Watercooling

[Bozo]

Is your motherboard hanging on a wall outside?

 

[ddrueding]

No, inside. It rains here sometimes

 

[Pradeep]

Since you don't use cases, just mount the motherboard as the bottom component in a chimney, with the board upside down. Any condensation can go into a 'drip bucket'.

Peltiers are WAY cooler then water cooling;0

Problem with peltiers that I have found is getting a reasonably cost effective way to power them. From what I remember the bigger peltiers need a significant amount of current at 12 or 24V. A switch mode supply of that capability ended up costing more than the peltiers by a significant margin.

Using the peltiers to drop the temperature of the water is another option. Again can have probs with condensation occuring. The last thing you want is water under the cpu.

 

[ddrueding]

At one point I read quite a bit about them, but have forgotten most of it. My CPU is currently drawing under 150W, so a 223W peltier could do a decent number on it. I have that much headroom in my existing power supply

 

[ddrueding]

The loop now includes 5 waterblocks, and the outgoing temp is still nearly the same as the incoming. 1/2"ID tubing with this flowrate is ridiculous overkill

CPU->NB->5870#1->5870#2->SSD->Pump->Radiator

 

[CougTek]

CPU->NB->5870#1->5870#2->SSD->Pump->Radiator

Why in Hell are you watercooling a drive that dissipates something like 2.5W?

 

[LunarMist]

Wouldn't that be warming the drive? :spiderman:

 

[sdbardwick]

I think he's referring to cooling the integrated OCZ 512GB PCI-E controller/drive combo.

 

[ddrueding]

I think he's referring to cooling the integrated OCZ 512GB PCI-E controller/drive combo.

That would be it. Specifically the controller on the LSI RAID card; it was around 80C without direct airflow. It never crashed, but I like to treat my data better than that.

At the moment, the only thing in the 80C range is the RAM, with the portion of the chipset furthest from the waterblock around 50C. Everything else is great.

 

[BingBangBop]

With DDR2 you could get RAM with water blocks built in. Can you do the same for DDR3?

Checkout www.koolance.com. They have RAM water blocks for DDR3.

 

[ddrueding]

With DDR2 you could get RAM with water blocks built in. Can you do the same for DDR3?

Checkout www.koolance.com. They have RAM water blocks for DDR3.

I've been considering those for a while, but I don't think I need anything so complex. Just a waterblock mounted to the top of the DIMMs, pulling away heat eventually, would suffice. I've been told that Danger Den will be coming out with something simpler.

 

[time]

Why worry about the RAM? Manufacturers usually quote ambient operating temperature ranges of up 65 or 70C - the chips themselves may run at up to 95C, or in come cases, even 125C!

Think about it, even CPUs can be rated up to 90C.

Surely everyone realizes that RAM heatsinks are just for show?

 

[ddrueding]

I suppose you are right, Time. The only thing it would really get me is removing that heat from the room (once I relocate the radiator).

 

[Pradeep]

Why worry about the RAM? Manufacturers usually quote ambient operating temperature ranges of up 65 or 70C - the chips themselves may run at up to 95C, or in come cases, even 125C!

Think about it, even CPUs can be rated up to 90C.

Surely everyone realizes that RAM heatsinks are just for show?

Whilst most are indeed for show, I thought the heatspreaders common in FB-DIMMS served a functional purpose to spread the heat load of the AMB chip, to avoid meltdowns?

 

[time]

Whilst most are indeed for show, I thought the heatspreaders common in FB-DIMMS served a functional purpose to spread the heat load of the AMB chip, to avoid meltdowns?

Well, that may in fact be a JEDEC standard for FB-DIMMs, although I have no idea whether or not it's mandatory.

I get readings of more than 90°C (194°F) for the FB-DIM modules in my Mac Pro. Isn't this a bit high?

No, temperatures in that range are normal for FB-DIMMs and are no cause for concern. Each FB-DIMM contains its own controller, called "Advanced Memory Buffer (AMB)". An AMB also contains an internal sensor which measures temperature directly inside the chip. The specified maximum temperature range of an FB-DIMM, measured by the AMB, usually lies in the interval between 95°C and 125°C (203°F .. 257°F). It will vary from manufacturer to manufacturer, a typical value is 110°C (230°F).

 

[Stereodude]

And yet component lifetime is shortened by high temperature...

 

[ddrueding]

And yet component lifetime is shortened by high temperature...

That is more the thermal cycling than the high temps on their own, no?

 

[time]

And yet component lifetime is shortened by high temperature...

I guess that depends on how high and what sort of component?

I'm aware that capacitors and fan bearings have a life expectancy inversely proportional to their operating temperature. As ddrueding points out, where is the evidence that this applies to semiconductors (within normal limits, obviously)?

Electromigration is all I can think of so far ...

 

[blakerwry]

In effect, he is trying to heat the water, with the heat from the water block.

...

Exactly! The closer the water and the water block come to the same temperature, the slower the heat transfer. The idea is to regulate the flow to get the maximum heat transfer from the water block to the water.
If the water enters the water block at 50F and leaves at 55F, thats good. If the water enters at 50F and leaves at 60F, that's better. The idea is to regulate the flow to get the maximum heat transfer while still reducing the CPU temp. At some point the temperature differential will stop climbing and the CPU temp will stop dropping. He might even need to increase the water flow.

Bozo, I think what you are describing may be a matter of efficiency vs efficacy (effectiveness). You might find there are diminishing returns when running the pump at faster speeds, but there are still returns.

If one were concerned about efficiency (as is often the case with Heating/Cooling systems), you would try to minimize operating costs (which are primarily a function of pump speed) while providing adequate climate control. The operating costs for a system of this size are negligible, so getting the best performance (at a reasonable level of noise) is the concern.

I like DD's initial example: What one molecule doesn't pick up, the next will.

 

[Stereodude]

That is more the thermal cycling than the high temps on their own, no?

No. Take a look at the Arrhenius equation and how it pertains to semiconductors and accelerated life testing at high temperature. This is a good paper explaining some of it.

 

[ddrueding]

Wow. 95% of that is over my head, but jumping to table 3 does give some good info. Thanks!

 

[time]

According to Table 3, MTTF for an example SRAM chip was a bit less than 6,000 years.

The Arrhenius equation is just a model, and for chemical reactions at that. It's relevance to real life electronics failure rates is a moot point. The sort of failures that it predicts well are too rare to be of much practical use.

According to the model, running your computer in a freezer at -20C instead of in a room at 20C ought to make it 16 times more reliable, so the chance of failure is almost eliminated.

Of course that's bollocks.

Check out We still have a headache with Arrhenius.

 

[ddrueding]

I'm already assembling the third version, but here was the config for the last few weeks.

 

[ddrueding]

Please let me be smart enough to not try to use the built-in waterblocks on a motherboard again. ASUS or Gigabyte, their quality is clearly not good enough and led me to cleaning Titan X cards with alcohol.

That said, I suspect the radiator is large enough 3x3 120mm.