There's one big catch to Pentium M and that is its inability to clock above around 2.26 GHz without more voltage. I followed overclocking of these quite a bit back in 2005 because I had a Pentium M notebook. You could pin mod the socket and run a 400 MHz FSB @ 533 MHz and get an instant overclock. So people would buy the 1.6 GHz/400 and run it at 2.13/533 and this was sometimes possible with stock voltage. You could pin mod VIDs too if you needed to. But if you up the volts, the power goes up too and bye bye power advantage.

There are some other little quirks to Pentium M, like its relatively weak FPU and SIMD performance. SSE2 is slower than SSE believe it or not. Core Duo improved SSE2 a bit, but Core 2 was where they hugely boosted it with major redesign of the CPU.

In other words, regardless of their power output, Athlon 64 and P4 can be quite a bit faster due to clock scalability. But you can't beat that 30W TDP of Pentium M for notebook performance per watt, especially against the horrible Pentium 4M.

Pentium 4-Mobile suck and swaaye is rigth it was relatively weak FPU and SIMD performance and yes the SSE2 was slower the SSE now the Mobile Pentium 4 the was diff story is just to bad that intel didn't do Pentium 4 Cedar Mill in Socket 478 package

or screw the Pm and as the socket A was a strong contender supposedly it did have a shut down temp sensor but, no chip set ever took advantage of it. I may sound weird for this but, you know how early on AMD made chips that ran on intel platforms i wish someone would make a few chips for older sockets like 478 or even s370 just small cheap efficient CPUs to give old platforms one last boost, obviously thats not gonna happen but, would be cool to see how or what new tech would be used with old plat form hard ware limits.

the s478 was super ceded by lga775 one of the most universal sockets ever made now it's out date my whole new platform witch will be out dated by another whole new platform.

Tetrium wrote:

This thread was bound to become a shooting match 😜

Yeah, it's bound to be an emotional topic but I think most of the forumers here are mature and will not stoop to the "OMFG, P4 is the suxxors N008!" type of comments.

The various views are quite interesting reading, answering my question indirectly as to why almost no one chooses to (re)build a s423/478 rig even though it's from the same era as Athlon/ Tualatin systems- a lot of it has to do with nostalgia and bad feelings over Intel's actions during late 2000-2001- the RDRAM debacle, over-pricing Tualatins and then terminating them, retarding the 815 chipset, persecuting VIA, pricing vs. AMD etc.

Also I suppose, the Athlon was the enthusiast platform in that era and as many are seeking to relive past days of glory- that platform becomes the default choice for them.

As for the Tualatin, probably the 1337ness of having a creaky old P3 exceeding 1ghz and beating newer platform is a factor. I know it was for me when I built my various 1ghz + P3 rigs 😁

Tetrium wrote:

Anyway, I've read quite a lot about P4's supposed sluggishness. What would it take to have a Northwood, for instance, become 'quick'? Low memory latency?

Clock-for-clock it's an underperformer compared with , errrr, anything really 😵 but you must remember it can generate many more of those clocks than anything else from that era.

So with a P4 it's about getting as much mhz as possible, followed by memory speed and then latency. Example, a Northwood 1.6B is nowhere in the same league as a Tualatin at 1.6ghz (or even stock at 1.4ghz) or an Athlon at the same clock speed. But the Northwood can go far above 3ghz, lets say a 3.0c model at 3.6ghz or so- at that speed there is no Tualatin that can be overclocked to match it. Or take my Prescott 4ghz system for instance- it beats Athlon 64 3500+ chips clean off and matches/ beats the Sempron 140 (Phenom II architecture!) in some benchmarks. It is very very competitive with my old Athlon 64 3000+ @2.4ghz/ DFI NF3 system.

I will reserve my comments on AThlon XP chips as I have not owned any and I prefer to speak from experience rather than just talking from third party sources. I have heard of massive clocks in the 2.5+ range from them, at which point they're bound to be pretty fantastic.

I heard the Tualatin were better than the eraly P4s before i guess it wasn't till 2ghz P4s till it started closing the gap. the Tualatin on the other hand supported MP where the P4 did not hence the xeon.

So if more mhx and memory clock and timing are key then on of my original questions still remain valid just how far can northwood clock and whats the max safe vcore. also how many NM is the fab process CPUz just says 14um so i assume thats hundreds of NM?

@swaaye:
Cheers for you to chip in 😉

cdoublejj wrote:

I heard the Tualatin were better than the eraly P4s before i guess it wasn't till 2ghz P4s till it started closing the gap.

Personally I see this the same way. I regard the P4-2Ghz as roughly the same as the P3-1.4Ghz. But the supposedly sluggishness of the P4 is a bit unanswered for me, though I've read this on other forums as well.
Could this sluggishness be somehow related to P4's relatively slow performance per clock tick?

leileilol wrote:

Loads of P4 mobos come with a SoundMax audio chipset which has somewhat interesting DOS compatibility, though no FM at all 🙁

Also, the PowerVR PCX2 works fine on Pentium 4. It made great testing convenience when I had a P4+PVR setup, plus it probably helped make Deus Ex run great on it

There's a WEE bit of nostalgia to me involving the Pentium 4 and 2000-2001 gen computer stuff... sure it's 'post-3dfx death era' but that doesn't mean you can't make a top of the line Intel 2000 p4 1.2ghz Windows 2000/ME Sb Live 5.1 Geforce2 rig. 😀

^Agrees^
I look at this hardware from a "What can I do with it right now" perspective.
P4 had some (lots-of actually 😜) disadvantages compared to XP and A64, but one advantage is that you could put 4 crappy memory modules in there and still come out with a nice amount of RAM.
XP and A64 rigs have the disadvantage that, when using lots of banks of memory, the memory bus has to be underclocked. So even if you had 4 memory modules of DS 256MB DDR-400 and an Athlon XP board with 4 memory slots (if these were ever made), you couldn't run them @ 400Mhz.
In theory P4 has a better more practical cooling solution, but the Intel stock coolers are crap.
With sA, I can mix and match heatsinks with fans to give me the perfect HSF thats both quiet and cools well. The s370 Intel HSF's can be fan-modded (for lack of a better word) but the ones with plastic mounting clip won't fit right with my Nehemiah. On top of that, you need to remove the CPU fan to actually mount that Intel bugger.

I still plan to build me a P4 rig (or 2, second one with SDRAM 😵 ) so I can "junk" some of those 128MB/256MB DDR-400 modules I have laying around 😜

@ Tetrium: It is related to the slow per clock performance and the (relatively) slow clock speeds of early P4s. Once they started cranking them past 2ghz though, they became competitive with anything else out there at the time.

On A64 systems- I think the underclocked memory only affects s754 systems. Both s939 systems I owned ran their 4 banks at DDR400.

For s478, you can get stock-looking coolers that are pretty cheap and use a standard 80mm fan on them. But I'm curious why you find the stock cooler noisy- it seems quieter to me than the s775 cooler- but then such things are very subjective I guess.

cdoublejj wrote:

So if more mhx and memory clock and timing are key then on of my original questions still remain valid just how far can northwood clock and whats the max safe vcore. also how many NM is the fab process CPUz just says 14um so i assume thats hundreds of NM?

Northwoods can clock pretty well, here are some commonly achievable figures that don't require exotic cooling, ultra expensive equipment or other such nonsense and which can be achieved at home 😁

'B' 533mhz - usually 3.3ghz or so, best models are D1 stepping 2.4 and 2.8ghz. 2.66ghz seems to be one to traditionally avoid

'C' 800mhz- usually top out at 3.6ghz, 3.4ghz should be np. Have seen some over 3.6ghz but never at 4ghz except for some "mythical" machines using exotic stuff. 3.0C is the one to go for, though anything even the 2.4ghz is ok, just that the lower the multi, the better your mobo needs to be to withstand high fsbs.

The most important thing from my experience is 1) a stout psu- don't underestimate the power draw just because these are old machines- my Prescott 4ghz brings a Tagan 450w to its knees, and this is a very good psu that used to power my Phenom II 3.6ghz system just fine; and

2) Cooling is critical. Overclocked even NWs at the limit have TDPs approaching those of Presshotts. The stock cooler will become inadequate real fast. Best cooler back then was the Zalman CNPS 9500 but if you can ghetto rig a newer cooler on, it will be better. For my 4ghz run, I used a CM N620 huge mutha with some mounting hardware from an Aeolus cooler- voila, even the mighty Prescott at 4ghz operating at ambient temps of 28c was tamed.

I think Red Hill did a pretty good job at explaining the problem with the Pentium 4:

"To understand pipelining, imagine putting out a fire with buckets of water. If you are alone, you have to walk to the well, fill the bucket, walk to the fire, empty the bucket, walk back to the well, and so on. This is how a 386 works, going to the RAM, getting the next instruction, decoding it, running it, going back to the RAM for the next one.

Now imagine have four or five people to help put the fire out. One fills the buckets, one empties them onto the fire, the others pass the buckets along. You are still only pouring one bucket at a time, but it's much faster. CPU pipelining works the same way. Instructions are loaded one after another into the RAM end of the pipeline and the CPU just takes them from the other end as needed.

All modern CPUs are heavily pipelined. As a rough guide, the longer the pipeline, the easier it is to ramp up to higher clockspeeds, but the more severe the penalty each time the program branches and the CPU has to empty out the pipeline and start again with fresh data. This was one of the most significant differences between the Athlon XP and the equivalent Pentium 4: the Athlon had a long pipeline and operated comfortably in the 2000MHz range. The Pentium 4 had a very long pipeline and easily reached 3000MHz, but took a longer time to switch between different tasks — so much longer, in fact, that it was slower than an Athlon in general use. But where the P4 was presented with a long sequence of tasks that can be executed one after the other without interruption, it could be astonishingly fast."

which is why my Pentium M 1.86 laptop runs faster than my Pentium 4 640 3.2ghz laptop that overheats all the time

intel tried to win the cpu wars by selling numbers, they failed... they only got their senses back when they went back to using the pentium m core and evolved it to the core2 cpus

shear mhz numbers don't mean everything for speed... this was proven with how inefficient the P4 was in real world use

there's so little justification to use a P4 when there's better solutions out there... I for one don't want a space heater nor a bunch of fans buzzing to keep the damn thing cool

I'm surprised they actually tried using the Netburst core in laptops at all. One of the biggest reasons why Apple dumped PowerPC was because IBM couldn't make a low heat, low power version of the G5 for use in the Powerbooks. Those P4 M laptops must have been really uncomfortable to hold in your lap for more than a minute at a time. I had a eMachines M6811 and that had a Athlon 64 3400+ and that got damn hot. I can only imagine having a Netburst in there.

I've used a couple of P4 gaming notebooks. They used a desktop P4 instead of a P4M. In order to cool that they put in a radial fan that sounds like a hair dryer. The temps aren't crazy but the noise is, although it's not much different than the hotter video cards. Video cards have gone way past CPU power.

My aging XP gaming machine still serves my needs, socket478 + a severly overclocked P4 1,6Ghz@2666mhz ... (9 years old by now)
Should probably exchange that P4 for a better one some time, but.....

And also having fun with an ASUS s478 board, where a Pentium M resides (on a 'slotket' thingy).
Tried Speedsys(DOS) just to see how low I could go, and 6x 100mhz bus gave a whopping score of 405.33 (barely beating Celeron 333mhz, heh). 😎

Strangely enough when going 12x 200mhz (2,4ghz), speedsys 'only' gives me 1622.12 - had somehow expected more (only 4 times as fast as that lowly Celeron...). But then again, Speedsys is not the best testing program for this I guess....
(CPU gets named 'Pentium III' 🤣)

bestemor wrote:

My aging XP gaming machine still serves my needs, socket478 + a severly overclocked P4 1,6Ghz@2666mhz ... (9 years old by now) S […]
Show full quote

My aging XP gaming machine still serves my needs, socket478 + a severly overclocked P4 1,6Ghz@2666mhz ... (9 years old by now)
Should probably exchange that P4 for a better one some time, but.....

And also having fun with an ASUS s478 board, where a Pentium M resides (on a 'slotket' thingy).
Tried Speedsys(DOS) just to see how low I could go, and 6x 100mhz bus gave a whopping score of 405.33 (barely beating Celeron 333mhz, heh). 😎

Strangely enough when going 12x 200mhz (2,4ghz), speedsys 'only' gives me 1622.12 - had somehow expected more (only 4 times as fast as that lowly Celeron...). But then again, Speedsys is not the best testing program for this I guess....
(CPU gets named 'Pentium III' 🤣)

Pentium M is a revised Pentium III core which is probably why Speedsys identifies it as a P-III.

Yeah, I guess that might be it.
Though the comparison 'points' on the Speedsys 'bar' stops abruptly(mid bar) at P3 933mhz.
So I figured anything above, incl P4, would still be called 'Pentium III'...

Well, at least I know I have speed control all the way from a Celeron 333 and way up, only a reboot away... 😁
Where does really a Pentium M @2768mhz rate I wonder?
(vs the fastest P4's and above)

i'm running a 2.66/522 P4 with an IBM netvista CPU cooler at 3ghz/600 OC at some 1.5v core.

bestemor wrote:

Yeah, I guess that might be it.
Though the comparison 'points' on the Speedsys 'bar' stops abruptly(mid bar) at P3 933mhz.

I imagine that it looks at the CPUID reading and interprets a PM as a Super Future P3. 😀

whats the nano meter count on the Pentium 4? how big is the die in NM?

EDIT it's 130nm with todays CPUs reaching 32and 22nm. Wiki says all P4s were 180 nm to 65 nm. wow 180nm is one big chip.

cdoublejj wrote:

whats the nano meter count on the Pentium 4? how big is the die in NM?

EDIT it's 130nm with todays CPUs reaching 32and 22nm. Wiki says all P4s were 180 nm to 65 nm. wow 180nm is one big chip.

The 180nm chips are Willamettes. With that fab process it's remarkable that they got to 2ghz at all. Northwoods are 130nm, Prescotts 90nm- without any of the usual attendant benefits 🤣

Edit: And I've also broadened the subject to include Athlon XPs, Athlon 64s and PIIIs. Basically anything from year 2000 and before C2D, since they're bound to be furious debates about P4 vs P3/ Athlon XP/ 64 seeing as they all co-existed at some point or other. 😀

cdoublejj wrote:

whats the nano meter count on the Pentium 4? how big is the die in NM?

EDIT it's 130nm with todays CPUs reaching 32and 22nm. Wiki says all P4s were 180 nm to 65 nm. wow 180nm is one big chip.

It's right here
Willamette was using the same die process as Coppermine/Thunderbird/Palomino
Northwood was using the same die process as Tualatin/Thoroughbred/Barton/Early A64's
Preshot was using the same die process as the A64 Venice.
Cedar Mill was using the same die process as some of the early Athlon II's!!

The fact that Preshot had trouble keeping up with the early A64's is enough indication of how bad that architecture really was. Even despite the fact that Intel had a better die production, so to say, they couldn't keep up with AMD at the time.

Reading all these posts, I'm inclined to go with a sub-3Ghz Northwood for building a retro rig comparable with Athlon XP 2000+ or so.