Good stuff Chadti99.
Nice to see the mod ticking.

Just checked here.
WB mode is possible regardless of CACHE WRITE CYCLE value.
On this board WT mode is same or faster than WB.
For Q1 specifically it matters which video card is used.
If MGA - 20.4 fps in WT and 20.3 in WB.
If Voodoo3 - 20.3 in both WB and WT. Perf diff is below the rounding error.
And so on2ith other cards.

I have seen similar issue with UM8881 based boards.
Accidentally solved it by increasing the CPU voltage.
I know, it kind of does not make sense, but somehow it affects things.
It is easy to try, so check it out.

If that does not fix it, then keep looking for better quality RAM.

If that does not do it, then some of the L2 cache chips are flaky.

---

In another news - almost fixed the Gigabyte 486AMS motherboard.
Fussy fellow.
Down to trying to make it see 1mb L2 cache.
Can you confirm if you used 64k or 128k tag chip ?

pshipkov wrote on 2021-12-31, 16:40:

Just checked here. WB mode is possible regardless of CACHE WRITE CYCLE value. On this board WT mode is same or faster than WB. F […]
Show full quote

Just checked here.
WB mode is possible regardless of CACHE WRITE CYCLE value.
On this board WT mode is same or faster than WB.
For Q1 specifically it matters which video card is used.
If MGA - 20.4 fps in WT and 20.3 in WB.
If Voodoo3 - 20.3 in both WB and WT. Perf diff is below the rounding error.

Was this at 60MHz+ FSB?

Which MGA card specifically?

Def interested to hear what resolved the issues with the GA board. When I get home I’ll confirm tag ram size.

Side note discovery with the GA486 board, the latest available firmware resolves the constant “Updating ESCD” for me but then breaks Rendition Verite compatibility with no video at boot 9/10 times. I must of been running the later firmware when I had it going with the Riva128/Voodoo2 combo.

Yes this is for 60/66 FSB.

Matrox Millennium 2064W R3 (but R2 is same 100%).
So far this is the best PCI card for DOS graphics on 486 class hardware.
Keep in mind that there are at least 2 BIOS versions available. The later BIOS version helps with Quake 1 frame rate.
Most likely the 0.2 fps difference we are seeing comes from MGA + right_bios.

About the GA 486AM/S.
That board looks like brand new. No scratches anywhere around L2_cache/ISA/PCI/BIOS/RTC sockets.
But apparently it didn't age well.
Apart from not liking EDO RAM (even with latest BIOS), there were plethora of issues with it:
- Very picky about RTC module. It was not able to retain BIOS settings between cold restarts and sometimes even between warm ones. Some specific Dallas DS12887+ 1837A module fixed that. Go figure.
- Very picky about RAM. Chokes on trusted modules. Was able to find ones that it liked.
- Very picky about L2 cache chips. It can do only single bank. Even with that limitation 256/512K is unstable. Tried very hard with chips. It is something in the board really. But 128Kb L2 cache is stable with BIOS settings on max.
- This is the first motherboard that does not like Matrox Millennium PCI cards. Also it does not like Ark2000 and other ones. Does not initialize them properly every time (no screen). But works just fine with S3 cards. Go figure.

In short - with the right RTC module, video card, 128Kb L2 cache and specific RAM modules - the board works just fine.
But took a while to find the formula.

pshipkov wrote on 2021-12-31, 16:40:

Can you confirm if you used 64k or 128k tag chip ?

The manual says you need a 64k8 tag ram for 1024k cache. I’m using a UMC61512AK-15 which I think is 64k8.

I dug out my MGA2064W Rev3 and updated the bios from 2.4 to 2.7(the latest found here under drivers) but no change in Quake score. I’m seeing 19.9 with L2 in WT before and after which is 0.1 lower than my Voodoo3. Is this the bios you are using?

Happy new year everyone!

Okay that was not the latest bios, found version 3.0 and updated but still same Quake score, is this the latest?
Side note SDD6.53 doesn’t seem to work well with this card, garbage on screen for any vesa modes, is there a better vesa utility for this card?

Found a 32MB 60ns EDO module that seems so be working well with 3-1-2 cache settings at 60MHz FSB on the cache modded LS-D.

Curious enough, anything below 64MB with L2 set to WB imposes a performance penalty with my board.

I am looking at the same manual. Both 64kx8 and 128kx8 work. Wanted to check if either one worked better for you, etc.

Back to LSD.
Stared a bit longer at your last notes and screenshots, compared them to my previous posts and ran some brief tests here.
1. The 20.4 fps in Quake 1 was achieved with CL-5480 graphics card. That's 0.1 fps higher than what i get with Voodoo3 and MGA (which is 20.3 fps).
2. There is one difference between our BIOS setups - i set DRAM SPEED to FASTER which allows cache timings to be 2-1-2. This is better for performance. That is with 50ns EDO RAM - the same module you use.
3. I set DRAM SLOW REFRESH to DISABLED. It has no effect on performance as far as i can tell, but if ENABLED Windows 9x starts acting up when running multiple tasks in parallel.
4. On the two motherboards here stability is better if i use the top most SIMM slot. No idea why, but more memory modules work in it more reliably.
5. L2 cache in WB mode is same or slower than WT regardless of memory amount. Some tests are insensitive to this which can be misleading.

About MGA BIOS versions.
I checked today and yes, this makes no difference on LSD at 180MHz, but i remember seeing difference on Asus PVI at 160MHz. The later BIOSes give something like 0.1 or 0.2 fps extra.

So i was baking today.
Not pastries and other eatable substances but one of those fast VLSI-200 286 motherboards.

Its health deteriorating gradually - not booting up every time and so on.
If i bend it gently here and there it would eventually wake up until go silent again.
At some point it permanently shut down.

I have been replacing components on it - capacitors and ICs with spares from other dead PCBs with the hope that it is a dead element that i will blindly fix.
Early today got fed-up with this weak progress and decided to put it in the oven.
Started with 125C for 5 minutes, then increased to 150C for 5 more minutes minutes, then 175C for 2-3 more minutes.
Then chillout time.
Then turned it on - lit up right away.

I have 2 more boards that i hope will benefit from oven reflowing.
Let's see ...

I am looking at the same manual. Both 64kx8 and 128kx8 work. Wanted to check if either one worked better for you, etc.

I have run into a board or two which did not want 128kx8 as TAG with 1024K total - it insisted on having 64kx8 as TAG.

On the two motherboards here stability is better if i use the top most SIMM slot.

I've run into this on some motherboards. Usually the first SIMM is closer to the northbridge chipset.

I checked today and yes, this makes no difference on LSD at 180MHz, but i remember seeing difference on Asus PVI at 160MHz. The later BIOSes give something like 0.1 or 0.2 fps extra.

In this sentence does "later BIOSes"refer to the Asus PVI board or the Lucky Star LSD board? If the LSD board, do you have a zip with all your LSD BIOSes? I only have one revision.

fast VLSI-200 286 motherboards...Its health deteriorating gradually

They certainly are of the right age for this. Cold solder joints aren't usually visibly apparent. The cruise control computer on my 1979 Mercedes stopped working about 12 years ago. A year ago, I decided to hand solder every single through-hole component on the PCB. It took about 3 hours and stunk up the whole house, but the cruise control works again. The stink was primarly due to this think lacquer paint Mercedes decided to plaster over the back side of the PCB and joints, probably for some sort of moisture protection.

I've had 386 boards of this era doing the same thing with cold joints. If you have a known good board of the same model, you can more systematically pin point the bad points with a multi-meter, but I don't think it saves a lot of time compared to simply reflowing every joint in a suspect region of the PCB.

Started with 125C for 5 minutes, then increased to 150C for 5 more minutes minutes, then 175C for 2-3 more minutes.

Did you use your kitchen oven, or a proper reflow oven? 175 C doesn't seem nearly hot enough to melt the existing solder. I suspect the expanding metal vs. fibreglass of the PCB just made the joint contact again, but new flexing could disconnect it in time.

I tried putting some dead graphics cards in the kitchen oven at around 245 C for some pre-defined amount of time, as proclaimed on youtube, but none of the cards were brought back to life. This was completely contrary to the youtuber's experience.

I've run into this on some motherboards. Usually the first SIMM is closer to the northbridge chipset.

Yes, this is the logical thing, but in this case the upper slot if further away from the chip.

In this sentence does "later BIOSes"refer to the Asus PVI board or the Lucky Star LSD board? If the LSD board, do you have a zip with all your LSD BIOSes? I only have one revision.

It was about the MGA BIOSes. I can see how it was unclear.

They certainly are of the right age for this. Cold solder joints aren't usually visibly apparent. The cruise control computer on my 1979 Mercedes stopped working about 12 years ago. A year ago, I decided to hand solder every single through-hole component on the PCB. It took about 3 hours and stunk up the whole house, but the cruise control works again. The stink was primarily due to this think lacquer paint Mercedes decided to plaster over the back side of the PCB and joints, probably for some sort of moisture protection.

Yeah, smell was noticed and had to explain why is this "thing" in the oven. The kitchen oven.

I've had 386 boards of this era doing the same thing with cold joints. If you have a known good board of the same model, you can more systematically pin point the bad points with a multi-meter, but I don't think it saves a lot of time compared to simply reflowing every joint in a suspect region of the PCB.

I started there, but it was time consuming.
Between sessions i kept forgetting what was tested the last time.
It was more time consuming than just start going through the PCB components and then the bake.

Did you use your kitchen oven, or a proper reflow oven? 175 C doesn't seem nearly hot enough to melt the existing solder. I suspect the expanding metal vs. fibreglass of the PCB just made the joint contact again, but new flexing could disconnect it in time.

I tried putting some dead graphics cards in the kitchen oven at around 245 C for some pre-defined amount of time, as proclaimed on youtube, but none of the cards were brought back to life. This was completely contrary to the youtuber's experience.

I was not sure what to expect really.
Also, didn't feel like checking online what others say.
Acted brave and irresponsible.

175C was actually more than enough.
Learned this first hand.
I put the motherboard on a piece of baking paper right on top of the grill.
So when reached this temperature the solder was so liquid that some of the ICs whose pins laid on the grill bars slid up. Had to heatgun and push them down afterwards. Some are still slightly tilted, but will at this point i am not going for aesthetics. This board got enough beating already. Happy that it ticks.

The next board that will go in the oven will rest on folded aluminum foil pads placed in its corners.
Lesson learned.

huh. If you plan on doing more, it might be a good idea to place the DMM's thermocouple tip right at the point of the solder to see how hot the solder is really getting.

The tissue I had when using the rolled up aluminium balls is that the PCB's are rather challenging to balance and you will probably want to ensure that the aluminium ball isn't touching a solder joint.

Do you have one of those convection ovens? I find the blower causes uneven heating to the PCB's. The edge near the blower got a bit discoloured, while the far end did not. If I were to do it again, I'd disable the blower.

Any chance that your baking skills fixed the issue with floppy disks not booting when a 286 CPU is installed?

Yeah, the oven smell of PCB's was mild compared to 1970's automotive PCB lacquer. There was some really funky chemical in there, and it didn't smell like burnt varnish or melting solder.

pshipkov wrote on 2022-01-01, 20:17:

I am looking at the same manual. Both 64kx8 and 128kx8 work. Wanted to check if either one worked better for you, etc. […]
Show full quote

I am looking at the same manual. Both 64kx8 and 128kx8 work. Wanted to check if either one worked better for you, etc.

Back to LSD.
Stared a bit longer at your last notes and screenshots, compared them to my previous posts and ran some brief tests here.
1. The 20.4 fps in Quake 1 was achieved with CL-5480 graphics card. That's 0.1 fps higher than what i get with Voodoo3 and MGA (which is 0.3 fps).
2. There is one difference between our BIOS setups - i set DRAM SPEED to FASTER which allows cache timings to be 2-1-2. This is better for performance. That is with 50ns EDO RAM - the same module you use.

Thanks for reviewing this again, I’m now running 212 cache settings on the LS-D and hitting the same 20.3fps at 180MHz. Changing the DRAM SPEED to FASTER was the key to lowering cache timings!

feipoa wrote on 2022-01-02, 02:17:

huh. If you plan on doing more, it might be a good idea to place the DMM's thermocouple tip right at the point of the solder to see how hot the solder is really getting.

The tissue I had when using the rolled up aluminium balls is that the PCB's are rather challenging to balance and you will probably want to ensure that the aluminium ball isn't touching a solder joint.

Do you have one of those convection ovens? I find the blower causes uneven heating to the PCB's. The edge near the blower got a bit discoloured, while the far end did not. If I were to do it again, I'd disable the blower.

Yes, i am going to improve on the process.
Read up on what kind of plastics are used in the various motherboard sockets. Their thermal properties, etc.
Then try few temperature ranges.
Oven is not convection one. I think it is some model that is supposedly creating even temperature inside the chamber.
Didn't get discoloring in any pieces at up to 175C in this first attempt.

feipoa wrote on 2022-01-02, 02:17:

Any chance that your baking skills fixed the issue with floppy disks not booting when a 286 CPU is installed?

Good point. Didn't think to check that. Did now.
Unfortunately Gotek (at least) is still not working, otherwise i could see myself on the Great British Baking Show.
The rest of the competitors: showing their greasy stuff.
Me: Here is my Dunson Electronics mobo with working floppy drive. Gents, sit down.

feipoa wrote on 2022-01-02, 02:17:

Yeah, the oven smell of PCB's was mild compared to 1970's automotive PCB lacquer. There was some really funky chemical in there, and it didn't smell like burnt varnish or melting solder.

I doubt the odor was harmful to carbon based life forms. They would never use such chemicals in the 70ies ...

---

@Chadti99
Glad we sorted it out.
Btw, I have one memory module that can almost do Q1 with cache timings 2-1-2 and DRAM SPEED = FASTEST.
So close, yet so far.

Too bad about the 286 and boot floppy...still. Is it possible that it never booted or did using a 486 upgrade module some how ruin this aspect of the motherboard?

I doubt the odor was harmful to carbon based life forms. They would never use such chemicals in the 70ies ...

Pretty sure that's sarcasm. Everyone in the house felt dizzy and wanted to throw up for the next 2 days. I bet whatever was used on that PCB was also used as a rodent poison.

@feipoa What size/value variable resistor would you recommend for the LS486e-D board to adjust between 3.45-4v CPU voltage? Assuming the one used on the UUD volt mod would not work here?

Also is there a particular temperature probe you guys recommend for monitoring CPU temps? I’d like to monitor temps the same way that CPU Galaxy did on the underside of the chip through the convenient screw hole provided on the LS486 boards.

Also how the hell did he run at 200MHz for so long without destroying his CPU/Motherboard with condensation? Humidity controlled room?

pshipkov wrote on 2022-01-01, 20:17:

4. On the two motherboards here stability is better if i use the top most SIMM slot. No idea why, but more memory modules work in it more reliably.

Good call on this and thanks again! My 64MB module is running stable at 212 cache and 180MHZ with it installed on the slot closest to the edge of the board. Ram speed at “faster” of course.

I did attempt to change ram speed via TweakBios, after booting, to “Fastest”, but it crashes after applying and exiting to dos. Was worth a shot I thought.

@Feipoa
Yes, that was a bit of sarcasm.

Pretty sure VLSI-200 floppy issue has nothing to do with CPU upgrade module damaging things.
Tested it on the board that never had CPU upgrade on it.
It is a BIOS issue. It always reverts to 360 floppy drive.
I was so sure floppy worked back then, but maybe memory is misleading me if both of us cannot do it today.

Speaking of baking. This thing is magical.
Second board today came to life - ECS VLB mobo that never lit up before is now working.
So good.

@Chadti99
Nice - about the 64Mb running stably at 2-1-2 !

I use this thermometer - works great:
https://www.usalab.com/professional-digital-t … BCABEgJHbPD_BwE

Who was CPU Galaxy and the video you mentioned ?
About running 486 CPU at 200MHz - 1-2 or more hours is fine. The CPU freezes. Quite a chunk of ice can form there, but that's ok. The problems start when we turn the power off.

That blue trimmer will be fine for your needs.

pshipkov wrote on 2022-01-02, 18:03:

Who was CPU Galaxy and the video you mentioned ?
About running 486 CPU at 200MHz - 1-2 or more hours is fine. The CPU freezes. Quite a chunk of ice can form there, but that's ok. The problems start when we turn the power off.

That blue trimmer will be fine for your needs.

Got it, with my current setup, at 5v, it doesn’t seem to be icing up, just beads of water that continue to grow. So maybe I need to run the peltier colder.

For the blue 5 KΩ trimmer would I solder it in parallel with the existing resistor at R16 or replace it?

Either replace the resistor or connect the trimmer sequentially.

If you run your Peltier at 5v then you can actually avoid condensation.
Several posts ago I posted about how I did it.