View Full Version : Temperature probes...
ctsoccer13
06-06-2009, 07:40 PM
I currently have a bigNG and used the thermal tape to mount the digital sensors onto the CPU and GPU waterblocks. I wasn't very pleased with the accuracy or the sensitivity of them, so I added an analog sensor to each. The analogs were bit more sensitive to changes in temperature, but not enough to accurately form a cooling curve.
With that said... What's the best way to mount temperature probes to a CPU and GPU waterblock?
CT
faster3200
06-06-2009, 09:58 PM
BP temp probe (http://www.jab-tech.com/Bitspower-Silver-shining-G-1-4-Temperature-Sensor-Stop-Fitting-BP-WTP-CT-pr-4345.html)
ctsoccer13
06-07-2009, 10:55 AM
Do these work better than the thermistors that came with it?
With a watercooling system, it's generally better to base your fan curves off the coolant temperatures than the component temperatures. Otherwise you'll have your fan speeds spiking up and down all the time.
Do these work better than the thermistors that came with it?
They are probably much the same.
I've never come up with a truly satisfactory method of measuring the cpu temp. directly, always to much lag between the measurement and reality. Come on motherboard manufacturers, stop forcing to run the likes of SpeedFan and give us a little direct 2 wire interface into the temp. information so our fan controllers can do something meaningful without being at the mercy of Windows ! At the very least this would give us another set of numbers to obsess about :D
At the moment I let Asus Q-Fan control my rad. fans, it's not ideal but it is a workable compromise and it's not pwm. I did build a little electronic circuit that connects to the mb cpu fan header and interfaces the Q-Fan signal, which is roughly proportional to cpu temp., with my t-balancer. This works quite well over a small temperature range but the t-balancer software is not sufficiently flexible to allow me to try the control philosophy I was after...
With a watercooling system, it's generally better to base your fan curves off the coolant temperatures than the component temperatures. Otherwise you'll have your fan speeds spiking up and down all the time.
Loop temperature can lag quite far behind actual cpu temperature, it can also vary over such a small range that control becomes an exercise in all off / full on. When controlling off the cpu a very small amount of averaging will prevent spiking, at least it works for me.
The control I really want, or should that be used to want, is a mix of cpu temp. and water temp. such that my fans respond in a timely fashion to increased cpu activity but they also respond to elevated loop temperature, for example, running prime the fan speed increases in response to the cpu but after stopping prime the fans run on until the loop temperature drops to something approaching idle levels.
It's been a while since I pushed a cpu to the absolute edge for 24/7 operation but when I did having the fans respond to cpu temp. was the only way to guarantee stability.
Loop temperature can lag quite far behind actual cpu temperature, it can also vary over such a small range that control becomes an exercise in all off / full on. When controlling off the cpu a very small amount of averaging will prevent spiking, at least it works for me.
The control I really want, or should that be used to want, is a mix of cpu temp. and water temp. such that my fans respond in a timely fashion to increased cpu activity but they also respond to elevated loop temperature, for example, running prime the fan speed increases in response to the cpu but after stopping prime the fans run on until the loop temperature drops to something approaching idle levels.
It's been a while since I pushed a cpu to the absolute edge for 24/7 operation but when I did having the fans respond to cpu temp. was the only way to guarantee stability.
Heh, that's exactly why I prefer to base my fan operation on loop temperature. Fact is, even if you crank the fans in response to CPU temperature, the increased fan power isn't going to affect the CPU directly anyway. So whatever temperature the CPU was going to spike to, it will still hit that temperature. Whereas elevated loop temperatures mean that either 1) something's been under load, or 2) your ambient temps have gone up, which means that there's something for the increased fan speeds to work against.
There's no perfect solution, but having the fans respond instantly to increased CPU load doesn't really help much, I find.