• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

Quad G5 LCS overhaul - pump replacement/PWM dependant?

Byrd

Well-known member
Hi all,

I've been restoring a Quad G5 - well, two actually - the rev. 1 single pump (Delphi) and rev 2 dual pump (Delphi) LCS assemblies in two machines. First I'm going to overhaul the dual pump model which is easier to work with and more efficient in cooling (I've rebuilt the rev. 1 but found the blocks leak, it's shelved and I've had enough - new O-rings required).

In my Quad G5 rev 2 setup, one of the pumps has failed - the PCB internally smoked out. Replacement, aftermarket Laing DDC PWM controllers are available:


... however, both pumps are sealed with thick epoxy and there is chance of junking the enclosure when cracked open. Both pumps also look pretty rough externally. I've also read of others using the DIYINHK controller in their Quad but apparently the pump runs at full speed and doesn't cool the loop properly, it needs to slow down for the liquid to pass through the radiator and cool.

... ideally, I'd like to replace the pumps with new - there isn't much available, and Laing DDC branded pumps are really expensive (AUD $150+/pump), I found these substitutes (reportedly PWM compatible but also with standard molex connectors):



I suspect running these off the Apple pump harness won't be great, I'd be happy to run them off 12V tapped elsewhere, with a third party PWM controller to slow the pump down.

Does anyone know can the system behave normally without PWM controlled water pumps - can I calibrate the cooling so not to have the fans on turbo? Currently, one working pump on CPU #1 the system boots and chimes but CPU #2 (dead pump) quickly overheats making the system freeze, but fan speeds do not significantly ramp up.

Thank you

JB
 
Last edited:

Byrd

Well-known member
Sounds like the answer is: no-one really knows.

I've enquired about the DIYINHK replacement pump PCBs for the Laing DDC, their answer was "it may or may not work" :confused: so I'm going to purchase a few SC-P60A pumps, fittings and try my luck. These pumps will fit well and have a port for filling; anything that can replace the horrible OEM parts will be a help.
 

CircuitBored

Well-known member
both pumps are sealed with thick epoxy and there is chance of junking the enclosure when cracked open

Not really. The DDC pumps have a fairly intelligent, modular design. The epoxy used on the pumps in the G5 is only there to stop leaks, it doesn't have a structural purpose. Once you remove the screws on the bottom of the pump the top comes right off. The pump tops are actually spring loaded so you likely won't have to pry at it at all. It'll just pop right off.

Replacement, aftermarket Laing DDC PWM controllers are available:

Sadly these will not work properly in a G5 (as far as I am aware). It's been a few months since I've done G5 LCS things but off the top of my head there is a PWM sense pin missing from the normal Laing DDC controllers that will send the G5 into thermal panic mode if left disconnected. You can actually replace components on the pump PCBs to get them working again but it is an awfully fiddly job. It's likely that one of your pumps has seized as the pump PCBs are fairly reliable.

Does anyone know can the system behave normally without PWM controlled water pumps

Yes and no. I believe you can trick the G5 into thinking the PWM is connected with a bodge on the wiring harness. I need to check my notes on this one. There are quite a few resources online regarding this that I need to find again. Please bear with me, I'm currently at work!

You can calibrate the fans using Apple service diagnostics but you have to have the original fans installed for the program to work, otherwise you'll get "fan overspeed" or "fan underspeed" errors from the tool. If anything in your cooling system is non-stock the calibration will fail.
 
Last edited:

Byrd

Well-known member
Thanks for your input @CircuitBored !

I've been into PC water cooling for years, the Apple LCS Laing DDC pumps are truly horrid inexpensive components not designed for long term use.

While I've only three of them as sample size, two PCBs failing isn't great - both smoked out the pumps themselves were clean. The single pump LCS has the non-PWM DIYINHK PCB installed, which also failed.

Yes and no. I believe you can trick the G5 into thinking the PWM is connected with a bodge on the wiring harness. I need to check my notes on this one. There are quite a few resources online regarding this that I need to find again. Please bear with me, I'm currently at work!

You can calibrate the fans using Apple service diagnostics but you have to have the original fans installed for the program to work, otherwise you'll get "fan overspeed" or "fan underspeed" errors from the tool. If anything in your cooling system is non-stock the calibration will fail.

That's my concern - nice new pumps, low temps but with turbo fans. Currently though with a failed pump it does attempt to boot without ramping up fan speeds before the red overheat LED on the motherboard comes up.

If you do find anything on bypassing or tricking the PWM pump, please let me know. I've had a good hour of searching online and couldn't find much, only that Apple's implementation of PWM in fans is different to x86 standards. There are several PWM controllers online I could consider to cheat the system, perhaps.

I'm quite looking forward to getting a reliable Quad. I've been wanting to get one going for years, and was fortunate to find a parts unit recently. New pumps with accessible ports to refill will make them easy to maintain and trust!
 

Byrd

Well-known member
3 x Syscooling SC-P60A pumps purchased, and G1/4 barbs. Anything to ditch the OEM pumps.
 

Byrd

Well-known member
After some trial and error I have the Syscooling SC-P60A pumps installed - they are a similar design and output to the OEM Delphi pumps but shorter in height, they are mounted on ~ 25mm high density EVA foam blocks and adhesive foam to reduce noise and vibration. The Syscooling pumps are of good quality and sealed design.

With bench testing, so far so good - the pumps work quietly and while bleeding the system for air bubbles can take time, having a small inlet port to funnel in coolant in is rather useful. I've used a small fan controller and found they keep pumping without issue @ 3 - 12V, but are quiet so will just leave at 12V unless the loop requires more slower water flow for adequate cooling (assume analogy of pumping a garden hose through a straw).

Next up is to run them off 12V in the Quad and see if I can calibrate the Mac without PVM sense lines; these pumps are PWM but I'm quite sure Apple's implementation of this won't be the same.

Fingers crossed - I'm just happy to be gone of the Delphi pumps which really are junk. I've watercooled PCs for years (in the early days that was getting parts from the aquarium for my Eheim pump setup!) and these pumps are simply the worst in terms of design and you could never be confident that they wouldn't leak.

Pumps.jpg
 

CircuitBored

Well-known member
This is really great work - welcome to the radical G5 Quad custom cooling club! Your blue LCS is the yin to my red one's yang.

Next up is to run them off 12V in the Quad and see if I can calibrate the Mac without PVM sense lines

I am somewhat certain that you will get an error if you try to calibrate the pumps without PVM but I am rather curious to find out!

I've been meaning to do some work on pinning down how the Quad's PWM/PVM system actually works for a while and I think this is the motivation I needed. The end result I'm aiming for is some way to use generic cooling parts with the Quad's odd electronics. The way I see it there are two possible avenues worth investigating.

  1. Software-based: The AHT/ASD software must be interfacing with a programmable controller somewhere on the logic board that stores the PWM-temperature curves. This is likely just a set of values that kick into gear when certain thermal conditions are met. The pumps only run at two (maybe three) different speeds with the default firmware so I imagine these are just fixed values that the firmware can "choose" from. Fan control is likely done by way of a PWM%-per-degree ratio but it's also possible that it's just a table of values that the firmware incrementally puts into use as the temperature rises. I need to establish how the calibration software interacts with the PWM control hardware and how the values are stored. My idea for a software work-around involves figuring out how to manually set the firmware's PWM values and then writing a custom calibration tool.
  2. Hardware-based: The G5s have weird PWM signals, this is a fairly well-known fact. If you put regular PC fans into a Quad they will not be properly PWM controlled, some will even run slower as the G5 gets hotter - not good! AFAIK the pumps use "normal" PWM but the inclusion of a PVM "return" signal is unusual. My hardware-based workaround plan involves developing a device (or set of devices) that connect to the G5's PWM and PVM headers and "lie" to the system about what is going on. This would take the form of a microcontroller of some sort that has its own fan/pump connectors and acts as an interposer for more generic PC hardware. Weird G5 signals go in, regular PC PWM signals go out (and also the same in reverse for PWM-sense). That's the general idea.
First up I plan to properly get to grips with how the thermal calibration software works and what it is actually doing on a hardware level. I also need to take a close look at PWM and establish how the G5s' signals vary from the PC side of things.

Again, great work! Thanks for the inspiration.
 

joshc

Well-known member
You are brave people. Brave indeed.

"Consult product manual or an Apple Authorised Service Provider". Who is going to be the first to try bringing one of these into an Apple store and see what happens? :p
 

Byrd

Well-known member
@CircuitBored - Thanks for the additional information; the only way these beasts will survive, unlike other vintage Macs, is with moves to replace what's in there now and try to understand things like Apple's PWM implementation. I'm at pains to replace cooling components but that's what has to be done. I'm not familiar with coding but you have my greatest respect and support for looking into PWM workarounds; to the uninitiated would something like an Arduino be appropriate to alter the signals?

There was a person who removed the entire LCS on MacRumors and replaced with a custom welded huge heatsink, which still ran hotter at idle than water, but I recall had no issues with the LCS being removed once properly calibrated. AHT might return an error but whether or not it cares when it stays cool is the mystery. I think I'm going to run these pumps off the optical drive +12V molex connector with 4 pin fan extension cables neatly routed through the case; the PVM harness is right there but can see voltages fluctuating if I patch the pumps off those.


@joshc - thank you - this is one of those projects that has taken far too long and still might end up a wind turbine almost capable of hovering off the ground :) I've another Quad G5 with single Delphi pump (bought for parts) that'll get a similar treatment if any of this works.


@cheesestraws - I agree. The G5 represents a massive shift in technology but also a lack of open design.
 

CircuitBored

Well-known member
My bet would be that this would be an enormously easier way of doing it than mucking with the firmware of the machine

It'll undoubtedly be a nightmare to investigate but the end result could theoretically just be a custom calibration tool that doesn't care about PWM-sense on the pump and fans. That way the end-user experience would be largely the same as booting the normal diagnostics disc. You are correct though; the hardware-based route is the path of least resistance.

to the uninitiated would something like an Arduino be appropriate to alter the signals?

This is the part I really need to start pinning down. I've never worked with PWM signals beyond scoping them out to establish when they're present and correct. There are a whole bunch of resources about Arduino PWM controls that I really just need to sit down and read. Arduino might not even be the right way to go. I have a lot of time off this week so I will get to work on developing my PWM education and brainstorming ideas based on what I learn. What I really need is for someone more experienced than me to step in and say "this gizmo will do everything you need, try it" just to get the ball rolling.

AHT might return an error but whether or not it cares when it stays cool is the mystery.

Oh, you can run custom cooling setups without calibration no problem. As long as the machine stays cool it will run as normal. Issues only arise once you start trying to get the G5 itself to handle the control of non-stock fans or pumps.

There was a person who removed the entire LCS on MacRumors and replaced with a custom welded huge heatsink

Sacriliege! A Quad must be liquid-cooled. The idiocy/compromises of that moment in history have to live on (he said, tempting entropy's embrace).

I just realised I'm going to need to build a custom harness for the pump module so I can scope it out while the machine runs... that'll be fun.
 

Krizzo

Member
Any progress on finding out about the PWM signals? I'm planning on replacing the CPU exhaust fans with standard BeQuiet PWM fans but if they don't react properly to the G5's PWM signal I will have to run them at a fixed speed via the Molex power.

But as I think of it, this is where you could use an Arduino. Read the PWM signal from the Mac during different CPU temperatures and make the Arduino output a voltage accordingly. No need for a PWM fan, just make the Arduino a PWM-to-voltage translator. Or am I being naive here?
 

CircuitBored

Well-known member
Any progress on finding out about the PWM signals? I'm planning on replacing the CPU exhaust fans with standard BeQuiet PWM fans but if they don't react properly to the G5's PWM signal I will have to run them at a fixed speed via the Molex power.

But as I think of it, this is where you may use an Arduino. Read the PWM signal from the Mac during different CPU temperatures and make the Arduino output a voltage accordingly. No need for a PWM fan, just make the Arduino a PWM-to-voltage translator. Or am I being naive here?

I ended up focusing on finishing off another project before starting work on G5 PWM. The other project is pretty much wrapped up for now so I will be getting started with G5 stuff in this coming week! Sorry about the slight delay.
 

Krizzo

Member
Okay I did a little research on the fans and scrolled through hours of forum posts. Most sources are PC people who want to mod the fans to work with PCs.

It seems like the PWM signal that is outputted by the G5 Logic Board is a +12V PWM signal. The standard for modern PWM fans however is 5.25V max. If you were to connect a fan that is wired correctly to the G5s fan header wiring, you would still fry your fan – as soon as the pulse width exceeds 5.25V.

I suppose it's the same for the pumps.
 

Krizzo

Member
So I returned to my Quad to do an overhaul of the lcs again. I was shocked to find a leaking top loop after letting the computer sit unused for 7 months. The top loop was full of white gunk and the whole lcs was still smelling strongly of vinegar, even after I flushed it multiple times back in March. When opening the cable ties, the loop exploded on me, seems like the vinegar was gassing. No harm done though.

I completely re-did the lcs, I flushed it multiple times with clean water again. Then I interconnected both pumps, creating a single dual-pump loop. The temps are now well under control, the computer is running way more quiet.

Next thing I did was replacing ALL the fans but the two CPU intake fans. I'd like to keep the intake fans because they look very G5 specific, you can even get a glance at those fans from outside the case, which has a very distinct cool look to it.

Since I replaced the fans I can not run the thermal calibration. I purchased two PWM fan controller boards like these: https://www.amazon.com/-/de/dp/B07LBLWYZF/ref=sr_1_1?__mk_de_DE=ÅMÅŽÕÑ&crid=1ALI0J09XP821&keywords=fan+controller+board&qid=1669854449&sprefix=fan+controller+board,aps,195&sr=8-1 . The fans are now controlled by temperature sensors that I placed at the CPU and northbridge heatsink. So I basically installed my own temperature controlled cooling system on top of the G5s.

The pumps and CPU intake fans still run controlled, even without thermal calibration. I can't explain why. The Quad is super silent now in idle, you only hear the pumps whizzing quietly. However, if the computer decides to lose its calibration data and lets the CPU fans and pumps run at full speed all the time, I'll have to control them via the PWM board as well. Which will be a pain in the butt, because it means I have to replace the pumps with standard PWM ones. That's 90 bucks for the pumps if I buy cheap ones and it means completely redoing the loop again...
 
Last edited:
I have had one or both pumps fail on my G5 Quad, and I'd like to replace them with a modern LCS solution. How are you powering and controlling the Syscooling SC-P60A pumps? I was originally planning on splicing the pumps into the original wiring harness, but am not sure if they are pin- compatible? Or are you using a separate LCS pump control circuit?
 
Top