Intel Pentium-m as a desktop reviewed

[Moloch]

Those look tasty, too bad checking monarch, they're expensive as hell.
I'll take a 155$ A64 3000+ oced to 2.5 ghz over a 400 doller 2ghz pentium m.

 

[Bjorn]

Komplett.no (norway) had some prices for the Pentium M:

2.0 GHz - 3629 nkr
2.1 GHz - 5343 nkr

Athlon 64

1.8 GHz - 1401 nkr
2.2 GHz - 2308 nkr

So the price difference is definitely not minor. So you'll pay twice the price, or more and you'll get a slower CPU. Though it consumes 21 instead of 30W under full load. Don't know what the difference is under idle but 3W is good enough imo. So, the Pentium M doesn't seem that interesting to me. At least not at it's current price.

 

[ANova]

Don't bother guys. The 855GME chipset is memory bandwidth starved and lacking lots of features. Wait for the Centrino 2 based chipsets this year as they will add support for all the latest technology including dual channel higher memory bandwidth, PCIE, SATA Raid, 533 MHz fsb, etc. Then you'll have a winner.

 

[Saem]

I suggest being careful when quoting, "max" power specifications from AMD pdfs. You can do some digging around, but last I remember reading, they call it max, but in actuality it's under a simulated load. Intel figures are theoretical maximums which can't really be reached because the CPU doesn't work that way since all units can't be working all the time.

 

[Accord1999]

Here's another link which tried to measure CPU and system power usage.

http://www.x86-secret.com/articles/divers/conso/conso-3.htm

The 2GHz Dothan was measured at 16W at full load.

 

[Guden Oden]

I suggest being careful when quoting, "max" power specifications from AMD pdfs. You can do some digging around, but last I remember reading, they call it max, but in actuality it's under a simulated load. Intel figures are theoretical maximums which can't really be reached because the CPU doesn't work that way since all units can't be working all the time.

I believe you got that the wrong way around. AMD max power is MAX power. Intel typically quotes typical power, which is some arbitrary code which they feel represents real-world max and not ACTUAL max. Or they've done this in the past at least, and done it a lot for the P4 to hide the fact that chip can draw a hideous amount of peak power.

 

[mczak]

I suggest being careful when quoting, "max" power specifications from AMD pdfs. You can do some digging around, but last I remember reading, they call it max, but in actuality it's under a simulated load. Intel figures are theoretical maximums which can't really be reached because the CPU doesn't work that way since all units can't be working all the time.

I believe you got that the wrong way around. AMD max power is MAX power. Intel typically quotes typical power, which is some arbitrary code which they feel represents real-world max and not ACTUAL max. Or they've done this in the past at least, and done it a lot for the P4 to hide the fact that chip can draw a hideous amount of peak power.

That is correct. All of intels chips (including the P-M) indeed will draw more power under some circumstances than their quoted TDP (it says that in the datasheets). The difference between quoted TDP and actual maximum power is rumoured to be somewhere along 25% (at least for P4), though that number is 100% unofficial. That is also the reason why you might get cpu throttling under some circumstances with a cooling solution which indeed still meets intel's quoted cooling requirements.

 

[Pete]

I thought AMD's power specs were the max for the entire process, meaning it allowed (quite a bit of) headroom for future, faster CPUs. I remember the A64 had a similar TDP to the Prescott, but reviews showed the latter far outdrew the former.

 

[Accord1999]

The difference between quoted TDP and actual maximum power is rumoured to be somewhere along 25% (at least for P4), though that number is 100% unofficial. That is also the reason why you might get cpu throttling under some circumstances with a cooling solution which indeed still meets intel's quoted cooling requirements.

TDP for Intel is set at a value slightly higher than maximum sustained power found using a suite of CPU intensive applications.

 

[T2k]

Don't bother guys. The 855GME chipset is memory bandwidth starved and lacking lots of features. Wait for the Centrino 2 based chipsets this year as they will add support for all the latest technology including dual channel higher memory bandwidth, PCIE, SATA Raid, 533 MHz fsb, etc. Then you'll have a winner.

Sure: the on-die memory controller on AMD chips.

P-M lacks some optimization which actually helps the aging P4 cores to be able to compete in video stuff against brutal Athlon64s, for example.

On the other hand AMDs need some reorganization in their execution units.

Another factor is even P-M will get a new memory architecture, it'll be most likely DDR2, which is pretty sux, considering its really high latency.

And not to mention that AMD doesn't sit home and waiting for Intel's buttkick..

 

[T2k]

FOrgot to mention : next-gen die will be definitely multicore from AMD and most likely from Intel too. Which means even an upgraded (faster bus, dual channel memory etc) P-M will be DOA against the upcoming desktop stuff.

 

[incurable]

FOrgot to mention : next-gen die will be definitely multicore from AMD and most likely from Intel too. Which means even an upgraded (faster bus, dual channel memory etc) P-M will be DOA against the upcoming desktop stuff.

Multicore is still quite a bit off, dual-core on the other hand, which we'll probably see in some form this year, is just as much part of the Pentium M roadmap as it is for other product lines. (Yonah, IIRC)

 

[mczak]

FOrgot to mention : next-gen die will be definitely multicore from AMD and most likely from Intel too. Which means even an upgraded (faster bus, dual channel memory etc) P-M will be DOA against the upcoming desktop stuff.

Multicore is still quite a bit off, dual-core on the other hand, which we'll probably see in some form this year, is just as much part of the Pentium M roadmap as it is for other product lines. (Yonah, IIRC)

Yonah will likely appear quite a bit later than the others in the market though (not before 2006). Probably because it's supposed to be built with 65nm technology, the other "first-gen" dual core intel/amd chips are all built with 90nm. OTOH, it seems to be a very promising design, not just 2 cores slapped together like the next-gen P4.

 

[T2k]

FOrgot to mention : next-gen die will be definitely multicore from AMD and most likely from Intel too. Which means even an upgraded (faster bus, dual channel memory etc) P-M will be DOA against the upcoming desktop stuff.

Multicore is still quite a bit off, dual-core on the other hand, which we'll probably see in some form this year, is just as much part of the Pentium M roadmap as it is for other product lines. (Yonah, IIRC)

Khm: "multi" includes/means "dual" as well.

AMD will announce it soon, Intel is way off to do so, as of today. Even if they do use the Pentium-M core as it is now - which I highly doubt -, it's already losing against A64 cores.
Intel needs a serious redesign on P-M to be able to compete with AMD's integrated memory controller, believe me.

When I predicted Prescott and Tejas will be canned, nobody believed - and voila, it happened las year. 8)
When I say AMD is working on something pretty cool, nobody believes me again...

BTW I'd love to see some competition from Intel because currently AMD is waaay better in any sector when it comes to non-mobile CPUs and I don't like when only one guy becomes the king of the hill. In fact we haven't seen too much in 2H of 04 from AMD, remember...

The sad fact is Intel needs a serious buttkick on the highest management level and I can't see it's happening. Remember: what they did during last few years? Burned zillions in canned projects, utterly useless products: Itanium, Itanium2, Prescott, Tejas, everything was nfucked up except their mobile parts.
I am one of the very few early Itanium adopters and I can tell you how shitty, crappy is that CPU, how it's only serves the late 70's-style computing ideas - which is fully contradicts today's general computing needs. It's actually ridiculous that such incompetent stupid project managed to reach even a design level, not to mention a manufacturing and retail level... It was a classic moneyburning project based on totally misunderstood costumer needs, market state and a really arrogant corporate crook attitude. Classic guinea pig how to fuck up zillions with a useless product. because of pompous but utterly ignorant management decisions.
Thanks God, Intel royally fucked up, they lost more than $10B on Itanium and now NOBODY is selling anymore on WS level - this is the last phase of the agony. I say thanks because - even though Intel still simply too arrogant and bigheaded to admit they screwed up - this is gonna be a big-big-big lesson for them, how NOT to deal with the market anymore.

I seriously think the Intel-dominated days are over, no more slow, shitty development and overpriced processors but at the same time I really hope there's not gonna be an AMD kingdom instead neither.

We need competition - that's what most of the rabid Intel and AMD fans should understand too.

 

[T2k]

FOrgot to mention : next-gen die will be definitely multicore from AMD and most likely from Intel too. Which means even an upgraded (faster bus, dual channel memory etc) P-M will be DOA against the upcoming desktop stuff.

Multicore is still quite a bit off, dual-core on the other hand, which we'll probably see in some form this year, is just as much part of the Pentium M roadmap as it is for other product lines. (Yonah, IIRC)

Yonah will likely appear quite a bit later than the others in the market though (not before 2006). Probably because it's supposed to be built with 65nm technology, the other "first-gen" dual core intel/amd chips are all built with 90nm. OTOH, it seems to be a very promising design, not just 2 cores slapped together like the next-gen P4.

Yeah, that's what I hear from Intel's camp as well.
I predict that first Intel multicore - OK, dualcore - will lose against AMD's first dualcore but second round is fully open...
8)

 

[London Geezer]

Burned zillions in canned projects, utterly useless products: Itanium, Itanium2, Prescott, Tejas, everything was nfucked up except their mobile parts.

Was that a typo or a pun?

*Nfucked up, the way only Nvidia can fuck things up!!*

 

[incurable]

AMD will announce it soon, Intel is way off to do so, as of today.

Personally, I don't care for announcements, I care for available products, so lets see when AMD can deliver their dual-core K8s (btw: I can't seem to get all those AMD codenames in my head.) and then start the clock for Smithfield, shall we?

Even if they do use the Pentium-M core as it is now - which I highly doubt -, it's already losing against A64 cores.
Intel needs a serious redesign on P-M to be able to compete with AMD's integrated memory controller, believe me.

Depending on the applications you test, of course.

When I predicted Prescott and Tejas will be canned, nobody believed - and voila, it happened las year. 8)

Yeah, well, only if you overlook the fact that Intel sells Prescotts by the millions as we speak.

When I say AMD is working on something pretty cool, nobody believes me again...

AMD has worked on many a cool thing in the past 3 decades, and it has an almost equally long history of not delivering these (on time). You can't fault people for being cautious under those circumstances.

We need competition - that's what most of the rabid Intel and AMD fans should understand too.

Hear, hear.

 

[hovz]

i cant think of 1 application offhand the pentium m wins in. im not saying there isnt one, but its probably extremely rare.

 

[T2k]

AMD will announce it soon, Intel is way off to do so, as of today.

Personally, I don't care for announcements, I care for available products, so lets see when AMD can deliver their dual-core K8s (btw: I can't seem to get all those AMD codenames in my head.) and then start the clock for Smithfield, shall we?

Here is your daily link: the Prescott w/ doubled cache, aka P 6xx-series: http://www.hkepc.com/hwdb/p4660-1.htm

This crappy new stuff clearly shows how desperate/stupid is Intel now.

Even if they do use the Pentium-M core as it is now - which I highly doubt -, it's already losing against A64 cores.
Intel needs a serious redesign on P-M to be able to compete with AMD's integrated memory controller, believe me.

Depending on the applications you test, of course.

NO, anywhere. Please, don't bring Anova's retarded comparisons about OCed P-M vs A64 on stock clock.

For the sake of this discussion let's stick with the basics: the usual base for a compariosn is typically the price.

Shall we?

When I predicted Prescott and Tejas will be canned, nobody believed - and voila, it happened las year. 8)

Yeah, well, only if you overlook the fact that Intel sells Prescotts by the millions as we speak.

Bullshit. Only OEMs buy Intels crap, my personal experiences that 70-80% of the people in retail buys AMDs nowadays.
In OEM Intel is fine - but it's rapidly changing as more and more big guy (HP) sell more and more AMD-based stuff.
Intel IS losing marketshare as we speak.

Edit: obviously it's not gonna change overnight but it IS changing - you know as well as I know. 8)

When I say AMD is working on something pretty cool, nobody believes me again...

AMD has worked on many a cool thing in the past 3 decades, and it has an almost equally long history of not delivering these (on time). You can't fault people for being cautious under those circumstances.

Facts are togh things, you know...

The last two promised change - 64bit cores and SoI - were successful at AMD.

The last half dozen thing - own 64bit platform, Itaniums, Prescott, Tejas, EMT64 - were completely fucked up at Intel.

Your turn.

We need competition - that's what most of the rabid Intel and AMD fans should understand too.

Hear, hear.

8)

 

[ANova]

I am one of the very few early Itanium adopters and I can tell you how shitty, crappy is that CPU, how it's only serves the late 70's-style computing ideas - which is fully contradicts today's general computing needs. It's actually ridiculous that such incompetent stupid project managed to reach even a design level, not to mention a manufacturing and retail level... It was a classic moneyburning project based on totally misunderstood costumer needs, market state and a really arrogant corporate crook attitude. Classic guinea pig how to fuck up zillions with a useless product. because of pompous but utterly ignorant management decisions.
Thanks God, Intel royally fucked up, they lost more than $10B on Itanium and now NOBODY is selling anymore on WS level - this is the last phase of the agony.

If this is so why then did NASA just build a supercomputer comprised of Itanium 2s which has overtaken Japan's Earth simulator?

Facts are togh things, you know...

The last two promised change - 64bit cores and SoI - were successful at AMD.

The last half dozen thing - own 64bit platform, Itaniums, Prescott, Tejas, EMT64 - were completely fucked up at Intel.

Your turn.

64 bit cores my ass. It's just an extension to the aging x86. SOI is nice but certainly nothing AMD created.

The Prescott and the Tejas were bad designs, that's one thing we can agree on. EM64T however, is almost exactly the same as x86-64. Hell Intel practically copied it, so I don't know how you can claim AMD's "64 bit" procs are a promised change and at the same time claim EM64T is a fuck up.