Build Log: The Manhattan Project

[Bastiaan_NL]

That's a nice idea, I'm interested in what that backplate will look like once it's finished!

And about that nun, I prefer the friendly looking types

[EHW Ai]

Getting there (just an update):

I'll probably go over it with GORD's one more time and call it ready for mounting. Going to track thermals and flow using this setup with the Hydro Copper block, then bump them against the readings I take using the Optimus Active block set.

[tictoc]

That is an epic backplate.  I cheaped out and just painted the backplates on the Radeon VIIs in my workstation.  Although, they don't really need much cooling on the backplate (gotta love HBM). 

Edited August 2, 2021 by tictoc

[axipher]

5 hours ago, iamjanco said:

Getting there (just an update):

I'll probably go over it with GORD's one more time and call it ready for mounting. Going to track thermals and flow using this setup with the Hydro Copper block, then bump them against the readings I take using the Optimus Active block set.

 

[Bastiaan_NL]

I love copper

It's almost sad you'll have to change it for the Optimus active backplate.. 

[EHW Ai]

Some KPE 3090 backplate eyecandy...

 

MP5WORKS block mounted:

 

 

After I've waited a couple of days and tightened the screws further, the screw holes will be filled with thermal putty so that they're flush:

 

 

Thermal pad between copper backplate and water block:

 

 

Last but not least, elbow grease:

 

 

 

[EHW Ai]

As a prototype, it's good enough for government work:

 

 

If I were going to mass produce this backplate, I'd make the following adjustments: 

 

• I'd shave a bit off the copper by the NVLINK cap upper left screw (right now, that's the only thing that didn't line up perfectly), and;
• If going full cover thermal pads, I'd leave out the slots and EVGA logo.  

 

Otherwise, it should work nicely with the Hydro Copper block until the Optimus Active block arrives. The NVLINK cap is secure, and leaving the one screw out won't hurt anything (except perhaps someone's OCD). It's really not off by much.

 

 

 

 

[Avacado]

18 minutes ago, iamjanco said:

As a prototype, it's good enough for government work:

 

 

If I were going to mass produce this backplate, I'd make the following adjustments: 

 

• I'd shave a bit off the copper by the NVLINK cap upper left screw (right now, that's the only thing that didn't line up perfectly), and;
• If going full cover thermal pads, I'd leave out the slots and EVGA logo.  

 

Otherwise, it should work nicely with the Hydro Copper block until the Optimus Active block arrives. The NVLINK cap is secure, and leaving the one screw out won't hurt anything (except perhaps someone's OCD). It's really not off by much.

 

 

 

 

Amazing work my friend. Nothing gets my heart pounding like a large hunk of polished copper. 

 

Maybe you could still mill the EVGA logo and vent, but don't do it all the way through. 

Edited August 11, 2021 by Avacado

[schuck6566]

2 hours ago, Avacado said:

Amazing work my friend. Nothing gets my heart pounding like a large hunk of polished copper. 

 

Maybe you could still mill the EVGA logo and vent, but don't do it all the way through. 

I was gonna suggest the same with the logo,maybe even off set it with a paint that matched the coolant.

[Bastiaan_NL]

Looks really good @iamjanco!
And same as mentioned above, engraving the logo would look good too.
 

[EHW Ai]

Great ideas, guys -- thanks!

 

 

[EHW Ai]

So OCD won out (the NVLINK cap alignment was actually a very easy fix, and was only off by a fraction of a millimeter). That screw is now installed:

 

[EHW Ai]

Spurred on by a comment over on OCN about providing better pix, I decided to hook up the new CamRanger 2 I purchased a few weeks back, to my Nikon D810 mounted on the Polaroid MP4. I didn't break out my studio flash equipment (yet), and while the lighting from the MP4 suffices for now, it's direct and harsh, and it exposes every little nook and cranny in the "polished" copper:

 

 

The MP4 setup:

 

 

 

Love the CamRanger 2, btw:

 

[EHW Ai]

Additional progress and some notes (caveats) about the backplate:

 

 

So I changed my mind midstream and decided to use the dual MORA3 setup to cool the bench. I set aside the three white radiators I have mounted to a CASELABS SMA8 pedestal for now, as things were just getting too awkward, given my workspace:

 

 

Anyway, about that backplate...  While it looks nice, my design is flawed because I should have shifted the MP4 further to the rear of the backplate such that the cooling ports on both items were aligned vertically. Had I gone with the bungee cord approach to securing the block to the backplate, the problem wouldn't exist as I would have been able to accomplish that easily. Hard-mounting the block prevents that though and also prevents using the card in the first slot of the motherboard because of the location of the ram sticks. 

 

Thought about bending some hard tubing and going that route, but doing so wouldn't really resolve the placement issues and figured why bother now.

 

...consider it a durr moment, one of my better ones. But hey, the backplate was my first backplate, a prototype afterall (when in doubt, rationalize), and will do until the Optimum PC active block arrives.  At least I was fortunate enough that I had the parts on hand to rig a connection between the backplate and the block. 

 

Next up: prep for leak testing and adjustments if needed. I'll use the dual pump setup that's mounted to the lower left of the bench frame to do that, then finish the two loops.

 

 

 

 

 

Edited August 21, 2021 by iamjanco

[ENTERPRISE]

It really is amazing work. Im jelly of this setup!

[EHW Ai]

5 hours ago, ENTERPRISE said:

It really is amazing work. Im jelly of this setup!

 

Thanks, E 

 

Then you might enjoy these as well:

 

 

[ENTERPRISE]

3 hours ago, iamjanco said:

 

Thanks, E 

 

Then you might enjoy these as well:

 

 

I love it bud, so authentic. Plus we all know that having EHW on your monitor makes it that much better 

[Bastiaan_NL]

Sorry to hear about the placement issue... That sounds like most of my first tries at anything, though the backplate looks really cool!
The bungee cord might be easy to mount the block, but the way you did it looks way better to be honest. I think that with a few fittings it wouldn't be a big problem.
Anyways, it looks like an industrial installation with all the stuff around it

 

19 hours ago, iamjanco said:

 

I've got those flexible arms for the fans mounted on my workstation just like you did, and I'm really happy with it!

 

[EHW Ai]

On 21/08/2021 at 17:04, Bastiaan_NL said:

Sorry to hear about the placement issue... That sounds like most of my first tries at anything, though the backplate looks really cool!
The bungee cord might be easy to mount the block, but the way you did it looks way better to be honest. I think that with a few fittings it wouldn't be a big problem.
Anyways, it looks like an industrial installation with all the stuff around it

 

I've got those flexible arms for the fans mounted on my workstation just like you did, and I'm really happy with it!

 

 

Thanks for the comments!  Glad those fan brackets worked out for you. For what they do, they do it well. 

 

Anyway, getting closer to running coolant through it:

 

 

 

 

Changes I made since my last post in this thread:

 

I took the AquaComputer Flow 400 flowmeters out from where I had them mounted near the dual pump setups connected to the two MORAs as they were just too close to the pumps. I might add them back in at a later date, but I'm really not too worried at this time because I'll take flow readings using the King Rotameters;

 

I removed the drain setups I had between the same pumps and the QD connections to the test bench. I really don't need them now;

 

I hooked up the CPU and GPU loops, and added temp sensors before and after the GPU/block combo; and one to the outlet port on the CPU block.

 

That's it for now. 

 

[EHW Ai]

So I've been busy troubleshooting a flow problem that drove me up the wall for an hour or two. Everytime I'd start up the pumps, flow would peak at about 1gpm, then drop to a fraction of that, regardless of which flow meter i had hooked up. Turns out the bitspower inline filters i had left in the loop were clogging up the works. Took them out, and flow bypassing the rads was about 2.1gpm, which dropped to 1.8gpm when i added the rads in series into the equation:

 

 

 

 

 

 

 

You can see the inline filter in that third pic.

 

Mind you, that was running two Watercool pwm d5s in series without running a pwm signal to them. Didn't bother connecting the pwm lines to 5VDC (could be they were running at 60%, but I'd have to verify that to be sure).

 

I'll split the rads apart now that i know flow is working properly and run the loop gpu block and rear block -> rad -> cpu -> rad.  Once i power up the motherboard, i'll be able to verify if more can be squeezed out of the pump setup.

 

 

 

Edited September 7, 2021 by iamjanco

[EHW Ai]

flow dropped to about .8gpm with the reconfigured loop. we'll see if I can't bump that up once leak testing is done.

 

pumps -> kpe hydrocopper block + backplate block -> rad1 -> cpu -> rad2 -> flowmeter -> res, through a gazillion convolutions.

 

 

more eye candy while the leak test runs its course:

 

 

 

 

 

 

Edited September 11, 2021 by iamjanco
added eyecandy

[EHW Ai]

So the latest iteration of the bench looks like this:

 

 

The following changes have been made:

• Swapped out the WaterCool Ind. Dual Top and 2x WaterCool PWM D5 pumps; replaced those with an AquaComputer Dual D5 Top and 2x NEXT D5 pumps;
• Swapped out the large KING 0-3.8GPM Flowmeter (Rotameter); replaced it with a smaller KING 0-2.0GPM Flowmeter;
• Tightened up some of the component level coolant lines, replacing a few that were TYGON; replaced those with 16mm OD Bitspower Crystal Link acrylic hard tubing;
• Added a support bracket to the bench for the QD disconnects between the radiators and the components being cooled by them.

While far more testing would have to be done to state anything conclusive, my gut feeling about the dual D5 pump setups (Watercool vs. Aquacomputer) is that they're similar in flow outputs (that's a guess based on my observations). Rearranging the coolant lines and repositioning the radiators did seem to give me a good deal more head pressure though, and air in the system cleared out far easier/quicker than it did previously. 

 

More pix follow:

 

 

 

 

 

Edited September 26, 2021 by iamjanco

[axipher]

The bench needs some black paint, but leave all the hardware bare metal

[ENTERPRISE]

Certainly a sweet setup. You are using the same res and pump combo as I, the D5 NEXT pumps are great.

[Bastiaan_NL]

Looking good @iamjanco 

It must be nice to have all those QD's. I decided not to go with them in my build, because even inside one of the 'biggest' cases I could find there is still not enough room.

Will you add some color to the coolant or do you keep it clear?