I recently received the Millennium Glaciator heatsink/fan and decided to test it out, head to head, against the Alpha PEP66T. This new HSF has been tested against several other units, but I have yet to see it put up against the Alpha PEP66.
When I received this in the mail, the first thing that impressed me was the sheer weight of it – this is a very heavy piece! I use an Alpha P3125 HSF on my PIII 600E and thought that it was heavy, but the Glaciator is far heavier than the P3125. In fact the weight of this is making me wonder if it would be suitable for a slocket application. I intend on testing this on a Celeron 600, mounted on a Soltek SL02A++ slocket. The Glaciator is designed with the AMD Socket 462 (“A”) in mind, but should work with the intel Socket 370 chips as well.
The Glaciator is shipped with a one page “Instruction Sheet” outlining installation of this unit. Also included is a small packet of Arctic Silver thermal compound – very nice touch. I recommend using this thermal compound when overclocking, as it has the best thermal properties available.
Where the Glaciator is designed for the AMD Socket A platform, I thought it might be a good idea to make sure it would fit on a slocket first, before removing my trusty PEP66T from my Celeron. I have a generic slocket that isn’t being used, so I did a test fitting to this first. The clip that comes with the Glaciator fit the Socket 370 perfectly, but I encountered a different problem.
The ZIF arm used on the generic slocket, as well as the Soltek I intended to use for this test, are made of metal, and have a bend in them to raise the end up off of the circuit board. This bend raises the arm above the top of the socket. The PEP66 is machined along this edge to clear the ZIF arm, but the Glaciator is not. I decided the easiest method to solve this, would be to straighten the bend in the ZIF arm out.
The Alpha also comes with rubber “feet”, that are used to keep the HSF level against the CPU die on an intel chip. The Glaciator does not come with these. If you plan on using the Glaciator with an intel CPU, you should get some of these “feet”, as they really do work well in keeping the HSF level.
Before disassembling the PEP66T from the Celeron, I ran several tests with it, using three different fans, both “sucking” air off of the heatsink, and “blowing” air through it. All tests were run with the same settings, for the same length of time. I allowed some time in between tests for the components to cool down. The test settings were:
Intel Celeron 600 @ 1008 MHz, 2.05v, fan plugged into the CPU fan header on my Asus P3C2000 motherboard. Ambient temperature was about 78F/25C at the time of the tests. All temperatures were measured using the Asus Probe.
Test 1: PEP66T, with 60x60x10mm Delta fan (unknown CFM), “sucking” air off the heatsink
IDLE Temp: 22 – 23C (idled for 5 minutes, to allow temp to stabilize)
LOAD Temp: 36 – 39C (running Prime95 Torture Test for 30 minutes)
MotherBoard Temp: 28C throughout test
FAN Speed: 5625 rpm (load)
Test 2: PEP66T, with 60x60x25mm YS Tech fan, 27 CFM, “sucking”
IDLE Temp 23 – 24C (idle 5 min)
LOAD Temp 35 – 38C (Prime95 TT, 30 min)
MB Temp 28C throughout test
Fan Speed N/A
Test 3: PEP66T, with 60x60x25mm YS Tech fan, 27 CFM, “blowing” onto the heatsink
IDLE Temp 22C (idle 5 min)
LOAD Temp 33 – 35C (Prime95 TT, 30 min)
MB Temp 28C throughout test
FAN Speed N/A
Test 4: PEP66T, with 60x60x25mm Delta fan, 38 CFM, “sucking”
IDLE Temp 22 – 23C (idle 5 min)
LOAD Temp 34 – 36C (Prime95 TT, 30 min)
MB Temp 28C throughout test
FAN Speed 6490 rpm idle, 7031 rpm load
Test 5: PEP66T, with 60x60x25mm Delta fan, 38 CFM, “blowing”
IDLE Temp 21 – 22C (idle 5 min)
LOAD Temp 31 – 33C (Prime95 TT, 30 min)
MB Temp 28C throughout test
FAN Speed 6617 rpm idle, 7180 rpm load
For test number six, I attempted the modification outlined in the article
“Quieting the Delta 38CFM Fan” by Ed Chapel. This involves silicone mounting the Delta fan up and off of the heatsink, in an attempt to quiet the noise this fan produces. While the noise reduction was minimal, there was a small change in temperatures, as seen in the next test.
Test 6: PEP66T, with 60x60x25mm Delta fan, 38 CFM, silicone mounted to heatsink, “blowing”
IDLE Temp 19 – 21C (idle 5 min)
LOAD Temp 28 – 33C (Prime95 TT, 30 min)
MB Temp 29C throughout test
FAN Speed 6490 idle, 7031 load
Now it was time to install the Millennium Glaciator, and test it out. This unit really is a work of art. Excellent machine work throughout. Anyone familiar with the Alphas will no doubt remember the circular flycut machines “grooves” in the base of the heatsink. There are only very fine parallel machine marks on the base of the Glaciator. Much smoother than the Alpha’s surface.
The Glaciator is constructed of 99.99 % pure copper, and weighs 780 Grams. The fan supplied with it is mounted down into the heatsink, as opposed to on top, and delivers 29 CFM, at a rated speed of 5700 rpm + or – 10 %.
I removed the PEP66T from the Celeron/slocket assembly, and cleaned off the old Arctic Silver compound with a Q-Tip and rubbing alcohol. I also cleaned the base of the Glaciator, as well. I still had the second set of “feet” that came with my PEP66, and I applied these directly onto the top of the Celeron, one each at the center of each edge of the chip, to provide stability for the Glaciator to sit flush. I put these onto the chip, in the event I ever decide to lap the surface of the Glaciator in the future (not that it needs it).
I then removed the Celeron briefly, so that I could bend the ZIF arm so it sits below the surface of the socket. This was quite easily accomplished using two pairs of pliers. I bent it almost perfectly straight, leaving just a small bend outward, away from the socket, to facilitate future removal of the Celeron.
That done, I reinstalled the chip, applied a fresh coat of Arctic Silver, and put the Glaciator in place. The clip went on quite easily, and holds the Glaciator in place firmly. I was expecting it to be a struggle getting the clip on, but it wasn’t. This should be welcome news to AMD users, where the cores of their chips are so fragile.
I put the assembly back into my PC, and fired it up. I let it idle for a few minutes, and then ran Prime95 Torture test for about 20 minutes to heat up the thermal compound. I shut the machine down, ate dinner, and resumed testing.
I took note of the fan speed when I initially started the Glaciator. It idled at 4600 rpm. After the initial running of Prime95, it idled at 4891. With use, the fan seemed to start loosening up, and picked up speed. I let it run Prime 95 again for a short time, and after that, the fan idled at 4963 rpm, where it has stayed since.
I was now ready for the final test. As the fan is inset into the Glaciator’s heatsink, there isn’t any way to “flip” the fan over to test for a difference between “sucking” and “blowing”. The fan is designed to blow down onto the heatsink.
Test 7: Millennium Glaciator, OEM fan, 29 CFM, “blowing”
IDLE Temp 21 – 23C (idle 5 min)
LOAD Temp 33 – 34C (Prime95 TT, 30 min)
MB Temp 28 – 29C
FAN Speed 4963 idle, 5273 load
So we see the Glaciator running virtually the same temps as the Alpha/Delta 38 CFM fan. The Alpha runs just a couple of degrees cooler, but the Glaciator runs much more quietly. I have a pair of generic (CompUSA) 92mm fans in my case, and with the case side panel on, you cannot hear it running.
When I was running the Delta with the 38 CFM fan, (with the side panel on the case) you could hear it outside my computer room, down the hall, in my living room, over the television. That’s loud. I think this is the most impressive part of the Glaciator, the level of noise it makes.
If the noise level of the Delta 38’s drives you crazy, but the temperatures it provides are what you need, then the Glaciator is the HSF for you. It kept my Celeron under 35C, and the temp dropped back into the low twenties fairly quickly after the load was removed from the CPU.
I don’t think you’ll get cooler temps at idle with air cooling, as all tests ran between 19 – 23C. Under load, the Glaciator more than holds it’s own. Despite putting out nine CFM less than the Delta (38 to 29 CFM), it kept the Celeron just as cool, almost to the degree.
This seem to be a very well thought out design, that performs as good as physically larger heatsinks and fans. The best way I could describe it, would be: “Golden Orb quiet, Alpha cool.”
This might just be the next “big thing” in air cooling.
Brian Berryman -aka- Mr B
Senior Forum Member, Overclockers.com