The NRZ N2 Waterblock Beta Thread

[thenrz]

I started designing about 25 years ago on old school paper and pencil but have evolved with the industry. I've used Catia since V4M and Unigraphics (now called NX) since version 11 back when GM/EDS still owned it. My shop does a wide variety of things but, for what I'm involved in, is mostly prototype stamping dies. We used to do a lot more production dies but that work has flown overseas with the majority of the industry.

Ah... Catia. The fine French Burgundy of drawing software. Heard and seen but never touched. With the China thing, my industry is actually kind of bouncing back to USA made even at a premium. The China molds are known to be unreliable at best. Even if you can pull parts from them, the wall thickness and detail oriented features are always all over the place. That leads to some really awful looking finished products. I have two China CNC molds on my Fryer machine right now. One is from a different shop that punched through a wall when they were fixing the China mistakes and we welded it up and now I get to hope that my inferences about the geometry are more aligned with the reality than the other shop's. Fun.

I just met with my favorite tooling rep who dropped by the shop, and ended up getting into a half hour conversation about tooling on my project. I think I'll end up taking him up on his offer and having the Seco guy come out one of these days when I am on my own time. I don't have a machine that can spin like acrylic likes at 18k rpm, so I need to try some things out to get a proper finish. That and the minuscule channels in the cold plate that are going to be either time consuming or terribly cut without proper tooling. I really need to try out some different alloys...

 

[thenrz]

Well, going to be doing a bit of a design change. I normally always research my information, but in the case of copper, I didn't do so and took someone on their word. They were wrong... 110 copper is a terrible material to machine. I need to take the coldplate and separate it into two parts with only the heat transfer base being copper, and the rest something else. So... there's that. I'm going to keep the same dimensions, but things are going to change here. The copper plate is going to be inlaid into some other metal and the original gasket will be used to seal to this plate which will essentially float within the larger frame but sealed upon the pressure of assembly. Ugh, but at least nothing else has to change (yet...). ah the joys of engineering.


EDIT:
Well, after a couple of hours of brainstorming, checking alloys and prices, modifying some drawings, reappraising sizes, and putting it back together, I came up with what I think is a really cool idea. The base of the block has been split into two pieces: an outer frame and an inner copper core. I am using a free machining copper allow now and actually saved some headaches since I can shorten my tiny cutters up for the slots. I might even get a slitting saw and do things on end. Or some other strategy I don't know about when it comes to production time.



View attachment 209086

Acrylic frame, free machining copper (145) core. Core seats and floats until the block is assembled. The pressure plate seals via gasket to the copper as before and is held perpendicularly by the acrylic frame. So nothing overall changes in the assembly except the coldplate is now a copper core within the frame. So guess what? Now there's two places that I can add RGB. Retail price just tripled.

 

[thenrz]

Finally making some progress on my fitment prototype. Had a line on line vertical surface around the inner core of the pressure plate, then I realized that the redesigned coldplate/frame compensated for that. So a quick undercut of a few thousandths and worked like a charm. My gasket grooves and socket head holes and recesses were literally perfect, thanks McMaster-Carr for all the wonderful CAD drawings. I have the acrylic top programmed, the mount is half done, and that just leaves the top aluminum plate. I should be able to get the mount cut on lunch tomorrow and the acrylic top after work... Would be killer to be able to get the top bracket done too and have the whole thing fit tested. 95% of my programs are all great, and now I can ramp up the feeds and speeds. I have been conservative, and it worked well because I could watch and see what was going to stay moving forward and what I needed to change. I will copy and repeat and start running two or three at a time for now after my first run with all the proper materials. Eventually I will be making fixtures and running 6-12, but that's a bit in the future.

Cutters came in as well. Have a selection of 0.5mm endmills and up for the slots. Eventually will use a slitting saw to dramatically reduce the time (10 hours right now for one coldplate. I expected that for now), but that will require pretty extensive fixturing to do it beyond a one off. Have a couple of pucks of copper 145 that I will cut up with the acrylic tomorrow when I have time, so things are moving forward and it's getting exciting.


Photos below! Don't know why they always attach sideways... Fitment model is all aluminum. Don't worry, no aluminum will touch a drop of liquid when I use the real materials.



View attachment 209101

View attachment 209102

View attachment 209103

View attachment 209104

 

[thenrz]

Fitment test complete

After some various modifications and revisions, the fitment test is complete.


View attachment 209157

View attachment 209158

View attachment 209159

View attachment 209160



Copper and brass for coldplate and pressure plate respectively are in and prepped for cutting. Arbor and carbide slitting saws are ready for slotting the coldplate. Getting the slotting program done was really an exercise in patience; I hadn't ever had to do this specific kind of operation before insofar as cutting parallel to the table so many times traveling down in Z... at least the entire slotting op will take 15 minutes as opposed to something like 8 hours with end mills.

Couple of notes: the only aluminum piece will be the mount. Eventually I want to cut the pressure plate and mount out of 303 stainless for a stainless edition, but that's going to be once I get things moving and finalized. The mount is sealed off from the liquid loop. The bottom frame around the coldplate is going to be acrylic. I have some done already, but they're all at work since I had to leave in a hurry yesterday due to losing track of time... So you'll be able to see everything going on inside the block. The slot you see in the side in both the photos and the drawing shot are for individually addressable LEDs. Those arrive tomorrow. Had to get small pitch (distance from LED to LED) strips so I can cut them up and wire them in. They'll be installed so there will be as much granular control as you'd ever want assuming you have the proper controller hardware.

All my finishing supplies, wet/dry sandpaper up to 5000 grit, polishing compound, felt and cotton buffer/polisher wheels for my dremel, and deburring tools started arriving today (thanks Amazon for weekend deliveries). I'm not going to spend much time getting the fitment model shiny and whatever, but presentation is important to me as it reflects build quality. Every part in this design is CNC machined and engineered with one specific goal: to perform as best as possible. My design paradigm takes MCM in Ryzen's layout into consideration as well as the ability to swap out only one part to target Intel's monolithic offerings. There will be about 11.2" inches (and up to 16" when my custom .75" diameter saws come in ~3 weeks) worth of surface area in a 1.5" square on the coldplate which will undoubtedly be able to move heat from the CPU to the liquid and out of the block with the best of the blocks out there.

I will have fully functioning, true to build prototypes this week. Hey Blaylock, you mentioned you'd test, so I'll need you here shortly. I have manufacturing notes and setups that will slowly transition into a production level operation. I have some nice pieces of aluminum set aside to make fixtures and softjaws for the speedy cutting of multiple parts per setup. Eventually, with what I have saved on the side, and assuming that I can sell these at a reasonable cost, I already have permission and space to, rent free, and with only paying utilities, setup my own machine at work and have my own workspace. I'd probably just go with something like a Tormach 600m+ since I will never need something big for this business. I've always wanted to do this for a living, and hopefully this is a segway into that!


Thanks for looking guys, and I'll leave you with a finished, revision 6, drawing:

View attachment 209163

 

[Blaylock]

I'm really liking how this is progressing. The mounting plate has a nice unique look to it.

P.S. I replied to your PM.

 

[thenrz]

I'm really liking how this is progressing. The mounting plate has a nice unique look to it.

P.S. I replied to your PM.

It's been a ton of work in a lot of areas, so I really appreciate that. There's a possibility, if this block even sells, that I could end up with my own workspace. There's so much more I want to do, but I've extremely limited on machine time the last week and probably a lot of this week. That all ends once I can generate some income and purchase time either here at work or at a colleague's shop. My medium to long term goal is to have a shop of my own. The upside of working with small parts are small machines. Small machines mean small spaces and not much in the way of expensive tooling or infrastructure.

I want to cut part numbers into each different part in the assembly since there will be the ability to swap out. On top of that, the space on the pressure plate between the standing features is ripe for something text or logo oriented. (and of course the acrylic for custom cuts...) The addressable LEDs that will go in the acrylic housing should backlight the text rather nicely, too. I could go deeper, and will, with the text since I'm using all carbide tools and the coated ones that I have cut the acrylic like it's not even there. Flood cooling on acrylic is compensating a bit for being maxed out around 8000 rpm since acrylic like to go fast fast fast. The finish is almost perfect, and some hand finishing off the machine will take it to the end.

Stay tuned as I cut copper and brass. I have a couple pieces of 303 stainless, too, that I want to try out, but that's another animal entirely compared to the free machining stuff. I really hope I have the acrylic bottom frame and copper coldplate cut today. Even better to get my slotting op done too. Copper today, brass tomorrow morning, and I have a complete beta block ready for pressure testing and some sample heat load experiments.

 

[thenrz]

Quick update:

Had to do a big revision to the bottom frame. Wasn't happy at all with the rigidity compared to the rest of the block. I mean, it would have worked just fine - and did - but it didn't mesh with my design goals. Was up until 4am modeling and programming, and hopefully today I can cut it. Brass and copper parts came out very well. Cut the brass a bit fast, and didn't use a nice TiCN coated end mill like I thought I had chucked up, so I need to make sure to do that next time. Just a lot of burrs, no other issues. Once the frame is cut and I get my testbed assembled again, I'll post some photos. Today should be the day for fin/channel cutting with my awesome slitting saw/arbor/vise/jaws/weird setup. I'm going for 0.050" deep on the first try and want to work my way up to .075" or even .100". No idea how copper is going to react to anything past .050" deep. Have .020", .025", and .028" solid carbide slitting saws on deck. Going to start with the .028" and see how it goes. I could squeeze in two more channels with .020", but i'm just not sure how the copper and such a thin saw (albeit solid carbide) will work together. Time will tell on that one... Couldn't pass up the deal on the .020"s either. The depth of cut is the real key here anyway after working my way through some math I haven't done since Calc 2 in college. Obviously the deeper I cut, the more surface area is exposed, that much is obvious, but there's also flow metrics in a deeper cavity. There should be more turbulence and therefore better conductivity.

Will post some photos this evening assuming I remember to do so. I had the block connected to my testbed last night and everything... that was prime photo material...

 

[trents]

My only input at this point would be to avoid obsessing over fine details of design goals that end the end would make minuscule, essentially non measurable differences in temps but would add much to production headaches and costs. I sense you are a perfectionist.

Your product design already has enough distinctive features to set it apart from the pack and avoid, I would think, any problems with patent infringement.

 

[thenrz]

My only input at this point would be to avoid obsessing over fine details of design goals that end the end would make minuscule, essentially non measurable differences in temps but would add much to production headaches and costs. I sense you are a perfectionist.

Your product design already has enough distinctive features to set it apart from the pack and avoid, I would think, any problems with patent infringement.

Looking at cutting fins into the coldplate is pretty much the most important aspect of heat conduction. I want to get it right from the get go so that my production tooling, programming, and setup strategies are only done once. I have considered doing a staggered design where I have three saws on my arbor with spacers, and the middle on is thinner than the outer two. Why? Because it looks cool. The more I look at things in the current state of this industry, the more I realize success means catering to that in more than a minor capacity. I need to set my product(s) apart. First is performance which will be understanding the sweet spot in depth of cut and other parameters of the conduction zone, and then making the package be something your rig can't be without.

...that being said, yes, you are right, I am a total perfectionist. I most definitely get caught up on little things. It's a bane and a boon all rolled up into one. Attention to detail is great for product design because I haven't had a part not fit yet, or done something silly and scrapped a part (okay, well one, but it was a trial anyway. My softjaws were all screwed up and I didn't notice). But sometimes I end up working for hours on something that 99% of people wouldn't notice. I notice, and that's what gets me.

 

[thenrz]

So, the second official test went splendidly. Couple of drops from within the outlet flow guide, but I knew it was going to happen. The gaskets are too small and it leaks from the glue seam; just sitting in water, there's not a drop, but under the pressure of the alphacool 1500lph pump wide open, the glue joint didn't fair so well. I am going to mill out the gasket slot tomorrow morning to accept the same gasket, .103 diameter, as the main coldplate->pressure plate interface. That one was bone dry for the whole hour of testing. So problem solved. I am going to order ready made gaskets so it won't be an issue at that point, but I like the idea of using the larger diameter material to keep 1000% sure under whatever pressure one throws at it. Also keeps things consistent.

There are lots of little things aesthetic things wrong with the prototype, but that's all been rectified within the models and programs. The mount wasn't designed to have something come up through it, so I didn't cut the inside fillets down to the proper size. Needed to use a quarter inch tool but used a half out of sheer laziness because at the time it didn't matter. So I undercut it by like .030" where as it will be basically flush and not gapped in later prototypes. There has been no sanding, finishing, or polishing either. A half hour with some fine grit and my dremel, and this would look so much better. Also, I didn't have softjaws machined for the new bottom frame that I designed, modeled, and programmed last night until 4am (lol), so there was a little bit of mismatch with the vise biting into the material a bit on the flip... but not bad enough that the thing didn't go together easily enough. The deck height is fantastic for the all important flow/turbulence curve, and I can't wait to get slots in the coldplate tomorrow to really see how she grabs heat.

There are also addressable LEDs that I have ready to put in after leak testing has ended. There are three different places where LEDs will be mounted, and since they're addressable, the themes are endless.

My power resistor testbed will be coming shortly. A lot of my parts that I ordered from all over the place are now heavily delayed due to the you who what virus...

Also, the contact surface is convex now. I used the information I received from AMD (Intel never responded) to craft what should be the perfect angle. It's actually a little but under, but i'd rather not see someone crank it down and start breaking stuff. So, I met in the middle.


The photos below explains everything better than words.

Assembly:

All the parts:
View attachment 209209

Coldplate+gasket:
View attachment 209207

Pressure plate+gaskets:
View attachment 209210

Mount:
View attachment 209211

Cap+barbs:
View attachment 209212

Contact surface:
View attachment 209213


Testing

Testing Overview:
View attachment 209214

All hooked up:
View attachment 209215

Block flowing:
View attachment 209217

Everything working:
View attachment 209216



Thanks for looking everyone! Stay tuned for a couple of changes, the slots cut, and then this thing goes on a motherboard.

 

[Blaylock]

Looking at cutting fins into the coldplate is pretty much the most important aspect of heat conduction.

This was going to be my comment as well. Fin thickness, fin density, and depth of cut not only determine heat conduction, it is also the main contributor to flow rate.

 

[thenrz]

This was going to be my comment as well. Fin thickness, fin density, and depth of cut not only determine heat conduction, it is also the main contributor to flow rate.

Indeed. I think my rather large advantage here is that I can go deeper than just about every other production part that because I am using a different slotting technique. It's more time consuming, more difficult to implement, and a lot less cost effective... but in the end my fins and valleys will have a .075" delta height. I will have to play with spacing between slots. Right now I am at .020", but I will have to see how this affect rigidity. I can always increase my stepdowns by a thou or two as I go from coldplate beta to beta and compare results. Flow rates should be massive with the depth of cut I am able to achieve.

Got the pressure plate's gasket recess milled out, and it fits the .103" gasket like someone designed it that way. I don't know if I will be able to get slotting done today, but I sure will tomorrow morning. I want to try to get two more coldplates done this evening so I can machine three different fin specs tomorrow, and another bottom frame as well. Maybe even another mount since I have really been able to reduce my machining time per piece. I really need soft jaws for the new bottom frame... so much to do.

 

[thenrz]

Went way overboard getting tool marks out... But I went ahead and just went up to 7000 and then a quick Dremel with some 421. Totally unnecessary, but why not? Not perfect, but shiny and I like to see the deformations from the nonflat surface.

Tomorrow morning is slitting day. Both coldplates you see in the photos will have slots: one with .025" slots making .015" fins, and the other .020" slots making .012" fins. Both will be .050" deep and trying for .075" (or if the solid carbide 98 tooth saws are happy, then .100" would be magical.) If I can get by with going so thin, then will lower the step and repeat until the fins stop being structurally coherent.

View attachment 209229View attachment 209230

 

[thenrz]

slotting complete

After an issue with my CAM programming (somehow, after doing this for years, I missed that my workplane was 90 degrees off in the XY), and having to go through and fix each slotting toolpath, I eventually rendered success. I was able to reach my initial target depth of .050" without any issues whatsoever, and the next one will go down to .100". I need to slightly adjust the pass pattern to avoid the raised wall on the subsequent deeper passes first, but that will be maybe an hour to draw and create two deeper passes per slot. My feeds and speeds were rather conservative, and the entire program took almost nineteen minutes with several quick stops for a visual check. I could easily drop that to 10-12 minutes without any issues.

So, in conclusion, the last hurdle has been surmounted and now my focus is on getting some hard numbers from the prototypes. I do have two complete sets made, and I hope to have a third by Monday or Tuesday... of course assuming that I can continue to go to work with COVID-19 and all that.


Some photos for your viewing pleasure:

View attachment 209232

View attachment 209233

View attachment 209234

 

[trents]

Those last pics were the ones we were waiting to see! Thanks.

 

[thenrz]

Those last pics were the ones we were waiting to see! Thanks.

I'm going to reduce the pitch and increase the depth on the next one. I wasn't sure how the copper would behave in this kind of operation, and I had completely underestimated just how nicely it took the fine toothed saw. There will be 5 or 6 more slots/fins and I will get 1.5x to 2x the depth of the slots.

UPDATE: 3 hours in at a full pressure test, new gaskets are doing the job. Things moving faster now. Working towards permanent fixturing and ordering higher volumes of material.

 

[thenrz]

Addressable LEDs doing their thing... completely modified the WS2812B strip I bought to fit. Worth all the ridiculous tiny soldering. May or may not also install LEDs in the bottom area, but that will be later on as that would require modification to run the wires... This is more of a proof of concept that I can use off the shelf parts to achieve the desired goal. Just ordered plenty of the standard 3-pin extension cables for ARGB motherboard headers or other control system that will have one end cut off and soldered in for the standard connection.

Should do an absolutely killer job of showcasing the block and accentuating the liquid moving in and out. Turned out to be a really satisfying outcome... and this coming from a guy who isn't necessary into the rgb stuff unless tongue is firmly planted in cheek.

Using an Arduino with the FastLED library to control them. Stupid easy...


Also... sorry for the hand. I don't have the right connectors yet, so I am holding three breadboard pins into the temporary female connector coming out of the block and taking a photo at the same time.


View attachment 209249

 

[trents]

When do you anticipate doing some temp performance tests?

 

[thenrz]

As soon as the mounting hardware is delivered. McMaster says tomorrow, but who knows with Ohio and the stay at home order. Springs, m3 screws, and washers. Mounts for am4, lga115x, and lga775 are done and ready. Just need McMaster to come through.

 

[Blaylock]

When do you anticipate doing some temp performance tests?

As soon as the mounting hardware is delivered. McMaster says tomorrow, but who knows with Ohio and the stay at home order. Springs, m3 screws, and washers. Mounts for am4, lga115x, and lga775 are done and ready. Just need McMaster to come through.

Looking forward to it!