Prescott... Why?

[ShaolinDragon]

From what Ive read, Intels Northwood cores performed a LOT better than Prescotts, and produced less heat, so why o why did Intel make this move? (Just wondering, Im getting an AMD64 anyway )

 

[ignitionxvi]

its cheaper

being based off of a smaller scale = more die's per wafer

also, it wasn't a LOT worse, it was about even, and improved at higher clocks.

 

[liquidtrance123]

its cheaper

being based off of a smaller scale = more die's per wafer

also, it wasn't a LOT worse, it was about even, and improved at higher clocks.

My question isn't why not the Northwoods, my question is why not the Pentium M's!!!???

such a better CPU all around:
Less heat.
Lots of Cache (maybe prescott can support more, not sure)
and fewer stages.

 

[xonik]

The Pentium M definitely has its strengths and weaknesses. Compared to the Pentium 4, it's either much better or much worse performing, depending on the benchmark. Here is some examples of what I'm talking about,

http://www.gamepc.com/labs/view_content.asp?id=dothandesktop&page=9

So I wouldn't be so quick to say that it's an all-around better CPU, but for certain purposes it has no peers.

 

[Dan_D]

The Northwood was close to it's limits on air cooling in terms of scaling the clock frequencies. Prescott scales better at higher frequencies and produces better performance after a certain speed, where the Northwood starts to run out of steam. That mark being between the 3.4GHz and 3.6GHz range.

The Pentium M does have it's strengths and weaknesses. But overall I think it's weaknesses lie in the platform it's using. The 855M chipset doesn't support any of the modern features we find so common place as it's a notebook chipset. It's single channel PC2700 memory interface is less than spectacular.

Even stranger considering that the Banias and Dothan are essentially P6 cores like the Pentium Pro/II/III were. Think of it as a Pentium III with SSE2 and Pentium 4's cache design. That's oversimplification as it's more than that and that's not entirely accurate. But for the sake of making it easier to understand the basics of its design we'll go with that analogy. The Pentium Pro's design was WAY ahead of it's time.

Basically it's amazing that in any category it can win on such a crappy motherboard chipset. You give it Dual Channel DDR 400 or DDR 2 667, PCI-Express x16, SLi, 1066Mhz FSB, SATA II and give the processor itself SSE3 instructions as well as EM64T and you've got yourself a powerhouse the likes of which an Athlon FX-55 might not match up against. If Intel did that, it might become the enthusiests choice and put AMD back down to being thought of as a knock of processor manufacturer that's cheaper and runs hotter. Because if they could get the Pentium M on a platform comparable to the Pentium 4's LGA775 boards I'd say that's about what you'd have.

I don't know why Intel really hasn't done more with that core to bring it to the desktop. Rumor has it that eventually a form of it will.

The thought of Dual Core Dothan's (Pentium-M's) gives me wood.

 

[xonik]

Yes, but by increasing the front side bus the power dissipation will go up substantially, into typical desktop numbers, I would suspect. Right now, the Dothan is doing well thermally because the FSB has been held back so far. Rumor has it that the 533 MHz bus Dothans will have significantly higher power dissipation figures (still impressive numbers, however).

The rest of the features will come as the core logic dictates. SSE3 and EM64T will have a minor impact for a while, and dual channel DDR won't have a very impressive impact, I'm sure. Seeing as how similar processors from AMD and IBM didn't benefit much from lots of memory bandwidth, I'd expect the same for a future Pentium M chipset.

But yes, ignoring any AMD advances in the meantime, a desktop-intended Pentium M would do quite well.

 

[liquidtrance123]

The Pentium M definitely has its strengths and weaknesses. Compared to the Pentium 4, it's either much better or much worse performing, depending on the benchmark. Here is some examples of what I'm talking about,

http://www.gamepc.com/labs/view_content.asp?id=dothandesktop&page=9

So I wouldn't be so quick to say that it's an all-around better CPU, but for certain purposes it has no peers.

Ok so basically the Pentium M's aren't that great at pure number crunching?

 

[uclajd]

So I wouldn't be so quick to say that it's an all-around better CPU, but for certain purposes it has no peers.

Dood, the Dothan was fighting with one hand behind its back: single channel, antiquated 333 chipset, limited FSB (I have yet to see a board that will do over 175), preposterously bad cooling, and it still was competitive with the top Pentiums and Athlons.

As far as the FSB increasing heat, these thing put out 21W at stock, one fifth of the top P4s, one fourth of the AMDs, and they are competitive or faster in many benches. How would minor tweaking of the chipset and FSB quadruple the heat of this chip?

So I will be very quick to say that if you put that chip on a modern board, maybe add an on-die mem controller, it smokes everything (the P4 would probably still be better at video encoding and multitasking).

Look at these Doom 3 numbers @ 2.3.

And the 2.0's with increased voltage are hitting 2.8 on air. If someone would make a board with a friggin 478 heatsink retention mechanism, these things would rule, even with the crappy mobo limitations. I'd buy one and strap my Vapo on it, run it at 3GHz+, and it would be all over.

 

[BitchBreaker]

Thats for damn sure! the lack of dual chan and the crappy chipset made it less then what it could have been..But never the less it kicked so much ass just at stock that ya really cant bitch....As for the pressy's...my sig says it all! got mine for $240 with shipping and all that crap..got it home,took it out of the box and placed it in my rig,fired it up and it posted first time at 250fsb with 1.5v iv been bumping it up slowly over the past 2 weeks with np,cooling is my only limiti right now untill my pump gets here..but 4ghz on stock air...there not that bad..and my 3.0c was DAMN close to this chips Wattage at stock..Pressy owns!

 

[Dan_D]

Yes, but by increasing the front side bus the power dissipation will go up substantially, into typical desktop numbers, I would suspect. Right now, the Dothan is doing well thermally because the FSB has been held back so far. Rumor has it that the 533 MHz bus Dothans will have significantly higher power dissipation figures (still impressive numbers, however).

The rest of the features will come as the core logic dictates. SSE3 and EM64T will have a minor impact for a while, and dual channel DDR won't have a very impressive impact, I'm sure. Seeing as how similar processors from AMD and IBM didn't benefit much from lots of memory bandwidth, I'd expect the same for a future Pentium M chipset.

But yes, ignoring any AMD advances in the meantime, a desktop-intended Pentium M would do quite well.

Dual channel DDR might not add that much to it, but a 800MHz or 1066FSB would. As far as wattage is concerned for small form factor owners that might matter. So long as my PSU can handle it I don't care if it draws close to what a Pentium 4 Prescott does.

 

[bountyhunter]

Dual channel should benefit this more then it would say an athlon xp or the pentium III.....the bus is quad pumped, so I'm thinking that means it performs better with extra bandwith. Since these things will work on the same motherboards that P4's will (they just need a different socket I believe) then ASUS needs to strap a m479 socket on its P5AD2-E motherboard and we can get this up to a 1200mhz bus. That would be amazing. I find it very odd that no companies have done this yet......

 

[xonik]

Do you remember the VIA PT800? It was a single-channel DDR400 chipset released at around the time of the Intel i865 and 875 chipsets. If you check benchmarks from that time, you'll see the PT800 consistently performing as well or better than the i875P/i865PE, despite its bandwidth inadequacy. That's right, even for Pentium 4s, memory bandwidth is not what it's cracked up to be.

 

[xonik]

Since these things will work on the same motherboards that P4's will (they just need a different socket I believe) then ASUS needs to strap a m479 socket on its P5AD2-E motherboard and we can get this up to a 1200mhz bus. That would be amazing. I find it very odd that no companies have done this yet......

That's not correct. Pentium Ms may fit in a similar socket as Pentium 4s, and share similar architectural features, but the Pentium M is NOT core-logic-compatible with Pentium 4s to a practical extent.

 

[Dan_D]

Not to mention that from a marketting standpoint few people would buy a 2.0GHz Pentium M instead of a 3.6GHz Pentium 4. Considering the Pentium M costs about the same. It just won't work.

 

[Stugots]

...but the Pentium M is NOT core-logic-compatible with Pentium 4s to a practical extent.

the Pentium M is more similar to a tualatin then it is to a northwood or a prescott.

 

[Dan_D]

the Pentium M is more similar to a tualatin then it is to a northwood or a prescott.

True.

 

[ameoba]

Not to mention that from a marketting standpoint few people would buy a 2.0GHz Pentium M instead of a 3.6GHz Pentium 4. Considering the Pentium M costs about the same. It just won't work.

Why do you think Intel's moving to model numbers instead of speed ratings as the primary CPU label?

 

[Dan_D]

Why do you think Intel's moving to model numbers instead of speed ratings as the primary CPU label?

Well it's to facilitate going to lower clock speed CPU's that are more effecient. But right now I don't think anyone would buy a Pentium M versus a Pentium 4 in a desktop yet. You have to overcome 20 years of "more MHz=Faster" I get people comming into my work asking why a 2.0GHz AMD Athlon 64 system costs as much as a Pentium 4 3.2GHz. (Not sure if that price comparison is valid it's just an example). Then I get this what does 3800+ mean? crap all the time.

AMD's been fighting it, Apple's been fighting and now so is Intel. But that kind of marketting won't undo what's been in peoples mind's for the last couple decades.

 

[uclajd]

So long as my PSU can handle it I don't care if it draws close to what a Pentium 4 Prescott does.

Great point. But you know that a 2.0 Dothan at 1.7v running at 4750 would put out...80W.

My point being, Intel has some wiggle room.

That's right, even for Pentium 4s, memory bandwidth is not what it's cracked up to be.

Based on the scores Dothans are getting in the 3DMarks (great overall, but really low in the tests that require bandwidth), the general consensus by really smart people seems to be that if the Dothan had a higher bandwidth platform, it would smoke the FX's like the last joint at a Reggae concert.

 

[Nexus]

This is the hope that Ive seen posted on Ars quite a bit. That the dual core direction will be taken with modified Pentium M's and not the over complex northwood>prescott line.
That intel is finally realizing increasing the pipeline to ramp "DA MEGA-HURTS" is not going to work out for them ad infinium.

That if they took the P-m's to the desktop, with its lower power consumption, they can pump up with the FSB, update the core with modern features and make it dual core, and design a desktop chipset with all the goodies were expecting (dual channel ddr-2, pci-e, ect.) They would have something to knock on AMD's door with.

Still, I dont think anyone knows for sure yet, I thought I saw somewhere the dual cores are being introduced with prescotts, which make this Intel supporter kinda sad.

I have to plunk some cash on a server for home here soon and my choices are probably going to be between a P4 northwood/875 chipset or an A64....

 

[nylint]

This is the hope that Ive seen posted on Ars quite a bit. That the dual core direction will be taken with modified Pentium M's and not the over complex northwood>prescott line.
That intel is finally realizing increasing the pipeline to ramp "DA MEGA-HURTS" is not going to work out for them ad infinium.

That if they took the P-m's to the desktop, with its lower power consumption, they can pump up with the FSB, update the core with modern features and make it dual core, and design a desktop chipset with all the goodies were expecting (dual channel ddr-2, pci-e, ect.) They would have something to knock on AMD's door with.

Still, I dont think anyone knows for sure yet, I thought I saw somewhere the dual cores are being introduced with prescotts, which make this Intel supporter kinda sad.

I have to plunk some cash on a server for home here soon and my choices are probably going to be between a P4 northwood/875 chipset or an A64....

The way the roadmap looks thus far - the first dual-core intel chips will indeed be presshot based (smithfield)

 

[Nexus]

yeah, like I said (typed?) I thought I saw that somewhere too, prolly a roadmap posted somewheres...

 

[FreiDOg]

The way the roadmap looks thus far - the first dual-core intel chips will indeed be presshot based (smithfield)

If Montecito doesn't get their first.

 

[DanIdentity]

If Montecito doesn't get their first.

I believe he meant the first dual-core desktop chips, as Montecito is the multi-core successor to the Itanium, and can't even run 32 bit code. But at any rate, Smithfield will almost definitely come before it.

 

[BitchBreaker]

I believe he meant the first dual-core desktop chips, as Montecito is the multi-core successor to the Itanium, and can't even run 32 bit code. But at any rate, Smithfield will almost definitely come before it.

I agree

 

[CrimandEvil]

Yeah first dual core is using Prescott although I won't touch Intel until Dothan based chips come out.

Dothan>Prescott

 

[BitchBreaker]

Yeah first dual core is using Prescott although I won't touch Intel until Dothan based chips come out.

Dothan>Prescott

Im touching my intel right now its all warm and stuff

 

[CrimandEvil]

Ewwww!

 

[matt9669]

Smithfield is about as elegant as the whole Pentium 4 design goal, which goes something like a caveman: ME CLOCK FAST!! ROAR!!

They slapped two Prescott 1MB cores together, hooked them to a shared 800MHz bus and hoped nothing exploded. That's about it.

Remember, Prescott was originally slated to clock past 4GHz, and even then it was just to hold the fort until the killer Tejas arrived, at 5GHz+. That's all ancient history now.

Dothan was designed for high IPC from the get-go, so naturally it compares very well even to the A64. But these things weren't designed for desktop use, and I'm sure there's more than one Intel engineer telling the execs "So what - it's a better chip!" Glad I'm not fighting that corporate battle . . .

Plus, I agree with xonik, especially with the software trend heading towards parallelism. Dothan isn't the best design for all cases - P4 actually handles vector FP math very well, though only scientific apps use such math extensively - Dothan is very efficient in sequential apps (i.e. current games, and most other software for that matter) but I wouldn't want a Dothan-based server.

Sorry for being long-winded, but my point is even if they wanted to, Intel can't just turn on a dime and make Dothan and its derivatives their mass-market desktop platform. I'd be willing to bet even an improved P6 architecture won't look that all that impressive on software a few years from now. EDIT: that's assuming software changes as the industry expects it to, but Dothan-based designs could easily rule the roost given otherwise . . .

 

[matt9669]

Montecito is the multi-core successor to the Itanium, and can't even run 32 bit code.

It can, but not natively - it has to emulate it. I don't even have to tell you when I say emulate, I mean it's slooooww . . .

If someone would make a board with a friggin 478 heatsink retention mechanism

I could be wrong, but I thought I read somewhere someone is doing this - can't seem to find it now tho . . . *doh*

 

[muddocktor]

If someone would make a board with a friggin 478 heatsink retention mechanism, these things would rule, even with the crappy mobo limitations. I'd buy one and strap my Vapo on it, run it at 3GHz+, and it would be all over.

The Aopen mATX mobo uses a standard 478 HRM. There is no reason I know of why the mobo manufacturers haven't produced a board using the i865 or i875 or even i915 chipsets for P-M except that they haven't seen the market for it and possibly pressure from Intel. After all, according to a few articles at Anandtech, the P-M and P4 are electrically the same and can use any chipset the P4 can. Here's a quote from his Banias article:

In order to feed the higher IPC execution core, Intel outfitted the Banias with a 64-bit 100MHz quad-pumped FSB, identical in design to the Pentium 4's FSB. The Banias' FSB is even electrically compatible to the Pentium 4's FSB, which is why any Pentium 4 chipset is able to interface with the chip as we saw at IDF with this E7501/Banias setup:

Source, also check out the pic with a Banias utlizing an E7501 chipset on an industrial board.

Even saddled with the crappy and outdated i855 chipset, some folks at the Extreme forums have been seeing remarkable performance out of overclocked Dothans. 2700 doesn't look like it's hard to do even on air, since the proc's TDP is so low to start with. BTW, Dothan is downward unlocked just like an A64, so the main holdup for high fsb numbers is the i855 chipset's performance on the few boards out there, so far. They seem to be topping out at around 166 fsb or so, even using a 6 multi.

 

[xonik]

FSB compatibility doesn't imply or guarantee full CPU compatibility. That's why the Pentium M is only compatible with a few current and future chipsets.

 

[muddocktor]

FSB compatibility doesn't imply or guarantee full CPU compatibility. That's why the Pentium M is only compatible with a few current and future chipsets.

How do you know this to be true with P-M and P4 chipsets? Do you work in the industry or for a mobo manufacturer or something? This isn't meant to be a flame BTW; I would like to know why P-M can't be used with a modern chipset that can run a high fsb speed. I know that it's not due to the P-M being based partly on the old P6 architecture because I used to have a P3S that would run at 180 fsb when pelted, much higher than what those guys at Extreme are getting with the present boards and P-M.

 

[xonik]

Okay, here's what's up. I may have erred by saying the Pentium M is not compatible with P4 intended chipsets. It is electrically compatible, but you can't have a motherboard that simultaneously supports the Pentium 4 and the Pentium M. This is because the pin assignments are different, and the two processors are physically incompatible with the same socket.

This is why very few Pentium M motherboards have been built. Like with other processors that lie "off the beaten path," the market is not big enough to warrant the creation of a motherboard. It would require a new motherboard PCB, which is costly and risky. I think we're lucky to find motherboards to support the lesser used processors like the Pentium M, AMD Geode, and the Transmeta Efficeon. As we speak, I am typing from a Pentium M SFF, and I wish that there were more Pentium M products out there for the enthusiast.

 

[muddocktor]

OK, I see what you're saying. I was interpreting you one way when you were talking about pin assignments. I know about the pin assignment issue (I checked that out when the Egg first started carrying the P-M procs) and that's a bummer for sure (and intentionally done by Intel most probably to keep from robbing P4 market share, since they meant the P-M for mobile use only). Still, maybe someone like Shuttle will see the light and design a SFF box that uses a modern chipset.

BTW, how do you like that P-M for desktop use and what mobo do you have? I've been thinking seriously about the Aopen board and a Dothan 1.6 or 1.7 and overclocking the hell out of it.