USER RESULT 11/22/00:
Read your review on the Alpha PEP66 for socket mobos. Just did the same mod myself and am pleased with it. My system:
- Abit KT-7 Raid
- Thunderbird 800 oc’ed to 896
- Alpha PEP66 with a Papst 33cfm fan blowing in
- 512 meg Crucial PC133 cas2 ram
- Guillimot Geforce 256 32 meg
- Guillimot Fortissimo sound card
- Ambient: 77.6 F
- Prime95: 113F
- Idle: 87F
SUMMARY 11/20/00: Modifying the Alpha PEP66 for use as a socket heatsink results in what looks like the best Socket A cooler.
Tom Leufkens wrote up a tip (HERE) about modifying the Alpha PEP66 for use as a socket heatsink. Being of curious mind, PC Nut kindly provided one to modify, so I proceeded to “Leufken” it per Tom’s excellent article.
What I did find, however, is that I had to cut slots in the shroud’s side; if you don’t do this, the clip is pushed down by the shroud such that there was no way to engage the socket’s lugs with tension. The shroud is aluminum and any decent metal snips will do the job (maybe even a good scissors).
The other cutting job is to remove the four nylon feet on the PEP66’s base. I used a utility knife for this – just make sure the blade is sharp and that none of your fingers are in front of the blade as you cut.
All told, this took about 15 minutes to do.
What is very nice about this mod is that getting the PEP66 on and off the socket is SOOO MUCH easier than almost any other Socket A heatsink. I have no qualms in saying that the probability of damaging your Duron or T-Bird is very low (although I must say that I use a needle nosed pliers to engage the clip).
I compared the two Alphas, the PEP66 vs the PAL6035, using an ABIT KT7 with a Duron 800 running 1000 MHz; both heatsinks were powered by a Papst 33 cfm fan (Swiftech has them). I ran Prime95 and recorded the following results:
|PEP66 Fan Blowing DOWN|
|PEP66 Fan Blowing OUT|
|PAL6035 Fan Blowing Out|
No question there’s a clear advantage for the PEP66. Note that orienting the fan to blow DOWN into the heatsink is important – if you look at the way air flows through the PEP66, mounting the fan to blow OUT will not get as much air over the base as blowing DOWN into it.
As you can see below, compared to the PAL6035, the PEP66 is larger; Alpha’s published stats also give the edge to the PEP66.
PEP66: .35 C/W
PAL6035: .37 C/W
The Duron at these settings is radiating 50.3 watts. Using a C/W of 0.35, you should expect a CPU temp rise of about 17.6C/31.7F over ambient (go HERE for more on this). In fact, the observed temp of 38C/100F is dead on to what you would expect based on the PEP66’s published stat:
Expected CPU Temp Rise = 50.3 watts x 0.35 = 17.6C
Observed Ambient temp 20.3C
Theoretical Performance = Ambient + Expected CPU Temp Rise
Theoretical Performance = 37.9C
Observed Performance = 38C
I was frankly surprised to get so close to theoretical – getting to within 5%-10% of theoretical using the on-board temp sensor would not be unexpected. For example, doing the same calculation for the PAL6035 yields an expected 38.9C vs 41C observed*. I am doing more work on heatsink testing and will be posting some lab tests shortly. Hopefully we can give you some more information to winnow out the duds.
If you are buying a heatsink for overclocking Socket A CPUs, IMHO a published C/W greater than 0.40 is not going to give you the cooling performance you need for maximum results.
Obviously, this assumes some accuracy in data published by manufacturers; those that sell products into the OEM market are usually not bad.
In summary, looks like the Alpha PEP66 is one great way to go for air cooling Socket A heatsinks. I’d like to thank Tom Leufkens for a great tip and PC Nut for providing the PEP66 to test out.
*Although pulling air off the board is usually warmer than ambient air into a fan blowing down, thereby affecting results adversely. Adding just 1C for this gets to about 40C, closer to what was recorded. If you want some more background on this, check out Burning Issues for some excellent How-Tos.