Scratch Build – Complete Posthuman

I'm back with another project! Building the CryoPC was a ton of fun, and I learned so much, that I couldn't wait to start a new build with the lessons learned. Looking for an excuse to build another PC, I found that it would be nice to have one in my workshop as well, as my laptop can barely handle heavier Fusion Projects.

I also had this idea floating around in my head to do a PC around the idea of the simulation hypothesis. For the uninitiated, it's the idea that we all live in a computer simulation that is indistinguishable from reality / is our reality. The most obvious way to realize is with a brain in a vat, that is connected to a computer. Except in my case, the computer is the brain, and it is in a vat of mineral oil. The "computer" would then house the radiator to cool the oil.

Looking for a stylistic direction to take this build in, after a very short search, Steampunk was the obvious answer. It fits perfectly with the SciFi/Fantasy nature and gives me the opportunity to try out a bunch of new techniques. In contrast to my last project, I'll be using little to no plastic parts, using metals like brass and copper in conjunction with wood for most of it.

For the hardware, I was limited by my budget (which is 0$, so I'm already a few hundred over), and the size of my vat. I could go with a bigger tank, but mineral oil is very expensive, so see reason number one, I went with a 10l fish tank. I was going to build the vat from scratch using acrylic, so I could get the exact size I want, but it would have cost over a hundred dollars here to get the materials, so I was lucky to find this fish tank that was close enough for a fraction of that.
But enough talking for now, here are some pictures of what I have so far:


I scored this used SFF build for a very good price, especially considering it has a GPU included (1060). Now I don't have room for a GPU, so I'll be selling that and the R5 2600, making back most of what I paid. What I'm left with is:
AsRock B450 Fatality
16GB Crucial Ballistics 3200MHz
500GB Samsung 970 Evo
500W PSU, though I changed that out for an SFX one.
For the CPU I'm going with a R5 4650G, as it seems to have decent graphics, and enough CPU power for what I need.



Probably the single worst thermal paste application I've seen in my life. It looks to be liquid metal but combined with the Crappy heatsink, it had maybe 3 or 4 tiny bumps that actually made contact. It's a miracle there was no thermal throttling. As I intend on selling the CPU, it took me almost half an hour to scrape off the thermal compound, and it still looks like crap. (If I was continuing to use the CPU, I would have just lapped it, but that would remove the markings, making it harder to sell)


On a more positive note, isn't this PSU just cute!




And of course, the first thing I did was to void the warranty, like a true modder. Removing the chassis, also reduced the height from over 60mm to just 40mm! I'll also be doing custom cables soldered directly to the PSU, so I couldn't care less about the stock ones. Funny enough BeQuiet is actually using a 3rd party fan in this PSU, although they have their own line.




I tried to see if I could fit the GPU in there as well, as I already have it, and while it does fit, it would be a bit cramped and make it hard to properly get the brain shape in there, so I scrapped that idea.




With only the mobo and PSU, I was able to make the brain quite a bit smaller, so it's just about 1.5x normal size. So just slightly Abby Normal. In the end, it will be a frame of brass wires, but just a simple mockup for now.



To get a better idea of the whole shape, I modeled it up in my new CAD software (cardboard aided design), where the accuracy and graphics are a bit crappy, but the controls are very intuitive.
The box in the back will house a 240mm rad, and the IO extensions, as the mobo will be completely submerged.
That's it for now. There are many more exciting parts of this build that I can't wait to share with you, but I guess we'll both have to be a bit patient

 

Awesome. Look up Legoman's oil pc. (F-it... LINK HERE.)
He ran into several issues with the oil:
1. It dissolves rubber and thermal paste, so O-rings need to be silicone, and the capacitor type like the ones in the PSU have rubber seals in the bottom. He had some come completely off the mobo.
2. It wicks up through wires and makes a siphon effect.
3. It STINKS. I can confirm, since I bought his damaged radiator.

So you need some creative sealing work like epoxying the cr*p out of the capacitors, isolated connectors, and tank seals.

 

Thanks for the tips, I'll definitely look into it!

 

Quite the build log that Legoman wrote. Took me a while to get through the 36 pages, but it was worth every minute. Here are my takeaways from his years of troubleshooting (and thanks @legoman666 in case you're still active 10 years later):

1. Sealing stuff is hard, especially when o-rings only work for a few weeks or months and then slowly start leaking. My solution: Avoiding seals as much as possible. Since I'm using a premade tank, there are no seals, it is injection molded as one piece (I should check the material it is made from, just in case). For the tubing, I was going to use copper anyhow, so I can just solder all the connections, even to the radiator, as long as I design it in a way that still allows me to lift it out. That should hopefully make sure there are no leaks. (The pump, a D5, will be in the tank)

2. Thermal paste dissolves in mineral oil. There are a few solutions mentioned in the thread, the most common ones would be no thermal paste / just the oil itself, which is certainly the easiest, but also seems to perform not ideally. It could probably be improved by lapping the CPU and cooler, but still not ideal. Using thermal paste and hoping for the best seems to work fine for a few days up to many months, but is a bit too temporary for me. The third solution would be to use a TIM that isn't dissolvable, like liquid metal. Not sure how I feel about this one, I guess I'll have to do some more research.

3. Oil will find a way anywhere and capillary action is a thing. Since my connectors will be submerged in the oil, with extensions going to the rear module, where I'll plug in everything, this is an issue I have to worry about. I would have never thought that the oil would find its way through the connector of a DP cable, but it makes sense that once in there, the fine mesh shielding is ideal to climb up. The only real solution I can see is to make sure there are no meshes to climb up, as a smooth cable should (TM) be fine. This means no sleeved cables going out, and for the DP cable, I'll have to either seal the crap out of the connector (kinda doubtful about the chances for success) or removing the insulation a little way up the cable and using epoxy to block the shielding mesh there.

4. Some capacitors will swell up and break. Since I only have an ITX board, it's easy enough to just epoxy the crap out of the few caps there, just to be on the safe side.

5. It stinks. I'm not too worried about that one, as I'll probably be using medical-grade oil (since that's what I can get around here and I don't need that much, so the price is not too bad), which should be a bit better smell wise. I'll also have everything pretty much sealed, so not much space for the smell to escape. And even if it does, this build will live in my workshop and not a bedroom, so it smells of paint, burnt things,... most the time anyhow, how bad can a bit of oil be.

Let me know if I forgot anything major. I'm definitely glad to be able to use the knowledge of the hivemind on this one instead of finding all these things out myself!

 

I like your approach!

NoX

 

So I'm in my workshop and just checked the material of the tank. It's Polystyrene (PS), which upon putting it into google gives me conflicting results on it's resistance to mineral oil. This source classifies it as good (best classification): https://alwusa.com/wp-content/uploads/2017/03/Chemical_Resistance_Polystyrene.pdf while this source calls it "limited": https://omnexus.specialchem.com/polymer-properties/chemical-resistance/mineral-oil-20-c

Just wanted to see if anyone here has slightly more chemical knowledge and has an interpretation, or if this just comes down to which exact mineral oil you're testing with?
Sorry about the lack of pictures, I'll post a more visual update tomorrow, or on Monday.

 

One way to find out: Put a drop in the tank. If it's 'styrene, there's a high chance of it clouding up from any chemical attack. Acetone for example, will turn it bluish-white, -if I remember correctly.

 

Guess so. I don't have the mineral oil yet and wasn't planning on getting it until I'm further along with the build (just to spread the expenses out a bit), but I guess I'll order it soon, so I can test it.

 

Here comes the promised update. Went to the hardware store and got some nice goodies!



Now that I have a lot of the materials, only time and lack of skill are stopping me. But I'm in no rush on this project and ready to learn new things. The only thing I've made from brass so far was a lighter in shop class a few years ago. Let's see how much I remember.






Not too much of the soldering apparently I'm just using regular 60/40 solder here and heating with a propane torch. It's definitely getting much hotter than needed, but the flux I got seems to only really work once I get it up to temperature. The reason for using 60/40 over silver solder (not sure if that translation works, just the regular higher temp soldering stuff is what I mean) is mainly lazyness. For any of the larger contacting areas it's also plenty strong. I ran into some issues though on the smaller joints, so I might try the hotter stuff after all. As with welding, a grinder and paint make me the solderer I ain't. (just maybe with sandpaper instead of paint, would be a shame with brass)



My fancy CNC mill also pulls double duty as my drill press


And a sneak peek at a related side project:



(not the actual working motherboard) Notice how the right tabs are not straight, that's because I broke the solder joints for that rod for the second time while assembling, maybe I should use a stronger solder than tin and lead...


But it assembles and gives an idea of how the mobo "tray" will look with the PSU on the back. I'll probably shorten the mobo standoffs a bit, but otherwise, I'm happy with that part.




Apart from that, I got my R5 4650G in, and after a bios update, it boots perfectly into windows, provided I'm still using a dedicated GPU. When using the IGPU, Windows can't find it's OS partition. It says invalid boot device, and when trying to reinstall it with a bootstick, it doesn't see the SSD at all, but the recovery partition somehow does work (otherwise there wouldn't be a bsod). In the bios everything shows up fine, and as soon as I put a GPU back in, no more missing SSD. I'm honesty at a loss, what would even cause this error. Officially the CPU isn't supported by the mobo (go figure, didn't even occur to me to check that, as 3000 and 5000 series are supported, and it has graphics outputs). There are a few people that got it working, especially with the newest bios, but not much luck for me. If anyone has a genius idea, feel free!

 

Alright, it's time to share some of the really exciting parts of this project (at least for me). I've been playing around with 3d printers for probably 8 years, but still, there has always been one 3d printing technology just too far out of reach, that being metal 3d printing.
However, today this changes. While I'm not announcing that I got my own metal 3d printer (I wish), PCBWay (a longtime sponsor of my YouTube channel) has expanded beyond just PCBs into other manufacturing services, one of them being metal 3d printing. So naturally, I begged them to print me some parts, and here they are! While they offer stainless, tool steel, and titanium as possible materials, I couldn't quite convince them of that and settled on aluminum. Here are the parts directly from them:


As you can see, the parts are very matt, as the surface is very rough. This is because of the technology, melting fine metal powder with a laser layer for layer. It's a bit like a very extreme sandblasted finish. They do feel incredible though, seeing a shape like this come to life in metal is really cool.
I was not going to leave them like this though, so I went ahead and started sanding... 3h later my fingers had lost all feeling, but the first part was polished up! Here it is compared to a raw part:





I love how the recessed parts of the design have kept the rough surface, with the other parts being super shiny. This adds a ton of depth to the parts. (This is also why I modeled the rivets as recesses instead of protrusions like they'd be in real life).
Here is a comparison between a finished module and all its stages:



Now as this is a steampunk build, I'd of course like the heat spreaders to be either brass or copper, but that's not a material being offered by PCBWay. I've researched various surface treatment options, and what I came up with is that while copper/brass plating aluminum isn't impossible, it is very difficult as the materials are quite dissimilar. Anodising would be a lot easier and I've seen very convincing results. Getting the supplies like battery acid locally is also not a problem, just the dyes prove to be difficult. I couldn't find any locally, and getting them shipped in costs ridiculous amounts in shipping as they apparently are classified as "dangerous substances". I hear that Rit's dye works well too, but sadly that's not a brand available here. Does anyone know of a brand available in Europe (Switzerland) that is similar?
For now, I'm leaving them just polished, but I'm keeping my options open.

And since I know you guys are curious, the price to get these parts metal printed is around $40 per part (in this case). This isn't cheap, but compared to what it used to cost just a few years ago, this is a bargain! And while the surface isn't super nice, the parts are very strong and can be processed just like any aluminum part.

I also continued a bit with the soldering saga, trying out brazing instead of using electronics solder. It turned out I did actually even have some lying around, go figure. It is quite thick stuff, so I struggled with getting a small enough amount on there, but it is much stronger, and once I got it hot enough it flowed even better than the tin stuff.



And after some cleanup:

 

NEAT! The 1st printer I'd ever contemplate buying would be laser sintering. Extruders are just too finicky. -juuust waiting for that price to drop...

Brazing is a much better idea. That's a rather pretty work surface. Decent stone is rare in the land of sand and mud, so it kind of hurts to see it used like this.

 

I probably shouldn't tell you then how I cracked the stone slab underneath when I didn't use the extra stone...
I would have used a firebrick if I had one, but this is the only thing I found lying around and I didn't exactly feel like going to the hardware store again

Also, I wouldn't be so sure that an SLM or SLS machine is less finicky, tough the pain might be more worth it.

 

Very cool

Concerning the Rit Dye... I dye my resin prints with it, but with metal, I am not sure...
I get it from Amazon.
Wouldn't they not deliver to Switzerland?

https://www.amazon.de/Flüssigfarbstoff-synthetische-Rit-DyeMore-Graphit/dp/B00U2IXMIW/ref=mp_s_a_1_4?dchild=1&keywords=rit+dye&qid=1621716986&sr=8-4

 

You'd think they deliver to Switzerland, as they deliver to Italy as well, which literally has the packages go through Switzerland, but they don't. No Amazon for us here. Thanks for the tip though!

 

How far do you live from the border?
If you are close by and shop regularily in Germany anyway, maybe you can register yourself a "DHL Packstation".
Have it delivered to the station (basically a wall of lockers, usually next to a supermarket) it in a city on the other side, pick it up and "smuggle" it home

 

Wow really cool the new RAM heatsink!
Interesting project, good work!

 

Thanks for the tip, though I don't quite live close enough. However, there are even businesses where all they do is you can ship them stuff and they take it over the border and ship it to you. It's possible to get almost anything here, but always a question of how much you're willing to pay to get it here.
At the point where I'm spending 20+ bucks on shipping to get a $10 bottle of dye, I'd rather spend another 10 more and get actual anodizing dye shipped in.

 

Time for another small update. While I've been super busy in the workshop, somehow I barely had time to work on this build the last two weeks. Here is what I have been able to do:





As you can see I completed brazing the mobo tray. Of course, after taking it apart and brazing it again, the holes didn't line up anymore, so I had to widen them a bit and soldered the standoffs in place. Using a Dremel , I cut the cross braces to match the outline and then bent and brazed that on as well. There is definitely quite a big learning curve to brass, but slowly I'm starting to get more comfortable with it.

Apart from that, I also got started on making a matching keyboard to go with this build. For that, I have to machine some 1.5mm brass plate (amongst other things). I certainly had a bit too high expectations of what my slow 2k rpm spindle could do with a 2mm endmill and broke two in short succession. With only one left I decided to maybe not slot the cutouts but use an adaptive strategy with a 4mm first and then go back in the corners with the tiny one. This worked a lot better and I was able to get a successful test pocket. For the complete keyplate, I decided to predrill some entry points. This way I don't have to helix down 90 times, which shaved over half an hour off the cycle time and probably doubles my tool life. Haven't had time yet to run the program for the whole keyplate, but even with the holes it already looks cool. Also, pretty descent chips form the drilling operation.





However, it is not all roses with my CNC. When doing the test cuts, I noticed that my spindle motor started smoking a bit when I was pushing it to the max 2300rpm. After a lengthy investigation, I noticed that it was going quite a bit faster than it should. This caused it to pull more like 1.5kw instead of the rated 1.1kw at even just 2k rpm under no cutting load. I can only imagine what it must have been at 2.3k while cutting.
The mismatch in speed was a fairly easy fix by adjusting a pot in the motor driver (was probably also me that mistakenly moved it ). Curiously the power draw is still a bit high. At the rated 2.3k rpm it pulls the whole 1.1kw under no cutting load, and with cutting, I can only get around 1.8k rpm until I hit the 1.1kw wall. And as soon as I go even 50w higher, it starts smelling again. I did take a look at the motor, but to my untrained eye, nothing stood out that would cause it to draw more power. With the motor removed, the spindle turns very easily and it's also well greased, so it shouldn't be anything external. Does anyone here have experience with DC motors? I'd prefer not being even further rpm limited than I already am, but just replacing the motor without knowing it's bad also seems a bit excessive.

 

Looks like you got the wrong brass for milling...
There are basically 2 grades.

MS63 (CuZn37)

This stuff is good for bending, soldering, welding.
It is soft material, and not really suited for machining.

If you cannot get anything else, go easy, like 0.3mm depth of cut and soil your mill in cutting oil/coolant while doing so.
Take the feed rate to your maximum.


MS58 (CuZn39Pb3)

This is milling grade brass.
You cannot bend or solder it (brazing might work), but it creates very nice little chips.

 

Do you have a pic of the motor (and/or the mill head)?