Dying GPU? Problems with Windows booting up? Problems with artifacts / distorted / garbage image display on your monitor? Welcome to the club of users who experienced problems with their graphics cards, myself included.
Graphics Card Failure – Classification & General Guide
Graphics Card Failure Symptoms
- Video TDR Failure BSOD when booting up from standby/sleep
- Image Artifacts during BIOS / Windows or Live Linux boot
- Failure to cold boot Windows completely with freezing screens, constant reboots/failures etc.
- In extreme cases: no image/signal output (completely black image)
Graphics Cards Failure Causes
- Software / Drivers bugs & compatibility issues
- Hardware Failure
- PCB board defects
- Solder joints issues
- Passive Component Failure (Capacitors / Inductors / Resistors / Connectors)
- Active Component Failure (GPU / RAM / BIOS / Power Supply Control / Transistors)
Graphics Cards Software / Drivers Issues Fix
- clean graphics card driver (re)install
- reverting or updating graphics card drivers
- reverting or updating your OS
- booting up Live OS from USB/HDD/CD/DVD
- fresh OS re-install (Windows, Linux, Mac OS)
Graphics Cards Hardware Problems Verification
- testing card with another PC & Monitor (just in case)
- booting up Live OS from USB/HDD/CD/DVD
- restoring backup OS image or fresh OS re-install
HOW I FIXED MALFUNCTIONING GRAPHICS CARD
with vertical stripes and image artifacts
Last week one of my older PCs, which is used mainly as a helper/verification machine and multimedia/internet on occasion, started showing strange display behavior out of the blue. To make party complete, add a recent Windows 10 update into the equation and then get totally confused whether it’s a software or hardware issue.
Problem started manifesting gradually. Initially, in the past year whenever I booted the machine from stand-by (sleep) mode, in 10% ~ 20% cases Windows BSOD occured with Video TDR Failure message and something about collecting and generating error report:
Video TDR Failure
blah blah blah + some error sequence code, which turned out completely unrelated/unhelpful in my case
I was frustrated, because solutions on the web and official Microsoft support forum suggested it was a driver/software issue.
What does VIDEO_TDR_FAILURE BSOD error stands for?
TDR stands for Timeout, Detection and Recovery.
It is a feature in Windows that detects response problems in the graphics card (e.g. when you boot up from stand-by/sleep mode) and resets the graphics card to recover functional desktop. By default, Windows waits for 2 seconds for a response from the graphics card and then resets the card if the response fails for whatever reason. TDR failure, thus, happens when Windows fails to reset the graphics card and restore functional desktop environment. Trouble is, this may happen either because of software or hardware related issues and this message alone cannot tell you which one is the case!
Later, there were very brief glitches which lasted for a fraction of a second, almost unnoticeable. With each day and cold boot problem progressively became worse. I started noticing weird pixelization distortion all over the place out of the blue, something I though was happening only on my old forgotten Commodore 64, when CPU and motherboard became very hot (because of poor thermal design and hours of gaming Manic Miner and Green Beret). While booting through BIOS and various test screens everything was fine (at first), but as soon as it started booting Windows 10 with recognizable logo, things started getting weird. Sometimes, like completely random, Windows booting sequence was constantly freezing, Windows was trying to do some internal magic behind the scenes (judging from the SSD/HDD activity LED indicator), but eventually, Windows 10 always ended reporting BSOD screen with familiar message about Video TDR Failure, at which point either reset or power-off buttons were left as the only option.
Then, after constant agony of countless power-offs and reset cycles, all on its own, Windows 10 did another reboot after another reboot after another reboot, until it eventually fully booted and succeeded! Huh, who would thought Windows 10 got so persistent and successful OS eventually!? Anyway, as I later discovered, this was more of a gambling luck, than anything, since the problem started happening more frequently, and became worse every day.
In one of many reboot-fail-reboot cycles, something suddenly changed and caught my attention. Windows 10 out-of-the-blue reported that NVIDIA 8600 GT Graphics Card hardware was disabled, because it reported a problem during boot sequence. I will spare you with all the details, since this event did not happen right away and all on it’s own (it would spare me a lot of trouble and unnecessary reboots and trials if it did), until Windows Repair Console was initiated (probably because of many reboots and unsuccessful startup sequences) and then I instructed Windows to try to detect and fix the problem on its own. It couldn’t fix the problem, naturally (because it was not software related, after all), however, it did detect that boot failures were related to the Graphics Card unit.
The biggest mystery that followed was the fact that (at least for next 2-3 days) computer started working “normally” with one (and only) GPU card disabled, and monitor (also one and only) still hooked up to it!? I thought if we disable one and only GPU card in the computer (like, in Windows Device Manager), that’s it – there won’t be any picture anymore.
Thing is, it was only disabled in Windows. Not BIOS, DOS, Linux or whatever other OS there might be laying there. Graphics card was now using Microsoft Basic Display Adapter driver, with all advanced features (bells & whistles) disabled. Sleep function was also gone (only power-off/shut down and restart were available in Start menu).
So, image was still showing up properly during basic system boot and tests, and yes, even Windows 10. It just wasn’t using all hardware drivers and features, and that came with a reduced resolution of 1024×800 or so (I raised it up to max 1280×1024 later, but that was still far from the native 1920×1080 Full HD for my monitor). It sucked, though, as everything was distorted, icons and windows were stretched, movie playback of few clips required pixel AR correction option checked in the media player, and so on…
Well, at least, I had my PC back for a while.
I also installed a local VNC server, so even without any picture displayed, graphics card inserted or enabled, I could remotely connect to it via LAN and see that otherwise PC is working fine. HDDs, CPU temps and MEMORY benchmarks were all passing with no errors, Windows was now happy booting up, even with garbled image, which gave me confidence that the computer was otherwise working properly. It has to be related to Graphics Card, but how and why?
Searching various online sources for tips about troubleshooting and quick repair, I already:
- Uninstalled Graphics Card from PCI-E slot, cleaned the honeycomb pins and put it back.
- Windows Repair session did recognize the problem at some occasions, but without any improvements.
- Booting live Linux distro (e.g. Ubuntu) experienced same image artifacts at some point, clearly indicating this was a hardware related issue, and not an Operating System one. This test alone was a clear sign to me that this was not a software/driver issue, but rather a hardware problem.
- Checking memory DIMM slots, CPU, and other cables and components haven’t revealed anything problematic, either.
There were hardware repair related suggestions, as well:
- PCB Solder Reflow – GPU area or entire card overall
Reflow can be used to “refresh” solder joints. My quick visual inspection of the card didn’t reveal any bad contact issues, but still, it could be non-visible problem. You can either use hot-air gun and gently hover over some areas, or put the card in the oven and “bake” it carefully. Both methods carry the risk of permanent damage and component overheating (but in general, hot air gun is better, since you do not heat entire board and all components at the same time, including non-electric contact bodies).
This PC was never used for gaming, in fact, GPU was never even benchmarked or stressed otherwise and even if I wanted to – it was only 8600 GT anyway, why would I want to do anything like that with 8600 GT? I mean, it wasn’t something like 8800 GTX, GTX 1080, GTX 2080 or similar. However, I recalled that at one point I disconnected GPU Cooler from graphics card’s PCB few years ago, as I was annoyed by the noise it was suddenly started making, and wasn’t too confident to repair it or replace it myself. It was fine, GPU was still below 70° C, and I didn’t actually care, because I never planned to stress it much with games or whatever, anyway.
Then came this gotcha.
Last and this year we had some heatwaves over here, and those heat waves stress not only us, fellow humans, but also our computers and their components. It is not the same when T ambient is 25° C or below, or 30° C and above! Remember that < 70° C from previous paragraph? Well, it was now close to 78-79° C! Poor thing simply couldn’t keep up!
Even without fan disabled, statistically, this ought to happen with non-solid low-quality electrolytic capacitors eventually. I actually have experience with motherboard repairs in this regard, replacing capacitors that were showing age and plaque.
I was suspecting that either GPU (Graphics Processing Unit) chip itself, or GDDR RAM memory was damaged somehow. I couldn’t verify GPU itself reliably, because as soon as I enabled Graphics Card under Device Manager, Windows 10 would freeze and enter endless rebooting sequence again and again – yeah, believe me I tried it, that’s how I know ;)
On the other hand, GPU is the most complex piece of technology on the graphics card itself, and it is usually very well protected (thermal shutdown and other safety mechanisms built-in). In another words, it should be rare occurrence that chip burns out, keeping all things checked (e.g. no overclocking, reasonable PC internal temperatures / cooling, no gaming or demanding GPU apps etc.).
There were some small utilities for testing and stressing Video RAM memory, and since at this point image was almost 100% of the time constantly distorted and useless, I connected through VNC LAN client and run it from remote desktop feature. To my shocking surprise, this tool run couple of cycles with 100% perfect health score! Hmm, so, it is not the RAM itself on the card, at least according to this software. It must be the GPU or output image compositing stage somewhere, then. What else could it be that was manifesting like this!?
It never occurred to me that power supply and voltage filtering section could be the problem! Voltage instability can cause memory corruption, image artifacts, freezing / BSOD and other symptoms.
Given the fact that all electrolytic capacitors age (yes, even the solid type ones, albeit at a several times slower rate), they are the usual suspects in any malfunctioning graphics card that displays above symptoms.
Since these cards came with multiple DVI-I output ports, I even got an idea to swap cable on both (plus mandatory reboot), but without any success. Both DVI outputs (digital DVI & analog VGA via adapter) were showing the same persistent problem.
And, just when I was thinking to throw it out, and put a nice upgrade as a replacement, I looked at the board and noticed three swollen ELCOs (electrolytic capacitors) right next to the “passive” heatsink fins! A moment of epiphany. Everything got in place, I had a clear picture about what happened and smiled. Hydrogen gas build-up occurred, probbaly because of the surrounding excessive heat (despite the fact those capacitors were rated at 105° C, the electrolyte solution was old and probably dried over time). Also, the fact they were standing next to the inductor that heats pretty well, wasn’t helping either.
OK, it is time for repair!
There’s a limited number of repairs you can perform at home with hobby tools. I was very lucky, and repair was straightforward and it worked! For complex repairs, professional equipment, spare parts and experience are a requirement, and repair is not always cheap.
You will need a decent soldering iron around, one that can probably go up to 300°~400° C. My 30 watts soldering tool barely managed to melt the PCB joints and it was excruciatingly painful thing to do it. Repair took couple of very long hours, instead of something like a 5 minute fix because of this.
I also recommend using Low ESR (low equivalent series resistance) versions with appropriate voltage specification and 105° C operating temperature (for longevity). It depends on the value of your graphics card, really. All modern cards with decent quality use solid capacitors nowadays. In my case, I used what I already had laying around: 3x 1000 uF / 105° C brand new capacitors, which were perfect for this aged low-end model. In case of some more expensive generations, I would even go with solid capacitors from established brands (Panasonic, Nichicon etc).
After repair, machine never reported Video TDR Failure message ever again, neither when booting up from sleep/standby mode, nor when booting up Windows.
I hope this repair journey I took will help someone else in similar situation.