Possible faulty video card?

[L'acquer]

Hello again. I've been having some problems with my pc over the last couple of months and it's been happening more frequently the last few days. Bear with me, since this is probably a long post.

First of all, my system:

CPU: Intel Core i5-3570K (OC'ed to 4.5GHz) [Stress tested with Prime95 Blend test for over 13 hours (max temp during test 77°C)]
Cooler: Cooler Master Hyper 212+
RAM: 2x4GB Corsair Vengeance LP 1600MHz
Motherboard: MSI Z77A-GD65
Video Card: MSI GTX660 Twin Frozr III (WHQL 344.75)
HDD: [1] WD Blue 320GB (Primary)
HDD: [2] WD Black 2TB
HDD: [3] Samsung F3 1TB
Case: Corsair Obsidian 750D (using stock fans. 2 intake on the front, 1 exhaust on the back)
PSU: Corsair HX750i (fairly new, got it as soon as it arrived here)
OS: Windows 7 Pro 64-bit

Okay, so my problem is, about 2 weeks after I installed the HX750i, I've been having some problems such as:

- White screen (which also hangs the PC)
- Nvidia driver crash
- Freezes momentarily

The last few days, the problems seems to worsen:

- Lost display
- Sudden hard restart

Now, this problem only occurs after I replace the psu from a Cooler Master GX 550W Bronze to the HX750i. However, I don't think it's related to the PSU (totally, at least) since these things only occur during:

- Playing games for a couple of minutes to hours (not quite sure why it's like that)
- Running 3DMark
- Watching videos using MPC (madVR + SVP which puts quite a bit of load on both CPU and GPU)

Based on the symptoms and usage, it seems highly probable that it's the video card (GTX 660) that's been causing trouble, but I've been monitoring temps all the time, even during gaming using Afterburner and HWiNFO and GPU temps hardly reach 70°C (max 72°C). It's possible that the rest of the parts on the card get hotter as well but it didn't happen before, so why should it happen now?

After the problems first appear, which is about a month ago, I've started to fixed the card fan speed to 70-80% speed. Then, the problem seems to go away, which stands to show that it's probably the rest of the components on the card that's overheating (VRM, memory IC).

The last two days however, it happened again but this time causing lost display and hard reset. This, turns my attention to the PSU. This morning, after getting a lost display, the system failed to boot. I've then started to resit the video card and RAM as well as replug all the PSU cables.

Checking the cables that come with the HX750i, I've decided to try and install Corsair Link software. The installation was painless and I've been able to monitor the PSU fan speed, temps and power in/out. I've noticed that the PSU temp was a bit on the high side (41°C...I guess it's high, fan wasn't even spinning) even when the system is idling. I touched the PSU, and it did felt quite warm.

Next, I decided to install the PSU upside down (fan on top), and ran the fan at a constant 60% speed. Temps drops to 33-35°C. So far, I've been able to run 3DMark Sky Diver twice and was able to watch a 1hr video without a hiccup but I haven't run any games yet. I'm still running the video card fan at a constant 72% speed btw.

Have any of you experienced something like this? Do you think the problem is gone now? I know it's not easily black and white but which components do you think is causing the problem?

Thanks for any help.

[quest for silence]

...mobo, PSU on a cardboard, GPU on a different slot than now (even if it's sub-optimal), just 1 stick of ram, mouse, keyboard, 1 hard disk (even an old one if you don't want to format one of yours, but not an SSD), no LAN/WiFi/Internet, factory BIOS settings (even if sub-optimal), fresh OS install, no Win updates, latest Nvidia drivers (not the whole bloatware, just the drivers): then run some Furmark torture tests with both the PSUs (since you have two), or other suitable stressing tools (OCCT, and so on).
Personally I'd do the first torture test with the old PSU...

[L'acquer]

I don't have the GX550W anymore, sold it. Nvidia drivers are fully updated without GeForce Experince and 3D Vision drivers. I did try the video card on another PCie slot but the system won't boot. There was a phase where they PC wouldn't boot at all, that's when I decided to replugged/resit the psu, video card and RAM.

[Jon Gerow]

The Corsair "flat" cables can be very difficult to fully seat, so I would first go through the build and make sure every connector is firmly seated into its appropriate connector. A loose connector can cause enough resistance to drop voltages to where some components will not function properly.

Then, as quest stated, I would field strip the PC to the essentials and try it on a non-conductive surface with minimal hardware. When you change out a PSU, you disrupt A LOT of hardware. You're removing and reinstalling connectors from a lot of different peripherals, so anything is possible.

[L'acquer]

After replugging everything and the system seems fine now (I guess), I thought it was the cables as well. The PCIe 6-pin and the 24-pin were a bit too easy to pull out, not just on the components side, but on the PSU modular side as well. This time, I plugged it almost as hard as I can, make sure it isn't loose or anything. The cables were really hard to be fully seated as you said.

I'm not partial to strip the whole system just yet. I guess the only way to know if it works properly currently is to put some gaming/benchmark load on it and if it persists, I'll strip it down.

Btw, is it normal for the PSU fan to not spin even when the PSU temps reach 41°C? Also, how accurate is Corsair Link software?

[L'acquer]

I played Far Cry 3 for more than 1 hour. So far, it seems like everything's good. Do you guys think it's safe the assume the problem is gone?

[quest for silence]

I played Far Cry 3 for more than 1 hour. So far, it seems like everything's good. Do you guys think it's safe the assume the problem is gone?

Personally I'd use some more stress (already named how).
About the PSU fan, looking at TPU review's data table, it should be normal: you may think to block the PSU back exhaust for a while, to let the temp rise, and then see what happens.

[L'acquer]

I thought Furmark has been known to kill Nvidia GPUs for the last generation? I've never run Furmark on my GTX 660 yet, only on my old GTX 460.

[yojo2]

Furmark serves no point anyway, it doesn't even tell if the card is fully stable. Its only use is to check the card's highest power draw.

[quest for silence]

it doesn't even tell if the card is fully stable

Why not?