dragonkn wrote on 2023-05-10, 14:26:

it's very very easy to damage Duron/ Athlon CPU Core, while installing a cooler

It's a Slot A cpu, you can damage the die only if you open it to change the thermal compound.

dragonkn wrote on 2023-05-10, 14:26:

it's very very easy to damage Duron/ Athlon CPU Core, while installing a cooler

Even if we were talking about a socket A CPU (which we're not), it's actually VERY hard - if not impossible - to damage the CPU in such a manner that it still works, but is unstable. The usual damage is cracked die --> bye bye POST. 😀

So, the CPU is 100% a tbird, AMD-A0900MMR24B A. However, everything I've read since about the MS-6191 shows that they are certified for tbirds, so I doubt it's a CPU compatibility problem, though obviously I'll check with my Pluto when it arrives.

I've tried the Voodoo 3 3000, the same random instability appears to be there still, perhaps not quite as frequent, but still definitely there. I'd imagine this is another nail in the "capacitor" coffin. Caps are ordered, but are going to take a while to arrive, so I guess I'll update the thread once the caps are changed.

I remember i got very same issue with one socket a platform.
I underclock cpu with jumpers on mobo and everything working ok ( i did 200 mhz less for athlon 2000)
for you 100 - 133 mhz less is a good test i think

There aren't any overclocking options on the MS-6191 unfortunately, and I don't own a goldfinger device (or clone), so I've no way of changing fsb or cpu speeds.

i got info from old forum that this motherboard and athlon are not stable with 133 mhz memory if you can try memory that are 100 mhz

The FSB is 100, the memory is running at 100 (even though it's a stick of PC133).

I think it's worth a try a memory that got a 100 mhz on a sticker, maybe from older pc 😀?

paradigital wrote on 2023-05-11, 10:36:

The FSB is 100, the memory is running at 100 (even though it's a stick of PC133).

Well, it is worth trying it. I did have a similar issue in the past with an Acorp 5ALI61 Super Socket 7 board. That thing just refused to work properly with all 'modern' PC133 SDRAM modules I tried (high or low density, didn't matter), even though the FSB was set at 100 MHz (and the behavior was very similar to what you are currently experiencing).
After finally switching to two good ol' 2 x 64 MB PC100 memory modules, all issues instantly went away and ran rock solid.

Unfortunately PC100 (128MB stick) made no difference 🙁

Looks like the capacitors will be here tomorrow though, so I can get them changed at the weekend and see if that makes a difference.

The board is just too worn out. Unless you have experience with replacing capacitors you may kill it. Learn it on a board you don't care about - e.g buy a worn out Socket A board just for recap practice. I would recommend always having a spare in case recap will not be successful. I have 4 boards with 440BX, Socket A, Socket 754. It makes things very easy to diagnose. Boards are also most valuable in the long term. There is plenty of CPUs, RAMs, graphics cards on the market.

AlexZ wrote on 2023-05-11, 18:00:

The board is just too worn out. Unless you have experience with replacing capacitors you may kill it. Learn it on a board you don't care about - e.g buy a worn out Socket A board just for recap practice. I would recommend always having a spare in case recap will not be successful. I have 4 boards with 440BX, Socket A, Socket 754. It makes things very easy to diagnose. Boards are also most valuable in the long term. There is plenty of CPUs, RAMs, graphics cards on the market.

paradigital wrote on 2023-05-09, 14:08:

Fortunately, I have both a decent desoldering station, soldering iron, hot-air station, and a decent multimeter, though not a cap tester/esr meter.

I'll order the relevant caps tonight and fingers crossed it'll resolve the issues. I need to re-attach an SMD cap on the Geforce 256 anyway, someone has done a previoulsy dodgy looking repair.

If he already has all this tools, i'm pretty sure he knows how to de-solder/solder.

Yes, I’m more than capable of through-hole and most surface mount. BGA is where I draw the line, I’m no Louis Rossmann.

I’ve recapped my dual S370 Tyan with success, so I’m sure I’ll be fine 😉

Check the CPU die, many S462 Athlons/Durons suffer from chipped die. This often occurs when bigger coolers appeared and the weight and dimensions would damage the die during installation process. That is why Intel went with Heatspreader on their P4 line.

aaronkatrini wrote on 2023-05-11, 19:12:

Check the CPU die, many S462 Athlons/Durons suffer from chipped die. This often occurs when bigger coolers appeared and the weight and dimensions would damage the die during installation process. That is why Intel went with Heatspreader on their P4 line.

It’s SLOT A, not socket A.

The CPU is physically fine.

My bad, didn't read the title properly.

Thought I’d cracked it today. Ditched the USB mouse for a PS/2 one and had a good hour or so of stability, but now it’s back to its usual self.

Oh well, caps should be here tomorrow.

Check the mosfets as well. I have an MSI MS-6167 - posted once, then a whiff of magic smoke. It smoked one of the chokes near the CPU slot. Pulled all capacitors and tested with ESR meter - they're still within spec. Replaced them anyway with a mix of Panasonic and Rubycon low ESR caps. Still magic smoke. No apparent short, perhaps an occult one, but I believe if I had a short it would not have posted in the first palce, and instead of a whif of magic smoke and some charring there would be a loud bang.

Here's my thread: Help diagnose and repair Slot A MSI MS-6167 motherboard - possibly bad choke, MOSFET or short circuit

MSI did something I consider unhealthy with these boards by putting the linear voltage regulators upright and slapping some dingy little tin heatsinks on top. Other manufacturers soldered them horizontally so they dissipate heat into the PCB, witch is in essence a huge hunk of fiberglass and copper.

So far I suspect one or more of the linear regulators on my board have gone bad and is letting too high a voltage go straight to the charred choke - not 100% sure of this since my test CPU still works, but I have no other working theory. Next step is to desolder the regulators and test all of them with a component tester.

Fortunately there are no hot smells or otherwise odd behaviour from this board, just the instability. It looks like the 6191 has less FETs than the 6167, so hopefully the heatsinks can cope with the heat dissipation:

paradigital wrote on 2023-05-12, 17:15:

Fortunately there are no hot smells or otherwise odd behaviour from this board, just the instability. It looks like the 6191 has less FETs than the 6167, so hopefully the heatsinks can cope with the heat dissipation:

IMG_1901.jpeg

Have you tried replacing the CPU's thermal paste? I noticed you're using an Athlon 900 - so it's very likely a Thunderbird core. These guys get pretty hot, and the thermal compound inside the CPU card has most likely turned into plastic.

Checking temps in BIOS is pointless, since most slot A boards do not have a CPU temp sensor, and those that do - simply have a temp sensor that sits in front of the CPU slot, right under the heatsink, and is by no means an indicator of the CPUs operating temperature. On my board CPU temp in bios read 33C, 40 with the case closed, while a temperature probe I put directly in between the heatspreader and CPU card read 79-82C with the case closed, despite my CPU having 2 fans!!!

I opened my 750MHz Athlon (thunderbird core) and noticed the thermal compound was dry and hard, and also to my surprise there was no thermal paste whatsoever between the heat spreader and the heatsink....

This is what my 750MHz athlon's CPU card looks like with the heatsink and heatspreader removed:

I did not clean the CPU! That's what it looked like when I removed the heatspreader.... dry paste and very little of it.

The cooling solution is made out of 2 components - an aluminum heat spreader and the heatsink itself:

The black part is the radiator and the silver one is the heatspreder. The radiator is fitted to the heatspreader either via a clip, or via 4 screws. Usually on the later slot 1 CPUs the heatsink is attached with screws, and on earlier ones it's attached with an easy to remove clip. My chip used screws.

As you can see from the pic above, they do not come with thermal paste between the heatspreader and heatsink - witch is fine on older slower chips, but for anything over 800MHz, there should be some sort of thermal compound there or at least a thermal pad to help transfer heat to the heatsink. I strongly recommend adding a thin layer of paste there, some cheap stuff since will be fine since there is a lot of contact area between the heatspreader and heatsink.

To clean and repaste the CPU, you will need to remove the fan (or fans, mine came with two), then remove the heatsink from the heatspreader, either via the clip or via screws. After that take a pry tool or flat head screwdriver and remove the plastic cover CAREFULLY. Be careful not to touch the PCB of the CPU when prying the plastic cover off. It's best to use a plastic pry tool or thick guitar or bass pick to make sure there's no way you'll accidentally scratch the CPU.

After the plastic cover is off, you will see two pressure clips holding the heatspreader to the CPU (and L2 cache chips on older slower models). Remove the clips using a small flat head screwdriver (do not touch the PCB with the screwdriver - you can accidentally scratch it) and then you can pull the heatspreader off, clean the old paste off the CPU die and heatspreader, then repaste. Personally I also like to rough up the contact surface between the heatspreader and die to get better heat transfer. Just cleaning the old paste off should be fine.

After repasting and adding thermal compound between the heatspreader and heatsink my CPU temp wend down to 57-58C

An overheating CPU could also be causing the symptoms you're describing. That or a failing MOSFET.