Was the P4 an 'engineering failure'?

[deimos3428]

Originally posted by: drizek
Do you know how much that thing cost you over the past 10 years in electricity?

I have no idea, as it's completely insignificant to most computer users. It's not even a blip on the radar in comparison to things that really consume a lot of electricity like my power tools, television, dehumidifier, microwave or vacuum cleaner. Are you trying to say that UNIVAC was a failure?

 

[dmens]

lol epic thread fail. pretty sure no one here is qualified to fully comment on P4 as an engineering product (me included).

Originally posted by: apoppin
the NetBust guys are back and evidently they have designed Larrabeast

are you ever sick of being wrong

 

[Arkaign]

P4 was a success with some setbacks and disappointments. I don't think anyone can argue that Northwood P4's were incredibly solid during their era. P4's even competed well in a number of benchmarks against the early single-core AMD64 chips, although AMD64 was indeed superior IMHO.

 

[VirtualLarry]

P4 was crap. Always.

 

[aigomorla]

The P4 was a failure in design.

The Netburst application theory was only possible at extreme mhz, which no average person could realistically attain.

The CPU's were HOT, and they wasted a lot of power.

Current C2Q's are nothing like P4's. As someone said, the entire P4 line was scraped, and the P3 adaption was matured to become the first C2D.

Yonah, the first C2D was a mobile processor off the P3 design. Also the first processor to shatter the sub 10 second super PI without the need to uber insane ghz.

 

[myocardia]

Originally posted by: alyarb
well that was the point i was making with my original post. but yeah, i've heard that "10ghz or bust" story before. pretty funny with y2k technology.

That was no "story": high-level Intel exec claiming the P4 would SURPASS 10 Ghz

 

[apoppin]

Originally posted by: dmens
lol epic thread fail. pretty sure no one here is qualified to fully comment on P4 as an engineering product (me included).

Originally posted by: apoppin
the NetBust guys are back and evidently they have designed Larrabeast

are you ever sick of being wrong

not of making fun of you and watching you react to something that i don't take seriously at all

You are probably too young to have been on that failure of a team anyway

 

[geokilla]

Way to bring back a dead thread.

Someone lock the thread!

 

[dmens]

Originally posted by: apoppin
not of making fun of you and watching you react to something that i don't take seriously at all

lol make fun of me because i am actually semi-qualified to comment on this topic, whereas you know nothing whatsoever. knock yourself out haha.

 

[drizek]

Originally posted by: Arkaign
P4 was a success with some setbacks and disappointments. I don't think anyone can argue that Northwood P4's were incredibly solid during their era. P4's even competed well in a number of benchmarks against the early single-core AMD64 chips, although AMD64 was indeed superior IMHO.

The problem wasn't just AMD64. Pentium 4s did not compete well against K7 either. They were usually faster, sure, but the difference in cost was huge, and the AMDs would overclock a ton(OCing a 2500 barton to 3200 was basically guaranteed).

 

[apoppin]

Originally posted by: dmens

Originally posted by: apoppin
not of making fun of you and watching you react to something that i don't take seriously at all

lol make fun of me because i am actually semi-qualified to comment on this topic, whereas you know nothing whatsoever. knock yourself out haha.

"half-qualified" ?


.. i have to agree; your assessment of your own qualifications seem about right
- it looks like you knocked yourself out

 

[Arkaign]

Originally posted by: drizek

Originally posted by: Arkaign
P4 was a success with some setbacks and disappointments. I don't think anyone can argue that Northwood P4's were incredibly solid during their era. P4's even competed well in a number of benchmarks against the early single-core AMD64 chips, although AMD64 was indeed superior IMHO.

The problem wasn't just AMD64. Pentium 4s did not compete well against K7 either. They were usually faster, sure, but the difference in cost was huge, and the AMDs would overclock a ton(OCing a 2500 barton to 3200 was basically guaranteed).

The difference in cost could be a factor, but it wasn't huge if you overclocked the NW as well. Northwood 1.6A, 2.0, 2.4B, etc, could usually overclock quite a lot. I'm not discounting the AXP by any means, it held a lot of value, dominated the Willamette P4s, and with the Bartons it remained at least competitive with the up to ~3ghz Northwoods. But there wasn't a giant price difference between stock CPUs. IIRC, the P4 2.4B was roughly the same price as the Barton 2500, and once overclocked, both performed about the same (if you got the 2.4B to ~3ghz).

Where AMD really had it going with Athlon XP was in the lower-spec chips. You could get an AXP 1700+ Mobile, 1800+, 2000+, etc, for a lot less than the 2ghz+ Northwoods that were out, and Intel didn't tend to continue making the lower-end chips, instead they just phased them out and released faster and faster chips at those same price points.

I built a large number of systems during that timeframe of '02-'04, and ended up using a majority of Intel setups due to small things like the IHS and availability of decent 845 boards for cheap. The only Socket A chipset I cared for was the Nforce2, I absolutely loathed the VIA chipsets, and the decent Nforce2 boards were as expensive (sometimes more so) than the decent 845 boards.

My conclusion is fundamentally this : one cannot rationally declare a 'winner' for that time frame of the primary P4 era. The P4 maintained good performance and marketshare, and the Athlon XP chips represented great value (particularly on the lower-end which Intel abandoned during that time other than the very weak Celerons) and very respectable performance. By the time Athlon 64 came on strong, P4 was definitely on the downward slope and the limits of the architecture. Even then though, the Socket T P4's continued to at least keep things going. A 3.4Ghz P4, while in no way my choice over a 3400+ AMD64 (loved the Socket 754 and 939 Nvidia NF3 and NF4 btw), at least managed to look respectable in quite a few benchmarks against the newcomer.

 

[Arkaign]

I think this would make more sense if we granularized the comparison to particular core gens.

Willamette = Failure (too expensive, not competitive with Athlon Thunderbird in most things)
Northwood = Winner (available in sub-$150 price points for most of it's run, highly overclockable, very competitive in almost all apps)
Gallatin = Mixed (good performance, but insane price)
Prescott = Failure/Mixed (too hot, no performance increase over Northwood, actually a decrease in some things, dumb socket change shortly thereafter)
Irwindale = Failure (too expensive, not compelling against superior AMD64)
Cedar Mill = Mixed (much less heat than Prescott, decent value for $, low-latency 2MB L2, but not really compelling against AMD64)

All of AMD's K7 products were winners outside of the initial K7 (horrendous 'Irongate' chipset mobos, crappy doomed Slot-A), and perhaps Thoroughbred-A (crap for overclocking, very quickly replaced by the excellent Thoroughbred-B, and then the Bartons).

 

[MODEL3]

Well, the original question was about if P4 was a 'engineering failure'! (Not failure in a general sense!)
I'm sure other departments, such as the marketing played their role also!
Can you say that it was also a financial failure?

All the answers for the above questions, can be different depending on your "point of view" and what you compering them with! (everything is relative!)

For the P4 era I liked (or better i should say "didn't dislike") only the 13nm era (A,B,C)

One of the lessons Intel learned from the P4 era (other than the GHz wall :laugh: ) is that: "performance improvements per TDP increasements" ratio is extremely important! (their words in an IDF)


From a 29W TDP PIII 1GHz (18nm) they went to a:
52W TDP PIV 1.3GHz (18nm)

So they nearly doubled the TDP (1,8X) while the performance was nearly the same
(Let's say that the 1GHz PIII had -5% performance in relation with the 1,3GHz PIV, and i would be extremely generous to PIV, saying that!)

So at least from this point of view, the P4 was an engineering failure (maybe failure is a strong word)

 

[Fox5]

Originally posted by: myocardia

Originally posted by: alyarb
well that was the point i was making with my original post. but yeah, i've heard that "10ghz or bust" story before. pretty funny with y2k technology.

That was no "story": high-level Intel exec claiming the P4 would SURPASS 10 Ghz

I'm sure it would have if they kept up with it eventually.

I remember hearing prescott was meant to hit 5Ghz, but instead maxed out at 3.8Ghz and extremely high heat levels. Had things gone at planned, a 5Ghz Prescott would have at least traded blows with the athlon 64, if not outright beat it.

 

[apoppin]

Originally posted by: Fox5

Originally posted by: myocardia

Originally posted by: alyarb
well that was the point i was making with my original post. but yeah, i've heard that "10ghz or bust" story before. pretty funny with y2k technology.

That was no "story": high-level Intel exec claiming the P4 would SURPASS 10 Ghz

I'm sure it would have if they kept up with it eventually.

intel didn't think so

they *dumped* NetBust because their engineers said it would take "luck" to achieve it

it looks like they are making the same big predictions about Larrabeast; i have little hope in market-driven engineering

 

[Arkaign]

Originally posted by: MODEL3
Well, the original question was about if P4 was a 'engineering failure'! (Not failure in a general sense!)
I'm sure other departments, such as the marketing played their role also!
Can you say that it was also a financial failure?

All the answers for the above questions, can be different depending on your "point of view" and what you compering them with! (everything is relative!)

For the P4 era I liked (or better i should say "didn't dislike") only the 13nm era (A,B,C)

One of the lessons Intel learned from the P4 era (other than the GHz wall :laugh: ) is that: "performance improvements per TDP increasements" ratio is extremely important! (their words in an IDF)


From a 29W TDP PIII 1GHz (18nm) they went to a:
52W TDP PIV 1.3GHz (18nm)

So they nearly doubled the TDP (1,8X) while the performance was nearly the same
(Let's say that the 1GHz PIII had -5% performance in relation with the 1,3GHz PIV, and i would be extremely generous to PIV, saying that!)

So at least from this point of view, the P4 was an engineering failure (maybe failure is a strong word)

Agreed, in any particular rational debate about this topic, one has to get into specifics to clarify 'failure' or 'success'. Thus, one cannot come to a rational general conclusion without serious caveats.

P3 at .18mu (and even at .13mu) was having ramping issues. P4 architechture offered a tempting path towards increasing clock speeds and pretty quickly, performance leaps. No intelligent person would say that the P4 Willamette was a glorious success, but it did pave the way for P4 Northwood, which offered far more performance than P3s.

It is arguable whether the entire P4 experiment was worth it at all, but to know the answer to that, one would have to examine how difficult it would have been to continually move the P3 towards the evolutionary state of Merom, etc. AFAIK, Intel never really stopped working on the P3 architecture, P4 simply offered a path towards offering competitive to class-leading performance for the time it took for Intel to ramp up high-IPC designs enough in clockspeed to be competitive.

A hypothetical P3 1.6Ghz would have been competitive with a 1.5Ghz Athlon XP, but not if the thermals were too high, if the yields too low, and so on.

It's also easy to look back with 20/20 hindsight. Intel and AMD have both made serious errors in judgment that must have seemed correct at the time. Chiefly : Phenom and Prescott. But both were learning processes that almost certainly yielded valuable lessons towards future projects.

 

[dmens]

Originally posted by: apoppin

Originally posted by: dmens

Originally posted by: apoppin
not of making fun of you and watching you react to something that i don't take seriously at all

lol make fun of me because i am actually semi-qualified to comment on this topic, whereas you know nothing whatsoever. knock yourself out haha.

"half-qualified" ?


.. i have to agree; your assessment of your own qualifications seem about right
- it looks like you knocked yourself out

haha you are too stupid to understand. no one here is a P4 architect or principal engineer, therefore no one is fully qualified to comment.

you on the other hand is fully unqualified since you obviously have no clue regarding either the circuit or the architectural issues of the P4. keep talking, you just look more stupid with each emoticon-laden post lol.

 

[apoppin]

Originally posted by: dmens

Originally posted by: apoppin

Originally posted by: dmens

Originally posted by: apoppin
not of making fun of you and watching you react to something that i don't take seriously at all

lol make fun of me because i am actually semi-qualified to comment on this topic, whereas you know nothing whatsoever. knock yourself out haha.

"half-qualified" ?


.. i have to agree; your assessment of your own qualifications seem about right
- it looks like you knocked yourself out

haha you are too stupid to understand. no one here is a P4 architect or principal engineer, therefore no one is fully qualified to comment.

you on the other hand is fully unqualified since you obviously have no clue regarding either the circuit or the architectural issues of the P4. keep talking, you just look more stupid with each emoticon-laden post lol.

i am too stupid to understand?
:|

How do you know what i have a clue about?


if we look at your posts, you come across as a very junior kind of engineer who supports his company no matter what and perceives even possibly negative comments about any part of it as a personal attack

You are the one that has been consistently rude to me for a long time with no basis other than you do not seem to appreciate my emoticon usage
- too bad; i really don't care for your continual 'lol' as some sort of a writing 'tic', but i was too polite to mention it until now


Now i do not agree with you that it requires a P4 CPU architect to comment here; obviously that will never happen
- i certainly do know what Intel themselves did; suddenly dropping NetBurst in a complete turnaround demonstrated it was a failure.

i have experts in other fields i have linked to that also call P4 a failure
- otoh you have provided nothing but personal attacks and insults instead of supporting your view

 

[TurtleCrusher]

Originally posted by: geokilla
Way to bring back a dead thread.

Someone lock the thread!

Maybe you don't frequent the CPU forum at night but it gets to be a ghost town. I think this topic has a whole lot of relevance today due to a new line of processors in the coming months. History is the best way of predicting the future and it looks like Intel has learned from their mistakes. AMD is learning right now in the enthusiast market.

It could have been a solution thread. For instance, if someone needed in a 3 year old thread to find the dual core driver for their brand new s939 Athlon X2 3600. I'd lock that myself if I could.

 

[dmens]

Originally posted by: apoppin

Originally posted by: dmens

Originally posted by: apoppin

Originally posted by: dmens

Originally posted by: apoppin
not of making fun of you and watching you react to something that i don't take seriously at all

lol make fun of me because i am actually semi-qualified to comment on this topic, whereas you know nothing whatsoever. knock yourself out haha.

"half-qualified" ?


.. i have to agree; your assessment of your own qualifications seem about right
- it looks like you knocked yourself out

haha you are too stupid to understand. no one here is a P4 architect or principal engineer, therefore no one is fully qualified to comment.

you on the other hand is fully unqualified since you obviously have no clue regarding either the circuit or the architectural issues of the P4. keep talking, you just look more stupid with each emoticon-laden post lol.

i am too stupid to understand?
:|

How do you know what i have a clue about?


if we look at your posts, you come across as a very junior kind of engineer who supports his company no matter what and perceives even possibly negative comments about any part of it as a personal attack

You are the one that has been consistently rude to me for a long time with no basis other than you do not seem to appreciate my emoticon usage
- too bad; i really don't care for your continual 'lol' as some sort of a writing 'tic', but i was too polite to mention it until now


Now i do not agree with you that it requires a P4 CPU architect to comment here; obviously that will never happen
- i certainly do know what Intel themselves did; suddenly dropping NetBurst in a complete turnaround demonstrated it was a failure.

i have experts in other fields i have linked to that also call P4 a failure
- otoh you have provided nothing but personal attacks and insults instead of supporting your view

hell i could have told you the same thing coming out of college that you have no clue about CPU's whatsoever, don't need to work at intel to tell you that. that much is obvious from what you write. and even if i were a "junior engineer" that still puts my knowledge way above yours don't it.

i could have told you that P4 taught plenty of valuable lessons and portions of it continue to be used in current and future CPU's, but that would be too much information for your feeble brain to digest.

and yeah you are annoying as fuck and i make it a point to correct your error-laden posts. and don't flatter yourself, you don't anger me a single bit, even with your rose emoticons, LOL.

 

[hans007]

i would say it was mostly a failure.

the only chip that was considered good was the northwood. and it still wasn't great just the athlon 64 was delayed and was only on servers for a while.

Even the athlon XP was fairly competitive. by the time prescott came out, amd was way ahead. the smithfield was terrible. and williamette was terrible.

Not to mention all the other failures in the beginning of the p4 era, like th ei850 and rambus . (i would consider those part of the whole "era").

I think the main benefit intel got out of building these chips was that they developed hyperthreading during this era. every other worthwhile thing from that time period was because of the bania pentium M and its descendants.

 

[TurtleCrusher]

My first true gaming PC was a P4 2.26ghz and a Geforce 2 Ti 4200 32mb. The memories of CS 1.3-6 going from 100+fps to a mere 20fps.

But that is nothing compared to my first midrange build of an Athlon X2 4200+ , 4GB PC3200 and a Geforce 8800GTS 320mb.

 

[Fox5]

It is arguable whether the entire P4 experiment was worth it at all, but to know the answer to that, one would have to examine how difficult it would have been to continually move the P3 towards the evolutionary state of Merom, etc. AFAIK, Intel never really stopped working on the P3 architecture, P4 simply offered a path towards offering competitive to class-leading performance for the time it took for Intel to ramp up high-IPC designs enough in clockspeed to be competitive.

A hypothetical P3 1.6Ghz would have been competitive with a 1.5Ghz Athlon XP, but not if the thermals were too high, if the yields too low, and so on.

Well, P4 was their big next-gen product, so you can't just remove it and say Intel would have been fine without it. Many of the challenging problems Intel had to solve to make the P4 were probably applied to future designs anyhow, and Intel's current offerings might not be quite as strong had they not suffered the limitations of manufacturing and circuit design.

I wouldn't say a 1.6Ghz P3 would be competitive though. A slow memory bus would have handicapped it against the DDR Athlon XP, and Intel architectures have always benefited from a large cache, which the P3 did not have. And in many applications, the Athlon's stronger FPU would have required more like a 1.7Ghz or 1.8Ghz P3 to compete with the Athlon XP, perhaps even more so.
Looking at some old benchmarks, the original Athlon with DDR could have up to a 25% clock for clock performance advantage over the Pentium 3. That's on par with the kind of advantage the Core 2 Duo has over the Athlon X2.
The Pentium M faired better, but even with its large cache could still get destroyed handily by the Athlon 64 in about everything but games.

There's no telling what kind of design Intel would have produced in place of the Pentium 4, but anything from the Pentium3/Pentium M line would have faired worse, especially in the server/HPC markets. They were relegated to mobile designs for a reason, Conroe was the first power-house chip Intel produced after Pentium 4.

 

[apoppin]

Originally posted by: dmens

Originally posted by: apoppin

Originally posted by: dmens

Originally posted by: apoppin

Originally posted by: dmens

Originally posted by: apoppin
not of making fun of you and watching you react to something that i don't take seriously at all

lol make fun of me because i am actually semi-qualified to comment on this topic, whereas you know nothing whatsoever. knock yourself out haha.

"half-qualified" ?


.. i have to agree; your assessment of your own qualifications seem about right
- it looks like you knocked yourself out

haha you are too stupid to understand. no one here is a P4 architect or principal engineer, therefore no one is fully qualified to comment.

you on the other hand is fully unqualified since you obviously have no clue regarding either the circuit or the architectural issues of the P4. keep talking, you just look more stupid with each emoticon-laden post lol.

i am too stupid to understand?
:|

How do you know what i have a clue about?


if we look at your posts, you come across as a very junior kind of engineer who supports his company no matter what and perceives even possibly negative comments about any part of it as a personal attack

You are the one that has been consistently rude to me for a long time with no basis other than you do not seem to appreciate my emoticon usage
- too bad; i really don't care for your continual 'lol' as some sort of a writing 'tic', but i was too polite to mention it until now


Now i do not agree with you that it requires a P4 CPU architect to comment here; obviously that will never happen
- i certainly do know what Intel themselves did; suddenly dropping NetBurst in a complete turnaround demonstrated it was a failure.

i have experts in other fields i have linked to that also call P4 a failure
- otoh you have provided nothing but personal attacks and insults instead of supporting your view

hell i could have told you the same thing coming out of college that you have no clue about CPU's whatsoever, don't need to work at intel to tell you that. that much is obvious from what you write. and even if i were a "junior engineer" that still puts my knowledge way above yours don't it.

i could have told you that P4 taught plenty of valuable lessons and portions of it continue to be used in current and future CPU's, but that would be too much information for your feeble brain to digest.

and yeah you are annoying as fuck and i make it a point to correct your error-laden posts. and don't flatter yourself, you don't anger me a single bit, even with your rose emoticons, LOL.

"don't it" ?

it sure looks like you missed English class as well as manners

There is nothing of mine that i have posted here that you have been able to comprehend, nevermind correct. There is nothing of value in any of your posts in this thread at all. Just your own personal opinion and personal attack after attack.

i do agree that P4 taught important lessons; most failures do.