Agilent ArticCooler report

[bmg]

The Agilent ArticCooler I backordered from Avnet-Marshall a couple of weeks ago finally arrived this afternoon. Naturally I wanted to compare it to my current Alpha PAL35T with 27cfm YStech fan. My system is a 933Mhz Coppermine running conservatively at 980Mhz with voltage set to 1.75 volts. My motherboard is an ASUS CUSL2. I use Artic Silver thermal compound.

Before installing the ArticCooler I ran a temperature test using Motherboard monitor 5 and Prime95 selftest. MBM5 sampled temperature every 10 seconds and I ran past the point where temperature had stabilized (to 70 samples in MBM5). The starting/max/average temp. was 84/111/107 degrees F using the Alpha. I then installed the ArticCooler and ran the test again. This time I got 84/105/100. I guess Kyle's ArticCooler review at [H]ardOCP wasn't a fluke! This thing really does work great! BTW, it's also quieter than the Alpha/YStech setup (admittedly a subjective opinion as I have no means to quantitatively measure the noise level).

Everyone that has one of these on order is going to be really happy! Just be carefull installing it as I had a little trouble installing mine. After I finally got the short side of the clip hooked under the socket tabs it went fine. Just tilt the heatsink over carefully until it's flat and then hook the long side of the clip under the socket tabs (easy).

 

[Apok]

send it to me

 

[bmg]

It looks like Avnet-Marshall is backordered again. BTW, I ran Prime95 for a much longer period of time and got 105 max and 104 average.

 

[Mikewarrior2]

Hey,

did you happen to try out that pad that comes with it?



Mike
P.S. I may try it out with the pad on monday when I get mine, just to see if it is better than AS.

 

[beer]

2cooltek got his Articooler...he said that basically it performs like crap. I am not sure if I like his testing methodology...it's linked off his front page, www.2cooltek.com

 

[bmg]

I removed the pad and just used Artic Silver. I'm pretty sure the pad wouldn't work quite as well. If you use the pad I believe the heatsink needs to get up to 65 degrees C for a short time for the pad to work well. Check out this link for info on the pad Agilent uses: link to agilent pad info

<added>
One other thing of interest in the Agilent article is the comment that the ArticCooler's .62C/W thermal resistance is measured using a 9x11mm heat source (very close to the size of the Coppermine). This thermal resistance can't be directly compared to that of another heatsink unless the size of the heat source stays the same. For example, I checked the Alpha web site and found that they appear to use a 12.7x12.7mm heat source when testing heatsinks like the PAL35T. That's 161 square mm for the PAL35T versus 99 square mm for the ArticCooler. The spec'd thermal resistance of the PAL35T is ~.37C/W. If the PAL35T were tested with a heat source closer to the size of the Coppermine (~9x11mm) it would have a thermal resistance substantially greater than .37C/W. A simple scaling of the PAL35T thermal resistance by the heat source area ratios would give .37x1.62 = .60C/W. I'm sure this isn't a very accurate way to correct a thermal resistance value for a different heat source area, but it's probably in the ballpark. Clearly a processor the size of a pinhead dissipating 25 watts is going to run a lot hotter than a larger processor also dissipating 25 watts. It's easy to see that the size of the processor die has a substantial effect on the thermal resistance from the processor to air. This is worth remembering when someone says that one heatsink is better than another based on the published thermal resistance numbers, and overlooks the fact the the measurement conditions used to come up with the numbers are different. The only way to really compare is to put the different heat sinks on the same processor (or processor sized heat source) and then take measurements under the same test conditions.

(Edit: It has been pointed out to me that the .37C/W number for the Alpha is heatsink to air, not junction (or heatsource) to air. I verified this from the Alpha web site. The thermal resistance of junction to heatsink isn't included in the Alpha number. The ArticCooler .62C/W thermal resistance is junction to air not heatsink temp to air.)

 

[2Cool]

However, the real world performance is also dependent on composition of the heatsink, and the copper base of the PAL/PEP series Alpha certainly give an advantage over the Arcticooler.

Also, the Arcticooler is tiny. Period. It is the only heatsink that I have tested that got very hot to touch. Efficiency is one issue, but that is only good for so much when there is just not enough mass and surface area to shed the heat.

The review heatsinks supplied to many of the review sites seem to have been &quot;optimized&quot; prior to sending out. Kyle reported that the Arcticooler he received had a &quot;mirror finish&quot; on the bottom. My production unit did not. This indicates his was lapped prior to his receiving it. I also suspect the rest of the heatsink was massaged a bit as well, and only the most cherry flowing fan chosen.

Color me cynical, but thats how I see it, and of course its only a personal opinion.

 

[CamaroGuy]

How the heck did you get it on the CUSL2? I couldn't get mine on. Now I need a new motherboard because when I tried with all my might to get the the metal edge on the back side under the notch on the socket, the metal went into the board and took out some traces. As it slipped down into the board, my processor also cracked.

-CamaroGuy

 

[bmg]

The copper insert in the Alpha heatsink is a definite performance booster, no doubt about it. However, that doesn't negate the effect of the heatsource area on thermal resistance. Thermal resistance can't be directly compared unless the test conditions are equivalent. (edit: It has been pointed out to me that the .37C/W number for the Alpha is heatsink to air, not junction (or heatsource) to air. I verified this from the Alpha web site. The thermal resistance of junction to heatsink isn't included in the Alpha number. The ArticCooler .62C/W thermal resistance is junction to air not heatsink temp to air.)

I got in 2 more ArticCoolers from Avnet-Marshall this morning so switched another of my systems from the PAL35T to the ArticCooler. Before the switch my idle temperature was 89 degrees with a max of 112 during Prime95 selftest. This system is identical to the first with the exception that this one is running a 64MB GeForce2 at 210/420 and the first was running a V5 5500 at 175. The temperature of the 2nd system has always been higher than the first (variations in the on-die thermal diode or video card power??). Anyway, when I changed the 2nd system to the ArticCooler my max Prime95 temp. went down to 109 degrees from 112. This isn't as great a drop as the first system, but it's still an improvement over the PAL35T. Interestingly, the surface finish on the second ArticCooler was noticably better than the first, almost a mirror like polish.

This ArticCooler install went much easier than the first. It still wasn't trivial to get the short end of the clip under the socket tabs. Don't push down with much force because if you look carefully you'll see the clip is down against the board, even though it doesn't want to easily slide under the tabs. Try applying a moderate amount of sideways force to pull the clip under the tabs without pressing down much. I found I could get under the top tab (processor socket is top, pci slots are bottom) without too much work and then pivot under the bottom tab. After that it's easy to carefully tilt over the heatsink with the left thumb while attaching the long side of the clip with the right fingers/thumb. One thing I did which seemed like it helped (although I'm not sure it was necessary) was make a couple of &quot;L&quot; shaped shims from some little pieces of metal and wedge them under the &quot;U&quot; shaped tabs on the short side of the clip. After the short side of the clip was in position under the socket tabs I carefully removed the shims (position shim so one side of the &quot;L&quot; is along side the heatside for easy access). The shims were .040&quot; thick. The metal I used was removed from an unused backplate (not sure what it's really called - the plate that covers up the pc case holes for unused pci slots). As CamaroGuy has noted in another thread, be carefull! I'd say Agilent should put some installation tips on their web site. I'm going to send them an email to that effect, and the more emails from folks the better. I think they've got a well performing product, but if people can't install it without damaging their motherboard and/or processor Agilent is going to get a bad reputation fast.