Maze 2.2 & Dual 120 watt Pelts - Results!

[Colin]

As many of you may recall, I tried dual 172 watt pelts a while back with less than stellar results. The pelts were suspect and replacement pelts were lost in the mail. In the meantime I decided to try two 120 watt pelts. With 100 watts less heat to dispose of, I had better expectations. You can read the dual 172 watt pelts thread here. Please note that the Photopoint server has been down for the last 24 hours so the pics may not be visible in the earlier thread.

The first experiments were conducted with a Duron. Partially for the absurdity of $600 worth cooling gear to cool a $39 CPU and also only loosing $39 if something went wrong. With what I have learned, I felt confident using my best AYHJA Bird. Watercooled this chip runs at 11 * 145 @ 2.04 volts.

The initial results are in. With an ambient temp of 27 C full load is 19 C with a cold plate temp of –4.7 C. The system idles at –11 C and water temps in the bong cooler stabilize at about 30 C. Water loss looks to be about two gallons per day running at full load verses a quart every 45 minutes with the 172 watt pelts. And the important stuff like the overclock? 1704 MHz, that’s 12 * 142 @ 2.04 volts.

Edit: Forgive me, I for got the acknowledgements in the original post. I would like to thank the guys at Danger Den for their products and superb support. You guys rock! I would also like to thank the members of the forums here for their encouragement and support though out my dual pelt experiments.

 

[Bender]

Very nice results. How much difference did 2 120 watt pelts make over a single 172? I would try a bong but I don't like the idea of losing water.

 

[Colin]

I never tried a single 172. Earlly this year I had my old AJFA 1 gig Bird running at 1392 with a 156 watt pelt. I believe the temps were close to zero or a degree or two below at idle with about 19C to 20C at load. Ambient was lower as it was winter and that CPU was only putting out 99 watts of heat. This one is putting out about 120 watts.

 

[Les56]

Colin (Jul 14, 2001 02:49 p.m.):
[url="ht

The initial results are in. With an ambient temp of 27 C full load is 19 C with a cold plate temp of –4.7 C.
quote]

I maybe being stupid but a temp difference of nearly 24c between cold plate and cpu is one hell of a temp gradient.
I apologiize for the abrutness of the above, and have i misunderstood ?(addendum)

 

[Colin]

Sure is. My results are not out of line from other T-Bird pelt setup I have read about. The coldplate is solid on the CPU. I pulled the block once to make sure the contact was correct. The springs are compressed all the way on the mounting bolts behind the motherboard.

 

[Les56]

My thoughts go back to "God" and the benefits ot a 15mm Cusil cold plate.
Although 8mm plates are considered adequate I have not seen any temp profile of any thickness of plate.

 

[Colin]

Seems to me one of the problems is the small surface area of the CPU at the center of the coldplate. The thermistor is at the edge of the coldplate about an inch from the core. Even though this is pseudo isothermal environment this qualifies as a secondary heat path. I am certain the temperature of the coldplate directly over the CPU die is much higher.

On of the main reasons I put a thermistor on the coldplate is to aid in diagnosis and powering up the computer. It was not intended to provide an accurate window on the coldplate temps near the core. To me, placing a thermistor so close to the core may cause the block not to seat properly and is an invitation to disaster. In the past I have relied on a relay to switch the pelts and pump on when the computer starts up. My current start up procedure is to start the pump, turn on the pelt PSU, and when the cold plate reaches 0C, start the computer.

 

[Colin]

It just occurred to me that I forgot one other person under the acknowledgements edit in my first post. Without Xjinn’s “Bong Cooler” this project would have been a real PITA. My sincere thanks to Steve Pignato for his cooling concept (i.e. gift to the overclocking world) and my apologies for not mentioning this at the beginning of this thread.

 

[Bender]

Is there any way to further reduce the amount of water evaporated by a bong without hindering its performance? Another solution could be to have a valve open to let in more water if the level got to low. Has anyone tried something like this yet. I would consider a bong if I could let the computer run 24/7 without refilling it with water.

 

[Colin]

Evaporation is the key to cooling with a bong. With a six gallon reservoir, I have no worries about running dry although I did have my eyes on a 17.5 gallon Igloo ice chest at Costco today. Not only does it have more water capacity but because of it's size, it will lower the height the pump has to deal with by 8" and improve the flow.

 

[Colin]

Temps have dropped 2C, probably due to the alignment of the particles in the thermal compound.

 

[Colin]

Update: Runnng stable at 12 * 143 @ 2.06 volts, 1718 MHz but with the increased voltage and speed, the full load temps are now 20C.

 

[William]

Bender (Jul 14, 2001 10:18 p.m.):
Is there any way to further reduce the amount of water evaporated by a bong without hindering its performance? Another solution could be to have a valve open to let in more water if the level got to low. Has anyone tried something like this yet. I would consider a bong if I could let the computer run 24/7 without refilling it with water.

i haven't tried this but if you could tap into a water line ala toilet valve you would be good to go!

 

[Colin]

The only problem with tap water is the build up of mineral deposites as the water evaporates. I would suggest filtering the water prior to adding it to your bong cooler.

Requested on another forum, here are the Sandra Bench Marks.

Memory - INT 615, FPU 652
CPU - Drystone 4825, Whetstone 2375
CPU Multi Media - Integer 9698, Floating Point 10764

 

[Bender]

Nice floating point Colin. Makes a P4 look like a toy.

 

[el]

Nice job colin. & Bender a P4 is a toy!!!!