smoke86 wrote on 2022-10-06, 21:42:

Update: I managed to do a bodge wire mod as described above, but nothing changed.
I think it may be faulty 83706 controller and further repair attempts won't make any sense.

Can you post a picture of the chip (after you soldered the wire).

I have got two Zappa boards and one of them has the exact same issue, not saving bios settings and time/date. On the working board with power off and cmos battery in, Pin7 - VBAT - measures 3.008V (one lead on pin 7 and one on ground). This is a new battery and taking it out of the board it measures 3.009v, so barely any drop while in the board and powering cmos memory.
On the other board that doesn't store cmos settings, pin7 measures 0.2V.

Searching around the web and Usenet I found other reports of the same problem, all on 430FX Intel Advanced boards (Zappa, Morrison, Endeavor) and all using PC87306.

I removed the wire as it changed nothing.

how about resistance between diode and pin 7, and pin 7 to ground?

i had some different problematic Intel brand boards, one that doesn't post, one that has that same CMOS issue, and one that has no working IDE. i guess may be reaching the end of its MTBF. 2 are zappas and 1 is a thor ATX, the thor doesn't work at all, and 2 of the zappas have different issues, 1 with cmos issues and 1 with non working IDE. i may transfer the PC87306 IC from the dead Advanced/ATX to one of the zappas to see if any changes.

Windows9566 wrote on 2022-10-08, 21:00:

i had some different problematic Intel brand boards, one that doesn't post, one that has that same CMOS issue, and one that has no working IDE. i guess may be reaching the end of its MTBF. 2 are zappas and 1 is a thor ATX, the thor doesn't work at all, and 2 of the zappas have different issues, 1 with cmos issues and 1 with non working IDE. i may transfer the PC87306 IC from the dead Advanced/ATX to one of the zappas to see if any changes.

Ironically, I have a PCChips M507 430FX motherboard that works great. One thing you can say for fake cache chips is that they don't go bad.

Repo Man11 wrote on 2022-10-08, 22:57:

Windows9566 wrote on 2022-10-08, 21:00:

i had some different problematic Intel brand boards, one that doesn't post, one that has that same CMOS issue, and one that has no working IDE. i guess may be reaching the end of its MTBF. 2 are zappas and 1 is a thor ATX, the thor doesn't work at all, and 2 of the zappas have different issues, 1 with cmos issues and 1 with non working IDE. i may transfer the PC87306 IC from the dead Advanced/ATX to one of the zappas to see if any changes.

Ironically, I have a PCChips M507 430FX motherboard that works great. One thing you can say for fake cache chips is that they don't go bad.

https://en.wikipedia.org/wiki/Survivorship_bias
most pcchips got thrown out while relatively new just for being pcchips

My PCChips theory is that the engineers had no illusions about them being bottom of the barrel products and designed them with wide tolerances such that whatever components were cheapest that week could be used... so parts wander 30% out of tolerance, no worries.... Mid level brands, the designers thought it would only get okay stuff, then crap happened and it didn't and they died early, top level brands, it works fine as long as high tolerance top price components stay within 1%, after that it gives up.

Despite that chip looking well soldered, I'd give the thing a "long cook" with a hotair gun, even a "paint stripper" type... reason is firstly the tin can migrate out of the solder in cold storage conditions and do weird semi-conductory crap, so a long heat bath will get that properly realloying again somewhat. Secondly, the design of flash RAM is such that when memory cells are fatigued and won't flip their bits when ordered, a restoration of performance is possible by baking them, letting the materials reset a bit. IDK how long that might give you with older tech, 50 more config changes before it screws up again? If it works it's a bit of a hail mary and you should set the system up as you want it with as few boots as possible, then leave it, don't be using it as your test rig, parts in and out all the time.

BitWrangler wrote on 2022-10-09, 05:03:

My PCChips theory is that the engineers had no illusions about them being bottom of the barrel products and designed them with wide tolerances such that whatever components were cheapest that week could be used... so parts wander 30% out of tolerance, no worries.... Mid level brands, the designers thought it would only get okay stuff, then crap happened and it didn't and they died early, top level brands, it works fine as long as high tolerance top price components stay within 1%, after that it gives up.

---

Despite that chip looking well soldered, I'd give the thing a "long cook" with a hotair gun, even a "paint stripper" type... reason is firstly the tin can migrate out of the solder in cold storage conditions and do weird semi-conductory crap, so a long heat bath will get that properly realloying again somewhat. Secondly, the design of flash RAM is such that when memory cells are fatigued and won't flip their bits when ordered, a restoration of performance is possible by baking them, letting the materials reset a bit. IDK how long that might give you with older tech, 50 more config changes before it screws up again? If it works it's a bit of a hail mary and you should set the system up as you want it with as few boots as possible, then leave it, don't be using it as your test rig, parts in and out all the time.

Both of the AT/Socket 7 PCChips boards I have use a jumper for the BIOS chip to choose between five and twelve volts, and my assumption for why they have this is that it allowed them to use both types of flash chips, giving them more options for using whatever ones they could purchase for the lowest cost. My hunch is that PCChips boards were prevented from having this CMOS memory issue because the I/O chip in question was probably too expensive for them to even consider using it.

Resistance diode - > pin7 441ohm, pin7 - > gnd shows nothing.
I actually resoldered all the side of 87306 to ensure no bad connections are there.

My I ask if you have any addon cards like ISA sound cards installed and hocked up to speakers or line in on some device?

I was troubleshooting the same exact problem on a SS7 board, VBAT had correct voltage on the soutbridge VBAT pin but still it would lose CMOS settings after some time.
Did the same things you did, reflowed the south bridge, bypassed cmos battery with a small resistor and diode directly to vbat with a CR2032, no difference.

I eventually unplugged a every PCI/ISA card one at a time until I got to the sound card.
Removing the soundcard and it worked fine and kept cmos setting until the next day, usually it would lose cmos with in 5-15 minutes.

Then I put the soundcard back, same problem, then I removed the cable from the line out on the soundcard, problem gone!

I tested two know good cards, SB16 and AWE64, same problem if line out was connected.

Seems to be a grounding issue so the potential between vbat and gnd got to low, odd but yea as long as I dont leave anything connected to the soundcard after power down its fine.

Might not be your problem but it could be I guess.

Nope, nothing except cpu and ram is plugged in.

smoke86 wrote on 2022-10-09, 18:51:

Resistance diode - > pin7 441ohm, pin7 - > gnd shows nothing.

define nothing? for me nothing means zero, and zero resistance to ground on pin 7 would be an obvious defect and something we can fix

Nothing means open line, no connection

smoke86 wrote on 2022-10-10, 04:48:

Nothing means open line, no connection

I specifically asked about resistance. Nothing resistance is zero, the opposite of nothing is all of the resistance 😀. If you are getting megaohms then the chip has blown battery backup power circuit. In normal operation 5V supply powers the SRAM, when turned off blown backup is unable to take over.
replacement "[1pcs] PC87306-IBE/VUL Peripheral RTC UART PQFP160Opens in a new window or tab" ~$15

>Resistance diode - > pin7 441 ohm

that also doesnt look good unless there is resistor between

rasz_pl wrote on 2022-10-10, 05:21:

that also doesnt look good unless there is resistor between

No idea, trace goes between layers inside of the the moterboard.
Does replacing the chip actually make sense in that case as there is no voltage on pin7?
I think I'll give it a try and bodge pin7 directly to the resistor before the diode. Will see then.

rasz_pl wrote on 2022-10-09, 04:40:

Repo Man11 wrote on 2022-10-08, 22:57:

Windows9566 wrote on 2022-10-08, 21:00:

i had some different problematic Intel brand boards, one that doesn't post, one that has that same CMOS issue, and one that has no working IDE. i guess may be reaching the end of its MTBF. 2 are zappas and 1 is a thor ATX, the thor doesn't work at all, and 2 of the zappas have different issues, 1 with cmos issues and 1 with non working IDE. i may transfer the PC87306 IC from the dead Advanced/ATX to one of the zappas to see if any changes.

Ironically, I have a PCChips M507 430FX motherboard that works great. One thing you can say for fake cache chips is that they don't go bad.

https://en.wikipedia.org/wiki/Survivorship_bias
most pcchips got thrown out while relatively new just for being pcchips

i have had really good luck with PCChips, have a M912 and a M520 that both work great.