RoozerXC wrote on 2024-03-07, 00:02:

<snip>
- repaste the CPU and heatsink together

You need to delid the CPU and then apply some good paste between the die and the heatspreader... Yes, the TIM does not age well and AMD did not solder the core to the heatspreader, like they should have done.

You can then place the heatspreader back on, but make certain it's put back on correctly.

There's also a risk of breaking off an SMD component when you insert the flexible razorblade to break the silicone bond between the heatspreader and the PCB. This happened to me and I had to do microsurgery to get the tiny resistor network soldered back in its place.

I ultimately used two thin copper heatspreaders (with paste applied between each layer) and went for a direct die cooling approach because I found that getting the heatspreader back on right was finicky.

anyways, good luck

If the paste pattern shows improper contact, then nothing else is going to fix that.

Is this the correct Compaq heatsink? Did this machine originally use some kind of shim or thermal pad?

If the HSF is not making contact delidding will only make that worse.

All really old Athlon64 CPUs (pre-AM2) have a thermal paste under the IHS. Delid and reapply fresh paste (or liquid metal).

The Serpent Rider wrote on 2024-03-07, 00:35:

All really old Athlon64 CPUs (pre-AM2) have a thermal paste under the IHS. Delid and reapply fresh paste (or liquid metal).

He said the heatsink is not making proper contact. Delidding won't fix that.

Socket 939 and AM2 (and later) are somewhat interchangeable, but not completely. I've tried AM2 coolers that would not properly fit a 939 bracket (couldn't make proper contact with the CPU), and one AM2 cooler I used made me realize that they changed the mounting bracket between 939 and AM2. The AM2 bracket uses four screws to secure it rather than the two used with 939. The four screw bracket is more rigid, so the HSF's clamping mechanism has less travel - this means that it doesn't clamp as securely as one made for the 939 bracket. Since the AM2 cooler was copper with heatpipes (made for a 6000+) I wanted to use it; but the 939 motherboard came with flexible push pin retainers and a plastic back plate. So I bought a 939 bracket that was secured with screws and a metal back plate.

The Serpent Rider wrote on 2024-03-07, 00:35:

All really old Athlon64 CPUs (pre-AM2) have a thermal paste under the IHS. Delid and reapply fresh paste (or liquid metal).

Fixed... There are many "post-AM2" that also use the thermal paste under the IHS.

kingcake wrote on 2024-03-07, 00:24:

If the paste pattern shows improper contact, then nothing else is going to fix that.

Is this the correct Compaq heatsink? Did this machine originally use some kind of shim or thermal pad?

If the HSF is not making contact delidding will only make that worse.

I have literally no idea, it's just some random heatsink that this machine came with when I first picked it up. It never came with a thermal pad or some shims of any kind

Edit: there's also no model number or manufacturer name too

mockingbird wrote on 2024-03-07, 00:23:

You need to delid the CPU and then apply some good paste between the die and the heatspreader... Yes, the TIM does not age well […]
Show full quote

RoozerXC wrote on 2024-03-07, 00:02:

<snip>
- repaste the CPU and heatsink together

You need to delid the CPU and then apply some good paste between the die and the heatspreader... Yes, the TIM does not age well and AMD did not solder the core to the heatspreader, like they should have done.

You can then place the heatspreader back on, but make certain it's put back on correctly.

There's also a risk of breaking off an SMD component when you insert the flexible razorblade to break the silicone bond between the heatspreader and the PCB. This happened to me and I had to do microsurgery to get the tiny resistor network soldered back in its place.

I ultimately used two thin copper heatspreaders (with paste applied between each layer) and went for a direct die cooling approach because I found that getting the heatspreader back on right was finicky.

anyways, good luck

The same issues apply to an older 3400+ that the machine originally came with too. I had purchased an AMD Athlon 3700+ as an upgrade on eBay a little while ago and I can't be assed to delid it with a razor blade since I would risk cutting myself, and due to my mental state I feel very uneasy with blades. Perhaps a delid would make a difference but I'm not risking it

Repo Man11 wrote on 2024-03-07, 00:38:

Socket 939 and AM2 (and later) are somewhat interchangeable, but not completely. I've tried AM2 coolers that would not properly fit a 939 bracket (couldn't make proper contact with the CPU), and one AM2 cooler I used made me realize that they changed the mounting bracket between 939 and AM2. The AM2 bracket uses four screws to secure it rather than the two used with 939. The four screw bracket is more rigid, so the HSF's clamping mechanism has less travel - this means that it doesn't clamp as securely as one made for the 939 bracket. Since the AM2 cooler was copper with heatpipes (made for a 6000+) I wanted to use it; but the 939 motherboard came with flexible push pin retainers and a plastic back plate. So I bought a 939 bracket that was secured with screws and a metal back plate.

Only difference is the heatsink has 2 metal latches, not screws (if you know what I mean basically). There are no screws involved to put the heatsink in place and/or to remove it from the CPU

The bracket in question. As I stated before, the HSF mounting bracket for Socket 939 isn't as rigid, so the HSFs for 939 have a longer throw on the clamping mechanism to make sure there is enough clamping force on the CPU. AM2 → have a shorter throw, which can be an issue. And some AM2 → HSFs have fins which will clear the bracket for an AM2 CPU, but will interfere with a Socket 939 bracket, which seems to be the issue that you have.

mockingbird wrote:

There are many "post-AM2"

Which has nothing to do with my statement.

Repo Man11 wrote on 2024-03-07, 01:42:

The bracket in question. As I stated before, the HSF mounting bracket for Socket 939 isn't as rigid, so the HSFs for 939 have a longer throw on the clamping mechanism to make sure there is enough clamping force on the CPU. AM2 → have a shorter throw, which can be an issue. And some AM2 → HSFs have fins which will clear the bracket for an AM2 CPU, but will interfere with a Socket 939 bracket, which seems to be the issue that you have.

Ahhh right. My frustration sometimes makes me mix things up and I get sidetracked a little bit so please excuse me

To my knowledge I don't have an AM2 bracket or heatsink at least as far as I know, and I'm still considering other solutions than just delidding or anything that might seem very hacky or risky (or could even cause injuries and additional damages). What im thinking is to maybe get a better 939 heatsink and bracket, but im not exactly sure what kind i would get in order to get stable temps again, since the OEM heatsink that came with this machine is probably subpar and terrible for the course given my recent "experiences"

RoozerXC wrote on 2024-03-07, 01:30:

The same issues apply to an older 3400+ that the machine originally came with too. I had purchased an AMD Athlon 3700+ as an upgrade on eBay a little while ago and I can't be assed to delid it with a razor blade since I would risk cutting myself, and due to my mental state I feel very uneasy with blades. Perhaps a delid would make a difference but I'm not risking it

This post conjured up this image in my mind, for some reason:

To test if the heatsink made good contact with cpu, you can do this simple test.

put a bean size pea size blob of thermal paste at the center of cpu. prefer the soft cheap paste like GD900 or mx4
clean the heatsink contact area, put the heatsink on, leave it 1 minute for the paste to spread then remove the heatsink.
if heatsink made good contact, the paste should spread thin evenly and should almost cover the IHS area.
if heatsink does not made good contact or there are gap, you will see a thick paste and it not spread out much.

#EDIT: change the size from bean to pea

RoozerXC wrote on 2024-03-07, 03:04:

Repo Man11 wrote on 2024-03-07, 01:42:

The bracket in question. As I stated before, the HSF mounting bracket for Socket 939 isn't as rigid, so the HSFs for 939 have a longer throw on the clamping mechanism to make sure there is enough clamping force on the CPU. AM2 → have a shorter throw, which can be an issue. And some AM2 → HSFs have fins which will clear the bracket for an AM2 CPU, but will interfere with a Socket 939 bracket, which seems to be the issue that you have.

Ahhh right. My frustration sometimes makes me mix things up and I get sidetracked a little bit so please excuse me

To my knowledge I don't have an AM2 bracket or heatsink at least as far as I know, and I'm still considering other solutions than just delidding or anything that might seem very hacky or risky (or could even cause injuries and additional damages). What im thinking is to maybe get a better 939 heatsink and bracket, but im not exactly sure what kind i would get in order to get stable temps again, since the OEM heatsink that came with this machine is probably subpar and terrible for the course given my recent "experiences"

You cannot have an AM2 bracket because they are only on Am2 motherboards and will not fit on 939 because they have four mounting screws rather than two. But you might have an AM2 heatsink, especially since some of them will not fit properly when used with a Socket 939 mounting bracket.

You ought to see an impression like this on the heatsink after you have mounted it and clamped it down; the thermal grease should be uniformly thin across the contact area. If it doesn't look like this, you need to investigate why it isn't making proper contact - most likely reason is some interference between cooling fins and the plastic mounting bracket.

It might help to post some pics of the CPU in the socket and the bottom of the heatsink assembly.

Also agree to do a spread test. But "bean size" is way too much. Do a pea. Using too much paste will give a false result because it will look like you have contact when you really don't.

kingcake wrote on 2024-03-07, 03:34:

It might help to post some pics of the CPU in the socket and the bottom of the heatsink assembly.

Also agree to do a spread test. But "bean size" is way too much. Do a pea. Using too much paste will give a false result because it will look like you have contact when you really don't.

Yeah, i was trying to use different patterns on the CPU but they all ended up being very messy to work with. I will try repasting it tomorrow since I'm taking a break from working on the machine enough times, and having me pull my hair out a lot

And sorry if i never had any pictures. I'll try to borrow a relative's phone to take some pictures of the CPU and the heatsink assembly as well since my phone is unusable atm

mockingbird wrote on 2024-03-07, 03:15:

RoozerXC wrote on 2024-03-07, 01:30:

The same issues apply to an older 3400+ that the machine originally came with too. I had purchased an AMD Athlon 3700+ as an upgrade on eBay a little while ago and I can't be assed to delid it with a razor blade since I would risk cutting myself, and due to my mental state I feel very uneasy with blades. Perhaps a delid would make a difference but I'm not risking it

This post conjured up this image in my mind, for some reason:

unnamed2.png

LMFAOOOOOOOO good one 🤣 I was actually being serious but that made me giggle a bit so it cheered me up a little, sorry if this seems to be getting off topic but I found that funny

If you're getting the overheating issue with both the original (3400+) CPU and the new (3700+) one, I suspect the issue is more likely to be that the heatsink is not seated properly. Of course, it could also be a number of other things too, like the motherboard sensor is not reading the CPU temperature correctly. Or perhaps both CPUs really do have dried thermal paste under the heat spreader (though I find that to be more unlikely, as I have quite a few s939 CPUs and have not run into this issue yet.)

So let's start in order.
To easily rule out that the heatsink contact is not an issue, just remove the mounting bracket from the motherboard and put the heatsink directly over the CPU. This, of course, would only work if you lie the PC down on its side so that the motherboard is horizontally flat so the heatsink won't fall or slide off.

And here's a PRO tip: thermal paste / compound should only be changed IF it's too dry to spread or has too much debris / foreign matter (like dust or hair) contaminating it.
If you've just applied fresh thermal compound on a CPU and it doesn't work or you need to swap that CPU for whatever reason, don't junk / discard the thermal paste from it. Take a CLEAN credit card, sharp-ish spatula (plastic preferable), or some similar "tool" (no razor blades I suppose, if you don't like them 😁 ) and scoop the "old" compound from the CPU. The put on the new CPU and spread. If it's not enough, add a bit more fresh compound. That's all!
This will save you from wasting quite a bit of thermal compound.
I have plenty of CPUs and GPU's done with such "recycled" compound and have never had overheating issues.

The only way thermal compound will dry out is if the device is allowed to reach high temperatures (be it due to poor cooling from a clogged heatsink / poor airflow or a heatsink that is not adequate for the job.)

Anyways, back to regular scheduling...
If you tried the above (running the CPU without a the mounting bracket on the motherboard and only have the heatsink on top of the CPU) and you still get high temperatures with both CPUs, then perhaps the mounting bracket wasn't the issue. Which moves us onto test #2: try some different thermal monitor apps like HWInfo, OCCT, and SpeedFan to see what kind of temperatures they show with the CPU idling on desktop. Post some screenshots, if possible. Some motherboards have "cooky" sensors that may show bogus temperatures (and voltages.) Hopefully this can tell us if that's really so.

If not, then perhaps you really did get the short end of the stick with both CPUs and both just need de-lidding and applying fresh thermal compound under the IHS. But let's see after you do the above two tests first.