Overclock.net banner

[UPDATE 28-Dec] [Other] A new perspective on AIO watercooling

618 views 9 replies 2 participants last post by  LaGsHoCk 
#1 ·
About a month ago I decided it was time for a new computer. I don't need anything next level powerful and I've always liked the IDEA of an AIO desktop, but weak hardware always held me back. When lenovo announed the Y910 I picked one up. 380W power supply, full desktop graphics AND standard desktop socket (I thought, more on that later).

FIrst thing first, this bastard was LOUD. Like jet engine loud. You can see below the only fans in this computer were the cpu and gpu blowers.



With all this full desktop hardware, I got the idea that I could being the noise level down a few notches with a watercooling loop. I placed an order with frozencpu and after three separate shipments and about a month of waiting I finally had everything I needed. At least, I thought I did. . . turns out that desktop cpu socket wasn't quite exactly standard.



So a normal waterblock was not going to work here. Luckily the XSPC waterblock I purchased uses a mounting system that allowed me to make this.



With that installed plugged in the R9 nano I picked up off ebay and sorted out the rest of the hardware.

Before



(a very long time) After



The fans have been wired up, along with the pump, to a cable made from a 6 pin extension cable. I also experimented with powering the pump and fans off the SATA port. This worked just fine, but I decided to go this route to allow for storage expansion down the line. And this way the left side of the case is largely unmolested.

I ended up butchering the right side cover badly enough that I decided to leave it off for now and revisit that down the road (read: never).

Overall I'm happy with how it looks, but there is a major shortcoming with this system as it sits now.

My original plan was to use an off the shelf DC motor controller to slow down the pump as there aren't any standard fan headers on this board. I quickly learned that this does not work AT ALL. So with the only PWM controlled fan header belonging to the CPU (and using a very tiny 4 pin connector) I started thinking about how I could use that.

As with most OEM computers this one throws a fit when the cpu fan isn't connected. My idea was that I was going to hook up the blue and green speed control wires from the pump to these pins and 1) utilize the motherboards factory setup to ramp up and down the pump, 2) hopefully do away with the bozzing noise that comes out of a mcp35x at full steam and 3) avoid the need for any complex trickery to get the system to boot without its precious cpu fan.

Unfortunately this didn't play out like I hoped. While the computer will still boot, albeit with a 'fan failure' screen and some ear splitting beeps, the pump is still running at full tilt all the time. Back to the drawing board. *UPDATE* This was resolved in post #6

I've got a 4 channel scope in the mail for an unrelated project, so my next step is to rewire the factory fan and watch the signals speeding through at startup. I'll program a microcontroller to reproduce these signals for me and skip the previously mentioned beeps. Hopefully along the way I'll figure out speed control for the pump as well, If I'm lucky I'll even be able to handle that with the same IC that is working on the cpu fan spoof.

That's all down the road though, for now I'll just turn my headphones up and make ample use of the 'hibernate' feature that I've never had a use for. I have no idea how to do any of the things I described above but it should be interesting to figure it out. I've seen an old thread on another forum that gave me a good starting point so I'm not flying completely blind.
 
See less See more
5
#2 ·
with a couple of 555 you can build a PWM controller for the pump
with a manual knob, but when set the rpm you don't need to change continuously

some mobo also want the negative pin connected otherwise they behave as you described

I would hook a simple fan to the connector so the mobo will be happy
a very slow pwm fan so it will never scream, there is always place for a cooling fan...
also
you can parallel connect a pwm booster (another pair of 555) to control the radiator's fans
 
#6 ·
I resolved the pump pwm and cpu fan error issues. The problem was, as it usually is, something shockingly simple.

So I made a circuit using a 5v trinket from adafruit (my goto microcontroller formy projects, eveything I make is simple so this board has plenty of IO for me and is compatible with most arduino libraries) to output a 25kHz PWM signal with an adjustable duty cycle. Credit to Marcelo Aquino, his code saved me a lot of time in reading through the ATtiny85 datasheet. Look HERE if you are curious.



Now once I verified everything was outputting what it should I wasn't getting any change from the pump but I was seeing a second square wave at the same 25 kHz but 2.5 V. With no other ideas I swapped the blue (control) wire for the green (sense) wire and behold! The pump was responding as expected to the PWM signal.



Channel 2 is the sense signal from the pump, I have no idea why it looks like that or is so noisy but it is likely either 1) I have no idea what I am doing with my new scope or 2) the signal is strange for the same reason the blue and green wires are switched. Everything is working so honestly I have very little interest in digging into that further.



At this point I thought "is THIS why using the CPU fan header didn't work before?" Sure enough it was, the picture above shows the motherboard outputting a 30% duty cycle (the 'standard' minimum per the document chimico9 showed me) and the pump has slowed right down to match. Also the boot error is gone as well. Yeah!! A stress test showed that as the cpu temps went up, the pump sped up and brought temps right back down to the low 40s. Perfect
smile.gif


The back of my computer is a wire nut cluster f at the moment, so the next phase of this project is to clean up the wiring. All thats left then is to figure out what to do about the butchered panel and I'll be able to call this one done.

I don't expect that any of this will happen soon as I can't SEE the wires while I'm using the computer. One of these days though.
 
#10 ·
Swapped out the R9 Nano for a 1070 mini. Since this is one of the GPU models that was originally available with this computer (the only model available currently) all of the refresh rate issues I was having are resolved.

This card is much larger than the Nano in every direction, and the large overhang of the waterblock isn't doing me any favors. It'll look great down the line when a few LEDs make their way in there so it was worth the extra effort to not remove it. On the plus side, there is so little room around the card that there was no need for adhesives to secure it in place.

I also cleaned up the, ahem, window. Down the road I plan on adding expanded metal mesh to break up the large open space.

 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top