Retired AM4 Workstation Case Swap - 1st Post

[BADCAP]

Hello everyone!

I'm a (very) long time visitor of TPU, though a relatively more recent lurker on the forums. In a past life I was a computer repair technician, electronics store clerk, and systems integrator (fancy talk for "I built custom computers"); I've been doing this for a while, and mostly from the hardware side of things.

I've wanted to share this shared hobby with others (gasp) on here for a while now, and figured there was no better introduction than just diving right in with (an admittedly simple) project.

On the bench today is a simple case swap: I recently built a new machine, and as I normally do, retired the old one to backup spare duty. Being a tinkerer and struggling (re: consistently failing) with ever being satisfied with any of my machines, I wanted to spruce things up a bit with my old workstation and make changes that might bring me closer to ever elusive contentment. My primary motivation was how well my new build came out and wanting to carry over some of the design elements to my spare.

Enter the only mostly retired workstation:

Gigabyte B550 Vision D (I'd have preferred Big D Visions)
Infineon hardware TPM
5900X with U12S
32GB 3200MT 14-14-14-34 Samsung B-Die G.Skill kit (not tuned or overclocked, chosen for intersection of stability and performance)
AsRock 6900XT Formula OC, blowout-free edition (repadded and given the Laird TPCM7000 treatment)
Corsair RM850X
WD Black SN850 1TB OS, SN770 2TB extended steam library, x2 Sandisk Ultra TLC 1TB 2.5" SATA windows storage space mirror
WH16NS40 BD-RW and stupid fast DVD-RW pulled from an old HP (imo old optical drives are waaaay faster, assuming they still work and their lenses are clean)
Fractal R5 (despite my griping, this is a good case, though just for a different use case than I was using it - great for something like a SOHO server, or quiet workstation with storage requirements)
x6 NF-A14 InductrialPPC-2000 PWM fans (not pictured: one mounted to side panel as intake, with dust filter)

I come from an era where cable management means you keep the cables from obstructing fan blades and they're routed for ease of access, case windows are for those willing to break out metalworking tools (and light it up with cold cathodes), bad fans were ones that had failed bearings or couldn't push air through a packed case, good fans were repurposed from the server world, and watercooling was only practiced by madmen with pond pumps and custom machined coldplates (unless you were loaded and bought a koolance rig). Subsequently, this build had none of those things. It was designed (kludged together poorly) with a somewhat dated perspective- it was the first machine I'd built in a number of years after having stepped away from the hobby for health and financial reasons. This was also a pandemic special, so it was done piecemeal and went through many iterations before settling on this configuration maybe a year ago. It was functional and performant, though always felt incomplete, even if composed of high quality parts.

Anyway.

I was predominantly bothered by three design issues with this machine:
1) Potential for damage to the GPU because it is gigantic, heavy, and mechanically connected at a right angle to the motherboard via the relatively fragile PCIe finger/socket. GPU sag causing issues down the road was in the back of my mind, though considering that this particular card is fully encased in metal (and the shroud/backplate assembly is screwed to its I/O shield) it probably wouldn't ever be an issue, though more likely just to damage the PCIe connector. Case flex, which caused motherboard backplane torsion, was also a contributing concern- this was a problem when I still had the drive cages in the front and only got worse after I removed them. Remember old server chassis with solid motherboard backplanes and crossbar supports? If only...

2) Thermals were acceptable, though they could have been better. Some areas of the motherboard, particularly under the graphics card and around the PCIe x16 slot were getting uncomfortably hot for me (50C after extended heavy whole system loads). Refer to my username and the 10 degree rule for capacitors to get an idea why I'd care. Also, my cable management could be better. A lot better. Existent would be a good start.

3) Difficult to transport. I really didn't feel comfortable transporting it without removing the GPU and CPU heatsink, especially considering their weight and their 90-degree orientation to the plane of the motherboard making for an excellent lever. I skipped on a LAN party and a vacation partially because of this and being spectacularly lazy and averse to effort (like how water is averse to traveling uphill). Seriously.

On my new build I mostly remedied most of these concerns by:
1)Vertically mounting the GPU. Mechanically decoupling the card from motherboard helps me sleep better at night. It also improved thermals compared to normally mounting it.
2)Using an AIO (barf). The pump/block on a simple AIO (meaning without a screen or other fancy bling) is certainly lighter and shorter than a large tower cooler, like the U12S.
3)A case with better cross-support, which these days means a PSU shroud and a motherboard tray that spans the full width of the chassis.

... which sets the stage for this case swap.

First part to arrive was the vertical mount, which I recently picked up on a deep discount. It's a discontinued phanteks model, though I'm fairly sure it should fit in the Antec P20CE I ordered since it looks suspiciously similar to the interior of the Phanteks P600S that the mount is compatible with. I suspect the two cases have shared tooling...

Assembled... I really like the design of this mount with the adjustable retaining clamp.

However, I don't like the thin PCB supported by two standoffs when 1.2kg of graphics card is going to be sitting on top of it...

...which is why EVA foam exists.

Much better. The area to the right looks like it isn't making contact, though that's just a shadow cast from an unintentionally beveled cut.

The fans on this particular card protrude from the shroud, making further adjustments to the foam pads on the clamp assembly necessary- you can see cuts at the edges of the pads I made to give the fans clearance. That was pretty much the best position for the clamp- any other would either have been directly over a fan, obstructed the 8 pin PCIe power connectors, or would have required modifying the base of the mount to extend the channels that the clamp screwed to.

Big heavy large huge 6900XTXH XXX supermassive chungus edition, a certifiable unit:

More posts to follow.

 

[BADCAP]

Continuing...

At first I just wanted to copy 1-to-1 what worked so well with my new build, which meant getting another Phanteks Eclipse P600S (no window), though they recently discontinued that model and I hadn't been able to find it in stock anywhere. I'm kind of oldschool in my views about equipment chassis in general- I'm of the mind that they need to be a metal box for the purposes of shielding and providing a good ground to all the stuff in it, so that automatically ruled out anything with a window. Mesh sides and front were also out because of my desire for it to be durable enough to transport without worrying about it getting dinged up. I also have a preference for filtered intakes. I also wanted to be able to fit that Phanteks vertical mount on account of liking its clamp so much.

Wow. Laundry list. Remember what I said before about ever elusive contentment? I guess I kind of do it to myself... lol.

So off searching I went, and happened to stumble on the Antec P20CE. From pictures it seemed that the P20 had a lot of similarities to the P600S in terms of design, and even some of the parts seemed very close in dimensions- made me suspect that maybe it shared some design or tooling with the P600. After some further eyeballing of the PCI slots and the vertical mount that Antec used to make, I decided that it was close enough and pulled the trigger on an open box return listed on ebay.

Gentlemen! Behold!

It came all the way from the magical northern lands of Canada, and without a scratch, dent, or import duty to boot. It smelled of maple, civilized politeness, and couth. How refreshing.

Now to see if I was right about fitment of the vertical bracket...

Survey says: "yes." Also pictured is a thermal probe I was playing around with. I have yet to find a better method of attaching it... kapton tape doesn't hold on the textured metal, gaffers tape would probably get crusty, so 3M electrical tape it is, though that too will eventually get smeggy, especially sitting right under the hot GPU exhaust. Maybe pinning the wire under the foam and bending the lead up?

So the mount has two screw holes (one to the upper right, next to the USB 3 header dangling there, and the other partially hidden by the clamp in the lower left), both of which are supposed to be used to anchor the flat base of the mount to the PSU shroud. I think some of the Phanteks compatible cases actually had pre-tapped holes that lined up with this specific mount, though no such luck here. Well, as Wonka says: "We are the computer-makers, we are the tapers of holes."

Protip: never ewaste old hard drives whole. Scavenge the magnets first, as they are useful for all kinds of things, like picking up metal shavings. Also, the platters make for good coasters. I don't know if it's a thing anymore, but in the past sometimes you could sell the boards on ebay if they were still good. Remember board swaps on dead drives?

The hole on the left falls over a cutout, so a no-go there. Oh well. One is better than none. I guess I could use a short bolt with matching nut and washer on the underside, though I surprisingly didn't find any in my screw cache. Maybe later.

To be continued.

 

[BADCAP]

Next up, I gutted the old system and did a quick clean, paying closer attention to the fans. I usually prefer quick cleaning fans carefully with a vacuum and brush attachment, while being extra careful not to allow the rotor to spin or to jar them too hard, lest I damage the bearing. For deeper cleaning my go to is a dry microfiber cloth and cotton swabs after removing the fans from the case. I've only had to resort to solvents like isopropyl alcohol or windex (ammonia based glass cleaner) in the past when cleaning smokers computers, or ones used in machine shops (they tended to develop an oily residue, my guess from from aerosolized cutting fluid).

These fans were pretty clean as the filters on the Fractal R5 are relatively fine. Mostly just powdered human stuck to the blades.

Before:

After:

I've been happy with Noctua industrial fans and carrying them over was more affordable then buying new, though I was seriously tempted by T30-120s, which I have in my new machine and just ooze quality. These will be fine, and while they are certainly louder at higher RPM, I find their lower pitch sound more pleasant than the T30s. I also know that they'll be more than capable of pulling air through the front filter and perforated metal faceplate of the P20CE, as they generated enough static pressure to produce decent airflow on the intake restricted Fractal R5.

The P20CE comes with 3 generic Antec fans in the front, which I removed to make room for three of the NF-A14 IPPC 2000s. I decided to re-use one of the stock fans as a rear exhaust- even if it isn't a retail quality fan, it is still PWM controlled, and I have two spares should this one die or start making noise. It doesn't have vibration isolating pads at the corners, though there's a simple fix for that: more EVA foam.

I just cut pieces of foam to fit the corners, pre-compress it a little, and run a narrow shank screwdriver through it, using the mounting hole on the fan body as a guide. Coincidentally, the Wiha precision Phillips size 1 fits perfectly on most fans.

Cleaned the CPU with isopropyl. Anhydrous or bust. (This bottle is kind of old, so there's no way it's still 99.9%- every time you open the lid, some moisture gets in- never leave it uncapped. Always decant the minimum needed into a non-flammable container, preferably with a small surface area. At this concentration it is very easily ignited, and the vapor is flammable as well. Alcohol flames can be difficult to see under bright light. Use outside or with adequate ventilation. /safety) I usually turn on my charcoal filter when working with it, or just go outside.

The high concentration makes cleaning up hardened thermal paste or oily residue left behind by crappy thermal pads a lot easier and quicker. I swear there's a difference from 90%.

Motherboard mounted and AIO installed.

AIO selection was a chore. I would have preferred the Arctic Liquid Freezer III 360, as that's what I have in my other system and I am very satisfied with it, though it wouldn't have cleared the GPU when mounted vertically. Between the hoses on the Liquid Freezer III coming out the bottom side, its offset mount, and size/shape of the pump cap it just wouldn't have fit. I did read that I could be mounted upside-down (hoses coming out the top), though that seemed kind of janky to me.

I ended up going with the Corsair Nautilus RS 360, which in many ways seems inferior to the arctic by my qualitative assessment. Maybe just trade offs. The mounting design makes for a waaaay easier and simpler installation, though I suspect it provides less mounting force than the Arctic- it has a bracket that slides on to the pump housing that lines up with and screws to provided standoffs that get installed on the stock backplate. It also lacks the VRM fan, the radiator is a lot thinner, the pump seems more audible (at startup at least), and the Corsair RS fans seem louder than the Arctic P12s. I haven't done any kind of quantitative measurement of any of this, so take what I have to say with a gigantic boulder of salt. This is just my opinion, man.

The ease of assembly and installation was surprisingly... easy. Quick. Not scary at all. Nothing like installing HSFs years decades ago when most CPUs were bare die and the possibility of chipping it was real. So little downward pressure is put on the motherboard during installation that I was comfortable installing it with the board already in the case. The other benefit was that it didn't have any kind of clearance issues due to radiator thickness or hose orientation.

I'm feeling tired and approaching critical lazy, so I'm just going to skip to the end.

tada.wav

I think it looks a lot better than before. The basement under the shroud is still kind of a PSU cabling nightmare, though that isn't impacting airflow and, like the messy basement I have IRL that I someday need to organize, I'm not going down there any time soon and can't see the mess, so fuckit.

Zero GPU sag. No more inadequate GPU support. No more worrying when I move my computer. For me, this project is already a success. Now to find another home for my optical drives...

Next post will be burn-in stress testing and temperatures.

 

[BADCAP]

Back after a slight break.

I did burn in testing for a few hours the other day, which it passed (survived without BSOD, freezing, hardware error, driver crash, or program crash = pass), though I didn't take any notes or collect screenshots, so I repeated a shorter stress-test this evening so I'd have something to post :-!

Setup:
Room temperature was between 21C and 22C. I wasn't using any kind of heating or cooling, aside from having pre-warmed my room with a couple of hours of gaming on a different computer preceding the stress-test.

I had the PSU plugged into a Kill-A-Watt watt meter so I could get a feel for the at-the-wall draw, just doing spot checks. The Kill-A-Watt I have does not perform logging.

Relevant EFI settings:
RAM set to EXPO profile with stock timings and speed for that kit (3200MT, 14-14-14-34-48)
RAM voltage manually set to 1.35V (Old habit)
FCLK manually set to 1600MHz
UCLK = MEMCLOCK
PBO on, manually set to 142W PPT, 95A TDC, 140A EDC
No boost override
Quick and dirty all core curve optimizer -20
VCORE, SOC, and dynamic VCORE set to "normal" (previous versions of the firmware seemed to juice the hell out of the CPU otherwise- kept setting it as a habit)
ACPI L0s entry (no L1)
Manually set fan curves

I customize a lot in the EFI, though those should be the main relevant settings for stress-testing, where factors that influence power and temperature matter.

Software:
Relatively fresh (<1month) install of W11 Pro 24H2, installed from 24H2 media. Core isolation features enabled, aka VBS (aside from firmware protection, which requires DRTM).
Furmark 2.5
Prime 95 (latest?)
HWiNFO64
Latest adrenaline, whatever version that is as of this post, with manually set (aggressive) fan curve.

Procedure:
I booted the machine and let it idle on the desktop for 5-10 minutes before starting.
I launched HWiNFO before starting the stress testing.
Prime95 ran with "Blend" torture test with 8 threads.
Ran for about 30 minutes before taking screenshots.

I took successive screenshots after scrolling the HWiNFO window to capture all the relevant sensors. I could have logged the data and put a fancy graph here, which I guess I could still do, though I didn't and probably won't.

I observed values on the Kill-A-Watt meter ranging betwwen 615W and 640W, ballpark. (I've previously seen this system go as high as 650ish under non-synthetic loads.)

EDIT:
System1 = rear 120mm Antec exhaust
System2 = x3 Noctua NF-F14 Industrial PPC (on a splitter)
CPU = x3 Corsair RM 120 daisy chained on the radiator
CPU_OPT = Pump (at 100% duty cycle)

Interpretation and notes to follow.