First post, by RacoonRider
- Rank
- Oldbie
Everybody lies.
- House M.D.
I've been reading various articles and forum threads, both in Russian and in English, on the topic of optimizing budget CPU coolers for Socket 462. A lot of myths wander around the net. People who claim to have tried them usually state 3-5 degrees Celcium improvement in CPU temperature during heavy load.
Of course, having a rig to mess with, I was tempted to try them out.
My setup is as follows (I mentionned it in several other threads already)
Ahlon XP 3200+
GA-7n400S (nforce 2 Ultra chip set)
Radeon 9800 Pro 128MB
1.5GB DDR2-400
Creative Audigy ES
ASUS PCI-N10 wireless PCI adapter
Seagate Barracuda 7200 250GB SATA HDD
Pioneer DVR-111 DVD-RW
Zalman 450GS PSU
Coolin solutions (Left to right, top down):
Arctic cooling F8 on a CPU
120mm cheap deepcool exhaust fan
LYF Slot exhaust turbine blower
Chinese videocard cooling system
Arctic cooling F8 intake fan
Myth number one: Copper core coolers are always better
I have a very fancy-looking copper core heatsink from Titan (no idea what model, they are all really similar). It's shiny and looks well polished.
I also have an Igloo 2640 heatsink. It's all aluminium, oxidized and very crude. Of course, I picked the Titan at first.
I was not concerned with it much when I had 2600+, but 3200+ is a hotter thing, so I decided to complete a stability test with AIDA 64. With all coolers at MAX (except for the slot blower, I did not have one back then), it showed me 77C max/ 74.8C avg after half an hour of heavy load. With maximum operating temperature of 85C it's no good idea to keep the CPU that hot. When (just in case) I tried Igloo with the exact same fan, the temps decreased to 72C max/70,3 avg. Not bad!
If we calculate the surface of heat exchange, we'll know exaclty why Titan is so much less effective:
Parameter____________Igloo_________Titan
Number of fins_________29x2_________21x2
Single fin surface_____945mm^2_____868mm^2
Total fin surface____54 810mm^2___36 456mm^2
So, being almost identical visually, Titan has only 66% of Igloo's heat exchange surface and the copper core can not compensate that loss of "raw power". However, Titan looks much better and Copper core certainly helped it gain its place in the market.
Myth number two: a heatsink should have cuts in its base parallel to it's fins
One website suggested I should cut the base of the heatsink promising a generous 5C decrease in CPU temperature under load due to effective removal of hot air from the gap between the CPU board and the surface of the heatsink. I just had to try that! So I went to the garage and made 4 cuts in the heatsink base and 4 additional holes for better airflow. I knew that it was unlikely that the hot air was hiding under the heatsink, but I thought that measure would help me direct the airflow closer to the middle of the heatsink, in other words, act like a vent. The 5C decrease in CPU temperature under load seemed unlikely as well, yet tempting.
Guess what? The effect was hardly noticeable, less then 0.8C average under load, yet positive. Why would one state their vents gave them 5C improvement, when mine sucked so much?
Myth number three: bigger is better
There was a page suggesting I should find a 90mm cooler with 80mm mounting holes and fasten it on the stock AMD cooler to get much better airflow. I grabbed this fan from a friend for a few hours, not sure about the model number.
Guess what? The temperature of CPU under constant load raised to 82C and kept going up. I shut the computer down to prevent any possible damage. The fan was all right, it's still cooling my friend's overclocked C2D. What was the reason then? Well, for starters, the rotary movement made by the fan creates centrifugal force driving the air from the center to the periphery. In other words, the fan blows much more near the edge than close to center. Besides, the central area is obstructed by the engine, so air barely moves there. And, as you might imagine, a larger fan has huge gaps between the fan shroud and the heatsink, where all the air goes.
Myth number four: higher is better
So how do you compensate both the fact that the engine obstructs the center of the heatsink and the centrifugal airflow without buying a new cooler? There is a website that suggests putting an empty cooler shroud between the cooler and the heatsink.
I made several shrouds for the job, including 2 90mm ones. However, I don't have a spare 90mm cooler at hand right now, so maybe next time.
A cooler featuring one 80mm shroud between the heatsink and the 80mm fan gives mixed results and will be tested further. During AIDA64 stability test the showed slightly higher temperature than the cooler without a shroud. However, when I left the program gathering data and played Portal for 2 hours (at 100% CPU load with occasional drops), the statistic tab showed 71max/68.5 average against 72max/70.3 average received during stability test without shroud.
Everybody lies
The ideas I tested here turned out to be either next to useless or false.
Marketologists are known to lie to increase sales. Bulky Igloo 2640 is (probably twice) cheaper to produce, yet a fancy Titan is more likely to gain attention of even an experienced buyer. It does not matter if it's better or not, just wrap it up the right way and sell it. Overclockers and experienced users often lie as well. If they mod something, they start with a certain assumption - we all have them. And most assumptions are false. But after grinding a heatsink for a good hour in the garage it's really hard to admit your defeat. People still want to show off on the net and they wrap it up as if they got much better results. Sometimes they even do it unintentionally, helping the results confirm their hypothesis, that way they lie to themselves as well. We all know better than to trust Internet much, and we should not forget to question ourselves before we do something stupid.