The new Koolance Controller, TMS-200!

[biggles1]

Me myself would never buy an atomiatic controller based on a software solution. Been down that road to many times. Suddenly the manufactuerer dicedes to drop development and you're stuck with whatever OS you have. And what about running another OS then Windows? Sure, quite a few out there can't see themselves running other OS:es today, but in just a few years the market will be very different.

This is beside the fact that you are totally dependent on the OS funtionality. Say for example that you have en electrical error, caused by lightning, faulty wiring in your building or just a bad power supply. You OS could easily become unstable and hang long before your pump dies. Then you are toast. An OS-independet device with micro controller on board (like Aquaero or mcubed) will not fail in this situation...

[ecat]

If you set your fans to CPU temps then your fans would be porpoising up and down as you navigate thru the PC applications and the CPU varies its temp drastically. I work with CAD and keep RealTemp running in the tray as I work I see now and again my cores go to 48°C then back to 33° in just seconds, if the fans were set to monitor this they would do the same and the water temp would actually rise as the fans lowered with the CPU cycle. If you set the fans to operate according to water temp than as the CPU heats up the water the fans ramp up and try to maintain the set equilibrium temp and will ramp up or down until satisfied. You are basically monitoring the water temp not the CPU.

This depends on how the temperature control is implemented:

If fan speed is directly proportional to temperature then the system behaves as you describe.

If you sample the temperature once per second, which is plenty fast enough, and you average over 60 samples then even a step change in temperature, running Prime say, will result in a gradual increase in fan speed. Short blips in cpu temperature will be masked by the sample size.

On my main system I use T-Balancer control with reference to a sensor mounted on the rad, it works well enough but with a large over-clock it can be a fine balance to ensure the cooling power (fan speed) is there when the cpu is placed under load - the water temperature just doesn't change fast enough, in effect acting as an averaging calculation with a large sampling period.

On my new system I've been playing with a diy controller. When the controller has been out of action I've had the 3 Nexus fans daisy chained off the MB cpu header (well within spec, I double checked ) ie controlled directly off cpu temperature by whatever algorithm is in the BIOS. There is little sign of porpoising. When I take the cpu to the max over-clock the fans respond in a timely fashion and the system is more stable.

So, for general use and a quiet system using water/rad temperature works well enough. For a compromise between noise and performance I do see direct control from actual cpu temperature as the way forward.

[alacheesu]

Nice product, but I rarely run windows so it's out of the question for me.

[Brodholm]

Nice product, but I rarely run windows so it's out of the question for me.

Yeah, I get you. But Dean said that it changes fan speeds after he disconnected the unit so there is still hope!

[fox3]

If you sample the temperature once per second, which is plenty fast enough, and you average over 60 samples then even a step change in temperature, running Prime say, will result in a gradual increase in fan speed. Short blips in cpu temperature will be masked by the sample size.

If I follow this logic to its conclusion then all I am doing is allowing the water temp to rise. Consider this: The CPU spikes and the fans do nothing being masked by sample size but heat was absorbed, the CPU spikes again and again heat is released into the water and the fans do not see this being masked by sample size. Continue this cycle and the temp of the water raises and the CPU spikes will become higher proportionately. Warmer water is not the best cooling solution IMHO.

[alacheesu]

Yeah, I get you. But Dean said that it changes fan speeds after he disconnected the unit so there is still hope!

Maybe, but I personally won't consider it until there is at the very least an API to work with for linux users. Aquaero seems to have a relatively mature open source utility, making it the better choice at the moment, despite its other flaws.

[ecat]

If I follow this logic to its conclusion then all I am doing is allowing the water temp to rise. Consider this: The CPU spikes and the fans do nothing being masked by sample size but heat was absorbed, the CPU spikes again and again heat is released into the water and the fans do not see this being masked by sample size. Continue this cycle and the temp of the water raises and the CPU spikes will become higher proportionately. Warmer water is not the best cooling solution IMHO.

Well the process does not quite conclude there

You get a blip, averaging ignores it. You get another a blip, averaging ignores it too and so the sequence repeats and the water temperature rises. This rise in water temperature results in a slight rise in cpu temperature, this time it is more than just a blip so it doesn't vanish in the averaging routine, the fans speed up, the water and cpu cool down.

I was using the idea of an averaging routine to mask out the possibly continual changes in fan speed that result from deriving the speed directly from the temperature of the cpu, a temperature which can jump around in normal ie not fully loaded desktop use. The good thing about averaging, or smoothing, the cpu temperature in software is that the algorithm can be written to behave in any required fashion.

It's worth noting that the water in the cooling loop behaves in much the same way as the averaging routine mentioned above. Short bursts of cpu activity cause small blips in cpu temperature. The water is flowing at speed so a short temperature blip has no measurable effect on the water temperature, which is exactly why there is no porpoising. Another blip and then another and slowly the accumulated effect does cause a temperature rise that the probes can detect, only then does the controller change the speed of the fans.

The interesting case is when the cpu goes from idle to sustained load. Using cpu temperature and a software algorithm this change can be identified quickly and fan speed increased accordingly. When controlling from the water temperature using a rad probe, or even an in-line probe, it takes a relatively long time before the temperature reaches a level that demands the maximum fan speed.

The problem with all this talk of controlling to cpu temperature is getting the temperature in the first place. Any answers ? I was thinking along the lines of measuring the power demand or changes in power demand of the cpu using something like a current clamp but I have no idea if this is possible with nominally DC current

[fox3]

Well the process does not quite conclude there

The interesting case is when the cpu goes from idle to sustained load. Using cpu temperature and a software algorithm this change can be identified quickly and fan speed increased accordingly. When controlling from the water temperature using a rad probe, or even an in-line probe, it takes a relatively long time before the temperature reaches a level that demands the maximum fan speed.

This is not correct. I ran Prime95 for an hour last night and graphed the water temp and fan rpm in real time. The water temp rose and the fans followed pretty much as quickly as the CPU cores rose in temp. The RES lagged behind the RAD then cooled quicker than the RAD as the graph shows which is a little curious to me as I would have thort the RES would have followed the RAD but here is the graph and you can see how directly the temps and fans followed the CPU temps. I started the run at 10:07 and ran until ~ 11:06 so you can use those times as CPU 99%:

[Snyxxx]

Nice Graphs fox3. I cannot wait to get my Aquaero up and running.

[ecat]

Very nice graph, looks like very nice software

Red, the line that goes highest, is the rad ? Blue, the other square-ish line is the res ? The two wiggly lines that look the same colour to me are the fans ? I'm colour blind btw Is it possible to get cpu temperature on the graph ?

[ecat]

My graph is not as professional as yours but to get the info I wanted I had to use Samurize to chart the values from SpeedFan and the T-Balancer



The delay I'm talking about can be seen in the chart. At 1) I started prime on all cores, the first visible sign of a temperature change at the rad occurs some thirty something seconds later at 2). A probe with better resolution may cut this down to 15 seconds, or 10 or less, and possibly introduce porpoising to the control as a result. It doesn't matter, the chart shows the rad lagging behind, the mass of the water acting to average/smooth out changes in cpu temperature.

The same effect but in the opposite direction can be seen between 3) and 4).

5) Was the result of luck. Windows decided to do 'something', the cpu trace shows one of the blips we talked about earlier, quite a long one in fact. As predicted rad trace shows no measurable effect, but the water in the loop must have gained some heat as a result.

Anyway, if nothing else the quality of the Aquaero software alone makes it look worthy of investigation, sure beats the hell out of hand drawing screens in Samurize

[Snyxxx]

Anyway, if nothing else the quality of the Aquaero software alone makes it look worthy of investigation, sure beats the hell out of hand drawing screens in Samurize

Thanks for your trouble to get graphs like this. I also have the bigNG / T-Balancer and have moved on solely for the software side of the Aquaero.

I feel your pain brother.

[Brodholm]

Any news yet Dean?

[NaeKuh]

guys sorry to tell you guys, but i wasnt able to pick one up this weekend.

i had an additional 200 dollars in fittings that needed to be bought for my bench station, so the controller might be on hold til after i get my classified. :X

:X

[DeanB]

Any news yet Dean?

So it seems that fans do not ramp according to temperature before windows starts, but auto-shutdown does work.

Also form the people saying it doesn't have a microcontroller, they are located on the backside

[DarthBeavis]

I am not keen on a program that runs on Windows to do shutdowns when something goes wrong. With my luck Windows would lock up and freeze, frying everything. Other then that I think it's a nifty product that can have it's uses.

it runs on a separate Linux partition

[Brodholm]

So it seems that fans do not ramp according to temperature before windows starts, but auto-shutdown does work.

Also form the people saying it doesn't have a microcontroller, they are located on the backside

So if i have undestood corectly. The fan speeds doesn't change before the os i is loaded. But the autoshutdown works even if the computer is hanged or in bios? So why can the auto shutdown take temperature reeding's but not the fans?

Well if auto shutdown is working even after a crash it is as safe as aquaero or mcubed. The only thing missing is shutdown due to rpm values. Are you giving this any planing?

And also, how does the fans behave if they are at, lets say 5 volt and the computer hangs. Does the controller continue to set 5 volt or does it change it back to 12volt? And does the same thing happen to the pump?

[DeanB]

These are all very good questions and suggestions. I am forwarding them all to the programmer for review. I'll have more answers to these questions later.

Also I want to point out that it is possible to run two of these cards, please see the tutorial posted here.

[Waterlogged]

These are all very good questions and suggestions. I am forwarding them all to the programmer for review. I'll have more answers to these questions later.

Also I want to point out that it is possible to run two of these cards, please see the tutorial posted here.

Dean, please forward to the programmer that Linux support out of the box would be greatly appreciated as well. I have 1 box I'm currently kludging together an autonomous cooing system and I'm looking to convert more of my boxes to Linux in the not so distant future.

[alacheesu]

Dean, please forward to the programmer that Linux support out of the box would be greatly appreciated as well.

I would be very interested in that as well.

[TedShred]

any plans for a model combing a 5.25" bay LCD display?

[DeanB]

any plans for a model combing a 5.25" bay LCD display?

I was asked that by someone else today as well. I think adding a display would require a substantial change to the hardware itself... I think would be a feature more likely to appear in the next version of the product. But I will request that functionality .

Linux compatibility is one of the most requested additions to the software, I'm really not sure what the plans for it are though.

[OdinEidolon]

I was asked that by someone else today as well. I think adding a display would require a substantial change to the hardware itself... I think would be a feature more likely to appear in the next version of the product. But I will request that functionality .

Linux compatibility is one of the most requested additions to the software, I'm really not sure what the plans for it are though.

Yeah, for us linux users its a pain with this kind of application, they are often good only under Windows. Once this become avaiable under linux, i'll grab it in a second.

[Brodholm]

This is very good news!

If linux support and rpm shutdown is implemented. I think that this unit is superior to the aquaero and the mcubed. It offers what they cant give. Aquaero is lacking power and mcubed is lacking in support.

It doesn't matter if it isn't totally independent when it is failsafe with the shutdown still working outside of the OS.

You should also try adding shutdown from gpu, cpu temp if this is possible. I know this is a PITA for the manufactures but it would be great. I mean the crystalfontz can do it =)

You should post some more information on your site. Things like each fanheader can take 2A and stuff like that. And a total of 6,5A etc.

Edit: Is it possible to control Cathodes and leds from this unit?

[JOCKTHEGLIDE]

how many fans can this control again?