p4 - heat question

CodeMonkey

Hi, I just recently got a new p4 1.8gz northwood and it seems to run hotter than I expect. I am using the retail heatsink and use I've clean the thermal pad off it and used some proper thermal paste. When idling, the cpu temp is around 45 degrees and under load it can get up to 55 degrees. The motherboard starts off at around 30 degrees so it might be reporting a higher temperature than it really is or maybe my apartment is a lot warmer than I thought. When overclocked using a cpu voltage of 1.55v, it can get up to 60 degrees under load. So I am wondering if anyone got one of those northwood chips recently and what temp are they running it at?

Gerry

I'm getting similar results to you, though not quite as bad. Using the standard intel cooler, I'm getting about 55c max on a 2ghz chip. At 2.5 ghz @ 1.78v, I'm getting about 63 max.

The problem seems to be that the heat isn't getting out to the heatsink, since the heatsink seems to feel quite cool. I believe the heatsink is making perfect contact with the chip, since the excess compound squashes out, its all very strange really.

JustHalf

It's possible the heatspreader isn't doing it's job properly.

Gerry

Highly unlikely on 2 separate chips.

CodeMonkey

I suspect these chips we have here are different to the ones everyone is raving about over the net. They all say the chip runs at around 35 degrees idling and 45-50 when under stress using retail heatsink fan like we are here. Ah well, it's meant to be good up to 75 degrees and it has thermal protection thingy built in to slow it down and give it a chance to cool when it gets to that temperature anyway. Heeh, Gerry you do watercoolers for people? Those alpha heatsink with the copper plate are meant to be quite good, maybe I should get one of those. Maybe I shouldn't worry if it runs stable...

Inspector Gadget

...out of curiousity, how did you clean the existing thermal pad off the HSF, and, secondly, why? I've heard of people who leave them alone doing a bit better than that temperature-wise...

Anyway, the relevant datasheet states that the 1.8GHz P4 dumps out up to 66.1W of heat running at a stock clock speed - this is quite substantial, a little less than an Athlon Thunderbird running at 1.3GHz at 68.3W (datasheet), or an Athlon XP 2200+ (datasheet) at 67.9W, for the sake of comparison, and although the P4 has a heatspreader, you'd want to be damn sure that your HSF is in good contact.

To this end, I'm wondering how cleanly you removed the existing thermal transfer material (gooey pad or paste) as cleanly as possible, and what solvents (if any) you used. Also, did you lap (i.e. rub on very fine sandpaper to ensure a perfectly flat, mirror finish) the bottom of the HSF before reapplying it? Thermal paste is all well and good, but what you're reporting could be caused by too much thermal paste (filling gaps around chunks of the original thermal compound you didn't remove, perhaps, or that the HSF has bent or deformed slightly when it gets hot - it happens more often than people think); too much thermal paste is nearly as bad as none at all. Also, please don't tell me that you've used baby oil or WD40 to clean the CPU and heatsink as some morons have suggested in various forums - these can hurt thermal performance as they leave a thin thermally insulating layer of gunk behind them. Methylated spirits is the only safe bet.

The other thing is that your system seems to be reporting very high temperatures on startup. If your internal case temp is really 30C, you should be able to open your case while the PC's running and feel that the air inside is (for a time) noticeably warmer than the air in the room. Remember, though, that the thermistors used as temperature probes in many PCs are only accurate to about +/- 5C; however, if you've got a high-end graphics card and a couple of warm drives in there it wouldn't be surprising that you're getting temperatures this high. However, as effective air cooling depends on there being as large as possible a temperature differential between the heatsink and the surrounding air, you may want to consider your ventilation; leave the side off the case for a while and see what difference it makes before deciding on fans or anything like that.

Anyway, that's my 2c. Good luck with it...
Gadget

CodeMonkey

Inspector Gadget, that's all good points. I did leave the thermal pad on at first and when overclocked from 1.8 to 2.4 with a voltage of 1.55, the temp peaked at around 65. Overclocked using default voltage, it won't even finish 3dmark2001.

After removing the thermal pad and using some thermal paste instead, it can run 3dmark2001 when overclocked with no probs with default voltage but it's not stable running games for extended period of time. So voltage was upped to 1.55 again and temp now peaks at 60. I guess the thermal paste did help a little.

As for how the thermal pad was removed, it was just peeled off and then using lots of kitchen towels and a little bit of water, the black gooey stuff was rubbed off until I don't see any traces of it on the kitchen towel. I was only doing this last night so no access to fine sandpaper or methylated spirit to give the heatsink a proper clean and I am not sure how to clean the cpu other than using the kitchen towel rub method without risking damanging the chip.

Tried leaving the case open. Under load the cpu temp gets around 55 degress but it'll eventually peak around 60 again. It just takes a lot longer to get there. The case can def use more fans for better ventilation as I only have 1 case fan in there at the moment apart from the powersupply drawing out air. Even leaving the case side-panel off, leaving the machine to cool for an hour or 2 before powerup I'd get around 43 degrees for cpu and 30 for motherboard. That's why I am not sure if ambient temperature if too high or if it's just not reporting the right temp.

Gerry

Well, I have another retail cooler at home, with the tim still on it, I can try that. I applied the tiniest amount of thermal compound possible, in fact I have tried with varying amounts of thermal compound.

Why? Well perhaps the intel mat isn't as bad as the normal pink goo on standard heatsinks, but usually these things are too thick, and do not have as good a thermal conductivity as a really thin layer of silicon based thermal paste. Also, if you take off the heatsink with a mat, you can't put it on again since the mat was moulded to the shape of the die, this gets messed up when you remove it and you now risk introducing air gaps.

I didn't lap the heatsink, but there should be no need.
Thermal interface mat was removed with white spirits, as I have been doing for a long time now.

The temperature differential for good cooling you talk about is only possible if the heatsink gets hot in the first place, and it doesn't seem to be doing that.

Are you sure the power figure you are quoting is for the northwood?

Inspector Gadget

...get a bottle of meths from your chemist, and give it a go. You probably won't need to lap it.

However, if I was in your position right now, I'd go for a HSF with a little more oomph. Might I propose the following pair:

Alpha PAL8942T
Coolermaster Fujiyama IHC-L71

They're both around 35 sterling, so that's about 56 euros, unfortunately.

Might I propose you read this? I found it doing a quick google, but look at the temps this guy is getting with a 1.8 Northwood!

Gadget

Inspector Gadget

Gerry:

Actually, it would appear not, sorry - Intel didn't think to provide a suitable datasheet for those :mad:. Having had a further nose around, a processor type search gives me a figure of 49.6W - a bit lower, alright, for the SL63X boxed 1.80A P4, running at a core voltage of 1.5v... that sound better?

Again, my apologies...
Gadget

Gerry

Erm, how much better is methylated spirits than white spirits? Surely theres no difference?

I've also tried the coolermaster fujiyama. From memory its only very slightly better. However I'm not using it much in my current experimentation, because its clip is far too tight, really far too hard to put on. Have to sand a bit off the clips so that the pressure is around the same as the retail cooler.

CodeMonkey

Did a little look around in other tech forums and there are people reporting very similar temperatures on their new p4 northwood chips. I'd rather find out why it seems to be over 10 degrees hotter than what the guy in that alpha PAL8942 guide in the first place before looking into more aggressive cooling solutions. And I wish komplett sells that, only seen the alpha on overclockers.co.uk so far and I hate those expensive deliveries for small items. Incidently, my roommate has a xp1900+ and it was getting around 75 degrees under load, his entire case was getting hot mad

SickBoy

Originally posted by Gerry
Are you sure the power figure you are quoting is for the northwood?

I was thinking that myself.

Anyway I have a P4 1.7 willamette not overclocked and the CPU runs under load at 62 - 63c. The HSF I use is a CoolerMaster Silent Heat Pipe (IHC-L71) which has good/great reviews. The system temp is 55c. The board I have is an ABit TH7II which has a third temp probe that can be attached to any component within reach and I've attached it to my hard drive. That's reporting a temp of 43c. The sides are off the chassis and I've added 2 case fans and it hasn’t made much of a difference. I've read on other forums that the temps can be out by as much as +/- 10c. I'm hoping there out by +10...

JustHalf

Originally posted by Gerry
Erm, how much better is methylated spirits than white spirits? Surely theres no difference?

It makes you blind faster.

But seriously, there's also isopropyl alchohol to use. I can't remember why I got that (I remember there was a good reason at the time).

Inspector Gadget

...if this is down to the "chemistry of cleaning" as applied to blocks of metal, the order of merit (in terms of cleaning ability)looks something like:

1. Unleaded Petrol. Best by miles. But needs to be washed afterwards in soapy water. (never, ever use Diesel!!!)
2. Isopropyl Alcohol - good, doesn't leave a residue unlike the rest, but hard to find sometimes and expensive, IIRC.
3. Methylated Spirits - Good cleaning power, but does leave minimal residue after evaporation. Washing in soapy water does the trick.
4. White Spirits - the least predictable of the bunch as it's a cocktail of different types of alcohol. The formulations used vary quite a bit and as such the cleaning power (though generally poor compared to others) is variable.

Strictly in mine own opinion of course... also, using a lint-free cloth would help immensely (definitely not tissue paper...)
Gadget

JustHalf

I got about 500ml of isopropyl alchohol for around 10 pounds in a chemist.

Inspector Gadget

Originally posted by JustHalf
I got about 500ml of isopropyl alchohol for around 10 pounds in a chemist.

My point exactly. About thirty times dearer than unleaded petrol, by a quick mental guesstimation...

By the way, I forgot paraffin in the above list. However, IIRC it leaves a bit more of a residue than petrol, but soapy water still does the job.

Gadget.

JustHalf

Hey, I don't call that expensive, considering how little you need to clean a heatsink

Gerry

Extreme nitpicking on the heatsink cleaning here... sigh

Sickboy, I think the th7-II is reporting very high temperatures, with 6 system fans the system temperature is still going up to 40c, and cpu temp is anything up to 62c. The asus p4s533 I put in the previous machine reported a cpu temp of 46c max, and surely both of them are reading from the p4's thermal diode. Somewhere between the two lies the actual temperature, I reckon. I'm going to turn off the fan on the heatsink, and see what temperature the p4 throttles at. Its supposed to start throttling around 75c, if the reported temp goes much over that before throttling , we'll know that its overreading.

codemonkey, are you using an abit board by any chance?

Inspector Gadget

Originally posted by Gerry
Extreme nitpicking on the heatsink cleaning here... sigh

You started it

Getting back (hopefully) to the point, though, I can't help but feel that your temperatures are still surprisingly high; while I don't have a huge amount of faith in thermistors as accurate thermometers, I can't help but think about my own POS system; I've got a Duron 800 running at 1GHz; at this speed, and with the slightly raised Vcore I needed to achieve it, I'm definitely exceeding 50W (as it should, in theory, exceed 50W at stock Vcore). It would be quite an achievement, in my book, to place a thermistor anywhere within about a centimetre of a CPU and report the temperatures as being 20C too low - unless, of course, it's stuck in the airflow of the cooling fan; however, this is not the case.

Now, MBM on my computer (by the way, what software are you using to read the temperatures? This may be relevant too) reports temperatures of CPU=26/Northbridge=23/System=20 (on average) once W2k has booted; if I really try and work it (MPEG encoding, or something like that), I can get the CPU up to 43 or 44C, and the case temps go up to maybe 26 or so. (By the way, I'm using a cheap Taisol CGK760092 heatsink - worth about 20 euro - and some ASII paste; nothing at all fancy, and no additional case cooling)

I realise that comparing a Duron to a P4 is a bit like comparing a Golf GTi to, say, a Porsche, in terms of performance; however as lumps of silicon that generate almost exactly the same amount of heat, I reckon the comparison is reasonably fair. The P4's heating curve will be less steep (because of the heatspreader, it'll take slightly longer to heat up) than the Duron's, due to the exposed (not to mention tiny) die.

All of this leads me to think that your system is running too hot overall; the hot system air is hindering the HSF. What else is in there - have you got a lot of RAM (this is becoming less of an issue with more recent, lower-voltage RAM types, but it could still be generating 20-40W of heat if you've got a lot of sticks)/a graphics card that requires its own HSF/one or more quiet or fast recent HDDs? How hot is the air coming out of the exhaust fan at the back of the PC?

Gadget

CodeMonkey

Originally posted by Gerry
codemonkey, are you using an abit board by any chance?

Nope, I am using the MSI 645E MAX2-L mobo. I had a look around yesterday for p4 heating issue and heatsink reviews and found that we are getting the same temperature range as a lot of people. It was a cooler night last night than sunday and after playing NeverwinterNights for 2 hours, my temp only peaked at around 56 so it's not too bad.

Also, take a look at this review of the CoolerMaster heatpipe http://www.tech-pc.co.uk/ihcl71-2.shtml. The idle temp for 1.6 overclocked to 2.13gz p4 is at the mid 40s and high 50s when under stress when using a retail heatsink. So I don't know, it looks like a lot of p4 northwood chips do run that hot. I might be getting one of those CoolerMaster heatpipe over the weekend as they seem to do as good a job as the alphas and itdirect.ie's website says they stock them.

Gerry

Originally posted by Inspector Gadget


Getting back (hopefully) to the point, though, I can't help but feel that your temperatures are still surprisingly high; while I don't have a huge amount of faith in thermistors as accurate thermometers, I can't help but think about my own POS system; I've got a Duron 800 running at 1GHz; at this speed, and with the slightly raised Vcore I needed to achieve it, I'm definitely exceeding 50W (as it should, in theory, exceed 50W at stock Vcore). It would be quite an achievement, in my book, to place a thermistor anywhere within about a centimetre of a CPU and report the temperatures as being 20C too low - unless, of course, it's stuck in the airflow of the cooling fan; however, this is not the case.

The p4's onboard thermal diode is accurate to within a few degrees I'd say. In socket thermistors underreport by 5 degrees or more, they also take a long while to reflect the change in temp when the system goes from idle to load, this being the time taken for the air under the socket to heat up.

Originally posted by Inspector Gadget



Now, MBM on my computer (by the way, what software are you using to read the temperatures? This may be relevant too) reports temperatures of CPU=26/Northbridge=23/System=20 (on average) once W2k has booted; if I really try and work it (MPEG encoding, or something like that), I can get the CPU up to 43 or 44C, and the case temps go up to maybe 26 or so. (By the way, I'm using a cheap Taisol CGK760092 heatsink - worth about 20 euro - and some ASII paste; nothing at all fancy, and no additional case cooling)

I've used several different pieces of software to read the temperature, the possible differences you are talking about come from different interpretations by earlier versions of mbm compared to motherboard bios readings.

mbm, winbond hardware doctor, speedfan, and the bios all agree on temps.

Originally posted by Inspector Gadget

I realise that comparing a Duron to a P4 is a bit like comparing a Golf GTi to, say, a Porsche, in terms of performance; however as lumps of silicon that generate almost exactly the same amount of heat, I reckon the comparison is reasonably fair. The P4's heating curve will be less steep (because of the heatspreader, it'll take slightly longer to heat up) than the Duron's, due to the exposed (not to mention tiny) die.

All of this leads me to think that your system is running too hot overall; the hot system air is hindering the HSF. What else is in there - have you got a lot of RAM (this is becoming less of an issue with more recent, lower-voltage RAM types, but it could still be generating 20-40W of heat if you've got a lot of sticks)/a graphics card that requires its own HSF/one or more quiet or fast recent HDDs? How hot is the air coming out of the exhaust fan at the back of the PC?

Gadget

I'm well used to a cool running system, my own p3 puts out only around 30watts, the cooling which I have blathered about in other threads keeps the cpu at 30c load, ambient idle, and the case at ambient.

I believe the different on the abit board is that the system temperature thermistor may be in a hot spot on the board, that is next to the power regulation circuitry, or next to the rambus ( which does get hot ).

afaik the sensor is built into the winbond hardware monitoring chip, so maybe this chip is in a different place than usual, which might explain these high system temps.

The air coming out of the back is not hot, thats the whole problem. It doesn't seem as if heat is getting from the cpu to heatsink, because I am used to seeing amd machines with the heatsink getting very warm. It is then a matter of getting cool air to the heatsink, and warm air away from it and you are sorted.

I think there may be a hot spot at the back of the heatsink which the air circulating through the case is not affecting. Dead air in other words. I'll investigate further this evening.

Inspector Gadget

...the reason I ask about the software is in case there's any risk of misconfiguring it (such as choosing the wrong chipset in MBM - having done it myself when I first built my rig, I know it's possible). Generally, as they're all requesting the same data from the same monitoring hardware over a simple I2C/SMBus interface, they should all agree, but it's possible to screw it up... I know, I've done it :rolleyes:

Thermistors are definitely slower to react than diodes; that's no maybe. However, I'm talking about sustained runs of many hours to make sure the thing gets good and hot (a thermal diode may be slower, but it's still only taking a matter of seconds longer to react).

Either way, if the CPU stays under 60C, it shouldn't adversely affect the life of the chip too greatly. Under 50C is better, but at the rate we're all upgrading these days, who cares if the chips are still intact in a decade's time (unless we're using 1GHz chips for firewalls and the likes the way we use 486's now? It's not impossible, I suppose...)?

Gadget

Gerry

I think I should be able to remove one of the hotspots with some ducting around one of the exhaust fans, we'll see what happens.