:mad: Be very wary of Thermalright products, especially the Ultra 120 eXtreme - quality control on their retail product is terrible. Here is my experience with the Ultra 120 eXtreme. All load temperatures used in calculations below were produced using TAT to generate the load, and thermal conductivity / resistance was calculated using the difference between the CPU case temperature (as read and reported by Speedfan) and the fan inlet temperature (as read using a digital meat thermometer but in case that wasnít precise enough I took other steps as highlighted below to validate my testing methodology).
1. My first heatsink performed out of the box at approximately 0.23C/W Ė a very far cry from the 0.10C/W that most review sites showed. This was reached w/ an Antec Tri-cool 120mm fan on high. That fan is rated at an ample 79CFM, and Antec is usually not bad w/ their ratings (even if that rating is a stretch, itís not different enough from the fan you used to result in that big of a gap in overall thermal performance). This was yielding a case temperature of 51C for an E4300 clocked to 9 x 300 MHz w/ stock voltage (approximately 100W TDP) given fan inlet temperatures of 28C (and an ambient temperature of around 24C).
2. Having read on a few forums that the IHS on the Core 2 Duoís was bad, and that the Thermalright bases havenít always been good, I hand lapped the CPU IHS and the Ultra 120 eXtreme base. I used an 8Ē x 11Ē sheet of glass with wet-or-dry sandpaper and worked up from 300 to 2000 grit until I had a mirror finish on it.
3. Lapping improved the performance of the heatsink to 0.19C/W Ė still nowhere near the performance Iíd seen it have at a few review sites out of the box.
4. Unsure if maybe Iíd been careless w/ mounting or lapping, I tried mounting it with various amounts of thermal compound from a thin line as recommended by Arctic Silver, to a BB in the middle, to spreading a thin layer out, to using none at all Ė the lapping job was good enough I could run w/ no thermal compound and only was 3C hotter. When that didnít work I tried different brands of compound, re-lapped it, and even tried increasing the mounting pressure (added a flat washer to each of the mounting screws so that the springs would compress more). Nothing improved its performance.
5. I added a second fan in a push-pull configuration, which made little difference.
6. I began to wonder if maybe something was wrong w/ my processor or test methodology, so I took the Ultra 120 eXtreme off, and mounted the original stock Core 2 Duo heatsink. After testing, that performed at 0.33C/W, which is exactly what Intel specs it out for. At this point I realized that I wasnít do anything wrong, my test methodology was good, and there had to be something wrong w/ the heatsink. Had anything else but the heat sink been at fault, then the Intel stock heat sink would have suffered similar issues.
7. I argued with the manufacturer over a week about replacing it, which they refused to do because I lapped it, insisting that the lapping had damaged it and voided the warranty. I offered them all of my test results (the actual data Ė not just the summary I present here) and macro photographs of the base to prove I had done nothing wrong. They still refused to replace it. DO NOT LAP YOUR THERMALRIGHT HEAT SINK UNTIL YOU ARE SURE IT IS GOOD.
8. Finally I asked NewEgg to replace it, and they did.
9. I received my new heatsink yesterday, took it out, and inspected the base. The base is not only warped, but poorly machined. While some heatsinks are indeed convex by design, they should then be convex across the middle of the heatsink to increase the pressure across the die area of the processor, but my base clearly bows outward only on one side. Worse, there are concentric circular tool marks on the base from machining that have resulted in visible ridges. If I run my fingernail quickly across the base it makes a distinctive Ďzipí noise. However, this time I was smart enough not to lap the heatsink.
10. Not to be discouraged, I tested the new heatsink. This one performed at 0.23C/W to 0.25C/W.
11. I emailed support and told them I am content to keep RMAíng these heatsinks until I get one that performs at or close to 0.1C/W, which is in line with what they seem to send review sites (or Iíll also accept an acknowledgement that people should only expect 0.23C/W from the Ultra 120 eXtreme, at which point Iíll promptly post their message at every site I can find).
12. Thermalrightís Taiwan support responded to me with the following message (and I quote them):
The first thing is that review is not tested by Thermalright.
There are many variables in each test.
I can not promise each heatsnk has the same performance (better or worse) with review one.
The base of extreme is just a little convex.
Therefore, it doesnít result in bending.
E4300, 9 x 300MHz, 1.325v (roughly 100W TDP)
Fan Inlet Temp: 28C
CPU Case Temp @ 100% Load (using TAT): 51C (approx 0.23C/W)
It is normal. Donít worry about this performance.
13. Bear in mind that 0.23C/W is about the same as a STOCK RETAIL Intel Pentium D heat sink. So Thermalright is telling me that it is normal for my Ultra 120 eXtreme to perform like a stock retail Intel Pentium D heat sink. Why did I buy it then? I have an extra retail Pentium D heat sink from my previous desktop that I could use.
14. I am continuing to fight with them over this, but unless you want to risk going through the same BS I had to, Iíd suggest looking at other competitive coolers such as the Tuniq Tower 120 (the only complaint Iíve heard is w/ the flimsy mounting screws), Scythe Ninja Plus B, or 3RSystems Iceage 120. If you do need better performance than those other coolers then Thermalright or one of the TEC solutions is the only way to go, but be very careful Ė Iím not the only poster to have issues with Thermalright, or to complain about their quality control. You could also always consider water cooling, but be aware youíre going to need a good rig to get into that 0.10C/W or lower neighborhood (dual 120mm fan radiator, good block, and relatively high pressure pump), and thatís going to run around $300.
Iíve done the following things to idiot-check my test results and verify that Iím not at fault:
1. Tested using the stock heatsink. Since the stock heatsink performs as it is designed to (0.33C/W) using the exact same test methodology, that proves the testing process I used is solid.
2. Re-did all my calculations using the idle temperature in place of the fan inlet air temperature. One could assume my temperature probe is inaccurate. Using the idle temperatures as read by the motherboard (or the temperature diode on the chip w/ TAT) in place of the air temperature results in a very optimistic result (since it assumes that at idle the processor draws 0W, which is not true, it biases the result in favor of the heatsink). Even performing calculations this way did not result in the heatsink dipping under 0.20C/W (w/o lapping).
3. Performed the above calculations for each sensor available and used multiple software programs (TAT, CoreTemp, Speedfan, and the software that came w/ my Gigabyte mainboard).
4. Tested with different thermal profiles (my E4300 at stock voltage and tried three different speeds Ė 9 x 200MHz, 9 x 300MHz, and 9 x 333MHz which produce roughly 65W, 100W, and 110W TDP respectively).
5. Tested with the CPU idle and set to extremely low speed and voltage (6 x 100MHz @ 0.8v) to validate my temperature readings (and make sure I did not have a stepping with thermal diode temperatures mis-reported in TAT and CoreTemp based on whether or Tjmax was 100W or 85W). At that speed and voltage the CPU case temperature should be nearly the same as the ambient air temperature. Testing revealed that TAT and CoreTemp were reporting temperatures correctly.
If you can think of anything Iíve done wrong, Iím more than open to listening to it, but I would be very very careful about purchasing Thermalright products.