PD2JK wrote on 2026-03-12, 18:31:

Tested out an Asus X800 XT AGP I recently got, turns out the previous owner did some BIOS flashing to PE.
Free from artefacts in 3DMark03, so I'm happy.

The attachment 39215 (1).jpg is no longer available

Funny enough I have the PCI-E version of the X800 made by ASUS, came with a recent Shuttle SB81P I bought that's a nice time capsule.

I wonder if the EAX800XT I have can do XTPE speeds.

Returned to repairing the Roland SC-55 I had on my bench a few weeks ago. This was the unit with the faulty resistors on the rear MIDI input.

The resistors are in a really awkward location right next to the giant plastic MIDI ports. I didn't want to use hot air or a regular soldering iron to try to remove them.

I'd been waiting on getting tips for my new soldering tweezers. After several weeks on backorder, they finally arrived.

Using these soldering tweezers for SMD removal is a dream. The two resistors came off with virtually no effort and no risk of damage to adjacent components. Pads cleaned up well too.

Soldered the new resistors in place and did the MIDI input test. It finally passed. Did some subsequent listening via the rear input and everything seems to be working fine.

I still need to replace the tactile switches on the front panel, but I'll save that for another day.

Wrapped up my 9500Pro repair. It got real challenging in the end 😁

And I found some DRAM that is literally sending mixed signals?

Fixed the AIO dating from 2012.
Lots of gunk I had to flush and then filling it to the max.
From 104°C max to 32°C idle and 54°C at full (i7-3820 - 130W beast)

Pretty happy as I discovered watercooling while looking for stuff 😀

tehsiggi wrote on 2026-03-13, 13:40:

Wrapped up my 9500Pro repair. It got real challenging in the end 😁

And I found some DRAM that is literally sending mixed signals?

Well that's certainly a new one. How does that slip past quality assurance?

Solved my crashing problem on my 3DFX Athlon XP Win98SE PC by replacing the SB Live with an Audigy 2ZS. It really was that bug. Now it runs smoothly again. That OS install is sure well-travelled. I was planning on a clean reinstall for obvious reasons, but it seems that will really not be necessary after all. Shame I can't use the Live Drive anymore...

Also, after 10 years of having it, I realise the cooler on my Voodoo 3 isn't hard to remove at all. I thought it was one of those that are impossible to remove without putting it in the freezer, but just putting a box cutter between the chip and heatsink after having removed the push pins and warmed it up with a a little use is enough. I know that thermal pad might have been way better than whatever thermal paste I use, but at least I can replace the whole cooler now with something better, I wonder, would one of those fancy PTM7950 pads be ideal for a Voodoo3? I think it should, considering these things like to run hot, and the pads get better with each thermal cycle.

Did more work on my SC-55 sound modules.

I replaced all the tactile switches (except power) on both SC-55 modules and the SC-55 mkII. Also replaced the 2200uF capacitor on the mkII for good measure.

Removing the switches was tedious since all them were soldered on with pins bent in random directions. Took a lot of patience in removing them.

After doing all this, I thought I was done with work on these units. However, in testing, I discovered one of the units has a wonky MIDI 2 input. It's the SC-55 that originally had the damaged resistors on the MIDI 1 input.

Unfortunately I didn't thoroughly test its MIDI 2 input originally. The internal test reports it's fine, however, in playing back MIDI files I get occasional stuttering/lag and dropped notes or instruments. MIDI 1 input works perfectly fine.

It seems partially dependent on which MIDI files I use. Some it's hardly noticeable, others it's more obvious.

The MIDI 2 circuit goes through a completely different path than MIDI 1. While MIDI 1 is more-or-less connected directly to the CPU via a single trace, the MIDI 2 input goes through serial-to-parallel conversion via this MB89251A chip, before going into the CPU via 8 data lines.

I tested the second SC-55 I have, and both its MIDI inputs are fine. However, it's also using a later firmware (v2.00) while the one with bad MIDI 2 input is using v1.10.

I don't know if this is potentially a CPU/firmware related issue, or if it's something wrong with the MIDI 2 input circuitry. I plan to test and compare circuits versus the working SC-55 and see if I can pin down any differences.

TheAbandonwareGuy wrote on 2026-03-14, 00:17:

Well that's certainly a new one. How does that slip past quality assurance?

No clue. It's only of the top side, which leads me to believe that top and bottom memory are being fed by different reels in production, which is cool to know.

I used the lunch break for measuring some more AGP cards and their consumption.

Now that I've measured it, I'm a bit baffled what they were thinking. Full size PCB? Okay.. fine. But they draw 3oW at max and at max 13W from 12V from the external power coming it. Which translates to 1A on 12V. I'm pretty sure they just wanted to use the exact same BOM as for the PCIe variant, where they feed their VRM from 12V for the GPU. And as they wanted to make sure the AGP slot with it's single pin for 12V isn't an issue, they just went with the additional external power input.

The 3.3V from the AGP slot are used for the memory and for rialto it seems. What a weird weird card.

Don't mind the soft ball as a card mount, I'm a professional.

I think it's great that some of you disassemble the vintage synth modules and refurb them with replacement parts to keep them going. I'd certainly do that if I had the hardware and knowledge. I wonder if anyone is making money doing that.

Got hold of some laptops and setup a Counter Strike LAN with my old Clan mates, we played the very first CS Beta 1.0 3.0 and 6.0. How much this game has changed from a little half life mod, video link below

https://www.youtube.com/watch?v=XOBVxrzkSUc

Spent the last 2 days messing around with my DOS rig. Playing and testing a few different games. Messing with the sound config a little bit. Moving some files around to make it a bit more organized. We're going to be moving in the near future and I wanted to get my "retro" systems sorted, Win98 rig is next. New place has much more room for my rigs and hobbies.

Can't go anywhere or do anything right now. I live in NE Wisconsin and Blizzard Elsa has dumped two feet of snow outside, and 40-50 MPH winds have pushed drifts much higher. So, I just fired up my DOS system and decided to ride it out in warm nostalgic memories. 🤣

I did a bench on my recently built Pentium 4 rig
It has socket 478 Pentium 4 Northwood CPU at 3.4 Ghz
Pixelview Nvidia GeForce 4 MX440 8x AGP 128MB
1GB Apacer/Hynix/Kingston DDR 400 RAM at 200Mhz
Western Digital VelociRaptor WD5000HHTZ 500 GB SATA (partitioned most of the alphabet in drive letters)
Operating system is Windows Millennium
I chose 3DMark 2001 for the DirectX 8 tests of this cards ability to run it but I believe it is originally a DirectX 7 card and PCMark 2002 pro for the CPU. Period correct (except for hard disk.)
I think its a great system for DirectX up to 7 but it runs as a qualified V8 and V9 too no doubt it'd have effects unavailable.
The bench of the SATA HDD leaves all the IDE drive's I've had for dead!
Perhaps one day I'll spot a DirectX 9 card but I am happy for now.
Today I expect the arrival of an IDE DVD Writer to complete the rig so I can finally close it up and say finished!
Good for IRC with an old mirc client, Protoweb browsing with retrozilla, playing videos and music lots of cool games too.
Although Windows Me supports 1024mb RAM I still used R. Loews memory patch to change the AGP memory address to better support a gigabit LAN card I have. Runs so fast and has tons of storage its like a dream come true!

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It isn’t a perfect system, but I’ve had a ton of fun already just putting it through its paces. I also found a small toggle push button that fits the case’s front panel. I soldered one of these buttons to the ATX-to-AT wire harness and routed the power switch lines to the front panel which gives the case a proper AT-style power button.

Final Configuration:

Biostar MB-1333AEA-G/Q (386/ISA board)
Ti 486DLC-40 Processor
Cyrix 83D87-40 Fasmath Co-processor
32MB Parity RAM (8x 4MB, 60ns, 36-pin modules)
Acer MIO-400KF Super I/O Card (2x serial ports, floppy controller, and Primary IDE interface)
Future Domain IDE-16011 Secondary IDE Interface Card
Diamond Speedstar Pro ISA Graphics Card (CL-GD5426 BIOS)
Realtek RTL8019AS 10BaseT Network Card (hosting a DTC Enhanced BIOS EEPROM)
Media Vision Pro Audio Spectrum 16SL
PicoGUS Hosting a Wavetable-PI
Voltage Blaster
Unbranded SD-to-CF Card (boot drive via a CF-to-IDE adapter in a 3D-printed mount)
Toshiba SD-M1202 (5x DVD-ROM drive – can read some DVD+R discs)
GOTEK Floppy Emulator (with FlashFloppy firmware)
DeepCool Black ATX Case
Generic 350W ATX PSU

So far, I’ve got installs of DOS 3.31, 4.01, 5.0, and 6.22 running on this system with full functionality. Windows for Workgroups 3.11, Windows 95, NT 3.1, NT 3.51, and NT 4.0 also have full functionality. Unfortunately, Windows 3.0 will not boot into Enhanced Mode on this system (I’m still troubleshooting). OS/2 2.1 and Warp 3.0 also crash randomly on this system causing HD corruption. I have not been able to resolve that issue yet.

Overall, this has been really fun since I didn’t build my own 386 or 486 back in the day. Does everyone need a 386DX/486DLC…probably not. But it brings back a lot on nostalgia from my high school days when I had a 386SX pre-built with 4MB of RAM. This system is way better than the one I had back then.

fosterwj03 wrote on 2026-03-17, 02:24:

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It is […]
Show full quote

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It isn’t a perfect system, but I’ve had a ton of fun already just putting it through its paces. I also found a small toggle push button that fits the case’s front panel. I soldered one of these buttons to the ATX-to-AT wire harness and routed the power switch lines to the front panel which gives the case a proper AT-style power button.

Final Configuration:

Biostar MB-1333AEA-G/Q (386/ISA board)
Ti 486DLC-40 Processor
Cyrix 83D87-40 Fasmath Co-processor
32MB Parity RAM (8x 4MB, 60ns, 36-pin modules)
Acer MIO-400KF Super I/O Card (2x serial ports, floppy controller, and Primary IDE interface)
Future Domain IDE-16011 Secondary IDE Interface Card
Diamond Speedstar Pro ISA Graphics Card (CL-GD5426 BIOS)
Realtek RTL8019AS 10BaseT Network Card (hosting a DTC Enhanced BIOS EEPROM)
Media Vision Pro Audio Spectrum 16SL
PicoGUS Hosting a Wavetable-PI
Voltage Blaster
Unbranded SD-to-CF Card (boot drive via a CF-to-IDE adapter in a 3D-printed mount)
Toshiba SD-M1202 (5x DVD-ROM drive – can read some DVD+R discs)
GOTEK Floppy Emulator (with FlashFloppy firmware)
DeepCool Black ATX Case
Generic 350W ATX PSU

So far, I’ve got installs of DOS 3.31, 4.01, 5.0, and 6.22 running on this system with full functionality. Windows for Workgroups 3.11, Windows 95, NT 3.1, NT 3.51, and NT 4.0 also have full functionality. Unfortunately, Windows 3.0 will not boot into Enhanced Mode on this system (I’m still troubleshooting). OS/2 2.1 and Warp 3.0 also crash randomly on this system causing HD corruption. I have not been able to resolve that issue yet.

Overall, this has been really fun since I didn’t build my own 386 or 486 back in the day. Does everyone need a 386DX/486DLC…probably not. But it brings back a lot on nostalgia from my high school days when I had a 386SX pre-built with 4MB of RAM. This system is way better than the one I had back then.

very cool, and the contrast with a more typical 386sx with 4mb ram of the time shows how maxxed out things can be made now 😀

gerry wrote on 2026-03-17, 10:37:

fosterwj03 wrote on 2026-03-17, 02:24:

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It is […]
Show full quote

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It isn’t a perfect system, but I’ve had a ton of fun already just putting it through its paces. I also found a small toggle push button that fits the case’s front panel. I soldered one of these buttons to the ATX-to-AT wire harness and routed the power switch lines to the front panel which gives the case a proper AT-style power button.

Final Configuration:

Biostar MB-1333AEA-G/Q (386/ISA board)
Ti 486DLC-40 Processor
Cyrix 83D87-40 Fasmath Co-processor
32MB Parity RAM (8x 4MB, 60ns, 36-pin modules)
Acer MIO-400KF Super I/O Card (2x serial ports, floppy controller, and Primary IDE interface)
Future Domain IDE-16011 Secondary IDE Interface Card
Diamond Speedstar Pro ISA Graphics Card (CL-GD5426 BIOS)
Realtek RTL8019AS 10BaseT Network Card (hosting a DTC Enhanced BIOS EEPROM)
Media Vision Pro Audio Spectrum 16SL
PicoGUS Hosting a Wavetable-PI
Voltage Blaster
Unbranded SD-to-CF Card (boot drive via a CF-to-IDE adapter in a 3D-printed mount)
Toshiba SD-M1202 (5x DVD-ROM drive – can read some DVD+R discs)
GOTEK Floppy Emulator (with FlashFloppy firmware)
DeepCool Black ATX Case
Generic 350W ATX PSU

So far, I’ve got installs of DOS 3.31, 4.01, 5.0, and 6.22 running on this system with full functionality. Windows for Workgroups 3.11, Windows 95, NT 3.1, NT 3.51, and NT 4.0 also have full functionality. Unfortunately, Windows 3.0 will not boot into Enhanced Mode on this system (I’m still troubleshooting). OS/2 2.1 and Warp 3.0 also crash randomly on this system causing HD corruption. I have not been able to resolve that issue yet.

Overall, this has been really fun since I didn’t build my own 386 or 486 back in the day. Does everyone need a 386DX/486DLC…probably not. But it brings back a lot on nostalgia from my high school days when I had a 386SX pre-built with 4MB of RAM. This system is way better than the one I had back then.

very cool, and the contrast with a more typical 386sx with 4mb ram of the time shows how maxxed out things can be made now 😀

Yeah, a maxed-out 386 build would have seemed crazy stupid back in the early 90’s. The added cost of the excess RAM could have instead paid for a well appointed 486 DX2 with VLB cards back in the 1993-1994 time period (my target era for the vintage parts in this build). A substantially better processor and bus would have made far more practical sense than the extra RAM given the dramatically high prices back then (these days, we act like RAM shortages never happened before).

Today, though, it’s just a hobby to build a vintage system that would have made no practical sense back in the 90s. Back then, I didn’t know anyone who had more than one family-owned computer in their household. Today, I have a maxed-out 486 platform and a maxed-out Intel Pentium MMX machine to say nothing of the multiple modern PC’s actively used in my household. It is a different world.

This wasn’t a cheap project though. I only had the ISA network card and case as spare parts when I began. I had to buy all of the other parts as needed. I figure this system’s cost came to about $800 US all-in. Not horrible, but well over twice the cost of a Pocket 386 with accessories.

I didn’t understand the PC architecture back then (being a newbie) as well as I do now. I really like the simplicity of the ISA system and configuring the cards now. That stuff (IRQs, DMAs, address space, and memory optimization) seemed so difficult early on, but now it really wasn’t any trouble getting things working and working together well. My biggest problems actually stemmed from trying to get modern parts (a SD-to-IDE adapter in particular) working correctly with the vintage parts. All of the vintage cards, themselves, seem to have no issues with each other once I correctly configured them.

fosterwj03 wrote on 2026-03-17, 22:23:

gerry wrote on 2026-03-17, 10:37:

fosterwj03 wrote on 2026-03-17, 02:24:

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It is […]
Show full quote

After many months of acquiring parts and testing configurations, I finally wrapped up my Ti 486DLC reverse-sleeper build. It isn’t a perfect system, but I’ve had a ton of fun already just putting it through its paces. I also found a small toggle push button that fits the case’s front panel. I soldered one of these buttons to the ATX-to-AT wire harness and routed the power switch lines to the front panel which gives the case a proper AT-style power button.

Final Configuration:

Biostar MB-1333AEA-G/Q (386/ISA board)
Ti 486DLC-40 Processor
Cyrix 83D87-40 Fasmath Co-processor
32MB Parity RAM (8x 4MB, 60ns, 36-pin modules)
Acer MIO-400KF Super I/O Card (2x serial ports, floppy controller, and Primary IDE interface)
Future Domain IDE-16011 Secondary IDE Interface Card
Diamond Speedstar Pro ISA Graphics Card (CL-GD5426 BIOS)
Realtek RTL8019AS 10BaseT Network Card (hosting a DTC Enhanced BIOS EEPROM)
Media Vision Pro Audio Spectrum 16SL
PicoGUS Hosting a Wavetable-PI
Voltage Blaster
Unbranded SD-to-CF Card (boot drive via a CF-to-IDE adapter in a 3D-printed mount)
Toshiba SD-M1202 (5x DVD-ROM drive – can read some DVD+R discs)
GOTEK Floppy Emulator (with FlashFloppy firmware)
DeepCool Black ATX Case
Generic 350W ATX PSU

So far, I’ve got installs of DOS 3.31, 4.01, 5.0, and 6.22 running on this system with full functionality. Windows for Workgroups 3.11, Windows 95, NT 3.1, NT 3.51, and NT 4.0 also have full functionality. Unfortunately, Windows 3.0 will not boot into Enhanced Mode on this system (I’m still troubleshooting). OS/2 2.1 and Warp 3.0 also crash randomly on this system causing HD corruption. I have not been able to resolve that issue yet.

Overall, this has been really fun since I didn’t build my own 386 or 486 back in the day. Does everyone need a 386DX/486DLC…probably not. But it brings back a lot on nostalgia from my high school days when I had a 386SX pre-built with 4MB of RAM. This system is way better than the one I had back then.

very cool, and the contrast with a more typical 386sx with 4mb ram of the time shows how maxxed out things can be made now 😀

Yeah, a maxed-out 386 build would have seemed crazy stupid back in the early 90’s. The added cost of the excess RAM could have instead paid for a well appointed 486 DX2 with VLB cards back in the 1993-1994 time period (my target era for the vintage parts in this build). A substantially better processor and bus would have made far more practical sense than the extra RAM given the dramatically high prices back then (these days, we act like RAM shortages never happened before).

I have some of what might be an original example of a maxed 386, a DX board, CPU was pulled and I haven't come across one. I got it in an eBay box lot several hundred moons ago, long before retro interest. It came with 8x4MB on it, 9 chip modules, so likely not a mid 90s last gasp. I do not know what it was used for, early PC CAD, netware box or what. Anyway, someone, probably pre-1992, needed a crap load of RAM for the time.

1992 was a bit of a pivotal year. 486 was released a few years earlier, yes, but 1992 is when Compaq started a price war that dragged 486 system prices down from the stratosphere, grazing space, to somewhere more breathable for home office systems. Basically by the end of 1992, 486 systems cost what 386 systems did at the start of 1992. In 1991, they were for sale, yes, but prices were extremely eyewatering. If you had shelled out $5000 for a high end 386 in recent preceding years, you were more likely to shovel $1000 of RAM into it vs paying another $5000 plus whatever for the RAM for the 486... which didn't have 486 killer apps yet, and until DX2 "looked about as fast" on old code. Plus at that point in time, many 486 took the same 30pin RAM, it was no biggie if you could swap it over later.

At this earlier time, VLB might have seemed a bit experimental, as did every other higher speed bus contender. Early adopters would have suffered from "ISA card on a VLB" implementations, meaning the components were not taking much advantage of the bus. Through your retrospectacles, it's easy to say VLB was the way to go from 92-95, but standing in 1990-1992, there was a lot of noise from competing technologies, so the choice was not simple.

Having said that, there's not much that's playable on a 386 that is improved by 8MB vs 4MB. 8MB maybe works better with some EMS hungry sound setups. So then 16MB is tending towards overly generous. 32MB is then not worth spending much to achieve, benefits are miniscule for retro. However, with some of the upgrade chips, "486 on 386" particularly clock doubled ones, you do start getting into the low end of 486 and maybe have the instructions to contemplate running things that are nicer on 16MB.

I got a Necroware Dallas RTC replacement for my IBM PS/1, but it doesn't seem to be going into the socket right. Is there something I'm missing here? The original Dallas chip goes back in like a breeze.

https://streamable.com/28nzdy

giantenemycat wrote on 2026-03-18, 16:09:

I got a Necroware Dallas RTC replacement for my IBM PS/1, but it doesn't seem to be going into the socket right. Is there something I'm missing here? The original Dallas chip goes back in like a breeze.

https://streamable.com/28nzdy

Could just be due to a difference in pin size and/or relative wear on the pins.

You could also try some Deoxit on the socket. It can help deal with any corrosion and also acts as a lubricant.

Pins are definitely larger. You could buy some thinner ones instead if you worried about it.

Or do like I did: use some force. 😀 The contact legs of the holder will bent a bit (so the original Dallas RTC will fit a bit loose after that). But I didn't care about that as I would never fit a Dallas RTC ever again on that board. You can also just desolder the holder and solder it directly on the board.

I'd say, even for a 386 or slow 486, more RAM helped with Windows (and GUI-based operating systems) and kept the system from swapping to disk as often. PC Mag had an article in 1994 about RAM availability vs system performance which found that 16MB of RAM was the sweet spot for OS/2 and about 20MB of RAM was the sweet spot for NT.