Windows 98 Second Edition in numbers
Since I still use Windows 98 Second Edition as my main OS, I thought I share some of the experience.
|Last install date:||2007-02-08|
|Last recovery from full backup:||2010-07-03 (virus infection)|
|Last registry recovery:||2016-06-15 (accidential HKCR wipe)|
|Virus infections since last install:||1 (IE6 vulnerability)|
Today mainboard swap to the replacement, since my mainboard ceased to work. No time to deal with it, as long as at least one of them works.
One month without problems, until today. Nothing serious, except that the sound output on the ESS Solo-1 sound card went from stereo to mono. Solder on one of the 3.5mm Jack socket pins came loose, probably due to the fact, that the manufacturer tried to save on solder by all means. After giving that pin, and all the others, some fresh solder, the problem went away.
Unexpectedly I found a schematic for the MS-8917. The mystery C1319 is part of an array of 10nF/16V 10% X7R capacitors. The C1102 is a 100nF by-pass capacitor. Not that I actually understand the purpose of either of these, since the schematic just shows them as connected to FBVDDQ (2.6V) and GND.
Also unexpected was to find the schematic for the entire mainboard (MS-7043), although for an older revision. Not that I need it right now, but may come handy, if the thing continues to refuse to cooperate with the VT6122.
The desoldering pump arrived. As expected, it came with a different plug (Type I). For that reason it also came with a for export only, travel adaptor; not that it will fit any Europlug compatible socket. As such, the rework tool needed some rework first. The power cable was on the a bit on the thin side considering the plug claimed a 10A rating, but at least there are three leads and the ground wire actually connects to the metal parts of the tool. As for operation, it sucks. For copper-zoned holes dual-wielding with the soldering iron is necessary, which is hard. One trace on the old mainboard died in the process, as the tip became abbrasive after a few rounds of desoldering.
Back to the business at hand, recapping the MS-7043. On 2020-10-12 I noted, that the 1000uF/6.3V capacitor near the south bridge has been replaced, but that was not true. I was inspecting the replacement mainboard and assumed that it was replaced on my mainboard, because I replaced all 1000uF and 1500uF capactors there. But today I stumbled upon a 220uF/16V, which was not on the BOM and it was low in capacity. This is not the first time that the two mainboards differ. In general the replacement mainboard (US) was equipped with branded components, whereas my mainboard (CZ) has some no-name components on it instead.
Having finally replaced all capacitors on my mainboard by now, I went for testing. On-board LAN makes some attempts at starting, but as soon as the speed is negotiated, the connection breaks again. Limiting the speed to non-gigabit values does not work at all. The components hidden inside the ethernet socket are suspect, but getting in in non-destructive way seems hard. WLAN and LAN PCI cards now work, though. So after installing and configuring everything, I could go back to normal PC use.
The caps arrived. Replacing the two 47uF/16V revived the hardware T&L capabilities of the graphics card on both mainboards, no change on the VT6122, though. I intent to replace all the capacitors, that were not replaced yet, but that has to wait a bit, see below.
The PCI-E graphics card is drying; the reason being, that it was covered with dried up traces of some questionable liquid. Since that is already literally dirty, lets assume, that it is solder flux or leaked electrolyte.
The MS-8917, well… I have removed and tested the remaining caps, and they were fine. For a test, I tried giving it fresh capacitors, but it still did not work, so I put the old caps back in. After some intense staring though a magnifying glass I found two spots with torn-off SMD components, that, according to the silkscreen, were ceramic capacitors. One is needed by a memory IC (C1102) and I might be able to figure out its capacity, because there are four memory ICs and the others has the exact same component layout. The other (C1319) is mystery, but the package is about the same.
In the end, I hoped to get the old mainboard to work at least. What seemed like easy to do, since it is just, well, recapping, turned out to be a test of patience. Apart from the northbridge, all heatsinking is done by the PCB, especially on the VRM capacitors. The solder melts on surface, but that's it. To get it to melt through the whole hole, the entire copper-zone has to heat up, on both sides (one for each lead). Getting the capacitor out works with some force, but getting one in makes for some great capacitor BBQ. The solder wick does not clean up the holes, because solder does not melt; the copper is driving the iron already to its limits. No way around a desoldering pump I guess.
Did not take long. Two capacitors in the neighborhood of the VRM PWM IC, that are supposed to be 47uF/16V, both give about half the capacity and too high ESR, dead giveaway. The replacement mainboard does not have it that bad, but these two caps are also out of specification.
For even more fun, I took more components out of the graveyard. One AGP graphics card (MS-8917), which had one leaky capacitor replaced, but is asking for two more. And one PCI-E graphics card, that I received for a fix or trash scenario, now asking for three replacements of bulgy caps. The repair BOM counts 141 capacitors… could be worse.
The ESR70 arrived. The spartan and neat packaging, and absense of the typical cushioning material gives the impression, that the money really went into the device.
Measuring the current major suspects did not reveal anything. So I am left with going through all components one by one.
And having done so much with these two mainboards, I felt like taking out the original mainboard, an MSI MS-6163, that once occupied this computer. It's fitted with a Pentium III in a slot configuration, running at 500 Mhz with 512 MB of RAM. Back around Xmas 2005, the computer just turned off while I was sleeping and left horrible smell behind. I have thought, that the processor had burned, because the fan failed, but stashed everything inside a drawer anyway. Looking at it now, seeing all those blown capacitors (ever heard of capacitor plague?), there is some hope to bring it back to life. Oh fond memories of playing RO over bluetooth connected to a wireless CDMA modem before I learnt, that there is something called WLAN.
While looking for more clues, on what could go wrong with a mainboard, I have stumbled upon a white papers for the PT880 north-bringe and the VT8237 south-bridge which both occupy my mainboards. They made me realize the fact, that both components suffer from under-utilization on my mainboard, but also that the south-bridge is responsible for the PCI bus, which is currently my primary suspect, as both the VT6122 and the AGP port relate to it. There is also mention, that the south-bridge includes an another Fast-Ethernet network controller.
The three components in the vicinity of the south-bridge that were questioned, were two capacitors and a voltage-regulator.
The bigger 1000uF/6.3V Teapo capacitor has already been replaced. The smaller 100uF/6.3V of the same brand passed the continuity test, but I have no means to measure it, yet. Reluctantly I have ordered an Atlas ESR70 to deal with these things in the future. The voltage regulator, an APL1084, is supposed to turn 5.0V to 2.5V to supply Vcc of the south-bridge, which it does and the resistors that make it do so, have the right values as well.
Continuing the search, the questioning fell upon the components that get hot. Given the cold weather here, I have compensated the lack of a thermo-camera by checking components that are hot to touch or heat-sinked. Checking the north-bridge is out; it gets hot, but not too hot and given the fact, that the computer boots, I can assume that it still works. The processor has already been tested, since I have a replacement now. What remains are voltage-regulators and power MOSFETs.
The MOSFETs next to the processor, that are part of the CPU VRM, are the only other heat-sinked components on the mainboard. Removing the heat-sinks required desoldering two posts for each of the two heat-sinks. The components are connected to the heat-sink by the means of thermal-tape, which for both mainboards looked like it has been boiling once or twice, though the big heat-sinks featured one MOSFET, that did not stick to the tape, additionally on my mainboard the non-sticky part was charred. Gettting that stuff off was tricky. It gets all slimy when hot, but I do not have a hot-air gun, so I had to scrape it off instead. The result is not pleasant and will need some more polishing. It has also shown, that the heat-sinks were not anodized. The MOSFETs themselves passed continuity tests, so did the doubler and PWM generator. The yellow capacitor, that has been sizzling between the two groups of MOSFETs the whole time, will have to wait for its evaluation.
The remaining MOSFETs on the board seem to do what they are supposed to. The voltage-regulators output the proper voltages.
Doing more thorough component checking revealed, that the replacement mainboard is defective for other reason than mine; there is a long scratch that tore off a ceramic capacitor, bent few legs of the VT6122 and shaved its edge. There was a burr on the near PCI slot as well. At the mainboard edge the scratch is so deep that one trace is interrupted. Both should be possible to repair.
The capacitor was donated by my mainboard for the time being. To fix the trace, I went for the botch wire approach to avoid damaging the mainboard more than it already is. My soldering skills are still pretty poor.
Now the replacement mainboard behaves like mine, properly freezing up Windows during boot.
The replacement mainboard has arrived. While it looks in better shape than mine, it has the same defect as mine, except that Windows boots, only complaining, that the LAN is not working. Even the same capacitor is an open circuit... what are the chances of that happening? Probably false positive.
After considering, that this repair is going to be expensive and time-consuming anyway, I have looked for a replacement board on ebay, found one and bought it. Taking into account, that it comes with a Pentium processor (as opposed to my Celeron), the price was just fine.
The fact, that the on-board LAN stopped working completely after replacing the capacitors, ended up being due to my poor soldering skills. Reflowing solder on one capacitor returned the on-board LAN back to its half-working condition, where it prevents Windows from booting. Since the VT6122-related circuit did not show any problems, I have moved to components around the processor. Where I thought, that I have a spare processor laying around, turned out, that the socket was different, so testing the processor is out. So at least before putting the processor back into its place, I decided to check all components obscured by it for continuity. Result, one capacitor does not charge (open circuit).
The datasheet did not provide many clues to me about what could be wrong, but certainly provides information about few components that should be right. The whole circuit is pretty tiny to mark it on board, so I have decided to track each and every connection from the IC and document it (read reverse-engineer that mainboard area). Of course with electronics, using an EDA program made sense, though I was not familiar with any of them. The typical first choice was Eagle, but that was paid in the past and since it was acquired by Autodesk, it became cloud-infested. In the end, I haved settled with KiCad, which was recommended to me by an acquaintance. Thanks to the introductory tutorial, getting started with it was easy. Thus I spend rest of the night creating the VT6122 component and then tracing each and every connection to it.
The whole-night capacitor replacement did not yield much difference. The mainboard worked as it did before, so nothing got fixed, but nothing broke either, except that the on-board LAN stopped working completely, so it no longer was interfering with the Windows boot. Since I could not find any datasheet for the VT6122 on the internet, I have asked the manufacturer, if they could provide one. While I got a prompt and friendly response, unfortunately they no longer support this ancient, 15 year old chip. At the end of the day, I found a leak of the datasheet after extensive google-fu.
At first the solder did not want to melt, but adding some lead-based solder fixed that. After figuring out, how to do the removal on the first capacitor, all the others were easy, except for one, which did not want to give up its spot. Soldering the replacements was another problem. For solder removal I opted for a desoldering wick, which was falling apart, instead of sucking up solder, and while I got the right capacitors in terms of capacity and voltage, the 1000µF/6.3V Nichicon capacitors turned out to be slightly thicker, than those, that were populating the mainboard before. In dense areas this means, that they are mounted at an angle, which looks comically at best.
On this day, after a restart, videos and graphics in general got slow. The "Direct3D HAL T&L" device was gone, too, only software vertex processing was available, which is, well, slow. Tried booting Windows 2000 (dual-boot for debugging), same, so a hardware issue. Since this was the third technical trouble this month (after the TV antenna fell apart due to corrosion and the ASUS DSL Router semi-bricking after an update), I was more than delighted.
Visual inspection of the graphics card did not show any obvious trouble. The mainboard, though, was riddled with puffy capacitors. After reading some websites on how to replace them, and a few restarts, both Windows 98 and 2000 stopped booting completely; they just hung around the device initialization phase. The culprit this time was the on-board LAN VIA VT6122. While the graphics did not bother me too much, this was a blocking issue, which prompted me to actually order replacement capacitors and a TS100 soldering iron (since mine was not up to the task). The capacitors arrived after few days, the iron after a week.
Stability (2019-08-06 ~ 2020-08-06)
|Starts:||51 (successful)||1 (failed)|
|Shutdowns:||31 (clean)||21 (unexpected2)|
|Uptime1:||38 seconds (min)||4 weeks 6 days 2 minutes 33 seconds (max)||6 days 4 hours 9 minutes 23 seconds (avg)|
Stability (2018-08-06 ~ 2019-08-06)
|Starts:||64 (successful)||3 (failed)|
|Shutdowns:||34 (clean)||27 (unexpected2)|
|Uptime1:||88 seconds (min)||2 weeks 6 days 1 hour 16 minutes 40 seconds (max)||5 days 12 hours 55 minutes 32 seconds (avg)|
Stability (2017-08-06 ~ 2018-08-06)
|Starts:||101 (successful)||6 (failed)|
|Shutdowns:||66 (clean)||25 (unexpected2)|
|Uptime1:||60 seconds (min)||3 weeks 1 day 17 hours 15 minutes 31 seconds (max)||3 days 14 hours 20 minutes 38 seconds (avg)|
Stability (2016-08-06 ~ 2017-08-06)
|Starts:||114 (successful)||4 (failed)|
|Shutdowns:||92 (clean)||18 (unexpected2)|
|Uptime1:||21 seconds (min)||2 weeks 3 days 6 hours 32 minutes 56 seconds (max)||3 days 2 hours 26 minutes 2 seconds (avg)|
Stability (2015-08-06 ~ 2016-08-06)
|Starts:||139 (successful)||3 (failed)|
|Shutdowns:||112 (clean)||24 (unexpected2)|
|Uptime1:||30 seconds (min)||1 week 5 days 9 hours 6 minutes 13 seconds (max)||2 days 1 hour 44 minutes 25 seconds (avg)|
- Measures time spend in GUI (from HKLM\RunServices till WM_ENDSESSION), only.
- Includes (in order of frequency): Whole OS stopping responding, power-outages, non-BSoD whole OS crash, BSoD whole OS crash (seldom).
- System + auto-started applications