Announcement

Collapse
No announcement yet.

IBM Thinkcentre S50 TYPE: 8086

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    IBM Thinkcentre S50 TYPE: 8086

    Hi all,
    Ok, this is my first post/thread, but not my first job. I've only recently discovered this forum, before that I've done a lot of recaps on older boards, but this one has me stunned.
    First, I did a search on this and a lot of other forums to no avail. Seems that no one had Thinkcentres in their workshop, which is cool for IBM, but unfortunately, these babies do fail and they fail big time - I did a full recap and I've still got nothing.
    Since I don't have the board with me now (will have in a few hours), I'll just explain the symptoms and what's done and then upload photos when I get the board.
    The board is labeled as MS-7024 (so it's a MSI) - socket 478, non-standard format. The board was filled with Nichicons HM (10v, 1000uF), 2 of which were swollen, near the southbridge (ICH5). Also, there were 3 large KZE caps near the ATX connector (25v, 820uF). These all (both Nichicons and KZEs) are replaced with appropriate Panasonic FM (all matching datasheets, of course). Symptoms: the computer fires up and stays on with Power-On LED and IDE LED both lit, fans running at full speed (with working processor, power supply and memory) - no beeps no nothing, just stays that way. It doesn't seem to matter if I pull out the memory and fire it up - still no beeps, same state. Unusual thing is that PATA attached drives are not spinning, nor functioning (appear as if they're not attached to power supply), but if I remove the IDE cables, they're functioning properly (sata drive is functioning properly when attached, though). Southbridge seems to be a bit hot for no apparent reason (does not have a heatsink on). The same thing was happening before I did the recap.
    Anyone ever had problem with these or knows what's going on? As I said - I'll post some pictures in a few hours.

    #2
    Re: IBM Thinkcentre S50 TYPE: 8086

    Try clearing the cmos. All measure pau voltages with a multimeter.

    Comment


      #3
      Re: IBM Thinkcentre S50 TYPE: 8086

      Forgot to say that I did that, does not help.

      Comment


        #4
        Re: IBM Thinkcentre S50 TYPE: 8086

        I damaged a trace while rcapping an identical board to this. I ended up with exactly the symptoms you describe.

        No Southbridge Heatsink? These boards NEED a Heatsink, the Southbridge runs hot. Mine was equipped with a large heatsink to which I added a powerful fan.

        Sounds to me like you've got a short or open circuit on that board.

        I ended up buying a replacement board for $20 off Ebay.

        Comment


          #5
          Re: IBM Thinkcentre S50 TYPE: 8086

          The board had the exact same symptoms before the recap, so... I don't think it's the trace. Will check, though. I must admit that some caps were tricky to get to and that I did burn the upper lacquer, but I got nowhere near the trace itself (copper is still well under the blue finish). I'll doublecheck with multimeter just to be sure, but I don't think this is the problem (as I said, I did quite a few recaps - there were a lot of them that looked much worse than this one and still functioning).
          Btw, you're sure it's the same board? This one has absolutely no space for a southbridge heatsink:
          http://publish.it168.com/2004/1206/images/n19773.JPG
          and:
          http://img01.taobaocdn.com/bao/uploa...EIT_013340.jpg

          Comment


            #6
            Re: IBM Thinkcentre S50 TYPE: 8086

            Jimmy:

            Braincramp on my part. I put a fan on the Northbridge Heatsink, as the heatsink was almost hot enough to burn your hand.

            I would still recommend that, as the heat tou remove from the northbridge wont find its way south.

            I still get temp readings of 67C out of Hardinfo on my other IBM M50 board
            ( I run Linux).

            I still think you have a short or open circuit since my board had the same symptoms. I would try reflowing the solder on the caps.

            You can also set the CPU fan to remain ON all the time.

            Other option is to switch to Northwood CPU which runs 15C cooler than Prescott.

            I have a Gigabyte 478 board running the Prescott and it runs way cooler -
            a reported 37C.

            My M50 board is similar to yours but a different model.

            Comment


              #7
              Re: IBM Thinkcentre S50 TYPE: 8086

              Originally posted by jimmy274 View Post
              These all (both Nichicons and KZEs) are replaced with appropriate Panasonic FM (all matching datasheets, of course).
              The ESR of the Nichicon HM caps is lower than that of the Panasonic FM's, hence the Panasonic FM caps might not be good enough.

              Comment


                #8
                Re: IBM Thinkcentre S50 TYPE: 8086

                It sucks that ibm (top of the line overpriced business pc) cant escape bad cpas.
                I saw one a while back. A p4 sff with all the caps byt the cpu bulging. They should have used polymers

                Comment


                  #9
                  Re: IBM Thinkcentre S50 TYPE: 8086

                  Just a few more things before I post the pics (still hadn't gotten around to do so).
                  The CPU is Celeron D 2.4, and even though it's a Prescott core, the heatsink doesn't get too hot, but I probably will switch to Northwood as soon as the machine actually posts.

                  @bigbeark: Even though bunch of people think that TDP of ICH-5 is too low to be cooled, I agree with you - if you touch the surface with your finger and fire up (first ground yourself of course), you can't hold it there for very long - it gets very hot. I did use some non-RoHS solder, so it might just be that somewhere there's an open circuit, will measure up. And yeah - M50 was also built in the same case, so there's no doubt that they have probably the same boards.

                  @momaka: Well, the Nichicon HM ESR is not that much lower, so I don't think that this is the problem.

                  @shovenose: I agree - idiots. They do tend to use polymers on newer thinkcentres for CPU VRM, though. In my S50 there's a bunch of Rubycon MBZ for CPU voltage regulation, so I'll let them slide... for now

                  Anyway, I do have the schematic of the board (I just don't think I should post it anywhere) and I do remember that one of the swollen caps was in charge of DDR Vcc, which seems to be fine now (output mosfet is somewhere on the + side of 2.6v), although I do need to check it with oscilloscope to see whether there's some significant ripple, just to be sure. Anyway, I'll recheck everything and post back with pics. In the meantime, keep brainstorming

                  Comment


                    #10
                    Re: IBM Thinkcentre S50 TYPE: 8086

                    My IBM xSeries 206 server was fitted all out with Chemicon KZG all over the board, they all failed.

                    I jammed Panasonic FM in as a temporary fix about 9 months ago, it's still running fine, so I'll use it till it dies.

                    Comment


                      #11
                      Re: IBM Thinkcentre S50 TYPE: 8086

                      Yeah, Chemicons tend to do that. This S50 board had 3 KZEs, and even though they usually don't fail, I still replaced them with Panasonic FM (I get bad vibe from Chemicons).
                      Temporary fix? Why? These Panasonics are great (I also used NHG series for some of my diy audio stuff, but they're not low-ESR) - I would even go as far as saying that I like them better than Nichicons - I've seen a lot of failed Nichicons (some of them being in top of the range tube audio amps - these were not HM, they were PW or PM made in late 80s through early 90s), and I'll probably stay with Panasonic as my main recap substitute.

                      Comment


                        #12
                        Re: IBM Thinkcentre S50 TYPE: 8086

                        Temp fix because they're the wrong diameter, so they're a bit wonky here or there.
                        I only had to get the server running for long enough to migrate the domain across to the new server.
                        Then it became mine.

                        Comment


                          #13
                          Re: IBM Thinkcentre S50 TYPE: 8086

                          what does it matter if it looks a bit wonky?

                          Comment


                            #14
                            Re: IBM Thinkcentre S50 TYPE: 8086

                            I use Panasonic and Rubycon for recapping. Excellent caps.

                            Comment


                              #15
                              Re: IBM Thinkcentre S50 TYPE: 8086

                              It matters to me, I like things to be absolutely right.
                              I use primarily Panasonic - they are good.

                              My PSUs are suffering from 'too good' though. I need to go back over them and get a scope.

                              Comment


                                #16
                                Re: IBM Thinkcentre S50 TYPE: 8086

                                New info guys. Traces around the changed caps are fine, I did a recheck and could not find the damage. However... The Vcc_agp and Vcc_ddr are pretty much shorted - When I check + and - terminals across some caps (used to filter Vcc_agp and Vcc_ddr), I get 40 ohms and 30 ohms (no fluctuation - hard 40.2 and 29.8).
                                So either I've got something shorted or something was shorted all along. Suspects - bunch of Rubycon YXA and some ZL (yes, I know - hardly these are making problems), northbridge (the blue finish has gone a bit yellowish under it), power control ic (MS-7) or the Gnd plane has somehow got shorted to something. I may be a bit suspicious about the newly installed Panasonic FM, but I really don't see how they could be causing the short.
                                Any ideas?

                                Comment


                                  #17
                                  Re: IBM Thinkcentre S50 TYPE: 8086

                                  Same ohms with leads reversed?
                                  Mann-Made Global Warming.
                                  - We should be more concerned about the Intellectual Climate.

                                  -
                                  Be who you are and say what you feel, because those who mind don't matter and those who matter don't mind.

                                  - Dr Seuss
                                  -
                                  You can teach a man to fish and feed him for life, but if he can't handle sushi you must also teach him to cook.
                                  -

                                  Comment


                                    #18
                                    Re: IBM Thinkcentre S50 TYPE: 8086

                                    Good point. Will check and post back.

                                    Comment


                                      #19
                                      Re: IBM Thinkcentre S50 TYPE: 8086

                                      Checked! Very weird results - the readings are mismatched by 1 ohm:
                                      1. Vcc_agp '+' to '-' (correct polarity): 41.1 ohms,
                                      2. Vcc_agp '-' to '+' (reversed polarity): 42.1 ohms.
                                      Vcc_agp is made from +5V and has one Ruby YXA 10v 470u on it's exit (and it's not him that's shorted, tried by removing it - good as new). Also, it's made by an SMPS - it has one switch (ic mosfet - has 2xmos inside - Nikos P2103HV), one choke and the before-mentioned output cap. It also has filtering all over the motherboard - some of it was done by one of the Nichicon HMs that died out. Also, plenty of filtering and bootstrapping caps all over the motherboard (the small 0805 SMD ones), which all appear to be shorted (probably all ok though). Double mos makes this dc/dc converter a huge pita...
                                      So... With the middle removed (all the caps for filtering), the beginning and the end are SMPS and Northbridge/Southbridge. I'm guessing it's one of these that's making a short. Hopefully SMPS, because it's really not a problem to change out a mos...
                                      Any thoughts?

                                      Comment


                                        #20
                                        Re: IBM Thinkcentre S50 TYPE: 8086

                                        No thoughts on this whatsoever? Anyone?

                                        Comment

                                        Working...
                                        X