jheronimus wrote:

What am I missing?

A POST diagnostics card, I guess 😀

Does your board also have SIMM slots (some versions I've seen around the net have them)? If so, try using SIMMs instead. Also, carefully inspect all the RAM slots for bent/oxidized/burnt contacts. If you have a multimeter, check for VCC (3.3V) in the DIMM slots (pins 84 and 168, among others).

If you have the means, check the content of the BIOS chip/update it to the latest version.

On a non-flexing surface (wooden board should be good), try powering the board while applying gentle pressure with your thumb to the chipset, see if that changes anything.

I know I have a Socket 7 board (but cannot remember which) that is extremely picky about the DIMMs it will work with, and only starts with very early ones that use low-density chips. If I remember correctly there are like 16 chips for a 32MB DIMM...

quicknick wrote:

A POST diagnostics card, I guess 😀

I actually have one, but I don't understand how to use it, because the manual seems to be written using Google Translate. Here's what it shows:

quicknick wrote:

Does your board also have SIMM slots (some versions I've seen around the net have them)? If so, try using SIMMs instead. Also, carefully inspect all the RAM slots for bent/oxidized/burnt contacts. If you have a multimeter, check for VCC (3.3V) in the DIMM slots (pins 84 and 168, among others).

No, it doesn't have SIMM, unfortunately.

quicknick wrote:

If you have the means, check the content of the BIOS chip/update it to the latest version.

Unfortunately, I don't have tools for that.

quicknick wrote:

On a non-flexing surface (wooden board should be good), try powering the board while applying gentle pressure with your thumb to the chipset, see if that changes anything.

quicknick wrote:

I know I have a Socket 7 board (but cannot remember which) that is extremely picky about the DIMMs it will work with, and only starts with very early ones that use low-density chips. If I remember correctly there are like 16 chips for a 32MB DIMM...

That was my first theory, too. However, none of my DIMM sticks seem to work, even the ones I've used on other 430TX/MVP3 boards.

Try a hotflash on another board to flash a new bios on the chip, might fix it and ain't that hard to do. (and get a programmer if you work with a lot of boards! It's awesome.)

Sooo, I've randomly tried a Pentium 133 and it works. Pentium 200 works too.

However, none of my MMX (200, 233, tried 5 CPUs) work. There's a thread about this issue, but my VRM doesn't seem to be damaged (at least, visually).

jheronimus wrote:

Sooo, I've randomly tried a Pentium 133 and it works. Pentium 200 works too.

However, none of my MMX (200, 233, tried 5 CPUs) work. There's a thread about this issue,
but my VRM doesn't seem to be damaged (at least, visually).

I'm not entirely sure that my VRM is damaged despite the boiled looks.
Theres instructions somewhere here on Vogons on how to measure a MOS-FET and mine seems to behave correctly.
Out of curiosity, what BIOS version are you using ?
The one thing I havn't tried is to downgrade the BIOS, and ATM my testbench is hosting another project.

Sorry to see you having the same problem, but maby that increases the chances of solving the issue =)

edit: Here it is:
SS7 motherboard not working

I've managed to start with MMX-233, but at a pretty high voltage of 3.4v instead of 2.8v. That's the only voltage that works, but obviously this means heavily overclocking the CPU.

I've tried three BIOS version:

- patched with K6 support from here — works
- 0108 (last stable version from Asus) — works
- 0112 (last beta from Asus) — wouldn't boot even with non-MMX. Had to do a hotflash just to revert to 0108.

Right now BIOS shows me the following:

So the CPU is really hot even with active cooling and the voltage is at least 3v. In this situation the motherboard can't even boot of a floppy, so this is definitely not a working solution.

bofh.fromhell wrote:

I'm not entirely sure that my VRM is damaged despite the boiled looks.

It looks very tired, I'd replace it - especially if I had problems with one of the CPU rails. That mobo is interesting in that the silkscreen says LT1084, which is a power LDO regulator. But instead you have a combo of LM431 and FD3055. This is not a switching regulator setup, don't let the nearby HIP fool you. This particular voltage rail is linear and so the power transistor will run hot - but will provide cleaner output with no ripple. With either LM+FD or the LT part. But if the hot part somehow gets outside it's spec, but not blown completly, it'll no longer provide correct voltage but will pass basic short/open tests.

In general, if you already tried a CPU in the socket so there's no risk this advice will fry a fresh, good CPU: Stick the MMX CPU in, and measure voltage on the output of the power transistor. Just be super careful not to accidently shove the probe and short the pins. Even a very brief short like that could kill the CPU. Anyway, you should see the voltage that you selected for the CPU.

On that particular board there are 2 MOSFETs: the FD3055 and NDB603AL. The NDB is most likely the main power transistor providing power to CPU core, it's controlled by the HIP and I assume there is a decent inductor below it, not seen on the photo. The FD powers only the auxilary rail for CPU I/O needed in MMX CPUs. So if normal CPUs work but MMX don't then really the only thing that could go wrong with this setup is the FD going bad. Other failure modes are so rare that I'd not even consider them but replace the FD transistor as the very first thing.

So, the voltage measurements should be done on the right-most pin of NDB and that should give you 3V3 for non-MMX CPUs and 2.8 for MMX ones. Depending on the jumpers in other words. The FD3055 seems to be wired opposite of what you'd normally expect but possibly because it has to share the same PCB space as LT1084, so on this one the tab is the output voltage. It should be 3V3 always, no matter what is set with the jumpers. Except maybe if core was set higher than 3V3 but that's not a setting any Pentium-class CPU should have.

EDIT: Well, one more failure mode is pretty possible, depending on how the power is routed from the PSU socket to the transistors. It could be that the main MOSFET died shorted, or that HIP is for some reason driving it ON at all times. This way the CPU ESD protection diodes would be routing the power from the CORE to the I/O and back through the poor aux MOSFET. That baby would get really hot, close to overloading. Not to mention constant overvoltage situation on the CPU core. So always check both power transistors.

Deunan wrote:

It looks very tired, I'd replace it - especially if I had problems with one of the CPU rails. That mobo is interesting in that t […]
Show full quote

bofh.fromhell wrote:

I'm not entirely sure that my VRM is damaged despite the boiled looks.

It looks very tired, I'd replace it - especially if I had problems with one of the CPU rails. That mobo is interesting in that the silkscreen says LT1084, which is a power LDO regulator. But instead you have a combo of LM431 and FD3055. This is not a switching regulator setup, don't let the nearby HIP fool you. This particular voltage rail is linear and so the power transistor will run hot - but will provide cleaner output with no ripple. With either LM+FD or the LT part. But if the hot part somehow gets outside it's spec, but not blown completly, it'll no longer provide correct voltage but will pass basic short/open tests.

In general, if you already tried a CPU in the socket so there's no risk this advice will fry a fresh, good CPU: Stick the MMX CPU in, and measure voltage on the output of the power transistor. Just be super careful not to accidently shove the probe and short the pins. Even a very brief short like that could kill the CPU. Anyway, you should see the voltage that you selected for the CPU.

On that particular board there are 2 MOSFETs: the FD3055 and NDB603AL. The NDB is most likely the main power transistor providing power to CPU core, it's controlled by the HIP and I assume there is a decent inductor below it, not seen on the photo. The FD powers only the auxilary rail for CPU I/O needed in MMX CPUs. So if normal CPUs work but MMX don't then really the only thing that could go wrong with this setup is the FD going bad. Other failure modes are so rare that I'd not even consider them but replace the FD transistor as the very first thing.

So, the voltage measurements should be done on the right-most pin of NDB and that should give you 3V3 for non-MMX CPUs and 2.8 for MMX ones. Depending on the jumpers in other words. The FD3055 seems to be wired opposite of what you'd normally expect but possibly because it has to share the same PCB space as LT1084, so on this one the tab is the output voltage. It should be 3V3 always, no matter what is set with the jumpers. Except maybe if core was set higher than 3V3 but that's not a setting any Pentium-class CPU should have.

EDIT: Well, one more failure mode is pretty possible, depending on how the power is routed from the PSU socket to the transistors. It could be that the main MOSFET died shorted, or that HIP is for some reason driving it ON at all times. This way the CPU ESD protection diodes would be routing the power from the CORE to the I/O and back through the poor aux MOSFET. That baby would get really hot, close to overloading. Not to mention constant overvoltage situation on the CPU core. So always check both power transistors.

Don't wanna hijack TS's thread, but if we have the same problem it should be OK.
I'd replace the "FD3055" in a heartbeat if I could find a replacement.
The "FD3055" seems to be discontinued since long, and I have no clue what a replacement would be called.

Measure the voltages on transistor pins first. You could be right and FD3055 got toasty but didn't actually die. In fact measure the voltages without the CPU inserted first, then with the CPU in place. Again depending on how the power is routed, you might get two considerably different readings if one of the MOSFETs died.

But I take back what I said about measuring the NDB - I mean the pin is correct, but don't measure there. It's a switching circuit, find the inductor connected to it (on these mobos there's probably only one anyway) and measure on the other end of the inductor. That other side should be connected to a big capacitor, or even several of them.