- Joined
- May 31, 2004
Having read a ton of great posts I came to three conclusions. The sp-94 with AS5 is an awesome heat guzzler. The dead spot significantly affects cooling. And blowing hot air from the hot heatsink onto a motherboard isn't a great idea.
As a result, I decided to use a 4" duct (dryer vent from the local hardware store) over the heat sink and the fan sucking heat from the sp-94 and blowing it directly out of the case. This means I don't have to have lots of auxiliary fans blowing air in and out. I have two 80mm panaflo fans on medium speed blowing on my hard disks and the two back fan openings are open to allow air to flow in as needed.
I used a spreadsheet analysis of airflow in the sp-94 heatsink to see clearly that blowing air in the usual way with one fan causes the air from the edge of the fan to be sucked inwards at high velocity carrying heat INTO the sink. The inward sucking is caused by the vacuum that forms under the fan hub. The harder you blow on to the edge of the sink, the greater the vacuum under the hub and the more air that heads towards the center. In essence hot air from the outer fins is being sucked back into the center of the heat sink making it harder for that air to cool the dead spot area. Therefore at least some of the heat transported by the evaporation tubes to the edge is being returned to the center area. Analysis shows that upon exit from the heat sink the air flow is pointing downwards towards the mother board, thereby heating the board. There is also a significant level of churning where high speed hot air from the edge is recycled back into the fan, again reducing the cooling effect.
To eliminate the dead spot, I attached two 92mm panaflo fans side-by-side and put the high speed center of those two fans directly over the center of the sp-94. Maximum air flow occurs at the center of the heat sink, instead of a minimum flow. The spread sheet analysis showed that the air flows quickly from the center and pushes on the slower speed air near the hubs which are now located outside the heat sink. (See my pics for a drawing of the air flow for one fan [on the left] and two fans [on the right] the length of the drawn line represents the speed of the air at that point. See how the air flow for one fan is actually towards the heat sink at high velocity for almost half the depth of the sp-94 and the exit air is slow from the center.) The reverse flow takes place if the fans are reversed.
Results: p4 3.2c overclocked to 3.6 ghz, pc3700 memory
With the standard intel fan and heatsink (which had a built in 3" duct to suck air in over the heatsink) my load temps were 58c CPU and 48c mobo, idle 38c and 42c (80mm fan on full 3000 rpm and lots of noise) stable under prime95. (And yes the CPU temp was cooler than the mobo, I assume because cool air was being brought in over the CPU from the duct.)
With the sp-94 and AS5 in place and a panaflo 92mm 59cfm fan blowing full speed inwards (2950 rpm) using the 3" duct provided on the system, the load temps dropped to 45c CPU and 44c mobo, idle 32c cpu, 39c mobo. I guess the heat from the CPU was heating up the motherboard and I had only had one 80mm fan blowing in, none blowing out.
Now, p4 3.2 overclocked higher to 3.84 ghz using pc 4200 memory (my pc3700 memory wouldn't go that high) stable on prime95.
Adding the twin fans sucking on the sp-94 and a 4" duct directly over the high speed center of the two fans my load temps are 38c/27c (room temp 22c) and idle 26c/27c. My hard drive temps have also fallen from 45c to 20c (which I assume means the temp measurement is a bit off for them). I can turn the fans down from full speed to 2206 rpm and the load temps go up to 41c/28c but the quiet is beautiful (certainly lots more quiet than a tornado would be) [note the dB of the two fans is not directly additive, I believe a second identical fan adds 3 dB to the noise. Hence the panaflo 35 dB fans combined make 38 dB fans at full speed.]
Here are some pics of my adpated fans. I just cut up some sheet metal to make two brackets to hold the fans together and used the supplied clips that came with the sp-94 to hold the whole thing on. Notice how the lower fan is sucking over the north bridge, and one of the air intakes of the sp-94 is sucking next to the memory. The memory and the north bridge are not warm to the touch even after a whole day of prime95.
My photos:
http://photos.yahoo.com/year1980a
As a result, I decided to use a 4" duct (dryer vent from the local hardware store) over the heat sink and the fan sucking heat from the sp-94 and blowing it directly out of the case. This means I don't have to have lots of auxiliary fans blowing air in and out. I have two 80mm panaflo fans on medium speed blowing on my hard disks and the two back fan openings are open to allow air to flow in as needed.
I used a spreadsheet analysis of airflow in the sp-94 heatsink to see clearly that blowing air in the usual way with one fan causes the air from the edge of the fan to be sucked inwards at high velocity carrying heat INTO the sink. The inward sucking is caused by the vacuum that forms under the fan hub. The harder you blow on to the edge of the sink, the greater the vacuum under the hub and the more air that heads towards the center. In essence hot air from the outer fins is being sucked back into the center of the heat sink making it harder for that air to cool the dead spot area. Therefore at least some of the heat transported by the evaporation tubes to the edge is being returned to the center area. Analysis shows that upon exit from the heat sink the air flow is pointing downwards towards the mother board, thereby heating the board. There is also a significant level of churning where high speed hot air from the edge is recycled back into the fan, again reducing the cooling effect.
To eliminate the dead spot, I attached two 92mm panaflo fans side-by-side and put the high speed center of those two fans directly over the center of the sp-94. Maximum air flow occurs at the center of the heat sink, instead of a minimum flow. The spread sheet analysis showed that the air flows quickly from the center and pushes on the slower speed air near the hubs which are now located outside the heat sink. (See my pics for a drawing of the air flow for one fan [on the left] and two fans [on the right] the length of the drawn line represents the speed of the air at that point. See how the air flow for one fan is actually towards the heat sink at high velocity for almost half the depth of the sp-94 and the exit air is slow from the center.) The reverse flow takes place if the fans are reversed.
Results: p4 3.2c overclocked to 3.6 ghz, pc3700 memory
With the standard intel fan and heatsink (which had a built in 3" duct to suck air in over the heatsink) my load temps were 58c CPU and 48c mobo, idle 38c and 42c (80mm fan on full 3000 rpm and lots of noise) stable under prime95. (And yes the CPU temp was cooler than the mobo, I assume because cool air was being brought in over the CPU from the duct.)
With the sp-94 and AS5 in place and a panaflo 92mm 59cfm fan blowing full speed inwards (2950 rpm) using the 3" duct provided on the system, the load temps dropped to 45c CPU and 44c mobo, idle 32c cpu, 39c mobo. I guess the heat from the CPU was heating up the motherboard and I had only had one 80mm fan blowing in, none blowing out.
Now, p4 3.2 overclocked higher to 3.84 ghz using pc 4200 memory (my pc3700 memory wouldn't go that high) stable on prime95.
Adding the twin fans sucking on the sp-94 and a 4" duct directly over the high speed center of the two fans my load temps are 38c/27c (room temp 22c) and idle 26c/27c. My hard drive temps have also fallen from 45c to 20c (which I assume means the temp measurement is a bit off for them). I can turn the fans down from full speed to 2206 rpm and the load temps go up to 41c/28c but the quiet is beautiful (certainly lots more quiet than a tornado would be) [note the dB of the two fans is not directly additive, I believe a second identical fan adds 3 dB to the noise. Hence the panaflo 35 dB fans combined make 38 dB fans at full speed.]
Here are some pics of my adpated fans. I just cut up some sheet metal to make two brackets to hold the fans together and used the supplied clips that came with the sp-94 to hold the whole thing on. Notice how the lower fan is sucking over the north bridge, and one of the air intakes of the sp-94 is sucking next to the memory. The memory and the north bridge are not warm to the touch even after a whole day of prime95.
My photos:
http://photos.yahoo.com/year1980a