From: nospam@needed.invalid   
      
   On Tue, 12/2/2025 8:40 AM, J. P. Gilliver wrote:   
   > On 2025/12/2 7:51:58, Paul wrote:   
   >   
   > []   
   >   
   >> And in totally unrelated news, I got another hint yesterday,   
   >> about what the problem is on my daily driver.   
   >>   
   >> I was typing away, I brought some window to the front.   
   >> The window stopped responding. I looked down, and my mouse   
   >> LED was off, and the shift key didn't work on the keyboard.   
   >> The bloody machine had turned off the +5VSB to peripherals   
   >> again.   
   >>   
   >> Now, normally when that happens, I'd be pressing Reset   
   >> and rebooting it. And the log would note a dirty shutdown   
   >> and no error recorded.   
   >   
   > Could you instead plug in a keyboard and mouse plugged into an   
   > externally-powered hub (possibly the same KB+M), rather than doing a reset?   
   >   
   >>   
   >> Well, this time, something different happened. A watchdog   
   >> timer went off. It seemed to be the NVidia driver that   
   >> was involved (daily driver uses a GTX1050 to drive the screen,   
   >> a low end video card). The driver actually recovered. And,   
   >> it seemed to send a report to Microsoft (there was network   
   >> activity).   
   >   
   > []   
   >   
   > So that caused the +5VSB to be turned on again?   
   >   
   I believe it did. As I went over and used Reliability Monitor,   
   to suck out that error code :-) While the network interface LED   
   was flashing and the error report was being sent.   
      
   Once the power drops on the I/O plate, no amount of   
   tomfoolery external to the machine, will cause data   
   to enter the dead port(s). The VCC on the USB I/O pad   
   has dropped, by the looks of it. I don't recollect any   
   other OS, doing something like this.   
      
   There *was* a working watchdog, during boot. When the   
   NVidia driver is coming up, and the Nvidia driver loses   
   contact with one of the GTX1000 series video cards, the   
   driver could recover. It was the Linux Nouveau driver,   
   that could not recover from that loss of communications.   
      
   For all of the movements of hardware that have been   
   used to try and figure out where the problem was, it   
   would appear there is something about the specific motherboard   
   that encourages it. I have an identical motherboard in   
   another machine, and it doesn't do that. But the other machine   
   doesn't have as much RAM as this one does, so the populations   
   of things are not consistent for a "test piece".   
      
   And for a good long while, this piece of shit "just worked".   
   It was not doing this from the beginning. It might have   
   been a year after getting it, that this started to happen.   
      
   I just think it is cool, that the watchdog is wired up again.   
   Whoever or whatever did it. When this first came out, it was   
   called "VPU Reset" and it was seen on an AMD video card. And the   
   notion of watchdogs is not new, and has been around forever   
   (like at my work).   
      
      Paul   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)