Graphics card shuts down during gameplay. Immediate action required.
Graphics card shuts down during gameplay. Immediate action required.
Hello, I purchased an EVGA GTX 670 2GB in December 2020 at a remarkably low cost (around $30). However, upon receipt, I encountered significant temperature problems. When the card was idle, temperatures reached up to 60°C, and during gameplay they climbed to nearly 100°C, causing thermal throttling. This led to poor performance, frequent stutters, and an overall disappointing experience. After cleaning the card and applying new thermal paste, things improved considerably—temperatures dropped to around 22°C at idle and 60-70°C while gaming. Eventually, I achieved smooth gameplay with stable temperatures.
However, once I hit at least 60+ FPS in most titles, issues resurfaced: random crashes, restarts, stuttering, and fluctuating temps. Despite these setbacks, the situation changed when I witnessed a GPU failure while playing CS:GO. The screen went black briefly, followed by strange color distortions on my monitors. At first, I dismissed it as an anomaly, especially since it was my first such incident. My Intel DQ67SW board also had performance issues, only supporting certain cards.
The problem worsened with a faulty motherboard, which prevented me from using my ASUS GT 440 or GTS 450. Eventually, I decided to upgrade to the GTX 670. After thorough cleaning and installing a new motherboard (ASUS P8B75-M/CSM), everything functioned normally. The GPU no longer crashed, and I enjoyed stable performance—especially with high settings.
Over time, I encountered another challenge: the Windows "Shut Down" command would force a restart of the PC, requiring manual PSU shutdown. Crashing became more frequent, prompting me to replace my power supply. Eventually, I upgraded to an ASUS P8B75-M/CSM motherboard and saved money on additional components. The system booted quickly, and performance improved significantly.
Some aspects remained unresolved: the PCI-E 6 Molex connection worked fine, but I tested various PSUs, RAM sticks, monitors, and even swapped out my Nvidia drivers. Notably, using the GPU in another PC helped confirm the issue was hardware-related. I also considered transferring the motherboard to a different machine for further testing.
Important observations: my system used a standard Molex connection, had only one 6-pin PCIe slot, and ran on a PSU with limited power capacity. The card itself was an EVGA model, but the cooler was stock Nvidia. RAM tests passed, though crashes persisted. One monitor would show only one color due to its single connection. Driving Nvidia drivers caused more frustration than clarity.
Some untested experiments included running the GPU on another PC and using a different motherboard, but results were mixed. I plan to continue testing with other components to identify the root cause.
I’m reviewing the hotspot temperature now. It seems one section of your card die is overheating while others stay cool. I’ll also check if the thermal pads were replaced correctly. This issue usually happens with VRAM on older cards during extended gaming, so consider switching to better thermal pads and a suitable cooler. Your detailed post is really useful—many visitors ask for help but often lack the necessary info.
I understand, I must rely on this. Right now I don’t have any alternative power supply with two PCI-E 6-pin connectors.
I thought it wasn't the issue but you're correct—it might be the problem, since the MOLEX rail isn't providing enough power for the GTX 670.
You can try the steps provided. If you need more details, let me know.
Additionally, reducing the card's speed causes delays before it fails under heavy configurations.
HWINFO64 displays all temperature sensors, including GPU temps, with options like "hotspot" or similar categories. Check for a section labeled "thermal pads" or comparable, and measure the gap between VRAM chips and the cooler. You can purchase thermal pads on Amazon at affordable prices, ensuring you select the right thickness.