I’ve just rebuilt my comp into a new case (Globalwin 802, Enermax 431W PSU, 3 x 120mm (112 CFM each) Nidecs etc.) and thought it was time to install my Swiftech MC370-4 copper pinned Peltier rig. For about 6 months I have been using a PelTEC Alpha PEP-66 with a 55W TEC, but thought I would try for something a little cooler and more stable. The PelTEC rig is stable enough and was superb value for money, but some people on alt.comp.hardware.overclocking manage to keep their min/max temps down to a range of 10°C which is pretty impressive – and that’s my target.
Also, I wanted to find a universal solution – something that could cool a slotket or socket mobo, an AMD, PIII or even P4. I hoped I had found the solution with the Swiftech MC370-4. But we shall see! The Alpha Pal6035 is also a contender, so I picked one up and will test that also. I didn’t bother with any of the Globalwin coolers, good as they may be; I’m looking for the best – not a cheaper alternative – they are simply not in contention.
So the computer was a fairly easily overclocked PIII 600e @ 800EB at stock voltage, on a P3V4X, Abit slotket!!! (bit more about this later), powered by a 430Watt Enermax, which delivers 15A on the 12V rail (more than enough).
The case cooling was turned off, so the box was allowed to warm up to a consistent and steady 30C (room ambient was 21.2C throughout). I used two TEC’s, an 80W Tellurex which was supplied with the MC370 and a no-name 55W supplied with the Alpha from PelTEC. These represent the highest and lowest wattage TEC’s you should consider for a single fanned solution for PIII’s.
Which brings me nicely on to the fans: I tested each cooler with each TEC and with two different fans. The first – a fairly standard and reasonably quiet 26CFM Sunon and the secondly – a not so standard and not so quiet 33CFM Papst (supplied with the MC370). All the fans were orientated according to manufacturers recommendations (ie sucking for the Alpha’s and blowing for the Swiftech).
For thermal grease, I used some cheapo junk I had lying around and used it throughout. Equally, the coldplate was not a 1/4 inch highly polished monster, but a puny 1.5mm offcut. I can’t stress this enough – the object was NOT to achieve the coldest temps, but to replicate as near identical conditions as possible for each heatsink. All heatsinks can be optimised, with thicker coldplates, faster fans, Arctic Silver etc, but that’s not the point of this test.
From left to right: MC370-4, PEP-66, PAL-6035.
The Swiftech MC370-4…..
The MC370-4 is also the heaviest of the heatsinks tested by a long way – those copper pins make this thing about twice the weight of the PEP66. The base is machined flat and very smooth, requiring little lapping. The mounting mechanism is by far the sturdiest here, with a screw down mechanism on each side, but a little tricky to get right (good thing I’m using a PIII; this cooler could easily crush a lesser chip).
The Alpha PEP-66…..
The PEP also has 88 thin blades arranged in four rows of 22, shaped like an aircraft wing in section. The fins are designed to have air draw across them rather than the more traditional blowing orientation. For a while now, the PEP66 has been the undisputed “King of the Socket/Slotket Coolers.”
The Alpha PAL-6035…..
For the most part, I followed the guide: Total Condensation Prevention on the OCtools website, with the addition of some neoprene to insulate the back of the slotket – I’m expecting some good sub-zero temperatures. This time, instead of using standard silicone, I stumped up the $20.00US for some Dow Corning 1-2577 RTV silicone conformal coating from OCtools; this stuff knocks spots off regular silicone sealant and is worth the cost in comparison to a new mobo. So, with a waterproof computer we are ready to test the rigs.
I should say at this point, that I don’t run hlt commands, nor should anyone who runs a peltier rig – it is pointless and dangerous. I let Motherboard Monitor run 200+ samples at idle and recorded the mean temp, then reset Motherboard Monitor and ran burnp6, again letting MM run 200+ samples before recording the mean temp. I rounded up or down any fractions to the nearest whole number.
For the coldest seen, it was simply that – the coldest temp that the machine reached (I discounted any momentary flickers as unreal). I should also take a moment to state that I did my level best to be unbiased in these tests. I had my private favourites, but I was determined to record exactly what I saw, and that’s exactly what I did. Here is what I found:
|You can clearly see which heatsink is coming out on top. As expected, the PEP did very well, surprisingly, the PAL has coped well with even the 80W TEC, but the clear winner is the Swiftech! 17C under full load – wow!|
The results of the minimum temp test show the clear winner as the Swiftech every time, but what is interesting is the battle between the PEP and PAL. The PEP is not the clear winner in all circumstances, and when the PAL is fitted with a 33CFM fan, it can match the PEP at cold starts.
However, it should be noted that the "minimum seen" is a very poor test of performance and reserved only for those nerds who run hlt commands with peltiers just so they can boast "…My machine runs at a gazillion degrees below zero… so long as I don’t move my mouse" 😉
This is the average temp recorded while running burnp6 and is the only real measure of heatsink/peltier performance. And it’s pretty obvious which one is the winner – the Swiftech again. It’s this graph that has made me a real copper convert. Just 17C under full load with an 80Watt TEC…stunning.
The curve of the graph illustrates the performance of the heatsink. The PEP performs well, but the PAL starts to show the strain of 80 watts – it can handle it at idle, but under load the graph starts to climb. It still did well, at no point reaching 35 degrees – not bad!
When I first tried the Swiftech, I was convinced I had wasted my money as it didn’t perform as well as the PelTEC PEP66. But having tried it with the fan mounted on top of the heatsink (PAL style) – it’s a definite winner. With regard to the PEP, it came close to performing as well as the MC370 in many of the tests and is outstanding value for money – PelTEC offer the kits at little more than the sum cost of the parts.
For those new to peltier cooling and experienced alike, getting it all from one place, even your 1/4 inch coldplates, at such low prices has got to place PelTEC as #1 for 1-stop shopping <g>. I have been running the PEP for nearly 6 months and have been extremely pleased with it so far. I feel a little sorry for it because it has been replaced – alas, no room for compassion in the hunt for more MHz 😉
It looks as though I’ve found my universal cooler – it may even do the business with the forthcoming Pentium IV if the form factor is close enough and it looks like it may be:
As regards copper coolers, this is my first experience with one and it seems that the qualities of this metal work quite well in this situation. Copper has a greater affinity for heat than aluminium and so provides a better conduit to "suck" it out of the peltier element. It also hangs on to it (heat) longer, meaning some quite aggressive fan therapy is required to keep the pins cool.
In the tests, I noticed it took much longer for the Swiftech to warm up and cool down than the other heatsinks. The PAL especially was very erratic, with temps bouncing up and down quite rapidly; this made me feel much happier about the copper cooler’s stability over it’s all aluminium counterpart.
Also the advent of the new "hedgehog" cooler might give the Swiftech a good run for it’s money. It is an all copper construction with solid pins, unlike the Swiftech’s hollow variety. PelTEC are offering them, pre-rigged under the title of “sledgehog”. It would be interesting to see how it compares to the Swiftech’s dual-metal construction – could it be the “Swiftech-killer”;, I wonder? (but that’s for another payday.)
With regard to the the 80 watt TEC, the Swiftech was the only cooler I felt could happily deal with the heat, even then I wondered if it was on the edge, as the pins did get quite hot. I shall be ordering a 72 watt TEC soon and I feel confident it will produce better results. Next project – optimizing the Swiftech, 72Watt TEC, 1/4 inch coldplate, Arctic Silver and possibly a small shroud like the PAL and the Hedgehog, to start with anyway, watch this space……..
The Swiftech MC370-4, 17°C with a 80W TEC, 33cfm Papst fan, running burnp6…what more can I say!
Note: The pictures show the MC370-4 with a 1/4 inch coldplate, this was NOT the coldplate used in the tests.
Reprinted from Martyn’s Website with his permission.