shevalier wrote on 2026-07-05, 17:10:

You have to experience this firsthand. :-) First, you buy an AGP DX10 graphics card (I have a 2600PRO), and only then do you rea […]
Show full quote

geordiepingu wrote on 2026-07-05, 15:35:

In other news, I have purchased a Sapphire HD 3850 AGP for this machine. I found one at a reasonable price, so I snapped it up. It should arrive next week.

You have to experience this firsthand. 😀
First, you buy an AGP DX10 graphics card (I have a 2600PRO), and only then do you realize it's useless.
It's underperforming for "relatively new" games and poorly compatible with older ones.
For such builds, an X800-type graphics card (or at most an X19xx) is optimal.

I know this, as I had not dissimilar cards 20 years ago. All for the benchmarks!

geordiepingu wrote on 2026-07-05, 17:19:

All for the benchmarks!

That is an appropriate quote for a signature here on vogons 😎

Progress has been slow as I've been away for a short while. Either way, I've managed to put another shift in. I've made my PSU cables and reassembled the case "internals" after painting them with cold zinc spray. The outer panels have not been fitted as I am going to sand and paint them (BMW Techno Violet Metallic...). I've also mocked up my water cooling hoses and have decided I need to go a slightly different route, but that's ok as this is all in the spirit of R&D... Finally I added some rubber "u-channel" trim to the cuts I've made to make them look a bit prettier and to stop the abrasion of cables.

Making the PSU cables was quite easy, but tedious. Over 100 pins crimped later... The irony was not lost in my making of coloured cables. Years ago, I bought a set of depinning tools (before the Hiper Modulars!!) so I could sleeve my cables in paracord to hide the colours.... Still quite nice to see the old coloured cables again, quite funny seeing them come out of a modern black PSU. They are definitely an upgrade over the previous cables; the old PSU had 18AWG for everything. This time, I have 18AWG for most cables; however, 16AWG for CPU power and PCI-E power. While on the face of the power requirements are fine for 18AWG, 16 will help keep things a bit more stable under heat as I push harder. I'm still in the process of cable tying them; however, I'm conscious that I'm not completely cable tying them down as I want some room to manoeuvre for the final assembly!

The large hole cutout in the mid-panel has allowed me to route the ATX and CPU 12v cables quite eloquently, for the sort of case we are dealing with.

You will note the addition of the HD3850 in there now. I am keeping it air-cooled as it will allow me to swap between video cards easily.

With regards to the water cooling hoses, I thought I would try the 45-degree angles to keep the hoses out of the way of the mounting hardware and to make sure I'm not applying pressure onto the other socket. It's more cumbersome than I thought, so I think I will swap them for 90-degree fittings.

For the other case panels, the plan is to sand them all down smooth and make the side window cut out in the side panel. Although I'm going for a fun colour, I'm not going super crazy on the finish. I will be using my automotive kit to do it, all in 2k, so it should be a durable finish. Just need to find some 2k plastic primer for the ABS where I've gone a bit far with the scotch brite pads, then it'll be 2k epoxy primer, 2k base, then 2k clear. Not going to go as far as making it look like glass, but it should be a pretty good finish when done. Attached a picture of my 911 engine when I painted that; that will be the sort of finish I'll be looking for.

Legendary thread 👍 . That retro re-cabling of the PSU is epic, first time I've seen something like that 😀 .

TELVM wrote on 2026-07-12, 18:22:

Legendary thread 👍 . That retro re-cabling of the PSU is epic, first time I've seen something like that 😀 .

Thank you 😀 Just trying to do the board justice!

I decided to be brave this morning and do some drilling and cutting.... Here's my mock-up. The rad fits. I've cut some tube to length for the pump, but not getting too far ahead of myself as I need my new 3d printed drive rails to arrive. I need to do some finishing to the holes, then the top cover is ready to paint later. Ditto the bottom cover; I have drilled and test-fitted the new feet this morning, and they work OK. I may fit some U-channel rubber trim around the fan holes in the top, just to get the grilles to butt up a little bit tighter. I used the 120mm holesaw, which gave me great accuracy but not much control over the size of the hole. The alternative would be to get a grille 3D printed, which I suppose would give me a better fit and finish, given I've got full control over the CAD drawings. Overall, I'm very happy with how that is turning out.

The 6mm 6-32 UNC screws are a bit too short to bolt the radiator on now. I have ordered some black cap head screws (9.5~mm long), with some 0.5mm washers to pack them out. The Silverstone mesh is 3mm thick.

A stroke of madness or genius - I have printed new fan mounts in TPU, which is obviously quite soft and malleable. The theory here is that the fan mounts will have some vibration-dampening properties and behave a bit like a gasket against the chassis, reducing turbulence from the fans compared to the OEM mounts.

Next up is the side panel cut out. I will be spending some time cutting and sticking CAD templates before I do actual cutting, unfortunately.

In the purchasing department, I've ordered a bunch of 90-degree fittings to tidy the waterblocks up

In summary, it's amazing what some contact adhesive and a printout can help you achieve.

Cut the side panel and prepped for paint. Doing the template was very fiddly and time-consuming; alas, it's turned out rather well. I hand-cut it with a jigsaw, finished some of the edges with a file, and then used a sanding flap thing on my die grinder. I've notched the acrylic for the handle too.

For the plastics, I sanded out the scratches to avoid filler. I then used my shot blaster to get a uniform finish across the whole thing. I should note that I am not looking for a show-car mirror finish on the plastics; I do want to preserve some of the faithful "mottle" effect. The metal panels have seen the shot blaster and then been surface finished with the DA and some 180 grit discs.

Unfortunately, I couldn't get a coat of primer on. The fluid knob for my paint gun has gone walkabouts, so I've ordered another to arrive tomorrow. With a bit of luck, I'll have the thing painted tomorrow.

The paint is all 2k solvent, apart from the plastic primer. So once the plastic primer has flashed off, everything will get coated in 2k epoxy primer. This should mean that when the 2k base coat goes on, everything should be the same shade - aka the plastics won't be a different colour. Then I have some 2k clear to go on after.

First coat of epoxy primer tonight... Not good. I don't think I cleaned the air cap properly on the gun, so it's caused some contamination (fish eye) in the paint. I'll have to let the primer cure, then sand it off and start again. I've just deep-cleaned everything in thinners; now I just need to order more. Trials and tribulations.....

.

If you ever start tripping Over Current Protection:

This feature of your case might come handy:

geordiepingu wrote on 2026-06-18, 18:11:

... Not a lot of dragons around seem to be dual-PSU variants; mine is ...

.

Another option available (to no holds barred tinkerers) would be to modify the mobo, adding a "P4" connector, to feed the CPUs with +12V juice (thus taking a great weight off the +5V rail's shoulders) ...

Vogons: It is possible to mod the A7N8X to use the 12V rail ...

Running socket A mainboards on modern PSUs

TELVM wrote on 2026-07-15, 23:50:

If you ever start tripping Over Current Protection:

Don't panic; everything is fine there.

Oooops, my mistake. Guess I need new glasses.

A day of triumph and setbacks again. I got the loop plumbed in and decided to test it (with all the components unplugged, of course). It survived a few hours and didn't leak, so I decided to plug everything in. The system belched into life, and would you have it, my CPU temps went from 58c to 38c. Magic.

I then spent 4 hours trying to find a set of drivers for the HD3850 that would play nicely with the 760MPX, where I eventually got a 3DMark01 score of 14400, which felt a bit pants. I then discovered there are much faster drivers out there, of which, by the time I discovered this, a bit of a setback occurred... But not before another triumph.

Being big and brave, I thought I should test my motherboard cut out and set the multiplier to 15x (15x150). 10-15 mins of soldering some tinned wire to the back, the system immediately belched into life. I had set the FSB to 133, so the system identified itself as having Athlon 2600MPs. I then ramped it back up to 150, increasing the volts for stability as required, getting it to 2x 2250MHz. Not bad for the first go. I think I will set the multiplier higher - perhaps to 17 (17x133 = 2800+/2261MHz). This will give me room to get up to 2550MHz via FSB speed. Hope the chips will go that fast, I don't want to lose the memory speed I've gained getting 150MHz FSB stable!

This is where I manage to get Catalyst 8.12 installed and stable. Except when I come to reboot the system, a strange oddity in the BIOS appears. Athlon XP 2250MHz. And then an error that I cannot install an XP with an MP, and such it will identify as an XP. Bummer. Looks like one of my L5 bridge mods hasn't lasted, and I can see which CPU looks suspect. Having re-visited the TDS for the MG overcoat pen I used to fill in the "ground plane", I haven't let it fully cure before overcoating with the conductive paint. What I suspect has happened is that over time, the thin layer of silver paint on top has shifted due to the curing process not completing on the substrate. With it being the sort of die where you have to scrape the solder mask and the pins are super tiny, the substrate from the overcoat pen wasn't that big to begin with, really. So I have started the L5 bridge mod on my spare 2500. You live and learn...

Still stoked I hit 2x 2250MHz on this build. I dare say, with my research on the net, that's serious for a K7D rig. Can't be too selfish, I've had other wins this week. Finally got my G4 Cube swapped over to an SSD and working.

I still haven't sorted the paint. I've been aching with the flu I've got at the minute, so I stayed indoors.

I decided to partially document my process for doing the L5 bridge mod, mostly so someone else doesn't make the same mistake.

The real lesson here is - ALWAYS read the Tech Data Sheet for any coatings! Be sure you have appropriate apparatus to cure, I have a fan oven in my kitchen, and occasionally my better half knows that she has to leave me alone to cook the chips...

This is a process suited to those who don't have access to microsoldering tools or the skills, and need something a bit more resilient than hoping some graphite doesn't rub off the CPU.

1. Identify if the pins are hidden beneath soldering mask - a microscope is very helpful for this. My setup is not fancy; a microscope that must be from the 70s with an iPhone torch held beside it will suffice.
2. Remove any solder mask for the pins you want to bridge. I used a sewing pin very gently to abrade the surface and verified it by comparing it with the other pins under the microscope.
3. I used plenty of flux and solder to clean the surface of the pins and ensure they are conductive.
4. Using some frog tape (the hard edges) pressed down firmly against the laser-cut valleys, I used an overcoat pen to fill the void with some acrylic lacquer (MG 419D-P-BK). I left this at room temperature for 10 minutes before baking in the oven for an hour at 65c. I wrapped the CPUs in foil, leaving the painted area exposed, to ensure the heat is concentrated in the right area.
5. After leaving the CPUs to cool down, I removed the masking carefully and reapplied some new masking for the pins I'd like to be conductive (just the one L5 pin closest to the centre of the chip).
6. Using some XeredEx silver conductive paint (XD-120) that I acquired from Amazon, I applied the paint on the bridge.
7. Cover the CPU in aluminium foil, leaving ONLY the conductive paint exposed. Bake at 150c for 5 minutes, leaving it to cool down in the oven naturally for a further 30 mins.
8. ????
9. Enjoy the Athlon MPs....

I've just re-done the dodgy 2500 and the spare 2500 using this method.