Heat Spreaders?

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Belgarion
Posts: 36
Joined: Fri Oct 04, 2002 5:46 pm
Location: Maryland

Heat Spreaders?

Post by Belgarion » Sat Nov 09, 2002 7:11 pm

Perhaps a better title would be something like "Why are P4 chips so much easier to cool than Athlon's?"

My dual Pentium II system is starting to show its age and I have been thinking about upgrading. Last January I upgraded my wife's computer to an Athlon XP 1700+. The performance is good, but the Volcano 7 HSF and stock Enlight PS sure make a lot of noise. You don't realize how much until you turn the system off and it gets eerily quite (unless my computer is also on).

After a little searching on quieter computers I came across this site (great site, BTW!). After reading about the issues with power dissipation, core sizes, heat spreaders, heatsinks, fans, etc., I'm still wondering why P4's seem to be so easily quieted compared to Athlons. For example, checking the wonderful tables at http://users.erols.com/chare/elec.htm, I notice that an Athlon XP 2000+ has a power comsumption rated at 62.5W typical and 70W maximum. The "equivalent" 2.0A (1.5V) P4 is rated at 52.4W "thermal design power" (maybe roughly comparable to "typical"?) and no maximum given. However, judging from the P4 entries a little higher up, the thermal design power seems to be about 75% of the maximum dissipation (at least for the older cores), which in this case would make the max around 70W, the same as the Athlon. Yet the P4 is considered easier to cool (I'm working only with worst-case max power numbers here for simplicity).

The obvious first reason is that the P4 Northwood core die size is larger than either current Athlon core (i.e., Palomino and T-bred), which we know from, for example, Anandtech (http://www.anandtech.com/showdoc.html?i=1635&p=2). Hence the P4 NW will have a larger area over which to dissipate the heat produced, i.e., a lower power per unit surface area (what's the proper term for this quantity? Areal power density?). One would think that the heat spreader would make this even better, but I read in another thread here (http://forums.silentpcreview.com/forums ... php?t=2037) that the spreader is really to help physical issues (no crushed or scratched cores) and even out temperature hot spots, but in fact increases overall thermal resistance.

With the reasons given, the heat spreader seems like a reasonable trade-off. But why then wouldn't an Athlon benefit from a heat spreader? If the P4 core can be cooled so easily even with the heat spreader getting in the way, shouldn't an Athlon with similar power dissipation also benefit by a similar amount cooling-wise by having a heat spreader? Is there a flaw in my reasoning?

I am just musing about this because it seems a shame that the Athlons, even (or especially?) the T-breds, appear to be so problematic for quiet computing considering they have such a great price/performance ratio. You can buy an Athlon and expensive HSF combo (like the SLK800 plus Panaflo) for less than the stock equivalent P4. I figure I'll probably end up going with a P4 for my next computer upgrade anyway (or perhaps a Hammer if AMD gets them out in time and they can be cooled ok), but it seems unfortunate that AMD doesn't provide much competition in this regard.

Melchior
Posts: 23
Joined: Sun Oct 20, 2002 3:58 pm
Location: Vancouver, BC
Contact:

Post by Melchior » Sun Nov 10, 2002 2:37 pm

I don't think that the Athlons are necessarily *that* bad for quiet computing-- certainly they've got higher power consumption than a comparable P4, but they can still be made quiet.

I've got an AthlonXP 1600+. I've replaced the stock CPU fan with a Panaflo 80MM L1A (undervolted to ~7.5v). The heatsink is still a stock unit. My CPU die temperatures are: 55*C typical, 58*C max. These temps are within a degree or so of the temps that I used to see with the stock CPU fan. The loudest components within this computer are the two undervolted Panaflos (cooling the PSU+CPU), and they're pretty damn quiet.

As far as heat spreaders go, I know very little about them, so I can't comment there.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sun Nov 10, 2002 4:02 pm

Keep in mind that the reported temps for P4 by any mobo are wrong. Way too low at peak. SOme of the engineering fellas I met at IDF in San Jose (ADI, Intel) say the difference between the thermal diode reading and the hottest spot on the die can be as much as 15C!

I have to say that I have seen P4 CPU temps on one of my low airflow, low noise rigs hover around 58-60C while I am pushing it hard & long, and thought OK, I am safe. And then the system locks up, crashes, etc. Only one thing I can attribute this to: the core was running much hotter than reported by the diode/MBM. The main thing with the P4 is because I know about its throttling, I am not conerned about whether it gets too hot. If it does, the thing just stops working, but doesn't burn up.

Note that Intel is not giving info about max wattage on their CPUs anymore. I've been in touch with some systems integrators who say that the latest P4s -- the 2.5 GHz and up -- are VERY difficult to keep cool when pushed hard. Makes you wonder. Makes you also wonder if the "wrong" diode positioning is exactly what they intended...

So to address the question, "Why are P4 chips so much easier to cool than Athlon's?":

I don't think are. They're just harder to burn, which makes them safer, and they allow for larger space for the HS, which does mean the average generic P4 HSF probably is more effective than the Socket-A equivalent.

crisspy
Friend of SPCR
Posts: 228
Joined: Sun Sep 29, 2002 9:05 pm
Location: Powell River, BC, Canada

Post by crisspy » Mon Nov 11, 2002 2:07 am

Very interesting info about P4 Mike. Thank you for doing our homework for us.

Check out Darek's Dirty Little Secrets if you haven't seen it before. He has very high quality dirt from a very technical developer's perspective. His take on the early P4's is scathing.

I don't hate Intel, but ca$holes are ca$holes. And Intel is pretty central to this industry, so I would hope for better.

Seems Intel did everything they had to, no matter how low, to get the P4 out the door at max MHz. The rule was Maximize Clockspeed at Any Cost. That was the T-bird era, and AMD was catching flack for high heat. Intel would want to make AMD look bad (opposite strategies wherever practical?) so...

Guesstimate News: Intel Appeases Hard-Core Cooling Crowd with Big Easy Fix

Could it be? I haven't been paying attention to P4 power draw specs, and I wonder how many folks have actually accurately measured the peaks. Even getting caught red handed would barely register with Intel loyalists. My take on how Intel's war plan unfolded:

-AMD maximizes real performance at any cost / by any method, it's the only way they can compete. High heat be damned.
-Intel leans on Top market reputation, prioritizes strategies based on contradicting AMD. Intel does it this way, so AMD must be wrong.
-Intel focuses on max clock speed at any cost. Assertion: bigger numbers = better & faster. Reality: maybe after several core revisions.
-first round P4's have small cache and comprimised implimentation, but High MHz vs. time of release.
-Intel puts thermal diode in cool part of chip, quotes average intead of peak power: processor looks cooler.
-Intel uses heat spreader, gains better performance from older-tech poor quality (non copper) heat sinks. Worse cooling with better heatsinks of today. Looks cooler, never really was.

What else should we have expected from 'Only the Paranoid Survive'? And it looks like AMD Inc. might be doing the Crash and Burn. Too bad, clawhammer would be interesting stuff. Does Intel hold a smoking gun? Did it benefeit anyone but them?

crisspy
Friend of SPCR
Posts: 228
Joined: Sun Sep 29, 2002 9:05 pm
Location: Powell River, BC, Canada

Post by crisspy » Mon Nov 11, 2002 2:36 am

Oh, and btw, if Intel has under stated the power consumption, and the diode -touted as the most accurate temps- is a -10-15deg lie, then they wouldn't dare rely on anyone else to supply a heat spreading function for the core. One spot of poor thermal compound, way too much, or a re-used plastic pad, and the core would pop. God forbid a dry heatsink. With their own thermal spreader they can cheat with impunity. Bet you dollars to donuts that's why it's there. AMD looks bad telling the truth and expecting the customers deal with it properly.

But we now have copper interconnects, so the chips shouldn't suffer heat migration anything like the aluminum ones did. Maybe scary temps are endurable, and Intel thinks that what we don't know won't hurt us. They could be right. We'll find out how they last in a few years.

If this is all true then standard practice is to run AMD maybe 15degC cooler than P4, and fret about it at that. Not a level playing feild.

Gekkani
Posts: 116
Joined: Sun Aug 11, 2002 3:26 pm
Contact:

Post by Gekkani » Mon Nov 11, 2002 7:15 am

MikeC wrote:Keep in mind that the reported temps for P4 by any mobo are wrong. Way too low at peak. SOme of the engineering fellas I met at IDF in San Jose (ADI, Intel) say the difference between the thermal diode reading and the hottest spot on the die can be as much as 15C!

So if thermal diode of P4 reads 15Cº below the real temperature (well the hottest spot) how many degrees reads an AMD thermal sensor (not diode) below real temp??

25ºC?

crisspy
Friend of SPCR
Posts: 228
Joined: Sun Sep 29, 2002 9:05 pm
Location: Powell River, BC, Canada

Post by crisspy » Mon Nov 11, 2002 9:26 am

No Gekkani, the whole damned AMD is screaming hot. :twisted: Not a single cool spot on it.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Mon Nov 11, 2002 9:41 am

the whole damned AMD is screaming hot.
Funny, but true, I think. The dif between my thermal diaode reader and the mobo socket thermistor varies on...

1) temp. when higher, the difference is greater
2) HS. The ones that don't provide much airflow down to the socket area show less temp difference, especially when idling

The actual temp reads are tpically (socket first, diode second)-
35C (socket) = ~42 (diode)
40C = 48
50C = 60
60C = 70~75
70C= 85C

The highest I've let it go is ~ 88C diode, which was reading ~73C socket. No damage.

My observation is that the AMDs can run really hot, and the diode (if read accurately) is a pretty good indicator of safety. I wouldn't let it go much past 70C though. Not just because of the CPU risk, but also what it does to the PSU just above it. In a low airflow machine, that heat builds up pretty fast.

ChiefWeasel
Posts: 262
Joined: Sun Aug 11, 2002 3:26 pm
Location: Worcester, UK

Post by ChiefWeasel » Mon Nov 11, 2002 10:02 am

Sorry going a bit OT, but people keep calling the thermal sensor a thermal diode? Im no electronics whizz, but i thought a diode just allows current to flow one way not the other - hence it measuring temp is a little confusing :?


Is it just a misnomer that has crept in on the sly?

crisspy
Friend of SPCR
Posts: 228
Joined: Sun Sep 29, 2002 9:05 pm
Location: Powell River, BC, Canada

Post by crisspy » Mon Nov 11, 2002 11:08 am

A thermal diode is a normal diode, or a transistor with base and collector tied together to form a diode. From Measuring temperatures on computer chips..., see the following explanation. I thought it was pretty cool. Hope it helps:
Analog Devices wrote:This technique, called "delta-VBE calibration," forces two different levels of current through the thermal diode junction and measures the change in forward voltage. The first current may be considered as a calibration current and the VBE forward voltage value of the junction is ascertained. The VBE value is then measured again with a second current. The change or difference in VBE is proportional to absolute temperature. It is independent of the junction's forward voltage or other differences due to manufacturing variations. The output from the DVBE sensor varies at approximately 2.2 mV/°C.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Mon Nov 11, 2002 11:18 am

thermal sensor a thermal diode
You could say a thermal diode is a type of thermal sensor, but there are many kinds of sensors. For the purposes of our discussion, the sensor is the one that's in the CPU socket of AMD mobos. It usually seems to be a thermistor -- a resistor whose resitance varies with temp. Thermal diode is the one embedded in the CPU.

ChiefWeasel
Posts: 262
Joined: Sun Aug 11, 2002 3:26 pm
Location: Worcester, UK

Post by ChiefWeasel » Mon Nov 11, 2002 11:51 am

ah i see, cool cheers guys :)

Belgarion
Posts: 36
Joined: Fri Oct 04, 2002 5:46 pm
Location: Maryland

Post by Belgarion » Tue Nov 12, 2002 7:58 pm

Thanks for the responses!

I don't need all that much speed at the moment, so I was thinking about something in the 1600+ to 1800+ range for an Athlon XP (that 1600+ price at newegg.com is great!) or an Intel P4 1.8A (the few 1.6A's still on the market are overpriced). After the replies, I'm thinking any of those will cool ok without too much trouble. I just read the other thread about "Athlon vs. P4" and that was informative, too.

The Athlon pricing is nice, but more important to me is stability, features (I want onboard USB2.0, Firewire, SATA, etc.), and platform longevity, because I tend to upgrade only every three years or so. Then again, I figured I'd upgrade my PII's before now, and that hasn't happened. The upgrade path for the Athlons is murky with the eventual transition to Hammer, and few Athlon boards meet my specs (no VIA chipsets). I'd probably go with the new Asus nForce2 board. On the P4 side there are some nice SiS 648 or i845PE mobos out now and any of those should take up at least the 3.0 GHz about to come out.

The question is, which path would be better cooling-wise for future upgrade processors? If I buy a high-end HS now with a slower chip, will the HS still have enough cooling power for a future CPU upgrade? It sounds like the higher clocked Athlons have small core size issues, while the faster P4's just run hot. It seems like neither path has a distinct advantage in that regard. I thought about upgrading my current system to dual P3 1GHz (slot 1) processors, but then I'm out the price of a mid-range P4 and still stuck with ATA/33, no USB2.0, no Firewire, etc. The cost of a USB2.0/Firewire combo card will cover about 2/3 of a new mobo that has both of those and more!

Hmm, another possibility if I wait until middle of next year, assuming AMD hasn't gone under already, is Hammer. Does anyone have any idea how hot those might run? I'm guessing pretty hot since it's supposed to compete with the Itanic, which itself reportedly runs very hot. Of course, judging by history, the initial units will run hot and it'd be better to wait for the first die shrink on them anyway.

Maybe I'll suck it in and buy a nice cool G4-powered Mac! Those G4 Ti PowerBooks are pretty sexy... :wink:

Post Reply