Are overclocked P4s really 'cooler' than overclocked Athlon XPs?

[aznskickass]

I know P4s have a larger die surface and all, but is the saying that P4s are 'cooler running than Athlon XPs' a misconception, at least in the wattage sense?

http://computernerd.com/cgi-bin/thermalcalculator.cgi

P4 1.6A @ 2.4GHz 1.7V - 91.4W
Athlon XP Palomino 1800+ @ 1.8GHz 1.95V - 90.7W

P4 2.26 @ ~3GHz 1.7V - 101.5W
Athlon XP Thouroughbred 2200+ @ 2GHz 1.85V - 89.9W

The P4 wattages are with thermal throttling disabled. Both AMD and Intel CPUs have voltages 0.2V above default.

Compare those wattage figures to the Tbird 1.4:
Athlon Thunderbird 1.4GHz 1.75V - 72.1W

Hmm. Overclocked P4s don't seem very cool running to me anymore...

So whats the deal with all this 'cool running P4s' that I've been so used to hearing since Northwood was introduced? Is the above thermal calculator accurate in its wattage calculations? Are there some factors that are not taken into account here?

Any comments/insight on this would be appreciated.

 

[Mikki]

This is all news to me, I'll have a look into it, but I can tell you for certain, P4's run ALOT cooler than XP's in the temperature sense, whether overclocked or not. Your 1.6a example, even at 1.71v with a stock hsf will run at about 38c or so. HTH 🙂

 

[Mikewarrior2]

this is where contact surface comes into play, as well as the IHS and other factors.

THe p4 has better thermal management, and with the larger contact surface in addition to the IHS, results in lower temps.

In addition, clock for the clock, the P4 northwoods are cooler. YOu've got a 3ghz Northwood matching heat output of a 2ghz Palomino.. 😛


Mike

 

[WarCon]

With the same exact setup except motherboard and processor. I had the same watercooler (etc.). The air coming from my blowhole while using a 1gig T-bird @ 1.466gig (2.05vcore) ~ 100watts was warm, while the air coming from my 1.6A @ 2.64gig (1.85vcore) wasn't.

Tragic side note. I am having to get a new processor because 1.85vcore really does do permanent damage to the overclockability of the processor. It is running ok at 2.133 now (1.75vcore), but it use to do 2.4 at this same voltage. So I won't be taking my voltage over 1.7-1.75vcore no more.

 

[Mikki]

Another case for the books, thanks WarCon for the confirmation. Sorry about your chip 🙁

 

[aznskickass]

Originally posted by: Mikewarrior2
this is where contact surface comes into play, as well as the IHS and other factors.

THe p4 has better thermal management, and with the larger contact surface in addition to the IHS, results in lower temps.

In addition, clock for the clock, the P4 northwoods are cooler. YOu've got a 3ghz Northwood matching heat output of a 2ghz Palomino.. 😛


Mike

Sure, surface area matters quite a bit, as seen in the Tbred/Palomino situation. But that doesn't explain why a Palomino 1800+ @ 1.8GHz will run hotter than a P4 1.6A @ 2.4GHz, as I'm pretty sure the Northwood and Palomino have similar sized cores.

As for the IHS, it actually increases the CPU temperature slightly, because of the extra thickness the heat has to travel through. All the IHS does really is protect the core. It has been proven before that taking off the IHS reduces temperatures by a few degrees.

As for the 'better thermal management', can you elaborate on what you mean? I'm talking both CPUs at full load here, so no thermal management should be in place such as shutting down parts of the CPU when it is not at 100% load.

As for the 3GHz NW, I used that as an example because it is what many people are hitting. Yes, I know clock for clock P4s are cooler, but we live in the real world, we don't run our P4s clock for clock to Athlons do we? 😛

 

[BlvdKing]

Could the 20 stage pipeline of the P4 have anything to do with it? I have heard that a cache miss results in the CPU sitting idle for awhile because of its 20 stage pipeline. Idle = not as much heat produced. This is part of the reason why Intel was reluctant to enable hyperthreading on CPU's using the .13 micron process... Increase in IPC = more work done per cycle = more heat.

I am hearing now that Intel may enable hyperthreading on its 3.0 GHz chip, so maybe they figured out a way to control the heat produced by a more active CPU on the .13 micron process.

Just throwing out some ideas...

 

[Naythn]

You should do a comparison with the Athlon and Pentium overclocked an equal percentage of their original clockspeed.

 

[MadRat]

VIA chipsets, the ones primarily used in Athlon setups, probably generate more heat than their SiS and Intel counterparts.

 

[Mikewarrior2]

Aznskickass,

are you trying really hard to try and be a zealot or something?

Did you not listen to Warcon's post?

One thing that also makes CPU heat calculators hard to measure is this: Intel chips are harder to get to "true 100%" load than AMD chips. Looking at both PDFs from both companies, AMD posts both "normal" max wattage and "max" wattage. its usually about 6-8W apart. Looking at intel pdf's, they only show the "Normal" max wattage. There is a greater difference between intel "typical" wattage versus max wattage.

Now, in order to max a CPU, you'll need a program or hardware device that pushes all parts of the CPU 100% at the same time. And no, prime95, rc5, folding@home, and even CPU burn do not do this for either CPU.

IE: Stock 2.26 "typical" = 56 watts; Computernerd calculator =~67W.

WHy the disparigy between intel typical and max wattages? Maybe PM or Wingznut can come in and answer that question. However, its entirely possible that in order to full max a p4 (since its architecture is different), you need highly specialized hardware/software not available to the public. So comparing "max thermal dissapation" could be a moot point.

In addition, let's look at some tbred temps that aren't crap/skewed socket-thermistor. Maybe the socket-thermsitor brainwashing has you confused for what actual tbred temps are: TechPC Int. Diode Tbred heatsink review. NOte that the heatsinks in this review are very clearly top notch heatsinks all the way around. And pay close attention to the temps, since they're internal diode temps.


Mike

 

[mechBgon]

Sure, surface area matters quite a bit, as seen in the Tbred/Palomino situation. But that doesn't explain why a Palomino 1800+ @ 1.8GHz will run hotter than a P4 1.6A @ 2.4GHz, as I'm pretty sure the Northwood and Palomino have similar sized cores.

Absolutely not. Go forth to AnandTech.com and research the die sizes. 🙂 The lower core temp of the P4 is quite natural considering its larger core size. edit: the difference is less than I thought, though, so maybe not such a big factor after all... details are here

I've read the technical documentation that ComputerNerd references. My impression is this: given a chassis which meets Intel's design requirements for airflow, and given that the P4 will throttle back if it's getting too hot, the documentation shows that for most apps, a heatsink which can dissipate 75% of the CPU's max output on an ongoing basis is all that's needed. Spikes will be absorbed, sustained overload will be dealt with by throttling. Make sense? Of course, a heatsink that can handle 100%+ of max wouldn't hurt either.