Nvidia GTX980, computer crashes with no BSOD (both DX11 and 9)

[Shosue]

I've been experiencing computer crashes for the past 2/3 weeks now. Everything was working perfectly fine while running the game in DX11 since Early Access, then suddenly, without any changes, I started having issues.

First it was a black screen, then being switched to the desktop with FFXIV being on focus but not appearing. Then I had the game completely crash once or twice. Now it's simply my screen going black, sounds still playing for about 10 seconds then my computer restarts.

I tried reinstalling my drivers (uninstalled with DDU, then proceeded to a clean install), that worked for about 3 days and then it started again.
Now I have to run the game in DX9 as DX11 crashes about 10 seconds after I log in. DX9 is somewhat more stable but still a bit iffy, sometimes it'll let me play for hours, sometimes it'll crash after 5 minutes.

Here's the event log error I'm seeing everytime my computer restarts:

[Judge_Xero]

What are your system specs? Hardware, Operating System.

Are you running your system overclocked? Sounds like an issue with unstable Ram.

[Shosue]

Right, apologies I forgot to post the specs.

ASRock Z77 Extreme3
Intel Core i5-2500K, 3.3GHz OC'd to 4.2GHz using preset OC settings
Gainsward GeForce GTX980 Phantom
12 GB Corsair Vengeance Pro DDR3 1600 MHz
PSU is a 500W 80Plus Bronze certified
Running on Win7 Ultimate x64 SP1

I'd also like to mention these crashes only occur with FFXIV, I haven't had any issues with any other game.

[Duuude007]

I've had this issue before, but without more information I couldn't confirm whether the symptoms match exactly.

In my situation the system's power would turn off due to sensors detecting critical temperatures in my computer.

Logic dictates that if heat is your issue, it could be due to poor air intake, poor exhaust, or issues with internal circulation/airflow, particularly when your CPU & GPU are working their hardest, such as during FFXIV.

If you determine heat is a factor, I would:

1. test all fans are running, not jamming with dust etc.
2. use canned air to blow all dust out of your computer (unplugged and outdoors, of course)
3. Minimize the tangle of cables inside and near the PC
4. Provide at least a foot of unhindered space in front of all intake and exhaust fans (desk cubbies are the absolute worst places for computers)
5. install some programs that offer early warnings for heat issues.
6. If necessary, replace exhaust fans with larger versions and/or ones that offer higher airflow. Ideally intake fans will be bottom-front, exhaust top-rear and possibly side(s).
7. More extreme, but you could even consider water cooling for cpu or gpu. But the cost will be significant.

I hope this is helpful.

[Shosue]

Thanks for your insight, I'm almost certain heat is not the issue, though.
Temperatures are usually pretty low, even in places where my GPU heats up the most it rarely ever goes above 70°C. Even during a very long stress test with FurMark, my computer didn't restart because of the heat.

[Raist]

At the very least undo your overclocks. You may already be in the realm of a 400 watt draw on a reasonable load, depending on what all else is pulling power. Start thrashing the drive, hammering the RAM, ramping up the GPU to full voltage and hammering it's memory, and thrashing the CPU to push past it's stock TDP rating may well push it to 435+ at times the way this game hits everything so hard at once.

Even though it is an 80+ model, it can still put a good bit of ambient heat in the case when you are riding it too close to peak output...and that can cause issues with some components you may not be monitoring like RAM, mboard power regulation, or assorted other components.

Your 12v rail(s) might also be getting tapped out as well. Should double check there is enough capacity for what you are running on that spec.---if you come up short on that rail, you are in for trouble...if not now, down the road as you wear down key components faster.

[Imoye]

Got exactly the same problem. The game will inevitably crash. I might quit because of it. I'm 99% sure it's a memory leak. I can stay in Idyllshire for 5 minutes tops before crash. Started happening more frequently after W10 upgrade, but this is an issue that has been present since launch. SE is yet to do anything about it.

[Train88]

Looking at the error log, something is requesting a windows power event. I would scan for backdoors etc using malware byte and run a "sfc /scannow" in an admin cmd window.

[Raist]

Bugcheck 278 - Video TDR Error (the vid card was trying to reset). Barring a driver problem....look to hardware. A very common cause of that can be a power problem.

The reference specs for that line of cards lists as a 165w card...bare minimum recommendation is a 500w PSU. That does not account for overclocking and such, which the OP admitted to doing.

Note also that the reference design has two 75 watt connectors, but the Phantom has one 75w and one 150w. Couple that with the draw from the port itself and there is potential for much more draw than 165 (like, if the card gets overclocked and such--which we know many vendors tend to push things past reference values all the time). In fact...the Gainward sight even states right in the specs that it is boosted (1304 boost over 1204 base on the GPU)...guessing that is why it has a 6-pin and an 8-pin connector (note, that means they want to have access to 225w of auxiliary power, in addition to what can be drawn from the port---we are talking potentially up near 300w in total).

http://www.gainward.com/main/product...3524.pdf?s=414

In testing, the reference cards with only two 6-pin connectors (150w) were shown to hit 175w often under extreme loads. The CPU has also been shown to hit as high as 125 at stock settings under heavy loads, above 160w when heavily overclocked. Factor in any additional drain for the motherboard/RAM if the FSB and such have been tinkered with.... and that 500W PSU may not be fairing so hot. It may not have enough headroom on the 12v supply, or it may have just enough wear on it to not handle the heavy thrashing the game is putting on the system.

Really need to verify the specs on that PSU were sufficient for the requirements of the hardware it is now being asked to support, and if it is questionably close within the breakdown of the rails it should be tested for reliability. Or, at least if the OP has access to a more robust PSU he/she may want to try hooking it up and testing it. Shouldn't need to completely dismantle the system to test it....if the cables are long enough it should be possible to leave the computer's case open and just swap the connections and do a test run.

Edit:
Out of curiosity, I plugged the basic settings in the PSU calculators at coolermaster and corsair. Coolermaster calculated 420w minimum based on reference values (ie: not including the CPU overclock nor the overclocked GPU). Switching the GPU to the 980Ti reference (250w, more in line with the Phantom's power profile) and it switched it to 511W (keep in mind this does not provide a means to compare a CPU overclock). While Corsair doesn't include memory in the options, it does have options for an extreme overclock on the CPU, as well as an overclocked GPU. Regardless if you checked for overclock or not, Corsair spits out for a 600W PSU MINIMUM. Their entry level bronze 500W PSU lists 456w available on the 12v rail...but it is recommending the 600W model instead, which claims 552w on the 12v rail. It is starting to look a lot like a power issue may indeed be in play here.

[Shosue]

Overclocks are undone, issue still persists. I'll try to hook my computer up with a better PSU, and I'll let you know how that works.
Thanks for your input!