The C0-C6 codes on Award BIOSes originate from the boot block that is supposed to decompress the runtime BIOS and hand over execution to it. To decompress the runtime BIOS, you need working memory, so the boot block has to initialize the memory controller first. This is performed using chipset-dependent code, which often detects L2 size as well as the memory size. IIRC, Pentium boards usually do not enable L2 cache at that stage after sizing the L2 cache, so I wouldn't expect L2 issues to be the cause.

I see that both of your boards both have DIMM and SIMM slots. While I guess that at your post count you are aware of the way to populate them, I will nevertheless lay out the basic rules of socket 7 memory configuration:

• PS/2 SIMMs must be installed in pairs (yeah, I know about the SiS 5571, that chipset is an irrelevant curiosity regarding this post).
• These boards are designed to interface PS/2 SIMMs at 5V operating voltage, while DIMMs are typically at 3.3V operating voltage. Inserting 5V SIMMs and 3.3V DIMMs at the same time can cause any kind of issues.
• Some boards have jumpers for the RAM supply voltage. 5V PS/2 SIMMs will fail at 3.3V supply voltage. 3.3V DIMMs can get damaged at 5V supply voltage.
• Just because there is a DIMM socket, it does not necessarily mean the board supports SDRAM. There are EDO DIMMs as well. I have yet to see a consumer board that support EDO DIMMs only, though. Some name-brand boards (HP, Compaq, DELL or some vendor in that class) do require EDO DIMMs, though.
• Modules can be "too big to be supported". Do not start testing with DIMMs above 64M on boards that have PS/2 slots as well.

Thanks for the feedback and confirming that it's probably not the cache.

I do have a single stick of 64MB SDRAM that I did test with both boards, but it didn't work. For the MS5145 I noticed on the jumper manual on RetroWeb states that SDRAM isn't supported, but in the scanned manual there are a couple references to using SDRAM. Not sure what to believe there.

Regardless, I mostly tested with various pairs of 72-pin RAM. I did make some minor progress with the MS5145.

With certain pairs of EDO RAM, it made it past the previous C6:C1 and got stuck on C5:C3. This still seems to indicate a RAM issue. If I installed a PCI video card and/or when using certain other pairs of 72-pin RAM, it would get stuck on C6:C1. And on a few occasions, it would get stuck on no POST code with the reset line stuck "on".

I also did voltage and ground testing on the 72-pin memory slots. Voltages on both boards reported 5.2V on the correct pins. Grounds seemed okay, except on the MS5145 board, pins 67 to 70 (presence detect) are all shorted to ground. From what I looked up, this isn't a feature implemented on all boards, so I'm not sure if these pins being shorted to ground matters?

Feels like minor progress, but not sure where to go next.

On MSI MS5154 V2, BIOS version 10/28/97-i430TX-2A59IM4AC-00 has the code that emits POST code 0xC6 in the bootblock (F000:E59A).
After this code is emitted, BIOS gets chipset's revision (if it is not stepping A0, it configures CKE to be muxed with MAA0 and CKEB with MAA1), then sets DRAM refresh rate to 31.2 us (which was previously disabled at F000:E0F6).
The previous POST code, 0xC1, is emitted at F000:E8A1 - it searches the ROM for the "*BBSS*" tag and runs the memory detection routine.
POST 0xC5 (F000:EACA) enabled ROM shadowing.
POST 0xC6 (F000:A6EA) again - cache is probed: there is some code that initializes DRAM and fills it with some constant values while L2 is disabled, then enables L2 and checks how exactly it is inconsistent with the DRAM contents.

I have a working Soyo SY-5BT but with a problem with a cache chip, the only way to make it work (now), is to disable the L2 cache from the BIOS, or alternatively use a CPU with integrated L2 cache.

Unfortunately, from what I've read, it's not possible to disable one of the two cache chips, to make it work half (256 KB).

I don't know what CPU you use in your PC, but if you use one up to 233 MHz, with multi maximum 3.5X, there is an error in the jumpers, currently you have a jumper in the BF2, and this activates the multi from 4X to 6X, you should remove it and try again, maybe 🤔 this could be the problem, and also check the other motherboard, maybe 🤔 they used K6-2 with multi highs.

Speaking of SDRAM banks, I use them in the SY-5BT already from the first boot, what I have on is a 32 MB with 8 chips of the single-sided Hyundai, at the end it says TC-10 I guess it's 10 ns.

PC@LIVE wrote on 2024-11-05, 09:06:

I don't know what CPU you use in your PC, but if you use one up to 233 MHz, with multi maximum 3.5X, there is an error in the jumpers, currently you have a jumper in the BF2, and this activates the multi from 4X to 6X, you should remove it and try again, maybe 🤔 this could be the problem, and also check the other motherboard, maybe 🤔 they used K6-2 with multi highs.

I'm currently using a Pentium 100 for testing. I set all my Socket 7 motherboards not in active use to default to Pentium 100, since it makes it easy to just grab the CPU and test.

Curious that both have problem with ram. How did you get them, any chance both came from salvage/recycling where someone could have bend them?
You could try pressing on TX chipset while resetting, just put something flat under the pcb so you dont bend whole thing while pushing with your finger.

Update on testing.

I tried the MSI MS5145 and it's again getting stuck on C6:C1, even without a video card and using the same RAM as yesterday. I then decided to check clock signals. Probably should have done this sooner.

The CPU clock was a bit 'fuzzy', not quite the clean signal I was expecting.

I then checked the input to the clock generator IC (SC653DYB). It was not showing a clean signal and at an incorrect frequency of about ~21.9MHz instead of the expected ~14.3 MHz.

Evidently the clock generator IC could still generate a workable CPU clock from this but the crystal is not outputting the correct frequency at all.

At this point, I'm going to replace the ~14.3MHz crystal and see if that improves anything. I don't have any on hand in the particular package used this by board, so I'll have to order some.

I also checked the clock signals on the Soyo SY-5BT and it seemed fine in comparison.

UPDATED:

This was a faulty diagnosis. I had my probe set to low impedance (x1). When I set it to x10, I was able to get a correct clock signal as documented in this later post.

rasz_pl wrote on 2024-11-05, 15:29:

Curious that both have problem with ram. How did you get them, any chance both came from salvage/recycling where someone could have bend them?
You could try pressing on TX chipset while resetting, just put something flat under the pcb so you dont bend whole thing while pushing with your finger.

I don't remember exactly where I got them, but I suspect it might be from systems I disassembled. I've had them for at a least a couple years in my repair bin, just hadn't gotten around to trying to fix them. Both boards are in great physical condition with no scratches or other visible damage. Not even any leaky or bulging caps.

I was kinda hoping they might both have the same fault which is why I was looking at them at the same time, but after initial diagnostics that doesn't seem to be the case.

I had a socket 7 board not posting. It was getting stuck on C6. Turns out the BIOS had bit rot. I installed a new BIOS chip with a fresh copy of the latest BIOS and she booted. Maybe try flashing the BIOS chip and see what happens.

zuldan wrote on 2024-11-05, 20:04:

I had a socket 7 board not posting. It was getting stuck on C6. Turns out the BIOS had bit rot. I installed a new BIOS chip with a fresh copy of the latest BIOS and she booted. Maybe try flashing the BIOS chip and see what happens.

That is good advice. I had previously checked / reflashed the BIOS on the MSI motherboard to no effect.

The BIOS's checksum on the Soyo board doesn't seem to match any of the ones one Retroweb. I reflashed it anyway with the 1B7 version, but that didn't change the symptoms.

Turns out there was another issue which I discovered and document in the following post.

I figured what the issue was on the Soyo board preventing it from successfully booting. Turns out it was a jumper issue.

The board features both AT and ATX power connectors. There are three total jumpers that need to be set correctly in order to use either AT or ATX. I thought I had set everything up when I first started testing the board. But apparently I missed the jumper labelled ATP1, next to the keyboard port. This jumper needs to be closed to use an AT power supply.

After bridging it, the system did successfully boot up.

This board still has an issue with the first PCI slot. Using a POST card with it, it produces nonsensical POST codes and LED readouts. For example, the Reset line shows as being "on" even with the system booting up. And then pressing reset turns the LED off. This is the opposite of normal behaviour.

I gave the slot a thorough cleaning with multiple rounds of Deoxit and scrubbing. Eventually I was able to get a video card to work with the first slot. But it's intermittent. When removing and re-inserting video cards, sometimes it works, sometimes it doesn't.

I inspected the slot under the microscope and there are no visible issues with the pins. I further tested things with both the multimeter and a logic probe. There were no obvious issues.

At this point, I suspect it just might be physical wear 'n tear on the pins itself. Perhaps they just are not making consistently good contact with the inserted cards. Part of the limitation of probing the slot is I'm using the tops of the pins, as opposed to how a card would contact the pins. What I really need is some sort of card with test points that I could insert and then test that way. Not sure if such a thing exists?

My next option is to try to replace the slot itself. If I can find a matching replacement from a scrap board, I might give that a try.

Before trying to replace the problematic ISA PCI slot (my mistake, all the time I taught for similar problem, but with ISA slot), why don't you just tighten its fingers?

analog_programmer wrote on 2024-11-06, 19:12:

Before trying to replace the problematic ISA slot, why don't you just tighten its fingers?

This is a good idea. This prompted me to do another inspection of the PCI slot. The pins themselves don't seem pressed in any more than any of the other slots. I tried prodding them to see if it was possible to bend them out any more.

When doing this, I did come across what looked like some sort of corrosion on one of the pins. It was this orange-colored gunk (like brightly colored rust). I suspect this might be the Deoxit having reacted with some corrosion.

Looking from the top-down provides a limited view of the pins, so I tried angling my microscope to see more of the pin. This is when I spotted corrosion on several pins. A couple of them were just +5V lines, but the worst of them was an Address line.

I ended up scraping the pins using an exacto knife to try to clean off as much as I could, then gave it another bath with Dexoit.

After doing this, the POST card starting reporting correct values! I re-tested video cards as well and it is consistently working.

This also reinforces why testing the pins with the multimeter and logic probe from the top didn't really help identify the bad pins, since I was probing them on a spot free of any corrosion. I'll have to remember this the next time I have intermittent issues with a slot like this.

Did some more troubleshooting on the MSI M5145 motherboard this evening.

I salvaged a 14.31818 MHz crystal from another motherboard and decided to try swapping the one on the MSI board.

Unfortunately, it produced the same unstable output as the original crystal. I'm not sure if this means both crystals are faulty, or if it's possibly a different fault on the board.

I also tested voltages. The clock generator is being fed ~3.4V (VDDCPU). The voltage on the crystal itself is ~1.6V. Not sure if this is normal?

I further desoldered and tested a capacitor just on the off chance that maybe there are bad caps despite the lack of physical symptoms. This is a 1000 uF capacitor and it tested fine. For now I'm going with my instinct that the caps on this board are probably good.

Shponglefan wrote on 2024-11-06, 21:44:

This is a good idea. This prompted me to do another inspection of the PCI slot. The pins themselves don't seem pressed in any mo […]
Show full quote

This is a good idea. This prompted me to do another inspection of the PCI slot. The pins themselves don't seem pressed in any more than any of the other slots. I tried prodding them to see if it was possible to bend them out any more.

When doing this, I did come across what looked like some sort of corrosion on one of the pins. It was this orange-colored gunk (like brightly colored rust). I suspect this might be the Deoxit having reacted with some corrosion.

Looking from the top-down provides a limited view of the pins, so I tried angling my microscope to see more of the pin. This is when I spotted corrosion on several pins. A couple of them were just +5V lines, but the worst of them was an Address line.

The attachment Soyo SY-5BT Motherboard microscope.jpg is no longer available

I ended up scraping the pins using an exacto knife to try to clean off as much as I could, then gave it another bath with Dexoit.

After doing this, the POST card starting reporting correct values! I re-tested video cards as well and it is consistently working.

The attachment Soyo SY-5BT Motherboard PCI slot 1 POST card fixed.jpg is no longer available

This also reinforces why testing the pins with the multimeter and logic probe from the top didn't really help identify the bad pins, since I was probing them on a spot free of any corrosion. I'll have to remember this the next time I have intermittent issues with a slot like this.

Nice! One more s.7 motherboard is saved from scrapping 😀

Reading about the problematic PCI slot I got the impression that its contacts were slightly bent and loose. A more thorough visual inspection is never superfluous before starting some repair job the hardest possible way and in this case it helped you find the real problem with corroded pins.

As for the other MSI board, are you sure there isn't a problem with some capacitors? I don't mean just a leaking electrolytic capacitors, but also cracked or missing SMD ones.

P.S. A very well equipped, clean and tidy workshop you have there. I wish I had one too, but where I live, people hardly bother to give their damaged computer parts, notebooks, TV sets, tablets and similar electronic devices for repair, except for some expensive smartphones, which are of no interest to me.

You can check c36 c37, but probing crystal is not that straight forward, you need high impedance probe - imo its fine. "MSI MS5145 V2 CPU clock signal.jpg" image looks fine to me.

rasz_pl wrote on 2024-11-07, 09:21:

You can check c36 c37, but probing crystal is not that straight forward, you need high impedance probe - imo its fine. "MSI MS5145 V2 CPU clock signal.jpg" image looks fine to me.

You are right. I realized when doing those measurements, I had my probe set to x1 instead of x10.

Once I changed it to x10, I got correct measurements. Looks like the crystal and associated clock signals are completely fine. That was a red herring on my part.

analog_programmer wrote on 2024-11-07, 08:14:

As for the other MSI board, are you sure there isn't a problem with some capacitors? I don't mean just a leaking electrolytic capacitors, but also cracked or missing SMD ones.

Oh yes, when I said the capacitors are probably fine I meant specifically the electrolytic capacitors. There certainly could be a problem with an SMD component. Though having gone through a physical inspection and a bunch of measurements, I've yet to find anything that appears faulty. Though, I do have more testing to do...

P.S. A very well equipped, clean and tidy workshop you have there. I wish I had one too, but where I live, people hardly bother to give their damaged computer parts, notebooks, TV sets, tablets and similar electronic devices for repair, except for some expensive smartphones, which are of no interest to me.

Thank you! I originally built this workshop for building arcade controllers, but it gradually evolved over time into general electronics tinkering. 😀

Couple more updates on testing the MSI M5145.

I revisted the BIOS chip and tried reprogramming it with all the available BIOS versions from The Retro Web. Unfortunately this didn't help.

During testing I do get somewhat inconsistent results. The V1 versions of the BIOS (including the original from the chip) cycle through slightly different POST codes, but still get stuck. Some times it would get stuck on C3 or C6. Mostly it's C5:C3 (with specific EDO RAM). And occasionally I get no post code (--:--).

One of the available BIOS's is an AMI BIOS. But instead of the C5:C3 I am usually stuck on, it produces a d8:d7. 😒

I also desoldered a couple more electrolytic caps (470uF) just to doubly check if there might be problems. But those tested fine as well.