit 3.6 to 4ghz

but overall the fact that the chip has HT and SSE3 makes me happy enough with the upgrade for the rest of the year

yeah they shoulda added another layer and lowered the voltage to like 1.25v with temps under 45c if they could

that would have been more stunning i suppose

 

[Sahakiel]

Longer pipelines are meant to increase execution resource efficiency. Taking a wild guess, it seems Prsecott should scale to ~40% higher frequency on the same process as Northwood depending on how many drive stages were added.
That said, the two new stages to the Athlon are also meant to increase efficiency, but I seem to remember the new stages adding functionality rather than extending existing stages. They basically allow better use of K7 execution units. If that's true, clock speed shouldn't scale too much relative to a K7 design on the same process.
It's all a question of balancing CPI and CLK period given the same x86 ISA. K6 had a very efficient core, but it was too short to scale very high. That, and I seem to remember it being an existing NexGen design rushed to market. K7 was most likely the original competitor to the P6 design given the similarities between the two cores. I have a feeling K8 was supposed to have a much longer pipeline than it has now, but thanks to lack of funding and a shrinking release window, it was decided to use the K7 design and beef it up.

 

[Zebo]

Originally posted by: DAPUNISHER

Originally posted by: Zebo
You said it redesign or pretty soon AMD will still be pushing 2700 Mhz chips while Intel is pushing 6000Mhz. I don't care what PR ratings and the reviewers say about parity no one will believe it.

AMD will have CPUs based on the 90nm SOI process in the 2nd half this year, and they did add a couple stages to the pipe so clockspeeds should get a nice bump once the process matures. I should think overall performance should continue to remain close between the two companies for some time to come.

I'm guessing you are intentionally being a bit facetious about the clockspeed discrepency, but understand your argument concerning the public just not understanding that it isn't the key indicator to performance. However, I do think you are failing to consider several factors that may substantially negate much of the need for concern over that lack of comprehension on the consumers' part. One factor is that Intel will be switching to a number system to indentify microprocessors and not promoting clockspeed as the way to judge performance if what is being reported is correct.The other factor I feel is pertinent is that Intel is now producing a line of CPUs that are incredibly powerful but don't have a high clockspeed, much like AMD. I've read that this has been a issue in getting non-corporate consumers to embrace the P-M. For instance, I watched the lady selling the new GateWay centrino tablet *it looks sweet* on HSN the other night and she continued to re-iterate that although it was a 1.4ghz CPU that it performed more like a 2.8ghz CPU *she didn't state which one * I wouldn't be at all surprised if her emphasis about the performance belying the clockspeed resulted in quite a few question about that statement from callers.

My point is that I believe clockspeed will cease to be a major handicap as it appears that Intel now has a practical reason for obscuring clockspeed in favor of a performance rating system like their numbering system will provide. Being able to label a P-M and a much higher clocked P4-M with the same series number for performance would help dispell the doubts consumers have about it's performance vs the P4-M wouldn't it? Certainly it'll take time for consumers to adapt but I'm certain they have a transitional marketing strategy *Intel's marketing division is damned good* that will help indoctrinate the general public.

Good points dp, and I was'nt being *that* facetious. Look what happens everytime a new iteration of an AMD chip comes out VS. a Intel chip.

In recent history there was the t-bird which was then replaced by palomino/t-bred operating at lower clock for same performance as t-bird. Then Barton came operating at lower clock for same performance as palomino/t-bred. Now A64's can operate at lower clock for same performance as barton. AMD has continually lowered thier MHZ while improving the chip(s) with every new iteration. I can really see them still "stuck" at ~3000 max in late 2005.

Now we have Intel, going in the opposite direction every iteration. The PIII was a kickass chip but could'nt keep up in the Mhz race so they redesigned arriving at PIV high MHZ long pipe, to quote anand "a laughable performer" relative both to it's older bro and the t-bird at the time. But over time with increasing the cache/bus and increase clock speeds it's been shown to be not only powerful but good for marketing now running at 3.4 northwood. Rinse repeat with Prescott once again they lenghtend the pipe "in order to increase clock speeds" and performance suffered at same mhz relative to older brother. I posit prescott will be at 5000 Mhz by next year and that's all that will matter, again.

Again among the enthusiast community we know a 2.2 Ghz A64 = 3.4ghz prescott/northwood more or less but it's a harder sell out in the world especially when there's more marketing saturation by the later. Hell even I know the diff and bought a 2.4C to OC last year at 100% more money when all the good week bartons seemed to disappear.

I guess what I'm trying to say is they both have different ways to get at the same performance...fortunatly for one it has a higher real Mhz number attached to it.

 

[mastertech01]

Originally posted by: FelixDeKat

Originally posted by: mastertech01
Overclocking Press Hot is the problem. Hopefully that will change soon.

No problems here - 2.8@3.5, 250fsb, 1:1 500.

Um no, not that it wont overclock, but that it wont overclock without getting extremely hot. Water cooling is a must to push the ceiling and then it can get flakey..
Hence my comment, Press Hot.

 

[Jeff7181]

LOL... this is funny...

 

[Soulkeeper]

Originally posted by: Jeff7181
LOL... this is funny...

haha yeah
i think we all need watercooling
maybe that will solve all our problems

 

[orion7144]

Originally posted by: Soulkeeper

Originally posted by: Jeff7181
LOL... this is funny...

haha yeah
i think we all need watercooling
maybe that will solve all our problems

It only got my2.8E from 3.4 to 3.5 but did lower my temps by almost 15c idle. I think some of the problems with the power distribution on the motherboards is affecting the NB more than anything. Even thow my NB is water cooled I have seen it's temp go higher than the CPU and the CPU is already at 58c under load. Granted I am running a Koolance system but 15c difference going from a Zalman 7000 HSF is pretty impressive. This is on my new IC7-G Max3 that replaced my max2 that fried just days after running the Prescott.

 

[FelixDeCat]

Originally posted by: mastertech01

Originally posted by: FelixDeKat

Originally posted by: mastertech01
Overclocking Press Hot is the problem. Hopefully that will change soon.

No problems here - 2.8@3.5, 250fsb, 1:1 500.

Um no, not that it wont overclock, but that it wont overclock without getting extremely hot. Water cooling is a must to push the ceiling and then it can get flakey..
Hence my comment, Press Hot.

Um no what? Those are my numbers w/sp34 +4000rpm fan. The CPU reads 52 in bios after restarting immediately after playing Vice City.

You can keep your UT benchmarks. If its not stable in GTA, its not stable!!

 

[Big Lar]

Just an informed heads up here, as I Had a 3.0e Prescott, and have a 3.0c before and after in the same rig/ same components/ same bios. Yes, the prescott does OC ok, but, I performed my own evaluations as to speed comparisons in Sisoft, both chips doing the same speed/ same rig, The Prescott, has its place for some things,( damn'd if I could prove this, but others state it), But, My northy beat it BAD in all tests I ran at 3750. Just my input.