GPU turning off without overheating

[Darrineth]

A few weeks ago I experienced a problem when the GPU would stop delivering signal and ramp up fans to 100% speed after being treated with heavy load but the rest of the system would seemingly run correctly as I was still able to hear sounds on headphones. I figured the GPU was overheating so I set up an aggressive fan curve and that solved the issue for some time. Something like two weeks ago the problem reemerged and the GPU would occasionally turn off even though temperatures were below 70 degrees Celsius. On Sunday I thoroughly cleaned the cooler and replaced TIM. I removed the old thermal paste using lint-free wipes and isopropyl alcohol. It is also worth mentioning that I was working on a antistatic mat with a wrist wrap that was properly grounded.

 

The GPU is running on low temperatures but every time I start a game or a benchmark I get a "NO SIGNAL" on my monitor after only a few minutes. I attached a screenshot showing a MSI Afterburner report on system statistics after running the FurMark for approximately 3 minutes - I got a "NO SIGNAL" after that. As you can see the highest reported GPU temperature is 68 degrees Celsius with 85% power usage.

System specification:

• i5 2500K @ stock
• Gigabyte Windforce GTX770; driver version: 398.82
• Asus P8H67-M EVO (I know it's a heresy but my old one died.)
• Mushkin Blackline 2x8GB 1333MHz CL10
• 730W PSU
• Samsung 840Pro + 1TB no-name HDD

I would really appreciate Your help with solving the issue.

[Muhammad Reza]

Was the Gigabyte Windforce GTX770 running stock?

[Darrineth]

I didn't manually overclock the card at all.

[Skyyblaze]

It could be that your PSU is faulty and doesn't properly deliver power to the GPU anymore under load.

[Darrineth]

Is there any reliable way to test and confirm that on my own?

[Skyyblaze]

2 hours ago, Darrineth said:

Is there any reliable way to test and confirm that on my own?

Well the easiest way would be to borrow a PSU from a friend to check or buy one in a store and ask for their return-policy beforehand. Other than that not really, you could look if your PSU has two PCI-E rails and swap in the second one, maybe the rail is just damaged and not the whole PSU.

[Darrineth]

Since my GPU is powered by 8+6 PCI-E connector and the PSU has 2 sets of cables with PCI-E connectors I did as you had recommended and plugged in the previously unused cable in the 8 pin connector. Thanks to that the GPU survived 30-minutes FurMark stress test which I guess confirms the idea that the problem was caused by a PSU. Even though the current setup is working properly I still am going to buy a new high quality PSU, this time with a single rail Thank you for your help.