Variable speed pumps and flow rate monitors

[Guest]

I just realized AVSIM doesn't have an Overclocking section ... probably should given how well a good OC works for flight simulators.

I'm looking for some input on flow monitor units and/or variable speed pumps.  So far I'm looking at AquaComputer flow rate sensor and Koolance (but NOT limited to):

AquaComputer Flow Rate Sensor

Koolance Flow Rate Sensor

My current pumps are EK DCP 4.0 PWM (800 L/h) which is variable speed - used for MB/CPU/VRM and EK DCP 2.2 pump for GPU cooling (400 L/h).  Both pumps are working well, but in my quest for safety and optimization, I'm looking to enhance or replace.

Software controlled is nice to have, but I'd also be fine with a front face plate or some control unit (mounted on case) that isn't necessarily software controlled ... either approach is fine.

Primary objectives:

1.  Obtain optimal flow rates for peak heat transfer (too high a flow can cause cavitation and/or reduce heat transfer, too low a flow reduces heat transfer).

2.  Provide alarm in case float rates change (pump starting to fail and/or leaks)

3.  Quiet pumps with low electrical line noise and minimal power requirements.

I'm discovering that 64bit flight simulators tend to produce more heat than 32bit ones ... guess I should have expected that.

I'm not tied to the above sensors/controllers and/or pumps, all recommendations welcome ... cost irrelevant.

Also looking for a good "easy drain" system for flushing lines/rads/blocks (and appropriate flush solution) ... suggestions?

Cheers, Rob.

EDIT: this setup will probably go into a Skylake X (August 2017) and Volta (Oct/Nov 2017) build I'll be doing later this year.

[DaveCT2003 1,820]

Excellent research on the cooling parameters.  I learned about cooling, flow rates, etc. with nuclear power but the same info applies across the board. I've read some absolute garbage concerning this in the Corsair forums (I didn't care enough to set them straight).

Just a thought regarding overall protection... why not just use a thermal shutdown?

Again, great research Rob!

 

 

[Guest]

Hey Dave,

Yes, Nuclear Power cooling all relates ... heat exchange ... in my racing days I had race car engine cooling, air to water heat exchanger, oil to water heat exchanger, oil to air heat exchanger, transmission pumps etc. etc. ... all benefited from very specific controlled flow rates (be it oil, water, or coolants).

Engine oil also benefited from very specific pressure values across the RPM range (too high and too low were both bad) ... setting up a dry sump (multi-stage pumps) with oil scavenge and getting crank case pressure correct ... was a little more complex than what one finds in a PC but same concept applies.

I do have Thermal Shutdown enabled but I believe that's only for CPU ... I'm running a mono-block that covers VRM, CPU, PCH ... but it would be nice to have some way to monitor the output and/or trigger a shutdown ... say a leak causes pressure drop and set a pressure threshold to trigger a shutdown before the leaky fluid destroys components.

Cheers, Rob.

[Ted Striker 338]

For Primary Objective 1 I'm thinking that if you maintain a constant load on your CPU and GPU with test software and maintain a constant air temperature to your radiator you should be able to find your optimum pump flow rate by varying the pump speed and watching for the minimum CPU and GPU temperatures.  I'm still on a Noctua air cooler and have no experience with water cooling in computers, only the power plant industry. So feel free to correct me if I'm full of hot air. 

 

This statement did raise my eyebrows though:

2 hours ago, Rob Ainscough said:

I'm discovering that 64bit flight simulators tend to produce more heat than 32bit ones ... guess I should have expected that.

How much more heat percentage wise are you seeing and is it on both the CPU and GPU?

Ted

[Guest]

Good point Ted, software controlled flow rates based on load might be better than adjusting to a rate at max load (single reference point).  So far I think the Aquacomputer pumps and flow meters are the only ones I've found that fully support that with software one can download ... looking for more options/opinions.

As far as temp difference on my setup, 64bit increases CPU, VRM, PCH on average by 5-6 F ... my radiator fans are static set, so I've pumped them up one notch to compensate ... also required some voltage setting changes.  GPU has increased heat also, but that's not related directly to 64bit, something else.

Cheers, Rob.

 

[Ted Striker 338]

I see what you are after now. Reduce the pump speed/noise when the heat load drops.

5-6 degrees increase from 32 to 64 bit is significant. I guess that is the cost of all the goodness released by the 64bits.

Ted

[martin-w 3,234]

10 hours ago, Ted Striker said:

 

5-6 degrees increase from 32 to 64 bit is significant. I guess that is the cost of all the goodness released by the 64bits.

Ted

 

6 Fahrenheit... what's that in centigrade? Something like 3 degrees? It's such a low temp increase as to be insignificant isn't it. Or am I missing something?

[Ted Striker 338]

5 hours ago, martin-w said:

6 Fahrenheit

You are right Martin. I am so used to thinking of CPU temperatures in Celsius that I completely overlooked the F Rob wrote.

I think Rob is more interested in user feedback on control and monitoring equipment for water cooling systems of which I have zero experience so I won't derail this thread any further.

Ted

[TechguyMaxC 516]

In a closed loop flow-rate monitoring isn't really necessary.  If your flow rate drops your temps will rise.  If its dangerous the system will under-volt and under-clock.  You should notice the temp increase if the machine is in the same room.  

I've used both fixed and variable rate pumps over the years.  I always end up maxing out the speed on the variable pumps anyway as the noise is never going to be louder than my fans once I've properly bled the loop.  

So here's the thing about flow rate: more isn't always better.  There is a point where you're just generating extra heat rather than removing it.  If you want some data and analysis to do some more research, check out Skinnee Labs: http://skinneelabs.com/ and Martin's Liquid Lab: https://martinsliquidlab.wordpress.com/

[Guest]

7 hours ago, martin-w said:

Or am I missing something?

Agree with you that the change in temp isn't that significant in my setup ... but I'm running 5 thick radiators thru a monoblock covering CPU, PCH, VRM.  So for others running fewer radiators, or air only, or CPU only cooling they may run into more significant temp changes.

1 hour ago, TechguyMaxC said:

So here's the thing about flow rate: more isn't always better.  There is a point where you're just generating extra heat rather than removing it.

Agree on flow rate, but higher flow rate doesn't generate extra heat?  High flow rates can cause cavitation (small air pockets and/or bubbles especially thru the narrow passages found in the various cooling blocks) which reduces the efficiency of heat transfer.  Optimal cooling efficiency will have a very specific flow rate and it will vary based on radiators used, blocks used, etc. etc. 

Flow rate monitoring is important ... temps can raise based on load ... this is normal so one can't assume that temp increase is due to a drop in flow rate.  Monitoring pressure provides a level of safety and the ability to find the optimal flow rate at a given load.  Since load is where it's most important, finding the sweet spot of heat transfer.

Back on topic, anyone have experience with Flow rate monitors and/or software that will manage pump speeds under load?

Cheers, Rob.

[TechguyMaxC 516]

Rob, higher flow rate does generate extra heat because the pump is working harder (spinning faster) to generate that higher flow.  I believe I've seen testing of this at one or both of the sites I linked.  

Also, I'm not referring to normal temperature increases as a result of changes in workload, I'm talking about abnormal temps (i.e. hitting your CPU's thermal throttle threshold).  These temperature increases have noticeable effects on ambient temperature, as well as performance.