Jump to content

Repurposing My Old Desktops Into BOINC Crunching Rigs

Windows7ge

As the years have progressed I've accumulated enough hardware to build a couple of complete towers. Right now I have the equipment to build two desktops using old parts (CPU/Motherboard/RAM/SSD/GPU/PSU/Etc) from PCs I had in the past.

 

The primary theme of the build(s) here as indicated by the title is repurposing my old desktops into BOINC rigs. In particular for the World Community Grid's OpenPandemics project to help fight COVID-19.

 

First and foremost an overview of the hardware I'll be working with:

 

IMAG0676.thumb.jpg.159d41cd51fcd1960adcb0e78951e908.jpg

 

Starting from left to right is one of two Rosewill RSV-R4100's:

 

IMAG0680.thumb.jpg.599dfd49e84ae2d12def13bf6684233c.jpg

 

The plan here is to make these rigs rack mountable. As more systems are added to the cluster in the future the ability to stack them in a rack will save on floor space and make for easier maintenance. The downside of this form factor for the application I have here is I could not, for the life of me, locate a short depth 4U enclosure with 3 front mount 120mm fans. I went as far as to explore Protocase as a possible solution (custom enclosure) which I knew would be expensive but when I saw their asking price for just one it was decided that I was going to have to DIY this.

 

I don't have access to a laser cutter, or the benders, or the drills needed to move sheet metal and have the outcome look good but what I do have is a 3D printer (Creality CR-10) and...

 

IMAG0681.thumb.jpg.5a80db898d8f3539629b5101d9f616ef.jpg

 

...filament is cheap :D. The plan is going to be 3D printing a fully custom front fascia. One that can accommodate up to a 360mm radiator.

 

In the middle is an old rig of mine from 2013:

 

IMAG0678.thumb.jpg.de03c3064965ac3b2d202b2a3f976b2a.jpg

 

The Intel Core i7 3930K

The ASUS Sabertooth X79

4x8GB G.Skill Sniper series RAM DDR3 1866MHz

 

The sketchy PSU, GPU (1/2 broken MSI Radeon HD 7770), SSD, Hyper 212 EVO w/ Noctuas, and the wooden board it's all screwed down to are not original equipment from my 2013 rig. These were mainly just to verify that all the core components still worked (and they do...mostly...)

 

Then on the right is my old rig from 2016:

 

IMAG0679.thumb.jpg.ef0f89db60910aced478ee407fa35306.jpg

 

The Intel Core i7 5960X

The MSI X99A-SLI PLUS

4x8GB G.Skill Ripjaws 4 RAM DDR4 2400MHz

 

The surprisingly reliable for it's age PSU, SSD, Corsair H100i /w Noctuas, and testbench missing a leg are not original equipment from my 2016 rig. These were mainly just to verify that all the core components still worked (and they do).

 

Now the plan originally included buying new PSU's. A pair of Corsair RM650x's. Unfortunately with the current COVID-19 situation PSU prices have gone up exponentially to the point where I'm not willing to pay almost double what a RM650x should be worth so I'm making do with the PSUs I have on hand. An old sketchy Rosewill RD400-2SB and my old reliable Kingwin Lazer seres LZ-1000 Watt.

 

This build log that I expect to take quite some time to complete is going to include cleaning off the layers of dust on the old hardware here, Prototyping/designing then printing the front cover of the server, rebuilding the computers into the rack mountable chassis, and possibly getting into some of the software setup.

 

This is going to be fun. :D More details to come tomorrow.

Link to comment
Share on other sites

Link to post
Share on other sites

8 hours ago, Windows7ge said:

-snip-

Ohhhh looks cool. What cabinet do you plan to use?

Link to comment
Share on other sites

Link to post
Share on other sites

2 hours ago, kkpatel87 said:

Ohhhh looks cool. What cabinet do you plan to use?

For the time being the rack that I have won't have the room to accommodate these boxes with all of my other servers but when I upgrade it it will likely be repurposed for housing these BOINC rigs.

 

The Rosewill 4 Posts Depth-Adjustable (22″ – 40″) Open Frame Rack 25U. First impressions. It's pretty sturdy. Tighten up the bolts and she doesn't budge. Thick sheet metal.

Link to comment
Share on other sites

Link to post
Share on other sites

Now when I decided to buy this chassis, I was taking a gamble. It had the general shape I was after and the insides appeared easily mod-able but I couldn't get any good information on how easily (or lack-thereof) the removal of the front fascia would be.

  1. Is it held in with screws? (The best outcome, easy to deal with)
  2. Is it held in with rivets? (Annoying but I could drill these out)
  3. Is it part of the body itself and just folded up? (the worst possible outcome, don't have a dremel and the only angle grinder I have would make a royal mess out of it)

However as it turns out:

 

IMAG0682.thumb.jpg.fd47f3462bc8a66247bf21c8c0555b1a.jpg

 

IMAG0683.thumb.jpg.e0f11356635fe3c71baf6589bb75fdca.jpg

 

IMAG0684.thumb.jpg.2e2436f3b632a2551571b3fbf0932b12.jpg

 

IMAG0685.thumb.jpg.0f3c5a784826bb0f34621263277b532c.jpg

 

It was only screws. Couldn't have come apart any easier. Lucky me. :3

Link to comment
Share on other sites

Link to post
Share on other sites

The 3.5"/5.25" tray is only held on with four screws. Two from either side.

 

IMAG0686.thumb.jpg.c8ed29f568c2796d5c005526f2d8661d.jpg

 

I was really worried that these were going to be rivets. Glad they weren't.

 

Got that out.

 

IMAG0687.thumb.jpg.4e49bf57ccc5004486904af169f4f21c.jpg

 

I went ahead and installed the 9 stand-offs I knew I was going to need. Both motherboards are standard ATX.

 

IMAG0688.thumb.jpg.6ad5906563182f561c4134017f624af0.jpg

 

Now we can start measuring and prototyping.

Link to comment
Share on other sites

Link to post
Share on other sites

While prototyping the new front cover the dimensions are going to have to go through a couple iterations before the fit is just right, but I'm able to make some fairly accurate initial measurements with a pair of very sketchily powered (because I haven't been bothered to buy a new battery) set of calipers.

 

Measuring the height of this 4U enclosure I'm seeing:

 

On the left: ~172.00mm

IMAG0690.thumb.jpg.ee54687acdade7536d00b64b51f84971.jpg

 

On the right: ~172.50mm

IMAG0691.thumb.jpg.04b67ab596b426bb0343072daee3f38f.jpg

 

It's clear the side that's short ~0.5mm has a slight dent on the lid. That can be pushed up which is good. Otherwise I'd have to go with the shorter of the two and deal with a slanted crack going along the top of the front cover.

 

Now my calipers are not long enough to measure ~17" so I had to get creative here. Let's just say the included server ears came in handy:

 

IMAG0692.thumb.jpg.b04a00d88fc9795718f7df321b230fb0.jpg

 

Measuring the length of both handles + 70mm I'm seeing almost exactly 425mm.

 

How deep to make it is fairly arbitrary. If it has enough structural integrity making it any deeper is just a matter of aesthetic. Looking at the lip that makes up the rim of the front cover there's a fine line that marks 16.30mm from the front.

 

IMAG0693.thumb.jpg.cebdabc2a305a5f89b2105d7b4073854.jpg

 

We can start with this for prototyping and can make the final design thinner if it's really not necessary.

 

So for now it looks like our initial dimensions will be  172.5mm x 425mm x 16.3mm. We're going to under size this a very small amount to compensate for slight variations in printing accuracy.

 

This is all I'm going to get done for today. Tomorrow I'll start to go over the actual design of the shroud, marking out screw hole size/placement, and making up some test pieces.

Link to comment
Share on other sites

Link to post
Share on other sites

Not the most entertaining part but since I'm going to be self tapping machine screws into plastic I need to measure the hole size I'm going to need. There's also a lip around the screw holes I have to account for if I want the panel to sit flush on the surrounding body.

 

These are the only two screws I have to account for:

 

IMAG0694.thumb.jpg.585ae73670bd40c8617e95e74a759340.jpg

 

When you get down to sizes this small you can't trust the accuracy of your printing software. At least not with a 0.4mm nozzle. So I just designed this block:

 

366202157_Screenshotfrom2020-06-1722-04-34.thumb.png.42cb1a19768170db6d3632b40bd29ac8.png

 

This will enable me to test-fit 3 things:

  • The radius of the screw-hole itself (how big I need to make the holes)
  • The size of the flashing that surrounds the screw holes in the sheet metal (these have to be compensated for)
  • The height of the flashing around the screw-holes. (If I minimize the height I can get a better bite on the face-plate when I screw it in.

And here's the printer working on the above 3D model:

 

IMAG0696.thumb.jpg.0d27dd49d73613c49b5cde57cd0710df.jpg

 

The filament I'm working with is PLA+. Assuming nothing goes wrong it should be ready tomorrow morning.

Link to comment
Share on other sites

Link to post
Share on other sites

Not much could get done today besides a whole ton of measuring. With the help of the tool I made I was able to measure:

  1. Small screw holes
    1. Radius of 2.375mm
    2. Height of 1.4mm
  2. Small screws
    1. Radius of 1.5mm
  3. Large screw holes
    1. Radius of 2.625mm
    2. Height of 1.7mm
  4. Large screws
    1. Radius of 1.875mm

I was also able to determine that the distance between screw holes center to center along the bottom are 130mm apart and along the sides are 70mm apart.

 

It seems Rosewill built this chassis to a remarkably simple spec. I like that.

 

As boring as all these numbers might appear they are extremely important for lining up the holes and making them the correct size when I both design and go to install the final 3D printed model.

 

On Friday/Saturday's agenda:

  • Create test pieces using all of the above specifications
  • Do measurements to test for the position of these screw holes front-to-back (because they're not centered)
  • Start the design of the front fascia (I'm thinking a honeycomb mesh would look really cool).
Link to comment
Share on other sites

Link to post
Share on other sites

With all of the measurements (and a couple more) factored in I was able to produce this test model:

 

880059294_Screenshotfrom2020-06-2014-16-32.thumb.png.3452f96f02c65bf9b0121d9b461c7544.png

 

Unfortunately my print bed isn't large enough to do this in one large piece so I split it into two:

 

1265175547_Screenshotfrom2020-06-2014-18-45.thumb.png.8ae15b926c77194806927b30de5b8e99.png

 

I plan to experiment with ABS glue to see if we can bond two into what was intended to be one piece.

 

The original print time on this was going to take >14hrs but with some changes to some settings I got it under 7:

 

Now all we can do is wait and hope my measuring was accurate enough:

 

IMAG0700.thumb.jpg.d1566b4bf4d2ae60f5494a1d29b41b4a.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

It looks like for all intensive purposes the test-fit was a success.

 

IMAG0701.thumb.jpg.86f94e056f51e49cb4cbb2386f40c4b5.jpg

 

First try too. There's a little bit of warp but the sheet metal isn't perfect either. I know if I try to modify this I'll probably make it worse rather than better so it's good enough for our application. The test piece does sit a little proud of the surrounding metal, I might try to compensate for that I might not.

 

Next up is composing a mesh design. I'm still very into the honeycomb idea but something like staggered triangles would also look cool. I'm looking to give it a Dell Server or HPE feel. I'll model and print off some ideas tomorrow. I'm open to other suggestions.

Link to comment
Share on other sites

Link to post
Share on other sites

This took some time to figure out since my choice of modeling software cannot do Hexagons. It seems to make a Hexagon out of rectangles isn't as simple as I thought it'd be. I thought I could make it out of triangles but it turns out that's not a supported shape in the software I use.

 

Finally I was able to figure out through trial & error to make Hexagons with a maximum width of 15mm you have to use a rectangle that is 13mm x 7.5mm and rotate 3 of them 60° to each other. The results look good though:

 

86317055_Screenshotfrom2020-06-2115-18-26.png.d49728fe4e69cf5f725ee9748a826015.png

 

The goal here is to get a visual of what looks better. Using walls that are 1.2mm or 1.6mm and having a boarder that is 1.6mm or 2.0mm.

 

And it's done:

 

IMAG0707.thumb.jpg.4d76027ff95ec99a485bfeb5cd15c42b.jpg

 

Holding it up to the chassis, getting a feel for it's rigidity, and the overall aesthetic, although 1.2mm walls looks fine up close I feel like for how large of a mesh I plan to make here it'd look aesthetically thin. That and going with 1.6mm will add more structural rigidity.

 

IMAG0708.thumb.jpg.36d27508aa9427bdef554c4d3aaa1ff1.jpg

 

Feeling this smaller piece in my hands it's quite stiff which is good and this test piece is only 7.5mm thick. I think what I'm going to do is step down from the original 16.3mm thickness to 12mm. This should offer enough strength in combination with 1.6mm walls to make for a stiff front bezel and save on print material & print time.

Link to comment
Share on other sites

Link to post
Share on other sites

I was definitely pushing the limits of my 3d modeling software with this one. Here is the complete first prototype:

 

9534812_Screenshotfrom2020-06-2122-59-03.png.5d8b9e7961c885a225136fe7689e32c2.png

 

This 1/2 of the bezel is simply a mirror projection of the other 1/2 of the bezel. So I don't have to design two separate parts.

 

1630472476_Screenshotfrom2020-06-2123-00-08.thumb.png.2324e008acf4f435503456b326cd97d3.png

 

This took longer to design than it should have because of the software limit. What I'm using really wasn't made for making shapes this complex but I made it happen and it's printing. I'd share a picture of that but it's going to take about 1&1/2 days to print assuming the power doesn't go out and it doesn't screw up so what I'd show you on the bed right now is basically nothing.

 

I'll update with a picture of the print progress in around 10 hours.

Link to comment
Share on other sites

Link to post
Share on other sites

Exactly what I feared and hoped wouldn't happen. The power must have blipped or gone out because this morning she's dead. 😭

 

IMAG0709.thumb.jpg.a28d1ea0dadff9eef599c51cda8224f2.jpg

 

I mean, it looks good though for the progress that got done.

 

Ugh, it's infuriating though. I don't think I have the time to reset this before work.

 

I'm going to fix this some time today and post another update.

Link to comment
Share on other sites

Link to post
Share on other sites

Alright, this time the powers not going to fail!

 

Say hello to the CyberPower OR1500:

 

IMAG0710.thumb.jpg.d89e2f854ed4b9bfb19da723ea1d1f76.jpg

 

I'm just "borrowing" this off one of my servers. I'll return it. Eventually.

 

This thing will keep the printer going for well over an hour if the power goes out again (which confirming with someone I live with did happen around 6AM).

 

Here we go again, for the 2nd time! :3

 

IMAG0711.thumb.jpg.9c8cbc4a9e6b2f8a6e493960fcc4d21c.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

So I have good news and bad news.

 

Good news is part 1/4 has finished after 36 hours:

 

IMAG0713.thumb.jpg.5b2f678861b568b99f1b73c29d9548ca.jpg

 

It is surprisingly rigid for PLA. It's going to work great.

 

Now the bad news.

 

I miscalculated the width of the right side when I made a slot to make room for the chassis threads and it resulted in this:

 

IMAG0714.thumb.jpg.f0284b69ee220afe6570e0499bd809bd.jpg

 

Right, is what it was suppose to look like. To be honest though. I think it's probably fine to leave as is. Kind of looks cool. I did clean up the 3D model in software to remove what caused this mishap so hopefully what comes out next time will look cleaner.

 

Next problem. EVEN THOUGH I MADE A TEST PIECE TO DETERMINE THE SCREW HOLE LOCATIONS...I still managed to mis-align the 2 screws that attach from the bottom by 1.5mm. I may be able to save this part though by just widening the holes. It's going to have support from 2 other angles so it will probably be fine. I again went into the 3D model and corrected that.

 

Now I've hit a roadblock though. All my tech that I usually use to transfer files to the 3D printer have just up'n'failed. I have no way at the moment to update the printer with the new files so part 2/4 is suspended until I get this figured out.

 

In the meantime though. Gotta say, this is gonna look sick:

 

IMAG0715.thumb.jpg.c153531553c9307b97a80ef1ea34544c.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

On 6/16/2020 at 5:57 AM, Windows7ge said:

The Intel Core i7 3930K

The ASUS Sabertooth X79

4x8GB G.Skill Sniper series RAM DDR3 1866MHz

 

Oh a chipset fan 

just wait and everything becomes a trend again 

Hi

 

Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler

hi

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

6 hours ago, Drama Lama said:

Oh a chipset fan 

just wait and everything becomes a trend again 

Probably, I have a AM3+ board with no chipset fan so it wouldn't surprise me to see modern boards have them again.

 

IMO, they're much more for aesthetics than functionality. Basic chassis airflow is enough to keep the chipset in check.

Link to comment
Share on other sites

Link to post
Share on other sites

Fingers crossed, assuming I didn't mistake old files for the new files we should be back up and running. Now printing 2/4:

 

IMAG0717.thumb.jpg.4f0dbbb3e05ee3b308698ed9d6f5453d.jpg

 

I used diagrams I found online to lineup the screw holes correctly for the 120mm fans. This looks really cool:

 

IMAG0716.thumb.jpg.750b760becad1a419a45664d52b359ed.jpg

 

There does seem to be a little bit of misalignment but 3D printing isn't perfect. I can widen the holes a little bit if I have trouble lining up a 3rd fan.

 

Today I'm going to get this part cleaned up. Next I think we're going to experiment with making some ABS glue. I want to see if we can rigidify this even further by bonding the 2 halves together.

Link to comment
Share on other sites

Link to post
Share on other sites

I'm loving this so far. With everything screwed down this thing is very rigid. Little flex to it and everything lines up. Airflow resistance is basically 0. There's an oddball gap in the middle but I can deal with that.

 

IMAG0718.thumb.jpg.9dd5650e3517b611d04c03e920ca9953.jpg

 

Next up I think I'm going to at least attempt to address these two gaps (left & right):

 

IMAG0719.thumb.jpg.206ffc5a81b8ae79969c937fa27c41f9.jpg

 

Those are there because the original bezel went over top that lip on the lid. I have an idea on how I can cover that up but it's not a critical component for what I'm doing so I don't plan on putting a ton of time into it if my idea flops.

 

So this is where we are at right now:

 

IMAG0720.thumb.jpg.a4099e1db82289936046d5ad855adc20.jpg

 

All of it has to come back out. The motherboard is filthy, it's due for a TIM replacement, etc. Bezel 3/4 is being printed now for the other system and at this time I need to clean up the 2nd printed bezel component.

 

Also this time the screw holes lined up. :P

Link to comment
Share on other sites

Link to post
Share on other sites

In regards to addressing the holes on the top/front (left/right):

 

IMAG0719.thumb.jpg.1d6e69f5518f1f4d79a6d19bce13eb5f.jpg

 

I thought about, and I thought about it, and just nothing I'm coming up with is a simple, quick solution to fix what is really nothing but an aesthetics problem. I may revisit this but for now I think we're just going to move forward.

 

One thing I was worried about was weather or not when using standard fan screws if the holes lined up well enough for the fan mount that bridges the bezel halves.

 

And it does:

 

IMAG0729.thumb.jpg.7abab15d172c5517e014627bc44d011f.jpg

 

I've decided against ABS glue. I don't think it's necessary and the slight misalignment I see when everything is screwed down tight means I'd have to apply it in place and that'd get messy. More work than it's worth.

 

I currently have bezel half 3/4 printed now working on 4/4. Going to cleanup this bezel and start dusting off the motherboards.

Link to comment
Share on other sites

Link to post
Share on other sites

Motherboards are clean:

 

IMAG0734.thumb.jpg.b54ec53bab1ac9fe2b6d22ec44692cae.jpg

 

Or at least as clean as I'm willing to make them (which is actually pretty clean, everywhere I could fit a Q-tip or dust rag).

 

Next up I needed to figure out how I was going to mount the SSD since I ripped out the enclosures original SSD/HDD mounting trays. When I removed them it opened up these two holes on either side (above the SSD):

 

IMAG0735.thumb.jpg.122e193d5d8366de8e45b43a15c2f8ba.jpg

 

I measured the distance between those two holes and it was exactly 70mm. Running off of that I mocked up a pair of brackets that will mount to these holes and give the SSD a secure place to sit:

 

ssd-brackets.png.6d10f94f271cd89383e2e27e5263c5cd.png

 

We'll see how they do when they're ready.

 

Bezel component 4/4 has just finished off the printer. I'll have time to clean it up tomorrow evening. Once that's done. We can actually start putting things together.

Link to comment
Share on other sites

Link to post
Share on other sites

After one full week of around the clock printing, the better part of a 1kg spool of PLA+, and about 12 hours of cleaning with a hobby knife we finally have two complete custom 4U bezels.

 

IMAG0737.thumb.jpg.579e14057e14c664da56b058c225232b.jpg

 

The SSD mounts also fit & work. Tomorrow we can finally start putting some hardware in these boxes! :D

Link to comment
Share on other sites

Link to post
Share on other sites

Alright, now we can begin.

 

IMAG0738.thumb.jpg.51578e811cd6e47e4f8a14b4a0954c62.jpg

 

I figure we can handle this one major component at a time. First up are the PSUs. As mentioned current PSU prices have scared me off from buying new PSUs at this time but units I have on hand will suffice for now.

 

For the 5960X rig I have the 1000W:

 

IMAG0742.thumb.jpg.5caeccc2ba1aa9be6429401ad766ae39.jpg

 

For the 3930K rig I have the 400W:

 

IMAG0740.thumb.jpg.816d005ccde7e064d258ce825df1d4a7.jpg

 

I'm not worried about the 1000W but the 400W is a little sketchy. However it handled the 3930K under load without issue during testing so it should be fine. Overclocking is not an option though:

 

IMAG0743.thumb.jpg.1ab815eaf3a0d5fa5be18be61eee5a67.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

I like where this is going, awesome job so far!

5800X3D -30 CO | 7900 XT | 2x32Gb Vengeance @ 3600 | MSI X470 | Sabrent Rocket 512 | 2x1Tb Vulcan T-Force | Seasonic Focus GX-750 Gold

Dell S3220DGF - curved 32" 1440p 165Hz

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×