Nexxen wrote on 2021-12-12, 09:15:

Maybe changing the crystal to match the 486 frequency?
25 x 2

I've tried all sorts of crystal frequencies.

pshipkov wrote on 2021-12-12, 09:30:

I recognize that hanging pattern very well.
On these fast VLSI boards it happens with memory modules that cannot handle 0 ws properly.

At 8.3 and 12.5 MHz, would 0 ws be an issue? I was thinking probably not. However, it does seem that this board doesn't run well with 3-chip memory modules. I tried 3 different sets of 9-chip modules, and the hang-on-idle problem disappeared. I'm not sure why you had optimal luck with 3-chip modules. I also tried 3 different sets of 3-chip modules and they all exhibited the hang-on-idle issue. Unfortunately, using 9-chip modules didn't resolve the inability of floppies to boot when a 286 CPU is installed.

Aside from the 3 I/O cards tested, I also tried floppy booting from an AHA-1522B SCSI card, but again, when a 286 is installed, it cannot boot. Curiously, the system also won't boot to a SCSI HDD on the 1522B controller with a 286 installed; 486 SXLC no problem. IDE and 286, no problem booting. This board has character!

?

What's weird is these symptoms are similar to problems you get when overclocking. Leads me to suspect something wrong with CPU supply voltage.

BitWrangler wrote on 2021-12-12, 14:29:

What's weird is these symptoms are similar to problems you get when overclocking. Leads me to suspect something wrong with CPU supply voltage.

But I've tried two different PSUs. Maybe find some non-Harris 286 CPUs to test?

Maybe, thought they were Intel design on better yielding process, so you'd think should be less marginal than intels would be, but who knows. I feel like there's something "analog" going on, spurious oscillation or something.

About memory configurations.
I have more 3-chip modules than 9-chip ones.
Used whatever ended up working for the 25 and then 27.5MHz base frequency. Very few modules can do it, so one cannot be picky.
Clearly remember the locking issues like what you hit.
Your solution with the 9-chip modules is a great tip.
Doubt 9-chips can withstand higher base freq.

BitWrangler wrote on 2021-12-12, 20:38:

I feel like there's something "analog" going on, spurious oscillation or something.

Could be. If you know specifics of what I should look for w/scope, I'll take a look. I also wanted to try another AMI BIOS from a VLSI board, preferable from the PC Chips (Amptron) version of my motherboard because the layout is identical. If anyone has it, are you able to upload it? Photo of board attached. Jumpers/manual here: http://www.uncreativelabs.de/th99/m/U-Z/30595.htm

pshipkov wrote on 2021-12-12, 23:54:

About memory configurations. I have more 3-chip modules than 9-chip ones. Used whatever ended up working for the 25 and then 27. […]
Show full quote

About memory configurations.
I have more 3-chip modules than 9-chip ones.
Used whatever ended up working for the 25 and then 27.5MHz base frequency. Very few modules can do it, so one cannot be picky.
Clearly remember the locking issues like what you hit.
Your solution with the 9-chip modules is a great tip.
Doubt 9-chips can withstand higher base freq.

I'm only shooting for 16 MHz FSB max, so are criteria is different. Unfortunately, 16 MHz is a no go for the SuperChip. I suspect the onboard 2x PLL can't handle it. Unfortunately, my 9-chips are only 70 ns, but seems perfectly fine.

I've finished the surgery on the SuperChip. I've put the TI SLC back on. Hopefully it still works. Also, I'm attaching a photo of the bottom of the Evergreen 486 SuperChip for anyone who is looking for it.

The surgery went OK. There was only one solder bridge, not coincidentally at the point of the thickest paste blob. Also, I learned best to get the solder paste right in the middle of the pads - there was some no-connects at the point you see the paste line deviate from centre on the SLC chip. All easily fixable with a standard iron.

Attached are a few more images in case I've missed something concerning the floppy issue.

Screen at power-on:

This is the screen it hangs on when trying to boot from a floppy and a 286 CPU is installed:

This is the screen visible when you press DEL to enter BIOS:

These are the only BIOS options:

If I enter diagnostic mode, I can run the floppy diagnostic and it passes:

In case anyone is interested, here are all the BIOS diagnostic options:

If the original CPU is 25MHz, and the module supports clock doubling, then it is obviously designed as an upgrade for a 12MHz 286, so you may be correct that the PLLs can't handle 16MHz. However, once had a Make it 486 with TI486SLC/33 and it worked just fine on my 20MHz board at 20x2=40MHz.

I don't know if there was really much point switching the CPU back to the SLC. More than likely the issue if the board, not the module.

How the hell does this board run at 27.5MHz with no ISA bus clock dividers?

edit: I think my CT386 board has the same 1990 core AMI BIOS. Mine has an extra option for advanced settings to control chipset registers. Is it possible your BIOS has options that can be unhidden?

Anonymous Coward wrote on 2021-12-13, 03:58:

I don't know if there was really much point switching the CPU back to the SLC. More than likely the issue if the board, not the module.

Because the SLC will function with L1 enabled, while the SXLC will not. I wonder if some DIP switch setting will let it run at 33 MHz...?

Anonymous Coward wrote on 2021-12-13, 03:58:

How the hell does this board run at 27.5MHz with no ISA bus clock dividers?

No idea. Maybe pshipkov can answer that.

Anonymous Coward wrote on 2021-12-13, 03:58:

edit: I think my CT386 board has the same 1990 core AMI BIOS. Mine has an extra option for advanced settings to control chipset registers. Is it possible your BIOS has options that can be unhidden?

On my list is to check AMISETUP, but I've a feeling it won't work with this old BIOS. Any other utility I can use to look for hidden settings? Perhaps CTCHIP or TweakBIOS ?

How this board runs at 27.5MHz with 0 WS ?
Like this.
The post wraps the whole story.
Videos included as well.
Don't have oscilloscope, so no idea what the ISA bus frequency is, but perf is unique.
So it does something right there.

---

These boards are just solid and very fast. I am pretty sure they can do 30MHz at 0 WS if there is memory to withstand that.
So all the magic is in the memory.
Tested a lot of SIMM modules, including some modern stuff like this 40ns set i bought from another Vogons member.
While it works fine at up to 25MHz 0 WS - no lights at 27.5MHz with it.

The best set is this VT one, but i have only 2 modules:

Along with it is this "USA" one. For the rig in the post i linked to i use 2 VT and 2 "USA" SIMMs. It is almost fully stable in a cold day without active ventilation:

Second best is this Viking Components set. Solid at up to 25MHz with 0-WS and no active ventilation, but a bit flaky at 27.5MHz:

Interesting looking memory.

It there isn't an ISA divisor jumper, this would imply that your ISA bus is also at 25 MHz? I didn't think VGA and I/O cards could handle that.

I tried the Gotek floppy emulator, but it too would not boot when a 286 CPU was installed. So, not sure how you got it booting.

The FPU appears functional on my SuperChip module, or at least Landmark Speed v2.0 outputs a benchmark result. I'm using Cx87SLC-v25QP for the FPU. These should be the 3 volt variants. Are they OK at 5 V without a heatsink?

It looks like I do not need to enable the FLUSH or BARB pin to benefit from L1 cache. Is this the case when external cache flushing circuitry is used? Enabling the FLUSH pin does not change the performance, but setting BARB reduces it by 20%.

I am able to clock the TX486SLC up to 33.3 Mhz. I'm not sure why I couldn't go above 25 Mhz previously - perhaps the 3-chip memory, or the SXLC was the issue. Taking it up to 33.33 MHz, I get a Landmark v2 score of 105.2 MHz ALU and 41.6 MHz FPU. This translates to a Landmark v6 score of 135.4 Mhz ALU and 36.5 MHz FPU.

Also, cachechk shows 25.1 MByte/s read for the L1 and 11.5 MByte/s for main memory read/write.

I tried to take it up to 35 Mhz, but Landmark freezes.

feipoa wrote on 2021-12-13, 03:54:

Attached are a few more images in case I've missed something concerning the floppy issue. […]
Show full quote

Attached are a few more images in case I've missed something concerning the floppy issue.

Screen at power-on:
286-Start_screen.JPG

This is the screen it hangs on when trying to boot from a floppy and a 286 CPU is installed:
286-Floppy_hang_on_POST_boot_screen.JPG

This is the screen visible when you press DEL to enter BIOS:
286-pre-BIOS_options.JPG

These are the only BIOS options:
286-BIOS_options.JPG

If I enter diagnostic mode, I can run the floppy diagnostic and it passes:
286-Diagnostics_Floppy_Test.JPG

This is a VLSI 200 series chipset - have you verified the ROM shadow jumper is correctly set on the board?

If your FPU is 3V, then running it at 5V might not be a great idea.
The SXLC2 is also a 3V part. Some people on here run them at 5V without issue, but maybe you got one that just couldn't handle it.

The FPU doesn't seem to get very hot, but nonetheless, I've put on order a non-v variant with the 33 Mhz designation. I've also put on order some TX486SLC chips with the 33 MHz designation. Let's see how fast this little SuperChip can go.

Indeed, the SXLC may not like L1 enabled at 5 V. I'm not sure how others got around this. Were there PGA-68 or PLC-68 interposer modules with a VRM onboard?

Re: ROM Shadow - there's but one jumper with 'shadow' in the description and it is 'Shadow BIOS' via J8. I've tried it with and without J8 installed. There's an unspecified jumper, J1, but I've also tried with and without J1. Anonymous Coward had located this jumper page for me a few days ago, http://www.uncreativelabs.de/th99/m/U-Z/30595.htm

In case anyone else is trying to get a 486 SuperChip running with FPU, these are the DIP switch settings I'm using:

1SW8 - on
2SW7 - off
3SW6 - off
4SW5 - off
5SW4 - on
6SW3 - off
7SW2 - on
8SW1 - on

These are the settings found in a photo of another module online which had the FPU installed, however I suspect only SW2 is used to enable/disable the FPU.

Anonymous Coward wrote on 2021-12-13, 15:56:

The SXLC2 is also a 3V part. Some people on here run them at 5V without issue, but maybe you got one that just couldn't handle it.

Do you know the thread?

feipoa wrote on 2021-12-13, 20:56:

Do you know the thread?

No, but I remember the guy put it in one of those 286s with the Sega Megadrive hardware built in.

BTW, there are no 68-pin 3v interposers for 286s. Even if there were, the support chips on your module would still require 5V.

Probably, but they might just work at 3.3 V anyway.

Here's some kind of upgrade module with an SXLC2-50 on it, which has an interposer. Any idea what this is called? Image is from archive.org.

They made these as upgrades for both 286 and 386 systems. If I had to guess, this is a clip on 386SX upgrade. It seems to lack enough glue logic to make it work on a 286.

These were made by all the major upgrade vendors. I'd have to compare with the one I have, but it looks an awful lot like my Improve Technologies Make-It-486.

Just seen this thread again. I'll try and dig mine out in the next few weeks and take another look. I've been meaning to ever since CPU galaxy did the video on enabling the L1 cache, which I now realise was never enabled on mine.
I know mine used to refuse to post at a 16MHz "FSB" but worked fine at 12MHz. Getting it up to 16 would be really nice.
I'll also order an FPU for it, because why not.
I can find a Cx87SLC-33QP, that should be fine, shouldn't it? No V, means its 5v and 33MHz is faster than I'll be running it, so is ok?
Cheers
Andrew