Queries about Pentium M/3 archietcture

[pincho]

hello all,

after reading that Pentium M is a heavly modified version on the Pentium 3 processor, i stand confused. Why is Intel adopting a prettyr old archietucture back to its system, and how can it possibly run better than later P4s and AMDs?

Apparently P3's overclocked can smash P4's, but why did Intel in the first place move to P4's, while pentium 3's were not being used to their full capability? sorry if this all sounds confusing, but just common logic of mine, newer>older

I heard somewhere that introducing newer technology such as Dual core, 64 Bit, to the Pentium M, will bring Intel as high as AMD X2's, even higher. But wat is it about the 5 year old Pentium 3 design that can beat todays technology? sorry again if i sound confuding

 

[pxc]

Why is Intel adopting a 5 year old archietucture back to its system,

Maybe you missed the news, but Banias (the first Pentium M core) came out nearly 2 and a half years ago. This isn't new. Asking why the P-M is based on the P3 is like asking why the K8 is based on the K7: because it was a successful design. Many people will say the Pentium M is a P3, but it really isn't. There are many similarities, but they are not the same.

Why is it faster? http://www.intel.com/technology/itj/2003/volume07issue02/art03_pentiumm/p01_abstract.htm

 

[pincho]

really? sorry, i only recently started becoming interested in Computer hardware, and only heard of Pentium M's/ Centrinos here in Australia at the end of last year, and since its becoming adopted in Desktops only recently...

yes im asking y it is based on P3 cause i would have thought it was based off P4..

 

[ameoba]

The P4 is not an elegant design - it's big, power-hungry and inefficient and designed around running at high clock speeds (fast enough to offset the inefficiencies). If this was driven by tech people or marketing people is up to debate.

With the Pentium M, Intel is applying many of the things they learned from the P4 to a more efficient design.

 

[lithium726]

Pentium Pro, 2, 3, M, (P6 architecture) and AMD's K6(although with very weak FPU performance)/K7/K8 all follow the same general idea: maximize work per cycle, and keeps clock speeds relitivly low. This allows for lower power consumption, because no matter how you slice it, the transistors on the processor are switching slower.

the P4 on the other hand, was a brand new architechture, netburst. it does less work per cycle than its P6 counterparts, and offsets this fact with higher clock speeds. the problem with this is that no matter what the architecture, at higher clock speeds, transistors are going to consume more power, and this means higher thermal charictoristics. Currently, with pentium4 prescott (90nm) based Penitum D's, thermal Dissapation charictorists are in the 130w areas. in comparison, the Toledo based Athlon 64 X2's are in the 110w areas and the Machester based ones are at 89w... a huge difference, especially for the performance they offer.

so, until intel can get thermal charictoristics down (which is mostly caused by power leakage in the current 90nm design, hopefully to be solved with the shink to 65nm) htey need something that consumes less power, runs cooler, and most importantly, can compete with the a64 X2, cause the Pentium D is doing a poor job of it (the p4's are having a hard time keeping up with the a64's too). This is pretty much why (IMO) they are ditching netburst in favor of the sucessful design based on the Pentium Pro.

 

[pr0nasaurus rex]

We was robbed.

The initial buzz when P4 came out was that the architecture would stretch to 10GHz. The hype machine kicked into overdrive, and everyone picked up 1.5GHz P4s, just to learn they were slugs compared to their 1.26 and 1.4GHz Coppermine P3's.

Pentium 4 was one miserable failure, and a great example of marketing triumphing over technological prowess. The Pentium M is as close to Pentium 3 than Prescott is to Northwood. I really believe that upper management at Intel finally realized that they screwed up, and tried to update and remarket their previous core.

 

[lithium726]

1.26 and 1.4GHz Coppermine P3's.

Tualatin. Coppermine only went to 1.13ghz, and that was recalled.

 

[pr0nasaurus rex]

Regardless, I think Pentium M is a major step in the right direction for Intel. All we need now is a fully-featured modern chipset for the Pentium M.

 

[pxc]

All we need now is a fully-featured modern chipset for the Pentium M.

The i915 chipset isn't bad and that's what ships in the "Sonoma" mobile platform. Gee, I have one!

 

[Flak Monkey]

Currently, with pentium4 prescott (90nm) based Penitum D's, thermal Dissapation charictorists are in the 130w areas. in comparison, the Toledo based Athlon 64 X2's are in the 110w areas and the Machester based ones are at 89w... a huge difference, especially for the performance they offer.

The difference is actually a lot more than that. AMD rates it's power consumption max power. Intel rates it as average power. Otherwise the x2's would only be about 18% cooler than the Pentium counterpart but as we all know the difference is much greater.

 

[DanK]

If you're interested in digging deep into CPU architecture, one site I can highly recommend is Ars Technica. Here's their article on the Pentium M; you also might be interested in The Pentium: An Architectural History of the World's Most Famous Desktop Processor parts one and two.

 

[pincho]

thanks a lot for the replies people. sorry that the Pentium M is kinda old, its just ive only heard of it recently. ill keep reading those articles. thanks again

 

[pxc]

The difference is actually a lot more than that. AMD rates it's power consumption max power. Intel rates it as average power.

No, not really. Intel's definition of TDP from the datasheet:
"The processor’s power is specified as Thermal Design Power (TDP) for thermal solution design. TDP is defined as the worst-case power dissipated by the processor while executing publicly available software under normal operating conditions, at nominal voltages that meet the load line specifications. The TDP definition is synonymous with the Thermal Design Power (typical) specification referred to in previous Intel datasheets. The Intel TDP specification is a recommended design point and is not representative of the absolute maximum power the processor may dissipate under worst case conditions."

That's not an average power, it's a maximum power (without throttling) that the processor will run at 100% load. It is possible to write a "power virus" that exceeds TDP, so that's why the disclaimer is added above. To cut through the buzzwords in the underlined section:
"under normal operating conditions": Tcase < Tcase max (throttling temperature)
"nominal voltages": VDD=VID_VDD (standard rated voltage)
"load line specifications": up to IDD max (maximum current) is supplied to the CPU

But yeah, AMD gives maximum power in their datasheets. One thing that's not explained in the AMD datasheets is how they arrive at IDD max.

One mistake that is often repeated is that you can get max power for a CPU by multiplying max current by voltage. In general that is correct for electronics (P = IE) with an ideal voltage and current source, but on a motherboard when IDD = IDD max, VDD < VID_VDD. IOW, P < (IDD max)(VID_VDD) on a processor because voltage drops at max current. That's not a problem, it's built into the VRM spec.

And to toss in one more thing to a long winded post: the power conversion on the motherboard and PSU also makes the P4 CPU look worse than it actually is. For example, if the voltage regulator section on the motherboard is 80% efficient and processor P uses 20W more than processor A, the power supply input to the motherboard will show an increase of 25W vs processor A. Further up, the load from the wall to the PSU will show a bigger increase due to its own conversion efficiency: say the PSU is 70% efficient, then that 25W difference becomes 35W. It's all part of the system power consumption, but it's more accurate to lay the increased power consumption where it belongs... not all of it is the CPU.

 

[Scali]

You have to realize that chip design is a dynamic business, not a static one.
What seemed like a great idea 5 years ago, may be the most stupid idea today, because so many factors have changed as the different technologies required for manufacturing processors have evolved at different rates, and you run into unforeseen obstacles.

The obstacle for the P4 turned out to be the exteme power leakage at 90 nm. Up until then, P4 was the CPU to beat. AMD got lucky because their older design which led to the K8 turns out to be a better design today. Intel's design from that era (PPro/P2/P3/P-M) is therefore also a better design today.

At the time, P4 was the way to go, because P3 had trouble scaling to higher clockfrequencies, and getting more IPC would mean a redesign of the chip anyway.
Intel figured that if they tilted the balance more towards clockspeed, they'd get more performance than when they would try to boost IPC. And they did. AMD chose the other route, and the P4 stomped all over the Athlon/Athlon XP for quite some time.

But today the balance has shifted, so Athlon64, built on the Athlon XP design, has now overtaken the P4 again. But Athlon64 may not be the ultimate CPU today, and will certainly not be the ultimate CPU in a few years. So the challenge for Intel is to build a CPU that gets even more performance from today's technology, and continue to scale better for the next few years. The key is to have a new design ready when the balance shifts, but with P4 the balance shifted sooner than expected.

 

[DanK]

What seemed like a great idea 5 years ago, may be the most stupid idea today, because so many factors have changed as the different technologies required for manufacturing processors have evolved at different rates, and you run into unforeseen obstacles.

The obstacle for the P4 turned out to be the exteme power leakage at 90 nm. Up until then, P4 was the CPU to beat.

...

Intel figured that if they tilted the balance more towards clockspeed, they'd get more performance than when they would try to boost IPC. And they did. AMD chose the other route, and the P4 stomped all over the Athlon/Athlon XP for quite some time....

That's not exactly the way I remember it. As I recall, the Pentium 4 had terrible initial performance, because it was designed for high clockspeed almost totally for marketing reasons, since the general public equated performance with clock speed. Indeed, if you read part two of Jon Stokes' history of the Pentium on Ars Technica, you'll find:

Perhaps the most common gripe about the Pentium 4's microarchitecture, called Netburst by Intel, was that its staggeringly-long pipeline was a gimmick — a poor design choice made for reasons of marketing and not performance and scalability. Intel knew that the public naively equated higher MHz numbers with higher performance, or so the argument went, so they designed the P4 to run at stratospheric clock speeds and in the process made design tradeoffs that would prove detrimental to real-world performance.

I was one of the original dissenters from this school of thought, and in my P4 vs. the G4e series I tried to make a plausible technical case for why the P4's designers had made some of the design decisions that they did. I ultimately managed to convince myself and not a few others that the P4's deeply pipelined design was, in fact, performance-driven and not marketing-driven.

That was then, and this is now. As it turns out, the P4 bashers were right. Revelations from former members of the P4's design team, as well as my own off-the-record conversations with Intel folks, all indicate that the P4's design was the result of a marketing-driven focus on clock speeds at the expense of actual performance and scalability.

The P4 didn't gain parity with the Athlon XP until the beginning of 2002, when the 0.13 micron Northwood core came out (and it's been back and forth since then). As far as the transition to 90nm, just about everyone got bit by that (including IBM; AMD moved to 90nm fairly late, after most of the kinks had been worked out).

At the time, P4 was the way to go, because P3 had trouble scaling to higher clockfrequencies, and getting more IPC would mean a redesign of the chip anyway.

So implementing an entirely new architecture is easier than cleaning up a design you've already got? The Pentium M redesign of the P6 architecture was Intel's first-ever processor to be completed on time and under budget.

 

[m0unds]

from tom-womack.net:

"Banias series

Banias is the core, designed at Intel's development labs in Israel, used for the Pentium M processor. It's a very sophisticated chip designed for high performance at low power consumption: an advanced P6 core (with enlarged L1 cache, equipped with SSE2 instructions and what seem quite sophisticated micro-architecture enhancements -- better branch prediction, micro-op fusion to send decoded instructions around in bundles, better stack handling), with elaborate clock gating so that most of the chip can be turned off at any given moment, fitted to a low-voltage 400 MHz Pentium 4 bus and a 1MB level-2 cache designed for low power consumption. Peak power consumption for any model is 24.5 watts.

Intel encourages manufacturers to pair it with the low-power 855 'Oden' chipset and an Intel wireless network device, by providing very substantial advertising support to the producers of this configuration, and not permitting any other configuration to use the much-hyped name Centrino. At launch, it was available in a variety of thin, light, powerful and expensive laptops, running at up to 1.6GHz and offering performance comparable to a 2.4GHz Mobile P4.

Dothan, released after some delay in May 2004, is the 90nm shrink of Banias. Essentially the only difference is a faster clock and a doubled cache size; unlike Prescott, this chip runs rather cooler than its 130nm predecessor. On SPEC benchmarks, the Dothan at 2GHz is comparable for integer work with a 3.4GHz Northwood or a 2.4GHz Opteron, and for FP with a P4/2666 - not bad for 21 watts peak power."


it's a whole history of core-ness.

http://www.tom.womack.net/x86FAQ/faq_cores.html

 

[Scali]

That's not exactly the way I remember it. As I recall, the Pentium 4 had terrible initial performance, because it was designed for high clockspeed almost totally for marketing reasons, since the general public equated performance with clock speed.

You can't prove that the high clockspeed was for marketing reasons. And initial performance of a new architecture is never all that hot. Both hardware and software need to be optimized to get the most out of it.
I simply don't agree with the popular Intel bashing that sites like Ars Technica like to do.

So implementing an entirely new architecture is easier than cleaning up a design you've already got?

Who said anything about easy? Intel has plenty of resources, they don't need to go for the easiest solution. They can take the solution that they consider best for the long term. So they ran into unforeseen problems this time... Get over it, next year there'll be a new architecture and P4 will be nothing but a memory. By then you're probably bashing AMD for not having a good successor ready for the Athlon64, and not being able to compete with Intels new architecture.

The Pentium M redesign of the P6 architecture was Intel's first-ever processor to be completed on time and under budget.

And is that the most important feature of a processor? I think not. I don't care how long it took to design a processor, and how much you went over budget, as long as I get good price/performance. Pentium-M isn't too hot in the price/performance range by the way.

 

[1c3d0g]

We was robbed.

The initial buzz when P4 came out was that the architecture would stretch to 10GHz. The hype machine kicked into overdrive, and everyone picked up 1.5GHz P4s, just to learn they were slugs compared to their 1.26 and 1.4GHz Coppermine P3's.

Pentium 4 was one miserable failure, and a great example of marketing triumphing over technological prowess. The Pentium M is as close to Pentium 3 than Prescott is to Northwood. I really believe that upper management at Intel finally realized that they screwed up, and tried to update and remarket their previous core.

That's the best way I've ever seen it put by anyone. Good job!

 

[Scali]

That's the best way I've ever seen it put by anyone. Good job!

But it's complete rubbish. As someone already pointed out, it was not Coppermine, but Tualatin, which was released much later. At around the same time as Northwood, if I recall correctly, and obviously Tualatin was no match for Northwood. Just like Coppermine was no match for the Willamette.
Sure, the fastest Coppermine or Tualatin may perform close to the slowest Willamette or Northwood, but that's no argument for lack of performance on the Willamette or Northwood side... On the contrary.

Other than that the P-M is not a replacement for the P4, but is aimed at a separate market segment. They are not even competing solutions (okay, there was a P4-M alongside the P3s for a while, to hold Intel over until the P-M arrived, but obviously P4 was never meant to be a mobile chip, just like P-M was never planned to be a desktop chip).

Both P-M and P4 get replaced in the next generation, so any argument about the P4 failing will go double for P-M, since it gets replaced after a much shorter lifetime, and with less succesful salesfigures.

 

[Worminater]

just like how the x800 series was based apon the 9700 not 9800

 

[lithium726]

But it's complete rubbish. As someone already pointed out, it was not Coppermine, but Tualatin, which was released much later. At around the same time as Northwood, if I recall correctly, and obviously Tualatin was no match for Northwood. Just like Coppermine was no match for the Willamette.

Tualatin was released before willamette. it was intel's first die shink to 130nm, they werent quite ready to do it yet with teh p4, as i recall they were having quite a few problems with it at that point. Tualatin was soon discontenued due to their outperforming the willy p4's

 

[Scali]

Tualatin was released before willamette. it was intel's first die shink to 130nm, they werent quite ready to do it yet with teh p4, as i recall they were having quite a few problems with it at that point. Tualatin was soon discontenued due to their outperforming the willy p4's

That's an interesting way to look at it.
Another way to look at it is that Intel (and many other companies) usually tries a new process out on 'less critical' components first, like budget processors or chipsets. In this case the P3. Makes perfect sense, since it's a much smaller and simpler chip, and therefore easier to get the 130 nm process underway and solve any problems.

Willamette had already been around for quite a while by then, and Northwood followed shortly after Tualatin.
And why do you assume Tualatin was discontinued because it outperformed the Willamette? That wouldn't matter for Intel, would it? Whether people buy Tualatin or Willamette, it's an Intel chip, so Intel gets the profit.
Didn't they just discontinue it because the demand was too low, making it hard to make profit on them?

Also, if Tualatin is so fast, then why is it so low compared to even Willamette in most of these benchmarks: http://www.tomshardware.com/cpu/20041221/cpu_charts-13.html#opengl ?
Usually the P3 Tualatin 1.2 GHz performs around the same as the 1.3 GHz Willamette. And that's the fastest Tualatin against the slowest Willamette.
I think someone has a distorted view of reality...

 

[lithium726]

eh, it was something like that. i think the whole " the pentium 3 performs better than teh pentium4" was not an image taht intel wanted for its flagship chip, you know?

yeah, i figured that tualatin was just a test for 130nm, but it was a rocking awesome test! haha, i still love mine.

we obvoiusly have different memories about 2000/2001 id look up some release dates, but im feeling lazy right now

Also, if Tualatin is so fast, then why is it so low compared to even Willamette in most of these benchmarks: http://www.tomshardware.com/cpu/200...-13.html#opengl ?
I think someone has a distorted view of reality...

cuase the p3 always had a weak point in gaming (at lesat towards the end of its life it did). in many things, tualatin was keeping up or outperforming willamette chips.

 

[Scali]

cuase the p3 always had a weak point in gaming (at lesat towards the end of its life it did). in many things, tualatin was keeping up or outperforming willamette chips.

There are more than just gaming benchmarks on that site, the Tualatin loses to most Willamettes in all but perhaps 2 or 3 benchmarks.
Your view of reality is more distorted than I thought, it seems to be permanent too.

Besides, isn't that the exact same reason why the P4 is considered a worthless piece of sh*t by most people on this forum today?

 

[lithium726]

There are more than just gaming benchmarks on that site, the Tualatin loses to most Willamettes in all but perhaps 2 or 3 benchmarks.
Your view of reality is more distorted than I thought, it seems to be permanent too.

i'll check it out.

im basing all i have said up to this on 4 year old memories, i havent recently benchmarked the dual tualatin rig i have next ot me.

edit: it would be really intersting to see what setup they are actually using, as it seems like they would be using the 256k version of the tualatin. they also do not include the 1.4ghz chip, which kept up better than a 1.3ghz celeron (highest tualatin core they have on there)

also, gotta consider that tualatin was a hell of an overclocker. i have two running in dual at 1.75 ghz, rock solid stable. according to sandra, its competitive with opterons in all but memory bandwidth. (which is something that really hurts them)

Besides, isn't that the exact same reason why the P4 is considered a worthless piece of sh*t by most people on this forum today?

i cant speak for others, but the reason i'm not partial to it was becuase it ran twice as hot as the tualatin(30w to 60w, comparing highest end parts... 1.4ghz/1.7ghz... 1.7ghz was top when they came out, right?), give similar performance, and was bundled with RDRAM if you didnt want to choke it horribly with SDRAM.

 

[Scali]

By the way, I looked up the introduction dates in Wikipedia:
Willamette is from November 2000.
Tualatin from July 2001.
Northwood from January 2002.

So Tualatin is not older than Willamette at all, which I already knew ofcourse.
But some people's reality is so dang distorted, it's dangerous to their health.

 

[lithium726]

So Tualatin is not older than Willamette at all, which I already knew ofcourse.
But some people's reality is so dang distorted, it's dangerous to their health.

"dangerous to my health?"

chill out. i was wrong, you were right. you dont have to be an ass about it

 

[Scali]

chill out. i was wrong, you were right. you dont have to be an ass about it

I'm just getting rather tired of hearing the ongoing stream of lies and misinformation to discredit Intel's products.