After playing around with my air cooled Duron 700 for a while, I finally decided to close my computer case and finish experimenting. By then it was running quite stable at 1008 MHz under the max 1,92V my Abit KT7-R could supply.
After closing my AOpen full tower case, I experienced a severe heat problem: My CPU temps went up from 46C to 58C. This situation pushed my speeds to sub-gig speeds, something I could not live with for long. I had two options; doing the fan blues, going watercooling, or settle for sub-gig. Yes only two. 🙂
I live in “remote” part of the world (Iceland), so getting a waterblock or indeed any other overclocking gear is expensive and takes a lot of time. What’s involved is finding a fair reseller that ships oversees, doing toll reports, and paying this and that. It actually never crossed my mind to walk that route, besides I didn’t have to.
I got access to my friends mill and, armed with a copper bar, I made my own two piece 65x50x16mm waterblock. This unit is similar to the one Danger Den makes and uses 3/8″ hoses.
I started experimenting with my new waterblock using 6-9C tap water. I got greedy and did the voltage mod on my Abit KT7-R too, resulting in 2,15v as max cpu voltage. This brought my Duron to a stable 1100 MHz at 2.15v with cpu temps of 20C under seti@home. This result motivated me to try peltier cooling.
At this point, I could not find any information regarding peltier cooled Durons on the web, so I had to take a guess regarding peltier size. This decision was finally not very hard to make, since local resellers had only 68w units available. This unit is rated: max temp delta 65c, voltage 12-15V, and max current 8,5A
So I ran out and got one, slapped it on a 6mm copper coldplate with my waterblock on top. I made a clip out of 1,5mm steel MIC welding wire to keep things in place. I decided to try this setup first and worry about condensation later.
I did have my old Sytron PSU lying around, so power was not a problem; this unit can supply 0-50V DC and a whopping 40A regulated – not actually a portable unit, with nearly 50 lbs of bodyweight.
The results were shocking, at least to me: My Duron never went over a gig in this experiment; it did not serve any purpose to try. This was clearly way too small a peltier. I tried voltages up to 16V, but I got lowest CPU temps at 13V. Even at stock CPU speed and voltages, I got higher temps with the peltier than without it. I was still using tap water at 9C for my waterblock and my coldplate was indeed hotter than my water block at all times.
This led to experiment number two.
I got more copper and made another waterblock in the same fashion, but designed for two peltiers (80x50x16mm) mounted side-by-side. I got another 68 watt peltier and made another bigger 6mm thick coldplate. This time I was really going to get some impressive figures. After careful installation of the two 68w peltiers side-by-side, I booted my computer at 800 MHz with 1,7v core and went straight into BIOS to check CPU temps.
Yes my temps were under zero, -4C or so. I went into Windows and started VIA HWM. Duh, my temps had climbed up to 2C just booting into Windows. I started seti and with big tears I watched my CPU temps climb up to 9C in a few minutes. “Something must have gone wrong” I thought, so I shut down my system and checked everything. Many times over and over again, I measured voltage at the peltier terminals, tried less/more thermal paste, more and more clamping pressure, voltage from 10-18V – but with no luck.
I came to the conclusion that one of my peltiers was not doing its job. I tried them separately, but they both gave same results as before. I did put everything back together to see what happened at higher speeds and voltages. The results were simple: At overclocked speeds, my dual peltier setup was no match for my waterblock alone with 9C water. I tried several speed/voltage combinations and the results can be viewed in the following table.
| Cooling | CPU MHz | VCore | CPU Temp |
| Coolermaster hsf | 1000 MHz | 1,94 v | 46 C |
| Water @ 9c | 800 MHz | 1,70 v | 15 C |
| Water @ 9c | 1100 MHz | 2,15 v | 20 C |
| Water @ 24c | 1100 MHz | 2,15 v | 33 C |
| 1 x 68W peltier | 800 MHz | 1,70 v | 13 C |
| 1 x 68W peltier | 936 MHz | 1,98 v | 30 C |
| 2 x 68W peltier | 800 MHz | 1,70 v | 9 C |
| 2 x 68W peltier | 1050 MHz | 1,98 v | 23 C |
| 2 x 68W peltier | 1100 MHz | 2,15 v | 32 C |
My peltiers got very good cooling with pure copper waterblock flushed with cold water. I used the same thermal paste for my peltiers as I used with good results for my CPU’s. Voltage was not a problem either. This leaves only one thing left:
My peltiers cannot transfer enough heat to get any lower temps than with my waterblock alone.
Regards from Iceland
Be the first to comment