|
Heatsink reviews are new on me, this is my first, so forgive me if my testing methods are off base and please tell me if they are. Here are the configurations I'll be testing.
• Swiftech MCX478 with 80mm Panaflo L1A (24cfm, 1900rpm, 1.2W max, 21dBA, 80x80x25mm)
• Swiftech MCX4000 with 80mm Panaflo L1A (24cfm, 1900rpm, 1.2W max, 21dBA, 80x80x25mm)
• Swiftech MCX4000 with 74mm YS-Tech TMD (36.8cfm, 5800rpm, 3.54W max, 39dBA, 74x74x15mm)
• Intel Boxed P4 Heatsink with 70mm fan (~25cfm, 2800rpm, ~28dBA)
Tested CPU configurations will be at stock CPU speed (2.26GHz, 133MHz front side bus) both load and idle and at 2.73GHz (160MHz front side bus) both load and idle (where applicable). Arctic Silver Alumina was used as the TIM in all cases.
Temperatures was measured using Motherboard Monitor 5.1.9.1 monitoring the Winbond 2 P2 Diode sensor provided by the Winbond W83627HF sensor hardware on the test motherboard, the ABIT IT7-MAX. The motherboard was home to a 2.26B Intel Pentium 4 Processor.
Ambient air temperature in the room throughout the testing was a cool 22c as measured by my pocket digital thermometer.
Idle temperatures were measured by making sure no CPU intesive tasks were running in Windows XP (using Task Manager) and measuring temperatures after a 15 minute period of no CPU activity.
Load temperatures were measured by loading the processor with Prime95 and running the torture test for 15 minutes, enough to make the temperatures top out at their maximum level.
Onto the first graph, performance at stock CPU speed.
As you can see all 3 Swiftech based solutions, including the MCX4000 with both fan combos, were more than up to the task of cooling our stock clocked 2.26B. Idle temps and load temps were identical across the board meaning the MCX4000 had cooling headroom even with the least powerful fan on test, the Panaflo L1A and consequently were more than up to the task of getting the CPU as low as it could go at idle. The heatsinks are on show here, not the fans and it's nice to see the temperature parity across the board to highlight the heatsink performance.
It's worth mentioning that the stock Intel heatsinks is more than capable of cooling the stock processor and it's very quiet with a supposedly temperature adjustable fan (although I didn't see it waver from ~2800rpm during testing).
Things should change when we push the processor a bit. Here's the graph of performance at 2.73GHz with increased Vcore, giving the heatsinks some work to do.
The first thing to note is that the stock Intel heatsink didn't seem to like cooling the 2.73GHz processor at 1.55V. Intel are using the same heatsink with the 2.8GHz Pentium 4 so in theory it should have cooled this processor fine, however temperatures would approach 65c load and it would fail Prime95 torture test. It never hung or appeared to throttle, but I've marked it as 0 on the graph since it wouldn't let the CPU do 15 minutes of load with Prime95. Maybe a voltage bump would have let the processor pass the test but temperature would have risen accordingly.
What is interesting to note is that the Swiftech combinations were happy with this speed and ran the tests without any problems. The voltage is low enough that the fan used didn't seem to make any difference with the MCX478 and Panaflo coming out the winner by the smallest of margins. The MCX4000 has a larger copper base, more surface area and a more powerful fan when paired with the 74mm TMD YS-Tech but that doesn't make any difference here, both sinks have masses of cooling headroom left.
You can go cool and quiet by pairing either Swiftech up with a fan like the Panaflo which is nearly silent.
|
