Ahhh, you have it; many thanks! I've heard of some people having issues with NSSI VGA extractions, so I'll keep my fingers crossed. I'm a little busy with other household matters at the moment, so I'm not sure when I can get to the BIOS. I've saved it. I opened it briefly with a HEX editor and at least he ASCII looks good. lol.

Been messing around for a while with FIC PA-2013 version 2.0.
My first K6 experience. Also, very frustrating one.
Some real janky stuff there.

Picture of the mobo for reference here: Re: AMD K6 3DMARK, aiming for stars.

Basically somebody gave me this board. I tried it - worked. Experimented a bit with overclocking. Went pretty well.
Worked great with Quadro2 Pro. Something i took for granted at the time. All good stuff.

Since i don't have enough interest in this class hardware - put it in the "to get rid of" bin.
2-3 weeks later read online something K6 related. Decided to check personally.
Pulled the board. Put the same Quadro2 Pro - silence. Tried other GF cards - same.
Will save the details but ever since then i tried to get this configuration working again but without success.
Recapped the board. Checked the videos cards in other boards - all good. Memory is fine, CPU is stable even at 600++ MHz.
For some reason its AGP health deteriorated.
At this point it is not completely stable even with GF256 and the most conservative BIOS settings.
TNTs, Rivas and other early AGP cards are fine.
PCI cards are fine too.

Info online suggests that the AGP of these boards is flaky, but still - it worked really well initially, including some hard overclock and tight BIOS timings.
Any ideas ?

pshipkov wrote on 2021-12-05, 23:09:

Info online suggests that the AGP of these boards is flaky, but still - it worked really well initially, including some hard overclock and tight BIOS timings.

That's the same experience I had with a FIC based SS7 board. Almost identical, but I was using an ATI AGP card (I think, don't trust memory). I had avoided AGP MVP3 boards because of this issue, but ultimately got a Tyan SS7 board ten years later. Stuck a Voodoo3 in it and called it done. I suspect you'd have no issue with a Voodoo3 as well.

Ok, so i am not alone then.
Yes, Voodoo3 and any other PCI card or pre-GeForce 256 AGP card is just fine.
It is really annoying, since the board has good characteristics otherwise.
6x112MHz - fine.
Can do even 750MHz or whatever on a narrow OC.

Would you be willing to share the even/odd BIOS images for your Dunson Electronics 286-16 (VLSI VL82CPCAT-16QC) ?

Are you still having problems with the floppy drive ?

Thank you.

Yes. I can boot from the floppy if I have an 486SLC installed, but not a 286. For a proper update, start reading here: Re: Evergreen 486 SuperChip - settings for DIP switch?

Tried this before with SXL2 running at 60MHz. Shared some details here.
The outcome didn't wow me but it was an interesting exercise.

Will be watching your progress.
Looks like you are already hitting the FPU issue.
The module here has FPU on it i was not able to make it work, still - you may find some of the notes there useful.

Looks like you are already hitting the FPU issue.

No FPU issue because the FPU is not yet installed on the module.

Next up - "cyrix.exe -i1 -r" - instant hang ...
After some time i realized that if the FPU on the upgrade module is enabled via the available switches the above command locks the system.

I have the FPU switch set to disabled on my module and my FPU is not installed.

At the end i pulled my daughter's microscope, quickly discovered and fixed the problems.
That microscope will be seeing more electronics action in the future for sure.

What was the problem?

Running the cyrix.exe command in combination with enabled FPU didn't lock anymore, but the FPU was still unavailable.

What action taked allowed it not to hang any more?

- write access to CompactFlash cards is flaky with the SXL2 CPU, works fine with mechanical HDDs.

Interesting. I'm OK to use mechanical HDDs if I run into this issue.

- adding more extension cards such as SoundBlaster, LAN, etc. makes the system unstable.

Holy crap.

About the FPU - will try it on few more "classic" 286 motherboards with the hope that the upgrade module was tested on them during its development, to see if something will make it "click".

Any update?

My 486 SuperChip module is a lot smaller than your upgrade adapter. Usually smaller is better. I have 8 dip switch settings, but no manual. That's 256 different variations - probably too many for me to try each one. I might be able to make 3 static and do 32 iterations.

Did your upgarde module have no issue with the DLC, that is, only had issue with the SXL? Another difference between our modules is yours is using 32-bit DLC/SXL, while mine is using 16-bit SLC/SXLC chips.

Does your module also contain a PLL clock doubler on the PCB?

feipoa wrote on 2021-12-11, 21:45:

Looks like you are already hitting the FPU issue.

No FPU issue because the FPU is not yet installed on the module.

I see.

feipoa wrote on 2021-12-11, 21:45:

Next up - "cyrix.exe -i1 -r" - instant hang ...
After some time i realized that if the FPU on the upgrade module is enabled via the available switches the above command locks the system.

I have the FPU switch set to disabled on my module and my FPU is not installed.

In my case the initial problem was faulty FPU that came with the module. For some reason it was hanging the system upon running cyrix.exe.

feipoa wrote on 2021-12-11, 21:45:

At the end i pulled my daughter's microscope, quickly discovered and fixed the problems.
That microscope will be seeing more electronics action in the future for sure.

What was the problem?

Few pins were shortened from the soldering in spots that were impossible to see with naked eye.

feipoa wrote on 2021-12-11, 21:45:

Running the cyrix.exe command in combination with enabled FPU didn't lock anymore, but the FPU was still unavailable.

What action taked allowed it not to hang any more?

Replacing the FPU.
While that stopped the crashes i was still unable to make the system to see and use it. Will be interesting if you can get it running.

feipoa wrote on 2021-12-11, 21:45:

- write access to CompactFlash cards is flaky with the SXL2 CPU, works fine with mechanical HDDs.

Interesting. I'm OK to use mechanical HDDs if I run into this issue.

Yea, this issue is not unique to this specific case.
I noticed on several other occasions that using SXL2 CPUs in 286/386 motherboards can cause such problems with CF cards.

feipoa wrote on 2021-12-11, 21:45:

- adding more extension cards such as SoundBlaster, LAN, etc. makes the system unstable.

Holy crap.

Yeppp.
Keep in mind that i was pushing the system hard - up to 70MHz on the CPU.
Cannot remember what the outcome was with lower base frequencies.
I was not interested in that. Even 15MHz base hurt overall perf significantly - especially the disk metrics.
So going lower than that was of no interest to me.

feipoa wrote on 2021-12-11, 21:45:

About the FPU - will try it on few more "classic" 286 motherboards with the hope that the upgrade module was tested on them during its development, to see if something will make it "click".

Any update?

Never proceeded with that. 😁

feipoa wrote on 2021-12-11, 21:45:

My 486 SuperChip module is a lot smaller than your upgrade adapter. Usually smaller is better. I have 8 dip switch settings, but no manual. That's 256 different variations - probably too many for me to try each one. I might be able to make 3 static and do 32 iterations.

Extra switches most likely mirror some of the cyrix.exe command flags.

feipoa wrote on 2021-12-11, 21:45:

Did your upgarde module have no issue with the DLC, that is, only had issue with the SXL? Another difference between our modules is yours is using 32-bit DLC/SXL, while mine is using 16-bit SLC/SXLC chips.

Does your module also contain a PLL clock doubler on the PCB?

Cannot remember what was the story with DLC chips. Can check if we decide to drill there. I am inclined to say no. Because DLC chips don't have issues with writes to CF cards on 386 motherboards.

Yes. There is a clock doubler.
The best i could do is with 70MHz crystal (17.5MHz base frequency) which gets doubled once through the interoposer and doubled twice with software flag = 70MHz CPU speed.

Drop back to 65 or 60. If you are having problems at 70MHz then why are you having problems with cards and not going to drop to 65 or 60?

This is not a competition of overclocking. You want reliability and *some* overclocking, not way out to very near edge of issues at extreme overclocking.

Cheers,

We are completely aligned here.

I put effort to go as far as possible on perf, but the end results must be a real computer.
The 70MHz number was the upper limit. 66MHz is where things worked ok - can perform tests, run apps, play games - but without the full HW gear.
For everything to work at 100% i had to significantly lower the base frequency.
Also, remember the no-FPU situation.

I could stay on this configuration and enjoy improved processor metrics over standard 286 CPU, but the system was disbalanced and compromised in multiple ways, which violates the "real computer" rule.

Full hardware gear working counts more than just kicking the CPU harder.

Specifically 286 boards:
In my experience working with 286 boards, these boards tends to not take that overclocking too well even with CPU upgrade module.
Things goes much better with cached 386 boards than these non-cached 386 boards due to tighter timing requirements required more thought when going beyond 25MHz. The cacheless 386DX 25 I had was low quality 386dx baby at board using 5 or 7 C&T chipsets, these chipsets had been stamped "25MHz" on them, and anything beyond BIOS settings defaults crashed.

The 286 boards tend to be primitive and not designed to have good designs in mind as they were running up to 12 MHz even 16MHz is very relaxed and worked on poorly designed boards, which is way more common than you think. When things got faster especially re-using 286 era design rules and board routing also using 386SX 20 and higher then things get more dicey.

Cheers,

Dont you think that year 1990 was the point in time after which 286 PCB designs improved significantly.
Not each and every mobo out there, but some were done just right - within the 286 specification and design limitations of course.
Some VLSI-200 and late Headland H12P based boards come to mind and cannot think of a single bad word to say about them.
Some late implementations (all the way to 1994) were based on 386 or 286/386 hybridchipsets with very high level of integration.
While not performance beasts they were just solid and no hustle whatsoever.
Most pre-1990 designs were a lot more like what you described.

My first PC was a 286 I think from around 1989-1990. A VLSI based one.
Now I haven't played much with other 286 motherboards, but taken all the
ones I've seen this VLSI one was by far the fastest and had perfect design.
It had 1MB onboard RAM and supported SIPP sockets for expansion.
AMI Bios & originally running a 12MHz 286. I think for its class it was perfect

My first 286 PC was had Headland "chipset" based motherboard.
Putting the word chipset in quotes because it was early/cheap PCB design with very low level of integration - lots of chips.
16MHz CPU "beast".
But soldered on board.
But burdened by truck-load of wait states.
Yet it was different back then.
Moving to entirely 16-bit platform was wild.

My first 286 I thought was an utter pile of crap because it was a 6Mhz actual 5170 and I'd been using a 12Mhz V20 that felt faster. This was just into the '90s though, so both machines were obsolete and got for cheap... the only exciting thing about the 5170 was the 1.2MB floppy. Within a year or two I had a 25Mhz Harris 286 which I think was on a headland board, and whoooo, that blew your socks off, though I'd got one of those baby baby AMD 386sx40 boards by then as an actual main windoze 3.1 box. I miss both the latter of those, lost in long distance move. I think I had looked into speeding up the 5170, but every single chip on that board was the lowest slowest version possible, so I couldn't just whack a 10Mhz 286 and crystal in it, don't think it would have even taken the 8, not sure whether all of the 6Mhz turtles came from IBM like that or whether someone had swapped some chips over (a lot socketted) from a cheaper clone, or maybe an XT they were trying to hotrod. I'll give it one thing... it was an actual upgrade from a 5160, it was still like watching paint dry, but at least now you felt like you had a hand held fan to help it along or something.

Well, big brand PCs are bad investment.
A fact.
So your story checks out.
Pay more, get less.

Noticing an interesting pattern.
If i let retro PCs idle for a while some of them can get stale and throw a tantrum after the long pause.
Looks like will need to do the car thing and fire them up on a regular basis to keep the parts lubricated.

Rubber seals on the bigger capacitors are exactly like that apparently, need cycling so the plasticizers spread around and keep it supple.