Broken IC on old Pentium 2 motherboard

[james_s]

This is actually quite interesting, back when I had a PII I didn't really know much about the technical side of the hardware. I remember having to manually set the CPU core voltage of an older system using jumpers on the board though. Never really occurred to me how the later stuff automatically took care of that but it makes perfect sense now.

[Rasz]

Im guessing those mosfets arent all that healthy after all, or the diode went bad. Its also possible replacement chip is a dud.
if you have a scope put a probe on pins 14 and 17, if not I would just replace mosfets and diode, maybe even the electrolytic cap behind the coil.

james_s slot1 was pretty sweet for cheapskates, in 1998 a strip of insulation tape on pin B21 (+optionally three VID pins for 0.2V Vcore bump) was all one needed to make $150 Celeron 300A ~equal (sometimes even faster due to 2x faster cache) to fastest Intel offering at the time - $670 Pentium II 450MHz.

[james_s]

The one I had right around that time was a PII-300 (not Celeron) which was one of the later ones that would run at 450MHz right out of the box. Didn't even have to mess with the voltage, IIRC it was just a matter of setting the FSB to 100MHz vs 66. I used that machine for years and it was rock solid, later I gave it to my mom and she used it for several more years.

[spilihps]

This is actually quite interesting, back when I had a PII I didn't really know much about the technical side of the hardware. I remember having to manually set the CPU core voltage of an older system using jumpers on the board though. Never really occurred to me how the later stuff automatically took care of that but it makes perfect sense now.

Yeah, this is something i never payed much attention to either. and i guess when a board failed no matter how, one would just throw it away.

Im guessing those mosfets arent all that healthy after all, or the diode went bad. Its also possible replacement chip is a dud.
if you have a scope put a probe on pins 14 and 17, if not I would just replace mosfets and diode, maybe even the electrolytic cap behind the coil.

james_s slot1 was pretty sweet for cheapskates, in 1998 a strip of insulation tape on pin B21 (+optionally three VID pins for 0.2V Vcore bump) was all one needed to make $150 Celeron 300A ~equal (sometimes even faster due to 2x faster cache) to fastest Intel offering at the time - $670 Pentium II 450MHz.

Yeah, it would be very likely that they are bad in some way. Please see these images from the scope.

Pin 14 and 17 (these signals are present at the mosfets also)
I took snapshots of the drains on the mosfets also,  they seem a bit sketchy?

[james_s]

That second one doesn't look right, looks like it's just noise. I don't have a schematic in front of me so I'm not quite sure what should be there but I'd expect something similar to the others.

Are you sure all of the diodes are good?

[magic]

It has to be the high side switch, so its drain is just the 5V rail with bursts of switching noise
Meanwhile the low side switch drain shows 1/3 duty cycle PWM.

The waveforms are plausible, I think, but 1/3 duty cycle is not enough for 2V output from 5V. So either the output voltage sense network is screwed up (check all resistors from Vout to FB/COMP) or perhaps some overcurrent limiting or other protection is kicking in

[spilihps]

Yes it may be something with the sense, since that is also at 1.2v even when i made VID 11111 (that should switch it off i think?)

Will try to trace it out

[spilihps]

Quick update, so just for giggles, i checked the PGOOD pin, and it is at 3.3v. It should be low if it thinks that vsen is out of bounds.. So for some reason it thinks that it is OK?

[james_s]

That's pgood output from that regulator and not an input from the main PSU? You probably know already but ATX PSUs have a power good output that tells the motherboard when the PSU thinks everything is good to go.

[spilihps]

That's pgood output from that regulator and not an input from the main PSU? You probably know already but ATX PSUs have a power good output that tells the motherboard when the PSU thinks everything is good to go.

Hello, no this is an output from the chip itself, from the datasheet:
PGOOD (Pin 12)
PGOOD is an open collector output used to indicate the status
of the converter output voltage. This pin is pulled low when the
converter output is not within 10% of the DACOUT reference
voltage.

[magic]

Looking at the datasheet, it seems that the reference DAC may be outputting wrong voltage. See what's going on at the FB pin.

You said you can pull out the CPU and set the voltage manually? What happens if you set it to something other than 2V?

At any rate, it looks like the chip may be bad.

[Rasz]

That second one doesn't look right, looks like it's just noise. I don't have a schematic in front of me so I'm not quite sure what should be there but I'd expect something similar to the others.

second one is +5V supply, and other than lots of ringing (might need caps replaced) looks ~normal for synchronous buck converter.

Yes it may be something with the sense, since that is also at 1.2v even when i made VID 11111 (that should switch it off i think?)

Intel spec might say 11111 is invalid/off, but both original and replacement chips datasheets treat all ones as default 2.0 Volt. You have to set something different to see if anything changes (duty cycle). Pull VID2 to ground and lets see if that 1.263v goes up at all. Measure FB and COMP pins before/after too, as magic says there might be something wrong with compensation circuit - one of small smd resistors/capacitors might of went poof together with the chip.

I would replace electrolytic capacitor after the Vcore output coil just in case.

[spilihps]

That second one doesn't look right, looks like it's just noise. I don't have a schematic in front of me so I'm not quite sure what should be there but I'd expect something similar to the others.

second one is +5V supply, and other than lots of ringing (might need caps replaced) looks ~normal for synchronous buck converter.

Yes it may be something with the sense, since that is also at 1.2v even when i made VID 11111 (that should switch it off i think?)

Intel spec might say 11111 is invalid/off, but both original and replacement chips datasheets treat all ones as default 2.0 Volt. You have to set something different to see if anything changes (duty cycle). Pull VID2 to ground and lets see if that 1.263v goes up at all. Measure FB and COMP pins before/after too, as magic says there might be something wrong with compensation circuit - one of small smd resistors/capacitors might of went poof together with the chip.

I would replace electrolytic capacitor after the Vcore output coil just in case.

Edit: For some reason my text got embedded in the quotes.

It sure did, setting 11011 gets us up to 1.44V


Also, i tried to take some measurements of the resistors / caps that is for the vsen network. However, difficult since i think i need to pull each one to get an accurate reading. No shorts at least.


I include before and after snapshots from the scope, ( I think i managed to not mix them up    )

[Rasz]

Those readings + PGOOD make me think its acting like VID4 is internally read low all the time. Pull all VID pins to the ground and lets see if it produces 2.00V.

[spilihps]

Those readings + PGOOD make me think its acting like VID4 is internally read low all the time. Pull all VID pins to the ground and lets see if it produces 2.00V.

No go, 1.266V

[james_s]

Has this IC been replaced already? It really sounds like something is not right.

[spilihps]

You mean that it was a pull? Dont think so, it was in a reel package, and the legs looked fine. However, i did recieve 5 of them, you think its worth a shot to replace it again? Before pulling every resistor on the sense 

[james_s]

Oh I just meant is this the IC that blew up that you already replaced, sounds like it is. You're sure it's exactly the same as what was there and not a variation with some different parameters? I'm not familiar with this specific family of part but I've worked with other power management ICs that have multiple slightly different part numbers that have various differences. I suppose you could try swapping in another IC, but I would first try every combination of the configuration lines, then check very carefully all of the parts in the feedback loop and if you haven't already, make sure all of the config lines have continuity all the way back to the IC pin and that none are shorted to each other by a solder bridge, etc. Also I would test it while driving a load, probably best not to use the CPU itself, but maybe a power resistor, shouldn't really matter but it's possible that it doesn't like an open circuit.

[magic]

I don't think it's the feedback network because power good also indicates that voltage is within regulation and it uses a separate pin to sense the output directly.

Maybe check if the voltage at the input to power good circuit is the same as elsewhere, but at this point I suspect that it is and the chip simply generates wrong reference voltage internally.

Maybe it's a different part, maybe it's fake, maybe you need to replace all capacitors on the chip's supplies, no idea

Also I would test it while driving a load, probably best not to use the CPU itself, but maybe a power resistor, shouldn't really matter but it's possible that it doesn't like an open circuit.

IIRC this was tried already and output voltage was the same with the CPU or without.

[spilihps]

Guys, i think i have received the wrong part.
I have recieved HIP6004E instead of HIP6004.
It seems to be variable from 1.050 to 1.825 instead     

Im so sorry guys that i took your time with this. Completely my fault. Will try to source the correct part from somewhere.

[Rasz]

Guys, i think i have received the wrong part.
I have recieved HIP6004E instead of HIP6004.
It seems to be variable from 1.050 to 1.825 instead     

  that explains everything. This is VRM 8.5 specification for FCPGA socket boards (Coppermine/Tualatin) https://www.tomshardware.com/reviews/intel-roadmap-news-10,252-3.html

[james_s]

Well it sounds like you've at least validated that the rest of the parts are ok so all you have to do now is source the correct IC and swap it. That's an even more subtle difference in the part number than I'd have expected, you'll have to check carefully that any vendor actually sends you the exact part you ordered.

[spilihps]

Guys, i think i have received the wrong part.
I have recieved HIP6004E instead of HIP6004.
It seems to be variable from 1.050 to 1.825 instead     

  that explains everything. This is VRM 8.5 specification for FCPGA socket boards (Coppermine/Tualatin) https://www.tomshardware.com/reviews/intel-roadmap-news-10,252-3.html

Yeah i cannot believe it! Thanks for the info. And again, thank you for all your input and knowledge about this. Will keep you posted when the HIP6004CB (What i think is the right  ) Found a seller on ebay in china. Probably will have to wait over a month again. Seem to be impossible to get hold of these in Europe?

Well it sounds like you've at least validated that the rest of the parts are ok so all you have to do now is source the correct IC and swap it. That's an even more subtle difference in the part number than I'd have expected, you'll have to check carefully that any vendor actually sends you the exact part you ordered.

Yeah your right, this probably means that with the right part dropped in, everything should work. If there is no other problems somewere else after the other blew up.

Yeah, the partnumbers were not that easy to catch. It seems like i want the one that is HIP6004CB and found a seller on ebay in china.
Also, thank you for all your help and patients with this. Cant belive i wasted everyones time with this silly partnumber missmatch.

[magic]

 

Do you think that old Pentium would stand a chance at 1.825?

[spilihps]

Do you think that old Pentium would stand a chance at 1.825?

Are you suggesting hard wiring it up to 1.825 and see if it works?