The manual for your I/O card mentions a couple of jumpers used to set the DMA Channel Select. By default it's on DMA 1, have you tried setting that to 3? It could be conflicting with your sound card.

It's the very last setting in the manual you posted.

EDIT: also, keep in mind that setting your NIC to IRQ3 can conflict with the default COM2 port interrupt. It could be a good idea to disable COM2 and LPT1 if you're not going to use those...

The IRQ's listed as "(Reserved)" in MSD are actually free to use (they are "reserved" for future use by additional interface cards). Sometimes the terminology used in these old systems is confusing...

The best rule to follow with non-PnP systems is to avoid assigning resources that are likely already used to any additional interface cards. Yes, that may mean that you need to make a little table listing resources to figure out what remains free to use if you happen to have a lot of cards you want to use...

TheMobRules wrote:

The manual for your I/O card mentions a couple of jumpers used to set the DMA Channel Select. By default it's on DMA 1, have you tried setting that to 3? It could be conflicting with your sound card.

It's the very last setting in the manual you posted.

EDIT: also, keep in mind that setting your NIC to IRQ3 can conflict with the default COM2 port interrupt. It could be a good idea to disable COM2 and LPT1 if you're not going to use those...

Bingo! Thank you very much. As soon as I've changed I/O's DMA to 3, I got the sound working.

Frankly, I didn't know that I should be looking out for DMA conflicts as well as IRQs.

jheronimus wrote:

Frankly, I didn't know that I should be looking out for DMA conflicts as well as IRQs.

IRQ, DMA, and I/O adresses. Welcome to the wonderful world of before Plug and Play. 😎

Although Windows 95 era Plug and Play was often nicknamed "Plug 'n Pray" or "Plug 'n Pay" because it was pretty buggy at times...

What are JP16-17 set on the motherboard? It should be set for DMA channel 3 if its for the ECP Parallel Port. That would solve the DMA channel 1 conflict with the Sound Blaster.

I don't think I've ever seen anyone have so many problems with a VLB 486 and keep on it. Good job! When you've resolved all your issues, try adding a VLB SCSI controller 😉

Chaintech was a decent manufacturer. Interesting that your board doesn't have a ZIF. That would be the first thing I'd desolder and replace if it were mine.

For VLB/ISA-only 486 boards, the Asus VL/I-486SV2GX4 is probably the best. It compromises on nothing, but needs the BIOS update for Am5x86 support.

feipoa wrote:

I don't think I've ever seen anyone have so many problems with a VLB 486 and keep on it. Good job!

Don't know if that's a good thing. 😀 It's been going pretty slow (too many "real life stuff" happening in the background), but I like working on this system. Also, I hope other people might find my post useful, 486 systems seem to get "trendier" by the minute.

I guess the main takeaway is that it's pretty hard to learn a platform older than a Socket 7 by building from scratch. It would be much easier to start with a working system and just add whatever I want. Besides, this build simply wouldn't be possible if I didn't have a lot of alternative parts. I planned on a VLB system, but in the end I used ISA I/O and VGA cards for testing, not to mention the numerous NICs, HDDs, floppy drives and SIMM modules I had to go through to find a working combination.

Chaintech was a decent manufacturer. Interesting that your board doesn't have a ZIF. That would be the first thing I'd desolder and replace if it were mine.

Frankly, I don't like this motherboard at all, but it's the best I could find cheaply. My 5x86 is still working at 3x33, the Turbo button connector doesn't fit (the motherboard only has two pins, but the case has a triple wire connector). I really hope to find another SIS471-based board and ASUS VL series is definitely at the top of my list.

If you find one for under $100, its well worth it.

jheronimus wrote:

the Turbo button connector doesn't fit

Obviously you don't know how to connect the turbo header, you need to connect two of the three wires and depending on which ones you'll get the high speed with the button pressed or not.

Soooo, it basically took me almost a year to finish this build. Here it is, in all its glory:

There are still things I need to do, like setting up the turbo LED indicator once I figure out the turbo speed (seems about 486DX@33, btw) or adding a network controller. The 5.25 drive isn't connected, actually — I just didn't have a nice-looking 5.25 bay cover, but I don't think I'll ever use 5.25 disks for data transfer on this system. But other than that it works fine.

The motherboard I'm using now is QDI V4P895GRN/SMT V1.1.

It's based on an OPTi 895 chipset. Doesn't support 5x86 natively (so it's AMD DX4 in POST), but it works correctly according to CHKCPU. 4xSIMM72, coin battery, ZIF socket, AMI WinBIOS — I really like this board.

Full specs are:

- AMD 5x86 at 4x33 MHz, 256KB L2 cache;
- 4x4MB RAM;
- Cirrus Logic CL-GD5428 with 2MB RAM;
- Sound Blaster 16 CT2230 with DreamBlaster X2;
- Roland MT-32 via SoftMPU;
- 1.6GB Samsung drive connected to a VLB controller;
- 16x CD-ROM.

Some benchmarks:

3DBench 1.0c: 64.2
Chris's 3D Benchmark (640x480): 20.3
PCP (640x480): 5.1
Doom timedemo: 35 FPS (2134 gametics in 2117 realtics)
Speedsys: 48.12

Doom result seems kind of low to me — there are 5x86/VLB systems on this table that score almost twice as much.

Very nice system, I really like that case!

Regarding the performance: I don't know how fast that Opti chipset is, maybe that is affecting the performance? Also, can you post BIOS settings such as memory timings and all that?

Very nice system, I really like that case!

Thanks!

Here are the BIOS settings:

WinBIOS. Excuse me while I throw up a bit in my mouth. 😵

jheronimus wrote:

- 4x4MB RAM; ... Doom timedemo: 35 FPS (2134 gametics in 2117 realtics) ... Doom result seems kind of low to me […]
Show full quote

- 4x4MB RAM;
...
Doom timedemo: 35 FPS (2134 gametics in 2117 realtics)
...
Doom result seems kind of low to me

They are kinda low, it compares to the 34,8 FPS I get when I put in an intel 486DX4 and run it at 66MHz (2x33). Did you run Doom with 1 level of green border?. What about your memory: 60ns or 70ns? / waitstates?

In Chipset Setup, set Auto Config to Disabled and then try tightening those DRAM and Cache timings. I have a motherboard with that same BIOS that defaults to the slowest possible settings when Auto Config is Enabled (different chipset though).

that horizontal desktop case definately reminds me of the systems i remember from my teen years! (1990-1995)
unfortunately that monitor is reminding me about how i dropped off two sony trinitrons of similar make to recycling center back in like 2008 or so..
dumb dumb dumb.. should have kept them both;)

TheMobRules wrote:

In Chipset Setup, set Auto Config to Disabled and then try tightening those DRAM and Cache timings. I have a motherboard with that same BIOS that defaults to the slowest possible settings when Auto Config is Enabled (different chipset though).

What options should I pick? Never done this before

Very nice looking computer!

As one user already mentioned - to be able to compare your scores in Doom1.9 to the ones from Anton's complang site you have to decrease the screensize by two(HUD + 1 green border -> 72.2% viewable area which of course is not directly translatable into a 138% speedup). Your score is low even for fullscreen which could prove that the Opti chipset in combination with slow bios settings is killing your framerate.

Regarding waitstates... the lower the better but practicality depends on the quality of the chips used. As far as I remember you're using a 33mhz FSB so the "smallest" settings should be feasible.
Your cache has 15ns(cache+tag) latency which should work up until ~60mhz. With your slow 33mhz bus there shouldn't be a problem... 25ns cache could even be adequate
(duration of one tic for 60mhz fsb -> 1 / 6x10e7 = 10e-7 / 6 -> ~0.167x10e-7 -> ~16.7x10e-9 seconds -> ~16.7ns latency minimum BUT there's always overhead and the tag chip has to be faster).

Overall - try the minimum value and then go up if instabilities arise.