Why did the display become a snapshot of the last time the monitor was plugged into the computer?


I left the story of the return of the dead home desktop computer with a puzzle. When I plugged the monitor back into the original computer, it showed a snapshot of the screen at the time the monitor was unplugged. The computer itself continued operating, but the screen never updated. The frozen image remained until the power was turned off to the computer, at which point no image appeared on the monitor at all.

The reason became apparent when I opened up the computer to see what the matter was. The act of unplugging the monitor from the computer the first time had jiggled the video card loose from its socket. The card was still seated well enough that the pins responsible for power were still connected. The card continued to pump the contents of the VRAM to the monitor, and the power keep the VRAM from losing its contents. On the other hand, the pins responsible for transferring data between the motherboard and the video card were unseated. When the motherboard (under the direction of the video card driver) tried to tell the card, "Hey, change that pixel from blue to red" the card never received the signal. That's why the image was frozen.

This particular computer doesn't use screws to secure the cards to the backplane; it just has a hinged cover that allegedly holds the cards in place. Clearly this design leaves a bit to be desired.

With the power off, reseat the card, then power back on, and everything is back in business.

Congratulations to commenter Darkstar, who was the first to guess the correct explanation.

And a final remark on the dead computer: The new video card is noisier than the old one, alas. The old one used a heat sink for cooling, whereas the new one uses a fan. And none of the video cards in the store mentioned on the boxes what they used for cooling. (Though the one I chose proudly proclaimed Quiet operation which I interpreted to mean that it was fanless, but which on further reflection was actually an admission that it had a fan, for if it were fanless, it would have said Silent operation.)

Comments (23)
  1. Stan says:

    I’m really surprised the computer wouldn’t lock up from that.

  2. Nawak says:

    Raymond, you should buy something like that:

    http://www.zalman.co.kr/ENG/product/Product_Read.asp?idx=137

    It’s not really expensive and can be adapted to another card when you update. I built a very quiet PC using parts like this.

    If your motherboard chipset has a fan too (this tends to be rare, thankfully), I also advise you to buy a Zalman fanless cooler for it (about 4$)

  3. Thom says:

    Back in the Atari 2600 days, before I grasped firmware and digital signals I used to take the covers off my game cartidges and while playing a game I’d short out various combinations of ROM pins with a quarter or jumper them briefly with a wire.

    Sometimes the 2600 would lock up or just sit there mostly frozen with a few bits of the screen blinking.  Other times the colors or graphics would change or some other odd effect would appear and the game would continue as normal but ith with a surreal feel to it.  Every now and then though the game would freak out and my player would no longer obey the normal rules of the game.  It was wierd, and a waste of time, but fun for a pre-teen in a pre-computer age.

    I did that several hundred times over and to dozens of cartridges but never did a cartridge or the 2600 die.  It’s amazing how tolerant some hardware can be of intermittent shorts.  

  4. gkdada says:

    You’re lucky you end up with a fried video card. Once I forgot to screw down the video card. The plugging/unplugging the cable jiggled the card enough to fry it. It was almost brand new video card too.

  5. gkdada says:

    I meant: You’re lucky you DIDN’T end up with a fried video card. </sheepish>

  6. If you’re interested in quiet computing, the best inexpensive video card cooler at the moment is the Arctic Cooling S1. I have one cooling an overclocked 8800GT almost passively. (I placed an undervolted Scythe Slipstream 120mm fan on the case floor, blowing upwards to help keep the ram cool.)

    http://www.silentpcreview.com/article793-page1.html

    It’s a bit on the large side, so check the dimensions if you have a tight case.

    I’ve used many more expensive coolers that didn’t work as well as this one. Basically the thing trades tightly packed fins for very open ones, and just makes the whole area bigger to compensate.

    The secret to this cooler is twofold:

    • the heatsink fins are spaced widely apart, so you don’t need to force air between them…the slightest breeze will cool it
    • the fins extend well beyond the video card, where there isn’t any PCB or components impeding the airflow

    I picked it up for just over $19 CDN recently on sale, and it has exceeded all my expectations.

  7. James Schend says:

    Lorne, I just picked up one of these suckers from MSI:

    http://global.msi.com.tw/index.php?func=proddesc&prod_no=1236&maincat_no=130&cat2_no=137

    I couldn’t be happier with it. The Radeon 2600 is more than beefy enough for me, and their heatpipe system works great even in my small-ish Dell case. (I hear it’s a pain to fit in some cases, if your PCI-E is parallel with the CPU cooler they might collide since this card is "deeper" than most. Wasn’t the case for me.)

    There’s a $20 rebate on it at NewEgg at the moment too. (No I don’t own MSI shares, I just love this video card.)

  8. Larry says:

    Had something along those lines happen a long time ago.

    When a customer’s PC was powered on it displayed the screen (less some corruption) that was displayed when the machine was powered off the previous day!

    It turned out that the printer (which was never turned off) put enough power on the parallel port (because of a ground fault) to power the video refresh.

    Sure enough, power cycle the PC with the printer disconnected… no ghost display.

  9. Torkell says:

    Haha, doesn’t say much for the case design.

    It’s surprising how tolerant most hardware is of things like that. I’ve seen two cases now where a network card was just seated enough to get power, but not enough for Windows to recognise it. Of course, to make debugging more fun the card would happily speak to the switch on the other end without needing to be told to do so.

  10. Wesha says:

    Raymond, are you’re trying to tell us you’ve disconnected the monitor without first shutting down the computer?

    Gee, now that’s unprofessional (for precisely the reason you’ve ran into: you could’ve ended up with fried video card and/or motherboard.)

    No wonder I was beating my head trying to figure out how the image could’ve stayed in the monitor’s RAM, especially after you unplugged it from power (I assume. Or did you not unplug it from power? Unprofessional x2).

  11. James says:

    Some years ago, I briefly had the rather weird experience of seeing a combination of the Windows NT 4 and Windows 95 desktops simultaneously, for similar reasons (the installation process never cleared the entire video memory, so the old desktop remained in there apart from the small corner used for the low-res and text installation screens – until the mode was changed, making the old data visible at first!)

    Apparently you get similar effects with FreeBSD’s console boot messages: under some circumstances, a reboot leaves valid boot messages in the area of memory used to buffer these things, so the next boot just carries on appending to the same buffer, which is duly displayed on the screen.

    Wesha: an image staying in the monitor’s RAM would indeed be surprising, with or without power remaining! (Perhaps possible on some modern LCD monitors, if they tried using their own frame buffer for scaling purposes; quite a departure from current systems, though.)

  12. Benoit says:

    I recently bought a GeForce 7600 GS with passive cooling. The fact that it was "Silent" was proeminently advertised the front of the box. Maybe you just got unlucky or another store would have do?

  13. poochner says:

    James, I used to see that sort of behavior with ancient character-mode display terminals.  Some of them didn’t clear the display core (actual donuts) on power up.

  14. /.er says:

    Wesha:

    I never unplug a monitor from a laptop without turning the laptop off first. Not even on those newer laptops that make a bing-bong sound when they autodetect a new display device and automatically adapt the desktop size. It’s too dangerous isn’t it? I don’t even use hibernate – what about all that state? No only a full shutdown will do…

  15. Worf says:

    Not video related, but hardware related.

    I’m debugging a crash in an RTOS I was using, and it always happened after some time – far too long to write it out via the serial port (unless you want to go through a gig of log messages).

    What I used instead was its log to memory buffer feature. When it crashed, I merely power cycled it, dropped into the bootloader, and read my buffer. Finding the wraparound was fun, but just ended up looking for a chopped debug line.

    This was fairly modern embedded CPU – PXA255 with PC100 SDRAM. It survived power loss, and lack of refresh plus system restart. And the data was more of less uncorrupted.

    (The problem was I exhausted memory – it had memory pools – I was allocating 3kB buffers, but the pools were 2kB and then… 64kB. Eventually, the system would run out of memory because I’d have enough packets that would exhaust the limited pool if they came faster than I could transmit them. Adjusting things so I had a 4kB pool made it ultra stable. And capped memory usage to around 2MB.)

  16. ender says:

    I’ve got a VIA Eden 5000 board (mini-itx, 533MHz C3 CPU), and when I was using Linux with X on it, I noticed that the last picture that was displayed before X was shut down would display (with slight corruption) for a few seconds when X was starting back up. The interesting fact was, that this happened even if the computer was powered off and disconnected from the mains for a few minutes.

  17. Dan says:

    My computer was making more noise than usual, so me and my dad opened it up to see what was causing it.  After cleaning out the power supply fan, the heatsink, and replacing a totally busted case fan, we realized it was none of those thing and I suddenly realized "Hey, doesn’t my GPU have a fan?"

    So yeah that was it.  After some experimentation my dad stuck some WD-40 down where the ball bearings for the fan were and that sucker became whisper quiet.  Maybe you can try it… just make sure it’s not going to leak all over the card (my dad disconnected the fan from the card while he worked on it and he even had a 30 year old power converter thing he hooked just the fan up to… the fan plugged into the card with a standard fan connector thing.)

  18. JS says:

    Heh.  You’ve learned why the first feature I look for in a video card is "no fan".

    I’ve done the oil-the-bearings thing; I’ve done the replacement Arctic cooling fan; none of these things worked for long.  Quiet fans invariably become loud fans in a matter of months.

    Incidentally, "no fan" is also a key criterion in picking a motherboard too.

    Then I prefer a case with a 12-cm fan, since the bigger fans are generally quieter (and they’re much cheaper to replace than northbridge or GPU fans).

    If I’m recommending a computer for someone who isn’t interested in building one… then Dell typically does an excellent job building a quiet machine.

  19. peterchen says:

    > You’ve learned why the first feature I look for in a video card is "no fan".

    I am happy that I am not alone with this – it makes me feel less obsessive :D

  20. DriverDude says:

    "Dell typically does an excellent job building a quiet machine."

    Agreed, except one Dell has a video card that has a small fan on a small heatsink. I replaced the heatsink with a much larger one without a fan. Totally silent and runs cooler than the stock heatsink – the large heatsink is in the airflow path of the case fan.

    The only downside of the larger heatsink is it blocks the adjacent PCI slot.

  21. Worf says:

    A quiet fan that’s turned loud recently is normally a sign of impending fan failure…

  22. bigfoot says:

    "I don’t care if it is smoking…it must be a software problem!!!"

    ….a conversation overheard in the pointed-haired boss’s office…. followed by….

    "Raymond, how soon can you have this bug fixed?"

  23. Igor Levicki says:

    Sorry kids, video card without large heatsink, without two power connectors and without huge turbine-like fan is good for 2D only.

    If it doesn’t take two slots it is not a video card — it is a whiteboard.

Comments are closed.

Skip to main content