Earlier this year, I built an inexpensive Q9550 system and I installed the factory CPU heat sink. Last weekend, I replaced the factory heat sink with a inexpensive (approx $15) unit purchased from mwave. The system was not over-clocked.

I probably just gave away my ‘novice’ system building status. But, I was amazed to see the average CPU core temperature (reported by realtemp) drop by ~20 deg C (from ~80 to ~60 deg C). Amazing! I’ll definitely use an upgraded cooler on any future systems.

Gary

Hi. Not wanting to burst your bubble, but you may have just installed the new heatsink better than you installed the factory version. Most people who run 'stock' seem to find that the factory heatsink is sufficient. Having installed quite a few now, I find that as you fit more, you do get substantially better at it. However, if it's working better for you, great.

I have installed a Xigmatek S-1283 on my AMD955, and got a big reduction in temps, approx 12-15c. Mine is clocked up from 3.2Ghz to 3.6Ghz tho.

Does this mean you're gonna enter the shady world of 'overclocking' now? :-)

regards, Gizbar.

---

A proud GPU User Server Donor!

Yeah, it's probably due to installing the new heatsink properly with the better silver thermal paste?

The chewing gum stuff that comes on the standard heatsinks is cheap and idiot proof, but it's not the best thermal joint out there.

My quad core is still running the standard cooler, but with a nice thin layer of silver grease. It also runs around 60C. When the temp gets up over 70 it's time to clean the dust from the heatsink again.

Ian

The boxed heatsink of my Intel Core2 Extreme QX6700 [Quad-Core] was very very '$%$&%/&/&&%$%&$'..

The CPU (130 W TDP) temp was very high and the noise of the fan was terrible.

After installation of the Noctua NH-U12P heatsink with Noctua fan @ ~ 1,300 RPM, I'm surprised how cool this CPU can be. And quiet..

Ambient ~ 21 °C, full load with 4x MB WUs: 38-42-38-42 °C [CoreTemp, Delta to Tjunction max]

Because of OC from 2.67 to 3.14 GHz the CPU temp increased ~ 3 °C [~ 150 W TDP].

---

hmmm, you may both be right. The new installation used silver paste and the original installation used the 'white gum'. I probably won't reverse the procedure to re-test silver vs. white, but if you think the new paste was the silver bullet (ha!), that's good to know too.

First I made a test.. took away the stock Intel thermal paste and used Arctic Silver 5.. nothing better.

Now with Noctua heatsink I use IIRC Arctic Cooling MX-2..

---

hmmm, you may both be right. The new installation used silver paste and the original installation used the 'white gum'. I probably won't reverse the procedure to re-test silver vs. white, but if you think the new paste was the silver bullet (ha!), that's good to know too.

Keep in mind that the gum stuff is really hard to get off the CPU. But it's very important to use a thermal paste cleaning solution to get all that wax stuff off before using a "real" silver thermal compound. If you don't, the cooling performance may suffer. Sounds like you did that if the temps dropped significantly.

Good job on doing it yourself!

Actually, the thermal paste (type, brand) makes almost NO difference as long as the HSF is properly installed. Google "thermal goop" and look at the link for dansdata.com (it's usually first, or near the top).

The main thing thermal paste does is make sure there are no air pockets between CPU and heatsink - that's what causes the heat problems - air is a lousy heat conductor. The actual composition of the paste is a small factor in this.

I use plain old Vaseline for my CPUs, and they all run comfortably cool (24/7 for SETI). It fills the air pockets perfectly, doesn't dry out, doesn't run out, and is trivial to clean up. And it is cheap.

---

Actually, the thermal paste (type, brand) makes almost NO difference as long as the HSF is properly installed. Google "thermal goop" and look at the link for dansdata.com (it's usually first, or near the top).

The main thing thermal paste does is make sure there are no air pockets between CPU and heatsink - that's what causes the heat problems - air is a lousy heat conductor. The actual composition of the paste is a small factor in this.

I use plain old Vaseline for my CPUs, and they all run comfortably cool (24/7 for SETI). It fills the air pockets perfectly, doesn't dry out, doesn't run out, and is trivial to clean up. And it is cheap.

I can not believe plain old Vasoline would work.
I would think the heat would boil or melt, evaporate it.
That might be why your getting so many "Error while computing" errors?

I do not know for a fact that this would not work. I do find it kind of funny/strange that someone thought to try Vaseline for a thermal past.

The number of errors his Opteron with dual GPU has is low considering the number of WUs it is crunching and the fact that most errors are VLAR killers, which essentially is an error by design in the GPU app.
I can imagine Vasoline would work perfectly. Especially if it is (as it should be) a very thin layer the heat insulation will be very small regardless of the thermal resistance of Vasoline (haven't looked it up btw). Possibly there is even no layer at all for the most part of the CPU as Vasoline becomes pretty much a liquid at higher temperatures...

---

I can believe Vasaline would work, not sure about the long term stabiliry, but oil is commonly used for cooling in industrial situations. Even if it liquified in the junction that would just leave a super thin layer right where it was needed, and possibly no microscopic air gaps at all.

The original thermal compound was simple silicone grease, probably chosen because it is chemically very stable. It's only later that silver has been added to the mix, it does improve the grease slightly, but if it's applied correctly the difference is hardly measurable. It improves the long term stability because even if the grease does dry out over time, the silver is still left junction and does it's job.

Dan (Dansdata.com) did some actualy experiments with various greases, and found that the most important factor was how thin and even the layer is, not the actual type of compound. From memory he only had a 10% heat increase when he used Toothpaste (!!!) as a thermal compound. It's not a good choice as it dries out quickly, but it worked.

Ian

Gmnewell:
Just as a side note, how are the CPU temperatures now that you've been crunching for several days?

Others:
Hmm... seems like WinTea and Ianab made some good points. Understanding the physics of it better reveals that 1) removing air gaps and 2) thermal conduction are the most important properties of any thermal grease du jour.

For someone who may tinker a lot or upgrade CPU at a later date, having a poor electrically conducting compound and one that doesn't "dry out" easily would be great properties to also have.

The whole reason why I recommend a silver infused compound is because it's physically a better heat conductor. And you'd have to be assembling a lot of computers for the extra cost (vs stock or Vaseline) to be a factor.

I think this whole tangent of thermal grease got started by Ianab. So, I blame him for a thread that won't die. :)

... Others:
Hmm... seems like WinTea and Ianab made some good points. Understanding the physics of it better reveals that 1) removing air gaps and 2) thermal conduction are the most important properties of any thermal grease du jour.

For someone who may tinker a lot or upgrade CPU at a later date, having a poor electrically conducting compound and one that doesn't "dry out" easily would be great properties to also have.

The whole reason why I recommend a silver infused compound is because it's physically a better heat conductor. And you'd have to be assembling a lot of computers for the extra cost (vs stock or Vaseline) to be a factor.

If you have a good surface-surface contact, then the extreme thinness of whatever thermally conductive layer means that whatever the material goop itself is that you use will have little impact provided that it does give a thermal contact. That's why the poorly thermally conductive 'sticky pads' can work well enough. There's the slightly better 'stick pads' goopy stuff that deforms and 'melts' at high temperature that then allows the goop to be squeezed out of the way and into a minimally thin layer.

The ultimate is to NOT use any thermal grease, jam, goop, or whatever, and to instead ensure a good direct metal-metal contact by lapping both surfaces of contact to be as smooth and as flat as possible.

Note that the CPU markings ink is a thermal insulator and protrudes above the smooth surface of the CPU metal heat spreader...


Aside: RAM chips get hot when used... So why do RAM DIMM suppliers stick big thick insulating labels over the chip packages!...

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Aside: RAM chips get hot when used... So why do RAM DIMM suppliers stick big thick insulating labels over the chip packages!...

They do make some RAM chips with attached heatsinks... but they are a minority of the items for sale. My only explanation for those metal plates over them are 1) product decoration and 2) larger heat-conducting area for cooling. Not very useful as a heatsink, but it's probably better at cooling than not having the metal plates at all.

I think the day of heatsinks on RAM as standard is coming. Many video cards that I've seen (including my own) already have heatsinks on the RAM chips. Even my RAID card has one.