POWER:
To answer the "dual rails" question: if you need two additional graphics card power inputs and you are using molex to GPU power converters, then you need a PSU that has the molex power running on separate "rails" from separate sources internally in the PSU - definitely NOT from two plugs on the same wire! Always best if you PSU has the correct line-outs for your graphics card or at very least don't run them both off molex converters. Why aren't you using the PSU line that is dedicated for your GPU - that is a high amperage line designed for running a graphics card. Molex wires are designed to spin up a few HDD's, so if you're using one of those for graphics, then don't run anything else on it, especially not the other GPU line!
HEAT:
I know that the Northbridge is all CPU integrated now, but are you sure that nothing else is overheating on the motherboard / graphics card? Something happening after prolonged gaming normally points at lack of airflow or cooling on one of the components (or that one power rail on your PSU eventually giving up trying to give more current than it was designed for).
COMPONENTS:
Stability, can normally be traced back to these things:
HEAT! The biggest culprit - usually one component, not necessarily on you standard GPU / CPU thermal monitors either - if you get desperate - run a graphics loop that you know will make it fail and then "
HEAT GUN!!" it
CPU - normally only if overclocked or overheating (badly applied thermal paste / damaged cooling) - should show up on temp monitors though!
RAM - damaged / not seated / dusty - check with memtest86.
PSU - underspec'd or overloading one of the power lines would contribute. Short of getting a manufacturer-stle test bench that will load up each rail and properly load-test it, just use PCpartpicker to make sure the power load is within spec and the load is as evently distributed across the lines as possible.
Motherboard - a lot more rare with a decent make, but normally heat-related or environmental: if you've been put it through significant temperature/humidity cycles - this doesn't fit your "30min" description though - it would be more random than that.
NEXT?
Think it through: what changed just before it got unstable? Airflow / new drivers / environmental (central heating OR AIRCON!).
If thinking though the issue doesn't highlight anything and if PCpartpicker says your power draw is within spec for the PSU... this is where go.... the first two are easy after that, more hassle:
1) leave memtest86 looping.... run a few loops without errors (overnight), should be ZERO errors.
2) a long-running CPU burn-in benchmark without any crashes - this should rule out most CPU or motherboard-related issues and reduce it to GPU-specific stuff (power or the graphics card itself).
3) choose (this is where there's more effort/hassle/cost in involved):
a) favours/cost: phone a friend or buy: a high spec PSU and try that - ideally a local supplier that will take it back if it doesn't resolve the issue (?).
b) hassle: strip the box: unplug EVERY component that isn't required for gaming - removes potentially unstable components and reduces power load.
c) cost+hope: long-running GPU benchmark and a heat gun...
As an ex- bleeding edge overclocker, I loves a heat gun
But I'd still leave that for last - even if you've got one already you are fishing for something that might not be there... try 3a or 3b before the heat gun!
I've ruled out:
Swapping the graphics card - you seem to have tried two and if you know those grpahics cards are stable elsewhere?
Software/drivers? - Soul search about what changed before - new drivers could load the graphics card in a different way or have introduced a memory leak. Are you capturing taskmgr stats and checking for memory leaks, etc in your graphics drivers immediately prior to the crash? If you can't run dual screen with a dashboard of all your system stats on the 2nd screen, then run this in the background (as a scheduled task) to poll every 5-10 secs: "tasklist /v > c:\logs\tasklist_Date_%DATE:~6,4%_%DATE:~3,2%_%DATE:~0,2%_Time_%TIME:~0,2%_%TIME:~3,2%.txt" then review the last files before the crash and see if RAM spikes (or if any random apps kick in like a dropbox sync / AV scan, etc - unlikely, but worth a sniff!).
edit: you seem to have upgraded the graphics - was this an attempt to rectify the issue or did the instability start afterwards? Just a thought that any software change for the new graphics card could have either introduced a memory leaks OR overloaded and damaged the PSU.
Facebook
Twitter
Instagram