Ignoring extreme liquid metal ones, there are certainly better TIMs than ceramique 2 but the difference is only likely to be a few degrees. The best at the moment is probably something from the Thermal Grizzly range

http://thermal-grizzly.com/en/

1 or 2 degrees, nevermind then.

What about adding a peltier pad between the CPU and the heatsink? Anyone have long-term experience with such configuration? What was the temperature differential? I have a 12 V peltier pad somewhere, but last I looked for it, it had completely vanished.

feipoa wrote:

Wait... It seems like the newer Freezer 64 Pro's have 4-pin connectors. So this thing is going to run full speed at 20K RPM? Won't that be loud? Anything with 3-pins?

Motherboards with 3-pin fan headers that support "fan control" do their thing by directly modulating the fan voltage. The 4-pin connectors will connect properly to and work with 3-pin headers without any difficulty.

Not that it matters since you bought a 3-pin HSF already, but... for the future.

Seems like a poster or two mentioned that the 4-pin fans run full speed and loud if you don't have a 4-pin connector. Did I misinterpret something? Seems to me that I'd want a 4-pin fan which runs acceptably quiet at full voltage.

I've never heard my fan change pitch. Does that mean my motherboard does not support fan voltage control? I don't recall seeing such an option in the BIOS nor in the manual. I'm using an ASRock 939Dual-SATA2.

You're right - that board doesn't have any form of fan control; just the tacho reading in the bios, so any 3 or 4 pin fan will run max rpm all the time. Does the Zalman you bought come with the Fanmate 2 controller which you could use to adjust any 3 pin fan to an acceptable level.

Yes, it comes with the manual controller, which is fairly useless for this setup. I suppose I could adjust it to be the quietest in which the Opteron 185 does not overheat and crash. Personally, I'd have liked to see one of these fancy heatsink/fan combo units with a thermistor on the heatsink, closest to the CPU, and have some built-in controller to adjust the fan speed. I have one fan like that from a Supermicro server case, but the darn thing speads up and slows down all the time that it drives me nuts. Its transitions are not graceful at all. I'm sure by now, there is an improved design.

I am starting to wonder if there is more to this issue than just the fan. Like, is the CPU going bad, perhaps from being run too hot for too long? Or could aging capacitors near the CPU socket cause the CPU to overheat? The board currently has 4x Nichicon 1200 uF/16V and 6x Rubycon MBZ 3300 uF/6.3V capacitors near the CPU. They are not bulging, but that doesn't mean their time isn't up.

I just tried to install some Windows XP updates. I've attached some images of the system at idle, just before starting updates. You can see it sits at 33 C idle. After 10 or so minutes of installing Windows updates, it has crept up to 74 C. And this is with the side of the case off and a larg table fan blowing into the chassis. The system eventually crashes when installing the .NET 4.0 framework updates. Everest showed 86 C just before the screen froze.

Since I've owned this MB, I've never had long-term stable performance out of DDR400, so I've had to run it at DDR333. But its gone 2 years like this without issue until recently. Ideas?

feipoa wrote:

Seems like a poster or two mentioned that the 4-pin fans run full speed and loud if you don't have a 4-pin connector. Did I misinterpret something? Seems to me that I'd want a 4-pin fan which runs acceptably quiet at full voltage.

The same tricks that some boards use to control speed of a 3-pin fan will work on a 4-pin. But if your board doesn't have such a feature then it doesn't apply.

Specifically regarding the AMD 4-pin fans - they have a self conflicting design that supports a dedicated PWM signal but *also* has a thermistor built into the fan. The always changing resistance of the thermistor shifts the RPM range of the fan. As a result, they cannot attain higher speeds until after the thermistor has heated up, which takes several minutes or longer. The CPU bakes while you wait for the thermistor to heat up.
I remember having heat problems with video encoding on a quad core because of this. It seemed really stupid to me, and it still does. CPU was very hot and the fan speed just crept up from 4000rpm.... 4050rpm.... 4100rpm.... <10 minutes later> 5000rpm...
However, if you use one of those fans on a motherboard that applies full power all the time, then the thermistor would help to keep them quiet. It's just a weirdly compromised design IMO. It will hold down the fan speed when cool, but it's laggy to respond to temperature changes because it takes too long to notice. It's like driving a train between stoplights.

On the 2 such fans I've had, I shorted out the path across the thermistor so that they'd become instantly responsive to PWM from the motherboard. With the parking brake off, they accelerate quickly and can spin faster at max than I ever saw them spin with the thermistor. It's possible that AMD didn't want to allow such high speeds, but neither of the fans had any long term problem with it (one is still active in my server). They're loud at max, but if they're running at max, it's for a reason (on a PWM board anyway).

I am starting to wonder if there is more to this issue than just the fan. Like, is the CPU going bad, perhaps from being run too hot for too long? Or could aging capacitors near the CPU socket cause the CPU to overheat? The board currently has 4x Nichicon 1200 uF/16V and 6x Rubycon MBZ 3300 uF/6.3V capacitors near the CPU. They are not bulging, but that doesn't mean their time isn't up.

It's possible, but I think it's unlikely.

If the CPU is developing a short circuit, it would heat up. Could it be bad enough to heat up this much and still work? I don't know.
Capacitors usually fail open, but there are occasions when they short instead (which is a lot worse).
If a capacitor is starting to short, it would cause the capacitor and the attached MOSFETs to heat up (and eventually be killed by the electrical strain). Since this circuit is powering the CPU, it would conduct heat into the CPU also. I don't know if it's plausible for this to happen on a still running board and produce a measurable CPU temperature increase like you're seeing though.

Is the 4/8-pin VRM +12V connector (or wires) getting hot? Do you have any other compatible CPUs that you could use to see if they run hot also, or if it's just this one? Do you have an infrared pyrometer or other means of independently checking temps? If so, see if any of the components in the CPU VRM area stand out as being extra hot.
If you have a multimeter, might as well check the voltage on the +12V VRM connector also, but I don't think that would explain this.

I'm using a Ryobi branded infrared temperature gun. I checked the regulators with it, they aren't hot (38 C).

I agree that the most likely cause is the CPU's cooling system and is why I pursued obtaining a better cooler first. I have those capacitors listed for my next digikey order, which may be months away. I wish I could find my peltier pad. MIght have to add that to digikey as well.

I do have another compatible CPU. It is an FX-60, which is essentially an unlocked version of the Opteron 185. Same 110 TDP though. I'll try the FX-60 after I get the new heatsink fan, that is, if the Opteron 185 is still baking. For now, I'm just going to watch the temperature guage and take it easy. I am tempted to turn off one of the core's in the BIOS, but I remember in the past, on a Dual PIII, when I removed one of the CPUs (put in terminator chip, of course), Windows XP converted the ACPI uni-processor PC and wouldn't go back to multi-processor PC when I put the other CPU back in. I had to reinstall XP.

Oi,dude,there was one good suggestion b4,that u ignored: scalp it
The old Optys and FX are not soldered and use TIM under IHS,the best solutionis scalp and replace TIM with good nowaday paste or mod cpu for direct die cooling.

Cheers!

Sorry, I was overwhelmed with responses all at once it seemed. I assume you are referring to this post,

The Serpent Rider wrote:

Step 1: Remove CPU IHS with a thin blade/razor. Be careful not to damage small elements near the edges. Patience is the key. Ste […]
Show full quote

I think it is a fair assumption to say that the CPU is over heating.

Step 1:
Remove CPU IHS with a thin blade/razor. Be careful not to damage small elements near the edges. Patience is the key.
Step 2:
Remove 10+ years old thermal paste from the core.
Step 3:
Apply some fresh and good thermal paste. I recommend good ol' Arctic Cooling MX2/MX4, but there are also better options.
Step 4a:
Put IHS back with adhesive.
Step 4b:
Modify your cooler for direct contact with the crystal. Some may fit even without any modifications.

With the mass of posts, I saw this and though, "hmmm, directions for reapplying thermal paste to the [top of the] IHS. OK, next comment". I didn't realise until you just poked me the directions were to remove the integrated heat spreader from the PGA body itself. Has this method been shown to reduce CPU heat significantly? I assume that the manufacturer would have the best method for getting the thinnest possible spread of past and using paste which tolerates heat very well with time. I'm concerned that by reapplying it myself, I won't get such a thin application of paste and that this lack of precision may not improve upon what is already there, albiet probably dry.

This begs the next question, what is the loss of thermal conductivity with hardened paste vs. non-hardened paste?

One other question is how are you guys applying the thermal paste? Are you dropping a small gob of paste, then using the heatsink to smash it out uniformly, or are you trying to spread the paste by hand, e.g. with a credit card, or by rubbing it around with your finger until it seems uniform? I am applying a small gob, perhaps half the size of my pinky finger nail, then letting the CPU pressure smash it down. Seems to me that if you try to make the paste uniform by hand that you may end up with more air gaps.

I generally go with the gob+pressure method, but then judging the right size for the "gob" is difficult. I imagine especially more difficult if you're doing it under the IHS (which I've never done, but I imagine it being similar to early Athlons/P3s in FCPGA.

feipoa wrote:

shiva2004 wrote:

feipoa wrote:

Then there is something like this, which is original AMD for Athlon X2. Specifies it is for up to 125 W TDP.
https://www.ebay.com/itm/AMD-Heatsink-CPU-Coo … joAAMXQlgtS1PEN

I had this cooler with an Athlon X2 6400 rated at 130W and it wasn't up to the task, when I taxed the computer for extended periods of time it overheats and even shuts down, and it also makes a lot of noise; perhaps it can cope with 110W loads, but I wouldn't use it in anything rated above 95W. Also, it isn't worth 47$, not even half of that.

Right now I use an Opteron 180, also rated at 110w, with an Arctic Cooling Freezer 64 Pro and it cools the processor well enough, in a cheap case with moderate airflow I've never seen it to go above 56-57° celsius and most of the time it' s barely audible.

Well, that's two points for the Freezer 64 Pro. I guess I'll look into this one.

In fact what I have is the older version, the Freezer 64, the design is a bit less refined (and weirder looking 🤣) but it uses a three pin fan connector.

Another thing, in many motherboards that don't support fan speed management you can use Speedfan with the automatic mode activated and a fan curve defined by the user to achieve it.

feipoa wrote:

With the mass of posts, I saw this and though, "hmmm, directions for reapplying thermal paste to the [top of the] IHS. OK, next comment". I didn't realise until you just poked me the directions were to remove the integrated heat spreader from the PGA body itself. Has this method been shown to reduce CPU heat significantly? I assume that the manufacturer would have the best method for getting the thinnest possible spread of past and using paste which tolerates heat very well with time. I'm concerned that by reapplying it myself, I won't get such a thin application of paste and that this lack of precision may not improve upon what is already there, albiet probably dry.

De-lidding is something that has come up in recent years a lot for overclocking intel core iWhatever CPUs since the thermal compound Intel uses has derogatorily (and rightfully so I might add) been called thermal toothpaste because of how terrible it is. Replacing that thermal compound with better stuff or even liquid metal (I recommend against that, unless you know exactly what you are doing) has been shown to reduce CPU temps significantly.

In this case however it is not about improving stock performance it is smply maintenance to replace thermal paste that has gone bad.

feipoa wrote:

This begs the next question, what is the loss of thermal conductivity with hardened paste vs. non-hardened paste?

Anecdotal evidence only. I am not aware of actual "research" into this. However, replacing old thermal compound is recommended when cleaning coolers on CPUs and video cards anyway, so I'm not surprised to see the issue crop up with the compound between the die and a heat spreader.

feipoa wrote:

One other question is how are you guys applying the thermal paste? Are you dropping a small gob of paste, then using the heatsink to smash it out uniformly, or are you trying to spread the paste by hand, e.g. with a credit card, or by rubbing it around with your finger until it seems uniform? I am applying a small gob, perhaps half the size of my pinky finger nail, then letting the CPU pressure smash it down. Seems to me that if you try to make the paste uniform by hand that you may end up with more air gaps.

Short answer, it doesn't matter, like at all.

Over the last few months a lot of tech youtubers did test various ways of applying thermal compound and the results were clear, that unless you use way too little it does not matter at all, how you apply it. Even applying "too much" (within reason) has no negative effect. The pressure of the cooler against the IHS will spread it out as needed and squeeze out excess compound. The only thing that changed the thermal results was the quality of the compound itself.

The only instance where I would still carefully spread it out to cover the whole surface is with direct die contact, just to be sure to have full coverage.

SpeedFan 'eh? I've downloaded it from here, http://www.almico.com/sfdownload.php and will give it a go.

I remember the silver coloured viscous "premium" heatsink compound I had. It wasn't liquidy enough and it wouldn't compress very well. When remove the heatsink, you could see that it didn't flatten uniformly. Because of this, I stopped using the "spread it out first" method. I find the ceramique 2 sufficiently creamy.

When the new heatsink arrives, I'll try to remove the heat spreader on the Opteron 185. This CPU has become somewhat hard to find and I don't want to ruin it. Aside from using razer blades, are there any other tricks of the trade to get the IHS off? Perhaps grabbing the IHS with a rubber-padded vice, then prying with the razers?

Koltoroc wrote:

Replacing that thermal compound with better stuff or even liquid metal (I recommend against that, unless you know exactly what you are doing) has been shown to reduce CPU temps significantly.

Indium works well, a bit on the expensive side and you do need to know what you're doing. 😀
(like gallium though it does not play well with aluminum and its alloys; unlike gallium and other alloys it's actually not liquid at any temperature that a CPU would reach and still be able to function but it's very soft and "flows" to fill the smallest crevices even before melting)

I wouldn't recommend liquid metal alloys unless you will be putting it on a nickel-plated surface. It does react with copper, just not as aggressively as aluminum.

If you're not a crazy overclocker or have a special need to for the absolute highest efficiency, it's a waste of time and money anyway.

What do you guys think are suitable PCIe cards for this system? I don't own any PCIe cards except for a crummy x300 in my garage system. What's a relatively advanced PCIe x16 card with XP drivers that would be suitable for an Opteron 185 / FX-60? Prefer not to have something with more than 512 MB of RAM. The AGP HD4350 is 512 MB and is eats into the system RAM. If I set the AGP aperature down to 32 MB, I can get 2.92 GB of system RAM. Purpose is mostly benchmarking, troubleshooting, curiosity....

If you want your system to be able to run Win98, go with the X850PE. Otherwise, you can go with fairly modern DX10/11 cards (although that's not so consistent with the age of the CPU). To be period correct, a Radeon X1950 XTX or Geforce 7950 GTX would be best for a high end build.

There are lots of used PCIe Radeon HD 5450 and HD 6450 cards on Ebay, which have official driver support for WinXP and usually come with passive coolers. These are the last "low-end" discrete cards produced by AMD before they decided to give that segment to its APUs.