Recently we decided that a 3800+ AM2 chip is definitely not enough to push the latest heatsinks and instead we invested in a brand spanking new testing setup.
|Processor||AMD AM2 6000+ Dual Core|
|Motherboard||Biostar TA690G AM2|
|RAM||Corsair XMS2 6400 2GB (2x1GB)|
|HDD||Maxtor DiamondMax 20 80GB SATA|
|Power supply||Jeantech Storm 700w|
|Graphics card||Onboard – ATI Xpress 1250 series|
As this is a new setup, we’ll also be re-running the tests on the stock AMD cooler supplied with the processor and also the ThermalTake Max Orb which scored highly on our previous test rig.
All of the testing is carried out outside of a computer case.
To test we simply boot the PC up with a freshly installed copy of Windows XP, and measure temperatures using Speedfan. The onboard temperature sensor is disregarded and instead the CPU’s own diode is used.
For idle testing, we simply let the testing rig sit doing absolutely nothing for 30 minutes and take the most representative temperature of the last 10 minutes. The same is used for the load testing, but instead of letting the PC do nothing, Orthos Stress Prime is used to load both cores to 100%.
As this motherboard also has a temperature sensor in the chipset, we will also measure the temperature of that in the same way as the CPU in order to get an idea of how effective the ‘collateral’ cooling is of the heatsink.
Due to the lack of consumer availability of 140mm fans, we figured that 120mm fans would do the job; albeit at a heighten noise level.
As the IFX-14 doesn’t come with its own fan, we’ll be used two of our own which should be indicative of the home user and of the maximum potential of the heatsink. First is the Zalward 120mm Golf Fan which pumps 48 CFM while making 17 dBA, while the high performance fan is a pillaged server fan called Nidec Beta V TA450DC which alone uses 10W of power while pumping 112 CFM and making a racket of 48 dBA.
In the testing, the Max Orb was on maximum power.
As you can see from the results above, the IFX-14 at idle and load CPU with the low CFM fan is as good as the heatpipe equipped stock cooler while making a whole lot less noise. However, the way that the IFX blows the air means that the chipset is left out of the equation and reaches a scorching 60C at load. In this case, the stock HSF – while louder – is the cooler choice.
When the much more powerful fan is introduced, the IFX-14 only manages to keep on par with the Max Orb in regards to temperature while the noise was unacceptable. While the figures show that the Max Orb should be much quieter, it is only slightly better. That said our previous testing has shown that the Max Orb at an acceptable volume level is only a couple of degrees different. Looking at the chipset temperatures, you can see that the flower style method employed by the Max Orb allows for a lot of motherboard cooling, keeping the chipset in check at only 25C at full load.
The additional backplate cooler that the IFX-14 uses barely got hot at all, even though there was little airflow passing over it.
It would appear that the IFX-14 requires a lot more airflow to cool the large surface area compared to the Max Orb. However, this is out-of-case testing. If this setup were contained in a case then you’ll find that the general air flow of the other fans would cool the chipset and also pass by the backplate cooler providing it with additional degree killing power.
There comes your next problem as a) this cooler has the extra backplate heatsink which may or may not fit in your system and b) a ridiculous height that extends by another centimetre above the Noctua NH-U12 which failed to fit in the Lian Li PC-C32.
For those wondering about passive temperatures, the heatsink at idle keeps the CPU at 37C, while at full load I stopped after 5 minutes as the heatpipes were too hot to touch and the core had reached 66C. This means that you’d easily be able to watch a DVD or similar with this heatsink running purely passive provided that you had a reasonable amount of case ventilation (a couple of low RPM 120mm fans).
I was personally shocked at these rather lack lustre performance margins, and I expected a whole lot more from the IFX-14 considering its excessive size and additional backplate cooler. I re-seated the cooler three times just to be sure, interestingly, that’s where we find our solution to the mystery.
Every time that I removed the heatsink, the same pattern of the heatpaste was visible. The centre was quite clearly thinner than the rest which appeared to have a much greater quantity. The processor had been coated with the same amount all over the core, meaning that you should be seeing a nice square shaped heatpaste impression. Instead we only see a circle where the heatpaste is pushed out by the domed middle and the edges are too far up to even make contact with the heatpaste. I grabbed a craft knife, and I was shocked at what I saw.
It’s nearly a difference of a millimetre of the core to the outside parts which is completely unacceptable, especially considering the amount of aluminium that is tacked on top which all relies on this piece of metal. This part of the heatsink is with a doubt the reason why this heatsink failed to amaze. Interestingly, it would appear that Thermalright haven’t learnt their lesson from the Ultra eXtreme which also exhibited the same symptoms. Some have said that it’s due to the heat pipes being soldered to the core after the core has been machined.
With this amount of base warping, the outer heatpipes don’t have the same amount of access to the CPU’s heat as the middle pair. This means that the base has to conduct the heat further before the heatpipes shunt the therms up to the fins above.
Fortunately, you can sort this problem out yourself… providing you have a piece of glass and varying grades of sandpaper. For the asking price of half a ton it seems VERY expensive considering as you have to finish the job that Thermalright failed to do themselves. The size issues would pale into insignificance if the performance warranted it.
UPDATE: In order to give an impression of how well this heatsink can perform, we did a review of the surface after being lapped. You can read it here.