Multiple BSOD 116

Maximum is just that, the maximum, as in maximum critical threshold. Having it just a few degrees away will not be healthy on your card.

Realize that MemtestCL - like any hardware test - is not an absolute 100% guaranteed discoverer of hardware faults. There can be a number that can exist that won't manifest in MemtestCL, especially since MemtestCL is designed specifically to test the VRAM, and does little to test the GPU itself.

OCCT just does pressure testing for the CPU, not the GPU. Because of its design it will generate a lot of heat on your CPU, so make sure your cooling for your CPU is optimal. In addition to OCCT, running Prime95 on Torture Test with Blend or Large FFTs for several hours (9+) is a solid option. It tends to be a lot more sensitive to CPU failure than OCCT.

Don't forget to run Memtest86+ for at least 7 passes on your RAM.

The card is overheating, and moreso, that fact that it has been overheating for a while could very well mean you've lost it. They are right about the information provided by NVidia, but they are misreading it. Again, it's a maximum temp threshold. Yes, GPUs can handle more heat than a CPU, but not by that much, and the closer you get to the threshold the more wear you place on the components. The threshold is just there for the card to say "TOO HOT" and shut itself down as a safety measure, but riding it close will be no different than if you surpassed it.

You may have bad airflow in the PC, or some close obstructions that's preventing proper flow from your GPU fan. The GPU fan itself may even be insufficient (often the stock coolers do end up being subpar!).

You can record voltages and temperatures using HWInfo and setting it to Sensors only at startup. We recommend you log two 30-minute instances, one during idle, and one during high load. Send them to us in their default format (CSV) and we should be able to take a look at em. Don't worry if your PC crashes midway, it'll be saved up to that point. If you don't select Sensors only at startup, it'll just send us a report of all your hardware, which will be pretty pointless.

Vir Gnarus said:

You may have bad airflow in the PC, or some close obstructions that's preventing proper flow from your GPU fan. The GPU fan itself may even be insufficient (often the stock coolers do end up being subpar!).

You can record voltages and temperatures using HWInfo and setting it to Sensors only at startup. We recommend you log two 30-minute instances, one during idle, and one during high load. Send them to us in their default format (CSV) and we should be able to take a look at em. Don't worry if your PC crashes midway, it'll be saved up to that point. If you don't select Sensors only at startup, it'll just send us a report of all your hardware, which will be pretty pointless.

Thanks for replying.
Im not sure about air flow. Antec900 has good airflow i think, but theres a lot of cables from the PSU i have to squeeze in somehow at the bottom of the case. The GPU is in the middle of the case, with 2 slots between it and my soundcard. There is nothing blocking the GPU fan and it looks to be spinning ok.

I will try HWinfo now like u suggest, but there doesn't seem to be a guide with it.

It's very simple, as long as you press Sensors only at startup it should only bring up a window displaying sensor readings. Then you can click the button to start logging and choose where to place and what to name your log file and then let it run.

Vir Gnarus said:

It's very simple, as long as you press Sensors only at startup it should only bring up a window displaying sensor readings. Then you can click the button to start logging and choose where to place and what to name your log file and then let it run.

Ok thanks, ill do it now, and post results.
For the high load one, i could do this while playing a game? such as starcraft2 maybe?

Yep, gaming is a great method, as well as benchmarks like Unigine and Furmark.

When i try and upload them, it says invalid file, i don't think it likes .CSV extension(wasn't listed).