Computer hardware question about CPUs and heat sinks

[davidm]

Maybe I should be posting this on a techie forum but questons on here usually get answered fast and I've no doubt that there's more than one person on here who can help with this.

I recently replaced the motherboard on a desktop computer. Since then it has shut off twice because of what it called a "thermal event". In other words, the CPU is overheating and shutting down the computer to protect itself.

This particular computer has a large heat sink and a large fan that mounts over that heat sink. It had been at least a decade since I last installed a CPU on a motherboard and I wasn't aware of things like thermal paste and thermal pads so I just slapped it together thinking that was enough. I've since read that either paste or a pad are necessary to conduct the heat away properly.

Pads seemed like a less messy alternative so I spent 3 bucks for a package of "phase changing thermal heat sink pads". Upon opening them I discovered 5 pads a good deal smaller than the surface of the CPU.

Is the smaller size an issue? Some websites say that the pads melt from the heat and fill in microscopic imperfections. When they melt, do they also spread out over the entire surface? If not, would there be a problem with using more than one so as to cover the entire surface?

Should I just buy some thermal paste instead? Can anyone with experience recommend types and brands?

[carnivorousplant]

I'd go with thermal paste.

[jz78817]

what type of CPU? if it's a Pentium 4, Intel Core, or Athlon 64/Phenom/Opteron, it should have a heatspreader on it and generic white heatsink paste should be OK. If it's a Pentium III/Celeron/AthlonXP with a bare CPU die, then a pad is much safer.

ETA: also, if this is an Intel Pentium 4 or Core system which has that heatsink attached by the four plastic pins, make sure the pins are all securely attached. sometimes they can pop out and the heatsink won't be in good contact with the CPU.

[Whack-a-Mole]

Use thermal grease (paste if you prefer) all the way.

Thermal pads do not have near the performance that the grease does. Thermal pads are best used on other things like memory chips which generally do not get as hot.

I recommend Arctic Silver 5 (http://www.arcticsilver.com/as5.htm) (no I do not work for them or own stock in them but I have built many computers and this is good stuff). There are other brands out there too but not all thermal greases are created equal. There are a couple other brands that measure up to Arctic Silver 5 but I forget what they are offhand.

When applying the grease first be sure to thoroughly clean the CPU and the heatsink of all traces of the previous grease or pad. Rubbing alcohol works well here and use a lint free cloth (coffee filters work well for this). Be gentle and take your time. It can be a bit tedious but attention to detail here pays off.

When applying the grease do not spread it all over the CPU. Instead put a tiny blob (maybe the size of a grain of rice) in the center of the CPU and set the heatsink down directly on top. The grease will spread itself under the heatsink giving the ideal coverage. Too much grease is not a good idea for a few reasons. You want the layer to be pretty thin. Spreading it manually can leave air gaps as well which is also not ideal.

Also consider your heatsink and what you are using your PC for. Heatsinks are definitely not all created equal. The stock heatsink that comes with an Intel CPU, while better than nothing, is about the worst there is. It is sufficient if all you do is browse the web, get email and so on but not very good for more demanding tasks. Up to you but consider a new, aftermarket, heatsink. Just be certain any heatsink you choose will fit both your motherboard, CPU and in your case. Not all heatsinks will work on all motherboards or CPUs for various reasons. Also, some heatsinks can be huge and will not fit in all cases. Check and doublecheck the specs if you do this. Make sure it doesn't cover memory slots you are using (happens with some).

[Whack-a-Mole]

Here's a manual (http://www.noctua.at/pdf/manuals/noctua_nt_h1_manual_en.pdf) (PDF...just one page) for applying thermal paste. These guys manufacture heat sinks and manufacture what is probably the best heat sink on the market today (I have one and it is awesome). Point being they know what they are talking about.

[Zeriel]

ETA: also, if this is an Intel Pentium 4 or Core system which has that heatsink attached by the four plastic pins, make sure the pins are all securely attached. sometimes they can pop out and the heatsink won't be in good contact with the CPU.

If it's one of those, save yourself endless headaches and buy an aftermarket heatsink with a screw plate.

When applying the grease do not spread it all over the CPU. Instead put a tiny blob (maybe the size of a grain of rice) in the center of the CPU and set the heatsink down directly on top. The grease will spread itself under the heatsink giving the ideal coverage. Too much grease is not a good idea for a few reasons. You want the layer to be pretty thin. Spreading it manually can leave air gaps as well which is also not ideal.

I can't agree--I get much better results from spreading a very thin layer of paste on the CPU's contact surface with the edge of a credit card. Especially if you're using a stock heatsink--I've only found the quoted method to work with the best of screw-down heatsinks that have tension springs and all that whatnot.

Regarding thermal pads: follow the package directions. One will do, if the pads are marketed as being for your particular type of CPU.

[johnpost]

use thermal paste. if applied thoroughly will give absolute coverage and can be cleaned off easily if it ever needs to be redone.

[Darth Panda]

I have always used Arctic Silver and found it to be quite good.

Here's a good guide as to how to apply, although it's a little dated the concept remains the same:

http://www.hardwaresecrets.com/article/How-To-Correctly-Apply-Thermal-Grease/274/1

[davidm]

I knew the Dope would come through.

Fortunately I work near an MEI Microcenter and I think they have Arctic Silver 5 so I can stop by there after work today for the paste/grease and I'll have a weekend project.

The processor is a dual core Pentium. I don't have the specifics or the computer in front of me at the moment so I can't say exactly what processor it is.

It already has a big ass heat sink. There is a metal (I assume it's metal) square attached on the bottom of it that is the size and shape of the surface of the CPU and fits exactly over it when the sink is screwed onto the board.

[jz78817]

folks, I really doubt he's overclocking. there's no point to any of these fancy thermal paste application methods. put a blob on there, attach the heatsink, and let the clamping force squeeze it into a thin layer.


Thermal pads do not have near the performance that the grease does. Thermal pads are best used on other things like memory chips which generally do not get as hot.

Phase change thermal pads (e.g. the "pink" ones) work as well as or better than paste; but only on bare-die chips where the pad has a chance to "change phase" and squeeze into a thin layer. They also have the advantage of adding a bit of a "cushion" to prevent chipping of the processor die.

[davidm]

Dup post. Sorry.

[bump]

The point of the thermal grease is to avoid the air gaps between the processor heat spreader/ceramic package/etc... and the heat sink. Air's a terrific insulator, so your heat transfer will be less than optimal.

Even the best grease doesn't compare to the conductivity of copper/aluminum, etc... so you want as little grease as you can get away with and still fill in the air gaps.

In the grand scheme of things, the paste's properties probably aren't all that important provided you have enough (but not too much) in the right places.

[davidm]

No, I'm not overclocking. It's a factory standard Gateway where the motherboard died and had to be replaced.

The Dope can be great because of all the different advice you get.
The Dope can be a PITA because of all the different advice you get. :p

[Whack-a-Mole]

Phase change thermal pads (e.g. the "pink" ones) work as well as or better than paste; but only on bare-die chips where the pad has a chance to "change phase" and squeeze into a thin layer. They also have the advantage of adding a bit of a "cushion" to prevent chipping of the processor die.

I have never seen anyone claim a thermal pad is superior to properly applied thermal compound.

I will agree that for most average users a thermal pad (the change phase ones) would probably be sufficient and they would never know the difference.

On the upside thermal pads are easier to use than thermal compound when doing an installation (thermal compound can be messy).

On the downside if you ever have to remove the heat sink thermal pads are a PITA. I have had to remove it before and I thought I was going to rip the chip out of the socket (fortunately I didn't but I have heard of it happening). The pad can sort of weld the heatsink to the chip. Cleaning off old thermal is noticeably harder than cleaning off old thermal compound.

Personally the slightly more difficult task of installing with thermal compound is not all that bad and you get somewhat better performance and have a much easier task if you ever need to remove the heatsink.

[davidm]

Okay, I've gotten two opinions on how to apply the compound. Spread it over the surface with a credit card, or just put a small dab in the middle. The small dab method seems to have more proponents.

[jz78817]

Okay, I've gotten two opinions on how to apply the compound. Spread it over the surface with a credit card, or just put a small dab in the middle. The small dab method seems to have more proponents.

because the force of the heatsink clamping down will spread it into a thin layer anyway. so there's no real reason to go through the bother of spreading it yourself.

[Whack-a-Mole]

because the force of the heatsink clamping down will spread it into a thin layer anyway. so there's no real reason to go through the bother of spreading it yourself.

Also, spreading it yourself you can never make it perfectly smooth. There will be grooves (which could leave air gaps under the heatsink) and thicker and thinner spots. You are also more likely to leave too thick a covering when spreading it manually.

Using a dab tends to avoid those problems. The issue here is too little a dab or too big a dab. I have seen some sites suggest a "pea" sized dab which is waaaay too big. Go with the manual I linked above or judge it to be about a grain of rice (maybe a little smaller).

That said I have done the manual spread method a few times and it worked fine. Personally I find the dab simpler to deal with and it provides better results (usually...any method can be messed up).

[Telperion]

On the downside if you ever have to remove the heat sink thermal pads are a PITA. I have had to remove it before and I thought I was going to rip the chip out of the socket (fortunately I didn't but I have heard of it happening).

I did that once on an old Pentium 4 CPU. I almost didn't want to turn it over to check how bad the damage was, but miraculously it had gone straight up and none of the pins even got bent. I never want to do it again, though; I figure I probably used up a lifetime of good DIY-luck on that move.

[BigT]

When I looked it up, I was told it depends on the processor which is the better method. I was also told that, if you want maximum cooling right away, you need to prime the heat sink by putting it on with a credit card and wiping off the excess with a coffee filter. Otherwise you will have to wait on the paste to set, which can take a couple weeks.

And, depending on your chip, it may run hot already, so every little bit helps. I know it was a big deal for my off-the-shelf unit, as it apparently had very little margin.

[davidm]

I notice that they also sell liquids for cleaning the old paste off. Are those worthwhile or is alcohol sufficient?

I have some alcohol wipes (for injections). It seems like those would work.

[Whack-a-Mole]

I notice that they also sell liquids for cleaning the old paste off. Are those worthwhile or is alcohol sufficient?

I have some alcohol wipes (for injections). It seems like those would work.

Rubbing alcohol works fine and is what I have used in the past with no trouble. If the "special" stuff is the same price as rubbing alcohol (I doubt it) and you are in the computer store anyway then sure...go for it. Otherwise rubbing alcohol can be found in most supermarkets and pharmacies and it is cheap.

Are those wipes lint-free? Lint-free cloth is best. As mentioned coffee filters work well for this and are cheap and easily gotten (most people already have some at home).

As to the cooling we forgot to mention case cooling.

Even a good heat sink will fail if your case cooling sucks. If the heat sink just dumps hot air into the case then you will continue to have problems.

Make sure there are are intake fans and ideally exhaust fans. You need to get the hot air out and cool air in.

Also clean up the inside of the case (cable management). If the inside is a rat's nest of cables all over that impedes good air flow.

The issue with more case fans is noise. With some effort you can find quiet (a relative thing) fans. I have eight fans in my PC and it is nearly silent but I spent a lot of time and money to get it that way. You need to know the case well to determine what size fans it can accept if you need more case cooling.

[Zeriel]

because the force of the heatsink clamping down will spread it into a thin layer anyway. so there's no real reason to go through the bother of spreading it yourself.

I've had... inconsistent results from this technique with the plastic friction connectors on recent Intel stock coolers.

[johnpost]

rubbing alcohol works fine, it has some water in it so it is good to let it air dry.

[Palooka]

The issue with more case fans is noise. With some effort you can find quiet (a relative thing) fans. I have eight fans in my PC and it is nearly silent but I spent a lot of time and money to get it that way. You need to know the case well to determine what size fans it can accept if you need more case cooling.What case can even hold eight fans and how could you make that silent?

[Zeriel]

What case can even hold eight fans and how could you make that silent?

Some of the larger Cooler Master full towers with 120mm low-RPM fans do wonderfully. My Cosmos 1000 has 6x120mm fans and the only thing making any appreciable noise is the video card's cooler--when just on the desktop and not in a game, it's dead silent.

[Palooka]

Oh. I guess I meant without compromising your cooling like that.

[Whack-a-Mole]

What case can even hold eight fans and how could you make that silent?

As Zeriel mentioned big cases can do it. Big fans tend to be quieter than small fans (they can spin at a lower RPM and still push a lot of air...high RPM = noise).

As I mentioned it was not cheap either. I really splurged on this case (http://i.imgur.com/dTBa8.jpg) (which is not my actual PC but identical in everything except the video card pictured) but honestly it was worth it. Best case I have ever had (Silverstone Fortress (http://www.silverstonetek.com/products/p_spec.php?pno=FT02&area=usa)). It is also huge but in my room sitting next to my desk it is no problem at all and actually kinda cool looking.

In the picture linked above:

1, 2 & 3: 180mm case fans (huge). Located at the bottom they draw cool air from the floor and blow "up". Heat also naturally rises. Each fan creates a column of air so one fan for hard drives, one for video card(s) and one for CPU. The fans have their own air filter that is easily slid out for cleaning keeping dust in the case to a minimum (dust cannot drop in from above when fans are on either).

4 & 5: Noctua NH-D14 (http://www.noctua.at/main.php?show=productview&products_id=34&lng=en) Heat Sink with dual 120mm fans (yes the heat sink is a monster and it is awesome and quiet...the fans are unique to Noctua and made to be quiet including notches in the blades that supposedly makes it quieter still (no idea if it is a gimmick but they sure are quiet).

6: Power Supply has a fan (Corsair power supply so well built and also quiet)

7: Exhaust fan (120mm).

8: CPU fan (this is the only thing that sometimes makes noise...if I really stress my GPU the fan ramps up in speed. Usually even then it is just noticeable but if it hits max RPM it screams...it has never needed to do that though).

All told in a dead quiet room at idle you can barely hear it. When gaming the fan speeds up some on the video card but even then it is minor and only noticeable if you pay attention and listen for it.

[Apex Rogers]

Oh. I guess I meant without compromising your cooling like that.

It doesn't compromise your cooling. 120mm fans can push the same amount of air as 80mm fans at much lower RPMs and hence much lower noise. Also, the original claim of 8 fans didn't mention 8 chassis fans, he might be including his CPU, video card, and PSU fans in the count and could have only 5 chassis fans.

[Palooka]

The Raven 2 is such a great case. I feel bad for people who get talked into buying the Corsair 600T over it. The cooling performance difference is huge and the RV2 (or FT02) is probably the only case that really can deliver top-notch cooling and near silence thanks to its three AP181 fans (or do you have the old honeycomb patterned ones?).

Have you thought about putting a Thermalright Shaman on your GPU? It's supposed to be the thing if you want to kill noise. Doesn't work if you have SLI though, since it's so big.

[Whack-a-Mole]

The Raven 2 is such a great case. I feel bad for people who get talked into buying the Corsair 600T over it. The cooling performance difference is huge and the RV2 (or FT02) is probably the only case that really can deliver top-notch cooling and near silence thanks to its three AP181 fans (or do you have the old honeycomb patterned ones?).

Have you thought about putting a Thermalright Shaman on your GPU? It's supposed to be the thing if you want to kill noise. Doesn't work if you have SLI though, since it's so big.

I have the fans with the honeycomb on them. My video card barely fits. Some of the newest, high-end Nvidia cards will not quite fit unless you remove the honeycomb (or cut out part of the one blocking the video card).

In my case I bought the ASUS ROG Matrix (http://www.pcgameshardware.com/screenshots/medium/2009/08/Asus-Rog-Matrix-GTX-285-00.jpg) video card. It sold at a premium because it is slightly overclocked out of the box but that extra performance was rather minimal (it can be overclocked further). The main reason I splurged there was due to keeping an eye on air cooling. I have done water cooling and did not want to do it again but knew I had to be serious about cooling with air given that this is a power gaming rig so will be stressed and cooking (the computer can heat my room by itself).

So, the main reason for that card was because it had a custom cooling system on it. I was disappointed that they did not seem to care about noise overly much but the main thing is the cooling system is sealed to the card. Most GPU coolers just blow the hot air into the case. This one specifically exhausts air out the back ("up" in my case) without heating the interior.

As mentioned, even playing a modern game with settings on high, the card only ramps up one notch to the next higher fan speed and at that speed you can hear it but only just. For fun I have manually put the fan to 100% and then it is loud as hell. Fortunately with that big ass fan blowing cool air right at the intake for the video card it barely breaks a sweat.

And yeah, the Raven/Fortress are simply awesome cases. They are expensive as hell and not small or light but I think the price is worth it given I can reuse it for years. Small/light are not a concern for me (I do not take my computer to LAN parties). I seriously considered the Raven and I forget why I skipped it in favor of the Fortress. Personally I like the more elegant lines of the Fortress but whatever floats your boat. Either is an excellent case.

[davidm]

Regarding fan noise; this box came with a big ass fan which fits over the big ass heat sink and it has always been noisy when the CPU is busy. When I purchased (wasted my money) the thermal pads I also bought a PCI slot fan to add some more air flow in the hopes of cooling the inside more.

[Whack-a-Mole]

Regarding fan noise; this box came with a big ass fan which fits over the big ass heat sink and it has always been noisy when the CPU is busy. When I purchased (wasted my money) the thermal pads I also bought a PCI slot fan to add some more air flow in the hopes of cooling the inside more.

Heat sinks definitely need fans (they used to be passive...no longer for a CPU).

Not all fans are created equal. It is possible, if you want, to replace the fan on the heat sink with a fan that is more quiet one. Hard to say if it'd be worth it but something you might explore.

Again, it is all about moving cold air in and hot air out. Most cases should have a place for an exhaust fan close to the CPU heat sink. It is a good idea to have one there at least (fans can be flipped over so they can be an intake or exhaust fan so something to note when installing).

Then consider intake fans to bring cold air into the case faster. Most cases have places for those too.

You will need to check your case to see where fans can go and what sizes it will take.

PCI slot fan is ok in a pinch but case fans are what you really want. Perhaps the PCI slot fan, if near your video card, will help grab some of that hot air and get it out of the case. I am not a big fan of PCI slot fans but I have used them. Found it to be a bit noisy but maybe things have changed.

[davidm]

Shoot, my local store doesn't have the Arctic Silver 5 in stock. I'll probably have to order it by mail and wait.

[davidm]

I'm getting this one. It should be fine.

http://www.microcenter.com/single_product_results.phtml?product_id=0247009

[davidm]

Damn. They don't have that one either. :(

[Palooka]

The choice of thermal compound doesn't really matter. Anything from a major name -- Arctic Cooling, Noctua, Thermalright, Turiq, Cooler Master -- is going to work fine. Just get whatever is cheapest. If your system is overheating at stock speed, having the best thermal compound isn't going to make the different.

[notsoheavyd3]

If it's one of those, save yourself endless headaches and buy an aftermarket heatsink with a screw plate.



Oh god do I second this bit of advice. (I have a dual core PC and it had those stupid push pins and they never really held the heat sink down properly. Everytime I'd just move my PC I had a fairly largely chance of one of them popping loose.) Oh, you might want to get the non-conducting grease. (I used Tuniq TX-2 on my current PC with a screw down heat sink. That works fine and won't pop loose like the intel one.)

[Rigamarole]

Pads seemed like a less messy alternative

You know what they say: you can't build a computer without getting some thermal paste on your fingers.

[Zeriel]

Heat sinks definitely need fans (they used to be passive...no longer for a CPU).

Again, entirely dependent on your case configuration, though generally true for most workstations (not always true for HTPCs, almost never true for servers or heavy workstations)

The choice of thermal compound doesn't really matter. Anything from a major name -- Arctic Cooling, Noctua, Thermalright, Turiq, Cooler Master -- is going to work fine. Just get whatever is cheapest. If your system is overheating at stock speed, having the best thermal compound isn't going to make the different.

I have multiple 4diex6core servers running 90% loads with heatsinks applied with cheap-ass off-brand white generic thermal paste, so I don't even think the major name is necessary.

[davidm]

Oh god do I second this bit of advice. (I have a dual core PC and it had those stupid push pins and they never really held the heat sink down properly. Everytime I'd just move my PC I had a fairly largely chance of one of them popping loose.) Oh, you might want to get the non-conducting grease. (I used Tuniq TX-2 on my current PC with a screw down heat sink. That works fine and won't pop loose like the intel one.)The heat sink that came with the original machine is a large screw down heat sink. It worked fine for years until the motherboard died for reasons unknown. The only issue is that I installed the new mother board without any thermal compound between the sink and the CPU. The sink and accompanying fan are fine.

[davidm]

I went to the store and found that they did have the "Arctic Alumina Thermal Compound" in stock in spite of what they told me online.

The instructions say to first spread a small amount over both the CPU and the heat sink using a plastic card and then to wipe both surfaces with a lint free cloth. They refer to this as tinting the surfaces and say that this is done to fill in microscopic imperfections.

It says to then apply a rice grain sized amount to the CPU and spread that out with a razor blade or plastic card.

It then requires a minimum of 25 hours for a break-in period, including several cycles of heating and returning to room temperature. I'm clueless as to how long to heat it and how long it takes to return to room temperature. I'm assuming on for an hour and then off for at least an hour or two for each cycle.

It isn't clear from the instructions whether or not I need to spread it out so as to coat the entire surface edge to edge. It just says "spread and smooth out the compound".

[Zeriel]

Oh. I guess I meant without compromising your cooling like that.

Wait, how is that compromising my cooling? This is the coolest-running PC I've ever built that isn't water cooled.

[Zeriel]

I went to the store and found that they did have the "Arctic Alumina Thermal Compound" in stock in spite of what they told me online.

The instructions say to first spread a small amount over both the CPU and the heat sink using a plastic card and then to wipe both surfaces with a lint free cloth. They refer to this as tinting the surfaces and say that this is done to fill in microscopic imperfections.

It says to then apply a rice grain sized amount to the CPU and spread that out with a razor blade or plastic card.

It then requires a minimum of 25 hours for a break-in period, including several cycles of heating and returning to room temperature. I'm clueless as to how long to heat it and how long it takes to return to room temperature. I'm assuming on for an hour and then off for at least an hour or two for each cycle.

It isn't clear from the instructions whether or not I need to spread it out so as to coat the entire surface edge to edge. It just says "spread and smooth out the compound".

Okay, all those instructions are pretty overcomplicated, if you ask me. Do the coat-and-wipe step if you're REALLY feeling thorough, then spread the stuff to cover the entire heat transfer surface edge-to-edge as thin as you can get it.
Power up that bad boy, monitor temps for a little while. If they look good, run Prime95 or something that uses 100% CPU in a controlled way, monitor temps. Do that for an hour or so, then power it down for an hour, etc. Repeat until bored, but I typically don't even do that much.

Note, also, that if you read the fine print on that, the break-in period is supposed to settle it in so that your average temp drops by as much as 1-2C. I don't think you're loading it high enough to worry about that.

[Dr. Strangelove]

I just use my (recently cleaned!) finger to spread the paste around. You should spread it around until you just about see the heatspreader under the paste--if it's totally opaque, you've added too much.

Once I've rested the heatsink on it, I slide it around a bit (before fixing it in place) to get things fully spread around. You should be able to slide and rotate the heatsink easily, but at the same time there should be enough suction that you can nearly lift the motherboard up with just the heatsink. You know how two flat surfaces will stick together if there is a thin layer of water between them? Same deal. I'm assuming here that your CPU has a heatspreader instead of a bare die.

I've assembled many systems with this method and never had a problem. But several times I've disassembled systems with thermal problems, and almost inevitably there is a problem with the thermal compound--too much, too little, or a poorly installed pad.

[davidm]

Okay, all those instructions are pretty overcomplicated, if you ask me. Do the coat-and-wipe step if you're REALLY feeling thorough, then spread the stuff to cover the entire heat transfer surface edge-to-edge as thin as you can get it.
Power up that bad boy, monitor temps for a little while. If they look good, run Prime95 or something that uses 100% CPU in a controlled way, monitor temps. Do that for an hour or so, then power it down for an hour, etc. Repeat until bored, but I typically don't even do that much.

Note, also, that if you read the fine print on that, the break-in period is supposed to settle it in so that your average temp drops by as much as 1-2C. I don't think you're loading it high enough to worry about that.There's no way to monitor the temp. :confused: Even if there was I wouldn't have a clue as to what is or isn't an acceptable temperature. The only way I know there's a problem is when it shuts down.

[davidm]

I just use my (recently cleaned!) finger to spread the paste around. You should spread it around until you just about see the heatspreader under the paste--if it's totally opaque, you've added too much.

Once I've rested the heatsink on it, I slide it around a bit (before fixing it in place) to get things fully spread around. You should be able to slide and rotate the heatsink easily, but at the same time there should be enough suction that you can nearly lift the motherboard up with just the heatsink. You know how two flat surfaces will stick together if there is a thin layer of water between them? Same deal. I'm assuming here that your CPU has a heatspreader instead of a bare die.

I've assembled many systems with this method and never had a problem. But several times I've disassembled systems with thermal problems, and almost inevitably there is a problem with the thermal compound--too much, too little, or a poorly installed pad.
Since the motherboard is attached to the case trying to lift it with the heat sink doesn't sound wise.

[Whack-a-Mole]

I have multiple 4diex6core servers running 90% loads with heatsinks applied with cheap-ass off-brand white generic thermal paste, so I don't even think the major name is necessary.

From what I have read (numerous pieces...no cite) is most thermal compounds work about as well as the next one. There are differences but usually it is a matter of a few degrees and generally not much to fuss about.

I have read differing versions of the break-in period. I have seen some swear by it and some say it is bullshit. Same with tinting the heatspreader. Personally I have never "tinted" and I have never had problems. I think that effort is only worth it if you are going for ubermax overclocking. I overclock but do not push it to the absolute limits of my PC.

Where the difference lies is if you are a power user. Someone wanting to push the machine to its limits. Kinda like when they build an ultralight plane or solar racing car. They literally toss any gram of weight they can wherever they can. Each incremental toss of weight is negligible but if you are anal about it that stuff adds up and makes a difference.

So too with CPUs. There is a diminishing return for the effort but if you want to push it then each little degree matters and they add up and you get lots of performance. It is a LOT of work though (run Prime95 for hours...tweak one of multiple settings, do it again, rinse and repeat if looking for max performance and stability...I've done it, it is a pain in the ass and I have given up to be happy with modest performance gains and not max ones). If that is your thing or you care go for it. If you are building a PC for your mom who uses it for email only then don't bother.

The difference between "good" and "bad" thermal grease, so I have read, is how it performs over time. Good stuff like Arctic Silver will last far longer than your PC will. Cheap stuff will do the job on Day-1 but diminish over time. After a year or two its performance sucks. Thing is modern CPUs will throttle their speed when overheating so many users never know the difference. It'd be the power user who would notice the slowdown in their machine.

Bottom line is most people will not notice but they will be buying a new PC sooner because theirs seems slow. A bit of work and attention and a little extra money spent on good stuff, I think, gets another year or two out of a PC. In the end it saves you money.

Also, really cheap thermal compound you have to wonder about. It is important the stuff is not conductive of electricity. Some cheap stuff might conduct some electricity. Not a big deal if you are careful on installation but if some squeezes out from under the heatsink there is a potential to short circuit stuff.

Personally I am a fan of doing it well the first time. More hassle, more expensive but once done I think it more than pays off down the road.

YMMV

[Whack-a-Mole]

@davidm

Do not let all the crap above I am talking about confuse or scare you.

It all depends on how much of a dork you are about getting the best performance from your machine.

Kinda like some audiophiles go to extreme measures for the last 0.01 decibel. For most of us the ear buds on your iPhone are fine.

If you can get Acrtic Silver then great. Otherwise sounds like for your machine most anything will do. You MUST have a pad or thermal compound between the heat sink and CPU but sounds like, for your purposes, any of them would do fine.

Likewise you can put a dollop on or spread the compound with a credit card. Either should work (I have done both methods and had zero problems either way).

Only when you really want the best setup you can manage and want to push your PC to its limits does the stuff we are talking about really matter.

From what I have read your problem is a lack of a thermal media between heatsink and CPU (you have to have that). Second to that would be case cooling. Get some case fans on there. That said no need to go nuts. An intake fan or two and an exhaust fan behind the CPU should suffice for most purposes. If you are playing modern games and cranking the graphics to the limit then you need to do more but from what I gather a casual approach to it all should suffice.

[Dog80]

Are there any studies on the effect of different types of pastes or application methods? I doubt they make any difference.

[davidm]

I did get Arctic Silver compound so I've got the good stuff. I did notice a loss of performance over the years which I couldn't explain. Maybe it was due to cheap factory thermal paste that degraded over time.

Regarding fans; it came with three, including the one over the CPU heat sink. I assume some are intake and some are exhaust but I don't know for sure. They worked for years so I assume they're sufficient. I did purchase a PCI slot fan to provide a little extra cooling. There's no place else to put a fan.

[notsoheavyd3]

There's no way to monitor the temp. :confused: Even if there was I wouldn't have a clue as to what is or isn't an acceptable temperature. The only way I know there's a problem is when it shuts down.

Actually if you've put in a new mother board you may want to check the manufacturer's web site for a utility program that will let you monitor the temperature. So for example my motherboard is an intel DP35DP and intel provides a utility program that lets me check hard drive health and various temperatures. (And in the case of the cpu lets me know if it's a "good" temperature.) I'm guessing other motherboard manufacturers do the same.

[Whack-a-Mole]

Are there any studies on the effect of different types of pastes or application methods? I doubt they make any difference.

I have seen various results. Mostly the differences, while there, are minor (a few degrees) and only matter for the person milking every last drop.

I have heard that some of the "poor" thermal compounds break down over time and lose their efficacy. I have seen PCs I took apart where the thermal compound seemed kind of crusty (not good). I know with my PCs in the past after 4 years Arctic Silver was still goopy (as it should be).

Still, that is anecdotal. Considering even the good stuff only costs $5-10 you might as well get the good stuff if you can (although considering what you get for that money, on a per weight basis, that is actually very expensive stuff but you only need a teeny amount so unless you are putting thousands together it is not worth fussing about).

[Whack-a-Mole]

I did get Arctic Silver compound so I've got the good stuff. I did notice a loss of performance over the years which I couldn't explain. Maybe it was due to cheap factory thermal paste that degraded over time.


PCs tend to degrade over time with use. Mostly I think it is Windows getting mucked up with junk over time. Usually a re-install of Windows will put the machine back to peak performance.

Although, if your CPU was overheating, Intel CPUs have been able to throttle their speed in a first attempt to avoid damage (heat is the bane of computer electronics). If that doesn't work they shut off to avoid damage as you found out.

Considering your CPU shut off it would seem there is a temperature monitor on the board else how did it know to shut off? You should be able to find a utility that will read that temperature and tell you. Search the web for temperature ranges for your chip. IIRC it is somewhere around 75C (167F) but not 100% sure on that.

Know that even working just under the max threshold still hurts your chip. Heat, over time, degrades the chip and the hotter it runs the shorter lifespan it will have (although you could run in the 40-50s pretty much indefinitely). 60s would be iffy and low 70s start looking for more cooling.

[Zeriel]

There's no way to monitor the temp. :confused: Even if there was I wouldn't have a clue as to what is or isn't an acceptable temperature. The only way I know there's a problem is when it shuts down.

http://www.alcpu.com/CoreTemp/

Acceptable is generally "around 40C" , safe but high is generally "around 60C".

[jz78817]

Are there any studies on the effect of different types of pastes or application methods? I doubt they make any difference.

What's critical is that you use one. which one you choose isn't such a concern. (http://dansdata.com/goop.htm)

[Dr. Strangelove]

Since the motherboard is attached to the case trying to lift it with the heat sink doesn't sound wise.

Oh. I always install the processor and heatsink first. In fact, installing in the case is usually the very last thing I do. Well, the point is that there should be a fair amount of suction force involved. If the heatsink detaches immediately when you pull on it, it means that either the paste didn't make it to the edges, or that there was so much that it just flowed around.

[davidm]

It's already installed in the case now and I really don't want to take it all apart again if I can avoid it.

Come to think of it, I can't install the heat sink with the motherboard out of the case since the heat sink screws go through holes in the motherboard and screw into the case.

[Zeriel]

It's already installed in the case now and I really don't want to take it all apart again if I can avoid it.

Come to think of it, I can't install the heat sink with the motherboard out of the case since the heat sink screws go through holes in the motherboard and screw into the case.

I haven't seen THAT kind of design since the Pentium 4 era.

[jz78817]

I haven't seen THAT kind of design since the Pentium 4 era.

Big box builders like Dell, Gateway, and HP use their own cooling designs and retention mechanisms; they don't adhere to the Intel references.

[Zeriel]

Big box builders like Dell, Gateway, and HP use their own cooling designs and retention mechanisms; they don't adhere to the Intel references.

Last Dell I took apart was a Precision 470, which just has the standard side-spring clips of the late P4-Xeon era.

[davidm]

Damn.

I applied the thermal paste per the instructions and put things back together. Now the thing won't boot. The blue LED on the power switch comes on nice and bright when I push the switch and the little green LED on the motherboard is on as long as it's plugged in, but nothing else is happening.

There are no beeps, no drive sounds, and none of the fans even come on. What did I do to it? Even if I fucked up the CPU wouldn't the fans at least come on?

I installed a PCI fan and that just plugs into a standard power connection. It's not doing anything. It's like I have power, but I don't have power.

[runner pat]

You were stuck on the ceiling a few minutes ago. How did you get down?

[davidm]

You were stuck on the ceiling a few minutes ago. How did you get down?Shh... Different thread. Party pooper.

[Zeriel]

Damn.

I applied the thermal paste per the instructions and put things back together. Now the thing won't boot. The blue LED on the power switch comes on nice and bright when I push the switch and the little green LED on the motherboard is on as long as it's plugged in, but nothing else is happening.

There are no beeps, no drive sounds, and none of the fans even come on. What did I do to it? Even if I fucked up the CPU wouldn't the fans at least come on?

I installed a PCI fan and that just plugs into a standard power connection. It's not doing anything. It's like I have power, but I don't have power.

I'd say check every single power connector on the board.

[davidm]

I removed the heat sink and lifted up the CPU. There was a small speck of something sitting on some of the contacts in the CPU socket. I blew that away, put everything back together, and now it boots into Windows! :D

Now I'll have to make sure that the heat problem is solved.

[davidm]

Well, it seems to be running great. The fan is much quieter than before. I installed Speedfan, a temp monitoring program, and according to it the temperature is holding steady.

I was having some weird keyboard problems before that had me considering a new keyboard, but those seem to be gone. :D

[davidm]

I spoke to soon. Now it's locking up at random times. I'll be working on it, surfing the web or whatever when suddenly the screen freezes and it no longer responds to the mouse or the keyboard and I have no choice but to shut the power off and reboot.

No sudden shutdowns or messages about thermal issues, it just freezes and requires a reboot.

And rebooting is now a problem. It won't always do it. This morning it was giving three beeps (apparently indicating a memory issue) and then just sitting there. After several tries with this result it suddenly booted.

Tonight it locked up on me twice. Both times I had to power down, and both times I had to try booting several times before it succeeded. But this time there were no beeps.

So WTF? It's a brand new motherboard, so as I see it there're only a couple of possibilities, the CPU or the RAM. It can't be the drive because when it fails to boot it doesn't even get to that point. The three beeps apparently indicate memory issues so I suppose that's the next thing I'll replace, although my gut tells me that it's the CPU.

I got a new keyboard to replace the flaky old one and, just my luck, it's defective. The left shift key emits a string of "|" characters when I press it. I can't win lately.

[Dr. Strangelove]

So WTF? It's a brand new motherboard, so as I see it there're only a couple of possibilities, the CPU or the RAM. It can't be the drive because when it fails to boot it doesn't even get to that point. The three beeps apparently indicate memory issues so I suppose that's the next thing I'll replace, although my gut tells me that it's the CPU.

Download and burn a bootable CD of MemTest86+. Run it for at least 24 hours. That will tell you if the memory is flaky.

Prime95 is also a good one, but that will only tell you what you already know: that either the CPU or memory is flaky. Well, there are other possibilities, but you're right that these are the most likely.

[Mr. Slant]

On an aging machine, never assume the power supply is good, either.

[davidm]

On an aging machine, never assume the power supply is good, either.I've considered that also. I suppose that a flaky power supply could be the cause of the kinds of inconsistent behavior I've been seeing.

Something did kill the original motherboard, after all. (Unless the motherboard was actually fine and the problem was caused by something else.)

[davidm]

Download and burn a bootable CD of MemTest86+. Run it for at least 24 hours. That will tell you if the memory is flaky.

Prime95 is also a good one, but that will only tell you what you already know: that either the CPU or memory is flaky. Well, there are other possibilities, but you're right that these are the most likely.The machine will not run for 24 hours without locking up. It doesn't seem like it will even run 2 hours without locking up.

[Quartz]

Just a thought, but have you got the heatsink on the right way around? I once cooked a CPU by putting the heatsink back on the wrong way round which left an airgap between the CPU and heatsink.

[davidm]

Just a thought, but have you got the heatsink on the right way around? I once cooked a CPU by putting the heatsink back on the wrong way round which left an airgap between the CPU and heatsink.I could be mistaken but I don't think it's possible to have it on wrong.

In any case, I originally had it on without thermal paste. It was definitely heating up then, but the sypmptoms were different. It didn't lock up, it just suddeny shut off all power without warning and when I rebooted it displayed a message saying that it had been shut down due to a thermal event. It's not doing that now so I think it's a different problem.

However, I am open to the possibility that there is still a heat problem, I'd just need to understand why I'm not seeing the same symptoms as before.

The speedfan software I installed shows that the CPU temp mostly hovers around 55% C, Occasionally going into the mid 60s. I'm having difficulty finding any specs as to what an acceptable range is.

The System Properties dialog says that it has an "Intel Pentium(R) 4 CPU 3.20GHz".

Speedfan currently shows:
LIST]
GPU:59C
CPU: 54C
Local: 33C
Remote 2: 49C
HDD: 34C
[/LIST]

[davidm]

My editing time ran out before I could fix that mess at the bottom of that post.

Speedfan currently shows:
GPU: 59C CPU: 54C Local: 33C Remote 2: 49C HDD: 34C

I'm not sure what some of the things on that list are.

[davidm]

I've got Prime95 running a stress test. Does it keep logs in case it locks up?

[GusNSpot]

Since I am not a neat freak and I also think a computer case is a terrible thing to spend $$$ on as they are so unnecessary, I have a solution that causes my computers to never over heat.....

Cool computing made easy. (https://picasaweb.google.com/Dragon43/ComputersOnTheWall#)

Bawahahaha :D:D:D

KISS is the answer..... YMMV

[davidm]

Prime95 has been running for over two hours. Both CPUs are at 100% and it hasn't locked up or failed any tests yet. :confused:

[johnpost]

I spoke to soon.

unless you've rebooted the machine three dozen times and run it for two months then that is often the situation.

[Whack-a-Mole]

Prime95 has been running for over two hours. Both CPUs are at 100% and it hasn't locked up or failed any tests yet. :confused:

Then it would seem the CPU is not your problem nor is heat a problem with the CPU (you fixed that).

Your problem is probably with the memory.

Take your memory sticks out and re-seat them. Be sure they are firmly in and locked down (there is a clip on both sides that when fully engaged means the memory is seated appropriately). Make sure they are in the right banks on the mobo (many mobos have the memory divided into banks, make sure they are arranged correctly).

Check your BIOS. There are various settings in there about memory. Usually the BIOS auto-detects the memory and has the proper settings but it may have something wrong. Check your manual and look at the settings and compare to what memory you have (there are a lot of timing settings and such...if you do not know then don't mess with it). You may be able to reset the BIOS to defaults though which can help. Again, if you are unsure of what you are looking at in there then leave it alone.

Update your BIOS. Check the manufacturer website and get the latest/greatest BIOS. Be careful when updating BIOS...it is not hard but you can brick your mobo (permanently disable it) if the update fails (e.g. the power goes out partway through the update...fortunately the update occurs quickly).

[davidm]

Then it would seem the CPU is not your problem nor is heat a problem with the CPU (you fixed that).

Your problem is probably with the memory.

Take your memory sticks out and re-seat them. Be sure they are firmly in and locked down (there is a clip that when fully engaged means the memory is seated appropriately). Make sure they are in the right banks on the mobo (many mobos have the memory divided into banks, make sure they are arranged correctly).

Check your BIOS. There are various settings in there about memory. Usually the BIOS auto-detects the memory and has the proper settings but it may have something wrong. Check your manual and look at the settings and compare to what memory you have (there are a lot of timing settings and such...if you do not know then don't mess with it). You may be able to reset the BIOS to defaults though which can help. Again, if you are unsure of what you are looking at in there then leave it alone.

Update your BIOS. Check the manufacturer website and get the latest/greatest BIOS. Be careful when updating BIOS...it is not hard but you can brick your mobo (permanently disable it) if the update fails (e.g. the power goes out partway through the update...fortunately the update occurs quickly).
Yea, I'm suspecting the memory also. I did have one or two memory error messages that I had attributed to the heat issue. Plus, at one point it was giving 3 beeps when it wouldn't boot, which supposedly indicates memory problems on intel boards.

The mobo is brand new. Of course that doesn't guarantee that it's up to date. I'll check the BIOS version.

What I don't get is that I can lock it up by surfing the web for a couple of hours, but apparently not by running Prime95. I wonder what the difference is?

[jz78817]

Yea, I'm suspecting the memory also. I did have one or two memory error messages that I had attributed to the heat issue. Plus, at one point it was giving 3 beeps when it wouldn't boot, which supposedly indicates memory problems on intel boards.

The mobo is brand new. Of course that doesn't guarantee that it's up to date. I'll check the BIOS version.

What I don't get is that I can lock it up by surfing the web for a couple of hours, but apparently not by running Prime95. I wonder what the difference is?

depending on what preset you use, Prime95 may not stress much memory. if one or more of your DIMMs has a flaw at a particular location, P95 might not ever be trying to read or write to that region. other programs (or the OS) could be. Do run Memtest86+ overnight on your system and see what happens. it's not foolproof but IME it's very good at telling you if you have a problem.

[Whack-a-Mole]

The mobo is brand new. Of course that doesn't guarantee that it's up to date. I'll check the BIOS version.


Almost never is new computer equipment up-to-date right out of the box. By the time they manufacture it, ship it across an ocean, distribute to warehouses then distribute to retailers who may have it on a shelf for weeks or months means most everything you buy for your computer will have out-of-date software drivers.

Indeed when I install something new I use the provided drivers and then immediately update. I cannot think of a time where there was not a new driver already out.

You never know but definitely worth checking. Only takes a minute to find out.

[davidm]

Download and burn a bootable CD of MemTest86+. Run it for at least 24 hours. That will tell you if the memory is flaky.

Prime95 is also a good one, but that will only tell you what you already know: that either the CPU or memory is flaky. Well, there are other possibilities, but you're right that these are the most likely.
I booted the CD and it came up and immediately started testing. I assume it's using default settings. Should I let it run or do I need to configure it?

[jz78817]

I booted the CD and it came up and immediately started testing. I assume it's using default settings. Should I let it run or do I need to configure it?

default is fine, just let it run.

[davidm]

It's running. Is there some end point where it indicates that it's finished, or do I just stop after a day or so.

By the way, THANK YOU to you and everyone else who's been so helpful in this thread. I hope that I'm able to repay the favor one day.

The SDMB is amazing. It's not even a technical board, but no tech board I've seen gets such a volume of useful responses so quickly.

[davidm]

Okay I read up on Memtest. It runs indefinitely and you stop it when you want.

Right now it's on pass #15 with no errors. With the frequent problems I've been having it seems likely would have found something by now.

So it seems like it's down to the CPU or the power supply. I suppose the video card or the audio card are also possibilties.

This all started when it was acting flaky about starting up and then eventually failed to boot at all. This was the point at which I replaced the motherboard.

Fixing the motherboard got it to where it would boot again. So presumably there was a mother board problem and replacing it corrected it.

So I see a few possibilities.


There is some other issue totally coincident with the mother board problem. This seems unlikely.
There was another issue to start with and that issue damaged the original mother board.
Maybe the problem started with the original motherboard and that motherboard problem damaged something else.
The issue was never with the motherboard but something about the new one (newer bios maybe?) is causing it to be slightly more tolerant of the real issue.
Maybe my ignorance regarding thermal compound damaged the CPU.
Maybe I somehow damaged something else when I installed the new motherboard.

[davidm]

I had another shutdown last night followed by a "thermal event" message when I turned it back on.

The shutdown was proceeded by a total freeze where it wouldn't take any input and the mouse cursor wouldn't move.

I turned it off when I saw the thermal message and let it sit a few minutes then tried to boot it.

The power light came on, the disk activity light flickered a few times then nothing. I tried several times, waiting a few minutes between tries. During these tries I was unable to shut it off with the power button and had to pull out the power cord instead.

Finally it did something different at bootup and beeped three times (memory error) before hanging, so I turned it off and removed everything but the bottom two gigs.

It still hung up on booting a few times, but with no beeps. Finally it successfully booted into Windows again. It ran a while and then hung up again so removing the memory didn't completely fix it and presumably the removed memory is good (it did pass the memtest86+ tests).

Here's a theory. I'm seeing behavior similar to what I saw before I replaced the MB, except that I didn't get the thermal even messages before. Maybe the problem has been thermal all along. Maybe the original MB just didn't report it and the newer MB is better at detecting and/or reporting thermal problems; maybe because of a newer BIOS.

Maybe the old MB wasn't really bad, but I just didn't give things time to cool down before giving up on it.

The one thing that I can't explain with that theory is the behavior of the power light with the old MB. The power button has a bright blue LED which lights up when there's power, When the machine started giving me problems booting and before it died completely (or at least I thought it had) that blue LED wasn't lighting or was very dim even when it was on and seemed to be running normally. Since replacing the MB it's always been very bright.

So, why did replacing the MB fix the LED issue but not the other issues? I've been wondering if the LED may indicate a power supply issue, but if there wasn't enough juice to power the LED, I would think that there wouldn't be enough to boot up and run the computer.

I keep thinking that there must be one point of failure (presumably other than the MB) that's affecting, and possibly damaging, other things and causing all the different issues I'm seeing.

[davidm]

Still getting thermal shutdowns.

Suggestions? Should I reapply the thermal compound?

[Whack-a-Mole]

Still getting thermal shutdowns.

Suggestions? Should I reapply the thermal compound?

Remember too much thermal compound is not good. You want it as this as possible to still do the job.

And again, getting air into the case and back out is as important as the heat sink. The heat sink cannot do its job well if there is hot air in the case.

Pay attention to air flow in the case. Both intake and exhaust fans. Visualize how the air moves in the case. You want good air flow past the CPU and video card. Remember, the video card can produce substantial heat too (depending what you are doing).

Also, clean up cables. If it is a rat's nest of cables in there that hinders good air flow. Do some cable management and clean things up. Clean off lint and dust too. Dust on the heatsink diminishes its effectiveness (and affects the fans).

[davidm]

The airflow is the same as it's been for years; plus I added a PCI fan. So, if anything, there is more airflow.

The cables are a possibility, since I did replace the MB they're not necessarily arranged as they originally were. However, I haven't put the side back on the case yet. Shouldn't the side being open help?

I cleaned out the dust when I replaced the MB.

[Mr. Slant]

Sometimes the fans can't do their jobs with the case side off.
They could wind up re-circulating warm air rather than getting the cold in and the hot out.
In your shoes, I would certainly see what happenned with and without the case fan.
I'd also buy some hardware (or even use some tape) that I would use to re-arrange the cables to maximize airflow.

On edit:
Also, just for kicks, I might take a powerful box fan and put it right up against the open side of the case. I'm talking about a fan that could provide stupid-high amounts of airflow. If you still have problems with that much airflow going across your board, then you've developed a defect in your mobo or cpu.

[Whack-a-Mole]

Actually having the side open can hurt.

A closed case should make a river of air flow past the CPU. When it is open this may not happen.

Depends but open case is not necessarily a cooler case.

Try pointing a floor fan at it with the case open (from some distance away as fan motors put out a fair bit of EM interference). See if that helps. It should at least tell you if the problem is air flow or a poorly mounted heat sink.

Personally I use tension screws on my heat sinks (generally a feature of higher end heat sinks) which means the pressure on the CPU is dummy proof.

Doing it without those is harder. The heat sink needs to be on pretty firmly against the CPU to work but not so firmly you crack the CPU.

[davidm]

As I said before, The heat sink has tension screws. It screws into the motherboard and there'd be no way to keep it on without them.

The idea that the open case could actually be a problem is interesting. It seems to be the one factor that's been there through all of this.

This all started when I installed a new hard drive. I wasn't having these issues before that. After putting in the new drive, I hadn't bothered to put the side back on. I wanted to see how it worked before putting it back on and then I just got lazy and didn't bother.

So all of the problems have occurred since I've been running with the side off. That's the one constant factor.

[davidm]

I've been running with the case closed and the cables as out of the way as is possible.

I'm still having problems. Sometimes it shuts down completely and on reboot tells me that a thermal event occurred. Other times it just totally freezes, won't accept keyboard or mouse input, and has to be shut off via the power button. Sometimes the power button won't even shut it off and I have to pull the plug.

I've replaced the mother board, the memory passed hours of testing with memtest86+, I've applied thermal paste properly (at least I think I did it properly), I've arranged the cables, I've cleaned out any dust, and I've run with the case both open and closed.

None of that has solved anything. A couple of times I thought I'd solved it but the problems eventually returned.

So I think it's down to the CPU, the power supply, or possibly the graphics or audio cards. Since I'm getting thermal events, it seems like it must be the CPU.

So I guess the next step is to replace the CPU. Anyone agree or disagree?

[Mr. Slant]

My career in PCs is 14 years old. I agree with CPU.

[davidm]

I originally replace the MB because it wouldn't boot at all. Replacing it at least fixed that. Could a failing CPU have damaged the original MB? Could a failing MB have damaged the CPU?

My theory at this point is that the original problem really was a motherboard failure and when I ran the new motherboard without proper thermal protection I damaged the CPU. :smack:

[Mr. Slant]

Your question 1: Yes, absolutely. Hot CPU burns mobo with 200 degrees, sure.
Question 2: possibly. If the regulators on the mobo fed the wrong volts to the CPU, it could happen.

Your theory: Plausible.

One argument against your theory is that most modern CPU/mobo combos have design features that should keep thermal fails from breaking stuff.
I would go ahead with your plan.

[davidm]

Can I put in a faster / more powerful processor than the old one? If so, how can I tell what would be compatible with my motherboard?

[Mr. Slant]

The best bet is to check the docs for your motherboard.

[davidm]

From what I can see, it already has the most powerful one listed in the specs. It's a dual core Intel Pentium D 820 Processor.

Since it's an older MB I guess I was wondering if there a faster version of the processor that will work on the board. Probably not, I guess.

[Mr. Slant]

Yeah, sometimes you're at the top of the range your mobo can support.
Even if faster chips on the same pin-out come out, your chipset might be too dumb to understand it, even with a flash.
Also, the voltages sometimes change between versions of the CPU.
I've been doing corporate work for a while, though, and most Fortune 500s don't go in for CPU upgrades, so I haven't done one in years.

[davidm]

I can get a replacement for the current one online for a pretty good price so I guess I'll just go with that.

[Fear the Turtle]

I have a related question: If you do get a blue screen - how do you read those .dmp files Windows creates?

[Dr. Strangelove]

I have a related question: If you do get a blue screen - how do you read those .dmp files Windows creates?

There is a program called WinDbg (available free from Microsoft) that allows you to load a .dmp file and, assuming you have the symbols, allows you to see exactly where the crash occurred.

It's not much use unless you're a driver or kernel programmer, though.

[Fear the Turtle]

There is a program called WinDbg (available free from Microsoft) that allows you to load a .dmp file and, assuming you have the symbols, allows you to see exactly where the crash occurred.

It's not much use unless you're a driver or kernel programmer, though.

That appears to be beyond what I'm looking for then. So how does one find out why a blue screen occurred, especially if you are not there when it occurs or it reboots before the blue screen info can be hand written?

[davidm]

I installed the new CPU and it still doesn't boot reliably. I had to try several times after I got it back together before it would boot.

The first time the blue power light came on and it beeped twice and then just sat there. After a number of tries it finally booted. Now, sometimes it boots and sometimes it doesn't. When it doesnt, the power light comes on the disk light flashes and couple of times and then nothing. There are no beeps except for that first time. When it does boot there are no beeps and it brings up Windows normally.

:confused::confused::confused:

I have a new motherboard. A new CPU. The memory passed an overnight run of Memtest. I've considered that maybe it's the graphics card and that it is booting but the display is just blank. The problem with this idea is that I don't see the amount of disk seek activity that I see when it boots normally.

I guess I should just go for a whole new system but this has become a challenge (and a learning experience). Obviously this is fixable. I just have to replace the right part.

So now I'm thinking that it must be the power supply. What else, that I haven't either tested or replaced, could intermittently interfere with booting?

Then there's still the question of the thermal events. Could that have come from the power supply. Could a wonky power supply send too much power and fry other things?

[Mr. Slant]

Yes.
Overdriving a circuit could cause thermal events.
I can't say that I've seen that before, but if you were one of my commercial clients in the past, I'd have already replaced your power supply anyway by now.

[davidm]

Okay, I re-seated the cable that connects the power button to the motherboard. Now it boots easily and consistently every time. :smack:

Since there are about half a dozen different wires in that bundle I have to think that that button sends signals to more than one system, so it's conceivable that sometimes some systems were coming up and other weren't

The bad connection obviously occurred when I installed the new motherboard so it's unlikely this was the original cause of my problems. It's also unlikely that it was the cause of the heat issues I experienced with the new MB. Hopefully I've corrected those problems with the replacement parts and this bad connection was the final issue.

[davidm]

Yes.
Overdriving a circuit could cause thermal events.
I can't say that I've seen that before, but if you were one of my commercial clients in the past, I'd have already replaced your power supply anyway by now.I didn't see this post when I posted my previous message. You think it would be a good idea to replace the power supply anyway, since it could fry the new parts?

Is there a reliable way to check the output of a power supply. I know about things like multimeters but I'd think something more sophisticated is needed.

I know it's unlikely but is there software that will check it? I know fanspeed shows some voltage readings but I don't know how to interpret them or if they're even relevant.

[davidm]

Okay, I re-seated the cable that connects the power button to the motherboard. Now it boots easily and consistently every time. :smack:

Since there are about half a dozen different wires in that bundle I have to think that that button sends signals to more than one system, so it's conceivable that sometimes some systems were coming up and other weren't

The bad connection obviously occurred when I installed the new motherboard so it's unlikely this was the original cause of my problems. It's also unlikely that it was the cause of the heat issues I experienced with the new MB. Hopefully I've corrected those problems with the replacement parts and this bad connection was the final issue.

Weird. I did the above with the front of the case off so I could access the power switch directly rather than by pushing on the plastic button on the case front.

I put it back together, the side and the front, and suddenly it was refusing to boot again.

So, thinking there was an issue with the "button" on the case (a piece of plastic that pushes the actual button) I opened the case again to give me direct access to the actual button again. Now, pushing the actual button no longer boots the system. The light comes on and I think one or two fans in the back come on. Interestingly, I notice that the CPU fan does not come on (it definitely came on when it was booting properly a little while ago.)

So maybe this is connected to the overheating. Maybe at times it was seeming to boot properly but the CPU fan wasn't on.

So it seems like it's down to the button, the cable coming from the button, the connection of that cable to the MB, or the power supply.

[davidm]

I reseated the cable that connects the CPU fan to the motherboard. It now boots.

Reading around the net, I discovered that many motherboards won't boot if they don't detect that the CPU fan isn't on (a perfectly reasonable precaution). Since it apparently wasn't seated properly, it wasn't coming on, and the MB correctly refused to boot.

Now I'm wondering if there's a problem with the fan connector. It could be connecting poorly and intermittently losing a connection and causing the thermal events I've been seeing. It could explain the intermittent booting (depending on whether the fan was starting), and the thermal events. It might even explain the freezes which might have somehow been caused by overheating.

So now I'm wondering whether to spend money on a new fan or a new power supply. Maybe I should just replace both. :confused:

[davidm]

Another thermal shutdown, while watching a video.

[davidm]

After months of performing perfectly, last night... nothing. I hit the on button and it lights up bright blue and I can hear the fans spinning, but nothing else. No drive seeks, no POST, no Gateway logo on the screen, and the only way to turn it off (the light and the fans) is to unplug it. The obvious conclusion is that the CPU is not waking up. Nearly everything is new - the CPU, the motherboard, the disk drive - and the memory had passed a stress test. So the only thing left is the power supply, which sort of makes sense to me given the symptoms.

So I'm picking up a new power supply at lunch since I work near a Microcenter. It has more watts than the original so I'm hoping that it will finally fix this problem.

This whole mess started when I installed a bigger hard drive. Maybe it was too much for the factory power supply to handle.

[Mr. Slant]

More like installing the larger hard drive coincided with the age and wear-based beginning of the failure of the power supply.
Larger hard disk capacity generally doesn't draw more power.

[jz78817]

More like installing the larger hard drive coincided with the age and wear-based beginning of the failure of the power supply.
Larger hard disk capacity generally doesn't draw more power.

it does if there are more platters inside, or has a higher spindle speed.

[Mr. Slant]

Granted.
I considered that before posting, but when most consumers upgrade their computer they're not actually increasing the power draw on the drive.

[davidm]

In any case it seems like it's the only thing left to replace other than the video card, and that seems to be an unlikely source of the problems.

[davidm]

I just bought a power supply. I have to sit home and wait for a repairman tomorrow (for an air conditioner) and they gave me a frigging 9 hour window for when he'll be there. So now I have a project to work on during my forced confinement.

[drachillix]

From what I can see, it already has the most powerful one listed in the specs. It's a dual core Intel Pentium D 820 Processor.

The D series chips almost always ran hot

[davidm]

I bought a 575W power supply. The original is 400W so there should be no question about having enough power.

But now I have another dilemma. My motherboard has a 4 pin processor power connector that looks like this:
http://support.gateway.com/s/MOTHERBD/Intel2/WMEOEMD945GCZG1/WMEOEMD945GCZG1mvr3.shtml

The problem is that the new power supply doesn't have a 12 volt connector that fits that.

I think I can use an adapter like this: http://www.playtool.com/pages/psuconnectors/12v4pinadapter.jpg
and hook it to one of the 4 pin Molex connetors coming off of the power supply. It seems like it should work fine as long as the ground and 12V lines are correct.

Am I missing something? Can I do that?

[yoyodyne]

The new PS should have an 8-pin that either splits in half or can be connected to the 4-pin with the other half in the air.

[davidm]

The new PS should have an 8-pin that either splits in half or can be connected to the 4-pin with the other half in the air.You're right, it does have one like that with yellow and black leads and I think it will snap on with half in the air. Not sure how I missed that. I'll just make sure that the wire colors match up. Thanks!

ETA: It slides apart into two pieces!

[davidm]

Up and running. Hopefully for good this time.