[POSSIBLE FIX] for Fatal DirectX Error 11000002 on Nvidia cards

[FromTheInternetz]

TLDR: Try setting your GPU core voltage to 100% with Afterburner or another OC tool. Nvidia cards might be unstable under recent drivers when factory overclocked without increasing the voltage appropriately. The same applies to manual overclocking, of course, but there it's kinda obvious – you wouldn't think that the voltage might be too low on a factory OC.

Make sure you are on 4.4 Beta if you are using Afterburner and are on a 10 series card, otherwise you won't be able to change the voltage settings. Also, DO NOT install the RivaTuner Statistics Server when installing Afterburner since that is known to cause issues with game (as do other overlays like the optional Discord overlay).

Hey guys,

I've had crashing issues since I've returned to the game two weeks ago. Four different kinds of crashes actually:

- Fatal DirectX Error 11000002
- "An unexpected error has occured…"
- a Windows crash saying the game has stopped working
- and a single blue screen

I've basically tried everything at this point. Won't go into details, but it involved a lot of reinstalling the game, reinstalling Windows, hardware monitoring etc.

I guess my tinkering had some effect, because I haven't had a Windows crash since then and also no further blue screens since 1 1/2 weeks ago. But the other two errors still remained.

So today, after another 11000002 I looked over my Afterburner statistics (as I always do after a crash, trying to find something that looks wrong) and noticed that the GPU has been hitting the voltage limit all the time. I'm not actually sure if this isn't just normal behaviour anyway on Pascal cards (who boost themselves until they run into a limit), but it gave me the idea to just increase the voltage and see what happens – something I actually hadn't tried before.

Since Nvidia cards have a certain upper limit of voltage that you can't go over, I just went straight to 100% in Afterburner which brought up the voltage to ~1090mV (from around 1060mV) under load. I then started the Stormblood benchmark, set it to maximum settings and ran it in a loop – for seven hours straight now, without a single crash (Update: I let the benchmark running over night, 17 hours without issues). Normally, it would crash after a maximum of 10-15 minutes, sometimes even in the first run.

Two side effects I noticed: One, my card boosts even higher than before (up to 2075MHz (2025MHz before), which is pretty high for a 1080 Ti). Second, it's actually cooler than with lower voltage. I can't really explain it myself, but there are posts on hardware forums describing why too low voltage may actually run the card hotter when it's overclocked.

My guess is that the factory overclock of the GPU is too high for the voltage that is set – ZOTAC even advertises their 1080 Ti Amp Extreme as the fastest out of the box. They either forgot to increase the voltage accordingly or it may be a bug in the newer Nvidia drivers that causes the card to try and go beyond the voltage limit or something.

I've had no problems in other, more demanding games (like Witcher 3) before this change, but that doesn't really mean anything since it might just be the FFXIV engine being very susceptive to errors (even if they are corrected by Windows or the graphics driver instantly, otherwise we'd probably see more blue screens and not just crashes to the desktop).

Since I didn't run the game itself yet, I currently don't know if this also fixes the other remaining crash ("An unexpected error has occured…") for me since I used to get the 11000002 far more often than that one when running the benchmark (it's the other way around in the game, more "An unexpected error has occured…" crashes than 11000002). But my guess is that they are related or even the same with the message of the crash just depending on which part of the game notices the issue first (for the programmers: basically a race condition). I will certainly update this post and let you know in the coming days.

So, if you want, give it a go – just make sure to monitor your GPU temperature after increasing the voltage. Might actually go up a bit for you instead of going down, it's apparently pretty rare that it goes down and indicates that there was quite a lot of undervolting before.

My specs, just as additonal information:

OS: Windows 10 CU (with the latest updates)
CPU: i7-4790k (running on 4.5GHz/1.2V)
GPU: ZOTAC GeForce GTX 1080 Ti AMP Extreme (running on factory overclock/boost)
RAM: 32GB DDR3-2400 (using XMP)
SSD: Samsung 950 PRO 512GB NVMe M.2 (the OS and FF14 are both on that)
Display: Dell S2716DG (27 inches, 1440p, 144Hz, G-SYNC)

I'm currently running the driver version 382.53 since it has been stable everywhere (aside from FFXIV before this fix) and the newest one isn't well received on Reddit (many people report issues).

I've tried the following other things before I increased the voltage:

- Disabling any overlays (like Discord, RivaTuner Statistics Server (comes with Afterburner, so you might want to uninstall that if you just installed Afterburner to lower the voltage – sorry, forgot to mention that))
- Setting the card to "prefer maximum performance" in the Nvidia Control Panel (you can also do this via a profile just for FFXIV so it doesn't apply all the time)
- Running the game in fullscreen
- Disabling my CPU overclock
- Reinstalling Windows
- Disabling game mode, streaming and all the other stuff that came with the Windows 10 CU
- Reinstalling the game (copied over an installation from a friend who has no crashes so I can ascertain that I don't have any corrupted files)
- Turning down the graphics settings to the lowest possible
- Uninstalling Dropbox since the current version has an issue under Windows 10 where it floods the event viewer (see https://www.tenforums.com/software-a...0-entries.html)
- Running the game under admin mode and Windows 7 compatibiliy mode
- Checking my RAM with MemTest86
- Switching out my 32GB RAM with a 16GB kit
- Several different Nvidia drivers (basically all that officially support my 1080 Ti) with several different configurations
- Disabling G-SYNC
- Connecting my monitor over HDMI instead of DisplayPort
- Checking all my system temperatures under load (all way down in the green, I have very good cooling and even applied liquid metal to both my CPU and GPU die, so I wasn't expecting issues there anyway)
- Switching to my old GPU (a 980 Ti – had the same issues, although it was also quite heavily factory overclocked, so my voltage increase might have helped there too)
- Updating my Mainboard BIOS to the newest version
- Installing all the current Visual C++ runtimes (you could do it manually, but I used https://www.sereby.org/site/All%20in%20One%20Runtimes) and running the DirectX autoupdater

[Metatime]

Everyone having nvidia card issues should really go to the nvidia forums and make awareness on this issue just like in the ati board. After enough players posted on the ati board, ati released drivers to fix the issue

[FromTheInternetz]

Everyone having nvidia card issues should really go to the nvidia forums and make awareness on this issue just like in the ati board. After enough players posted on the ati board, ati released drivers to fix the issue

I sent them a bug report instead (already did that a week ago), didn't think about posting on their boards – didn't get a reply yet though.

I might create a thread on their boards if I find the time. If I do, I'll link it here so everyone can contribute.

[Velhart]

Everyone having nvidia card issues should really go to the nvidia forums and make awareness on this issue just like in the ati board. After enough players posted on the ati board, ati released drivers to fix the issue

I talked to a representative and sent as much information as I could regarding the issue. Suppose to hear back within 24 hours about their investigation. Hopefully it leads somewhere.

[TalonWeaver]

I've been having extremely similar issues but do not, as far as I know, have a factory overclocked card. I know my game has been causing some weird Dx11 errors lately, basically since downloading the latest Nvidia drivers. Very similar crashes to what you've explained, though more recently Windows will also give me a notification that the Nvidia driver kernel has crashed and restored itself, too.

[Velhart]

This is the response I get back....

Hello ???,

Thank you for your patience and cooperation. I had this checked and issue is being going on for quite sometime and most commonly, this seems to be after Windows CU update: http://forum.square-enix.com/ffxiv/t...1000002/page11

You may have to contact the game developers and report the issue.


Best Regards,
NVIDIA Customer Care

[Declio]

Thanks, i'll try this. The updated version of Precision XOC has a voltage slider (finally). I know pascal won't allow much, but maybe that little extra push will stabilize some of the offsets from the factory/manual overclocks.

I'm on a 1080ti/4770k using driver version 378.92 and crash with the DX11 error occasionally and CONSTANTLY in Susano specifically. I have to run in DX9 mode if we farm that fight. I will report back here in a few days to see if the voltage slider helped at all.

Has SE officially acknowledged this at all? I know AMD pushed an actual update out to address this specifically but it's clear Nvidia is going to shift the blame to SE.

edit: crashed quickly again after upping voltage. I'm going to have to play on DX9 until SE acknowledges/resolves the issue.

[Xeon]

Sorry but drivers do not control the voltage or the clock speed of the card. Speed and voltage settings are manufacturor specific on their O/C models and these features are controlled and can be edited directly through the BIOS. Sure, you can control clock speed and voltages with MSI afterburner, but the base values are set in the BIOS code and are not affected by drivers outside of throttling.

Slapping your GPU to use 100% vcore at all times can cause hardware failures. I did this on an AMD card with no overclocking. Just a lift to 100% voltage and it blew up one of the VRM chips on the card. It can also essentially "break your fans" too as they're running harder than intended. The only cards currently available that are built to run at 100% vcore and fan operation are the newly released mining cards, using lower voltage, fluid bearing fans that can sustain 100% speed 24/7.

If you suffer damage by applying 100% voltage at all times in afterburner, and it malfunctions, and you do not have a mining specific card, you have essentially voided your warranty. The mining cards share the same core and memory as the gaming cards, and in part, were brought to market because miners are blowing up their standard version GPUs due to running them at 100% voltage and higher fan speeds, and the obvious market shortage aswell.

If you need to "up your voltage" in this way to run a game, especialy this game, which doesn't seem to utilize over 40% of GPU power on a modern GPU, you need to be in contact of the manufacturer of your card if you've got things like adequate case cooling.

A safer alternative would to just up your fan speed in afterburner, or to set up a tighter speed curve. This should help you if running at 100% voltage does. If you need more voltage to run at stock or factory overclocked speed, usually this is because of a heat issue, because greater temperature usually equates into more needed voltage to run at said speed.

You also don't have to have a 10 series card to adjust your voltage. This is also manufacturer specific, and the higher tier manufacturers like ASUS, eVGA, ect have put in hardware that allowed the voltages to be tweaked on many older cards.

[Xeon]

And I know this information is accurate, because the guy above tried this and crashed even quicker...or quickly, or something.

Most of these DX11 issues are with the game. Some of them can be with the persons system, as I see alot of semi-old systems with 1080tis in them? Makes no sense to have a $70 cpu and a $500 GPU.

If you suspect that it is your GPU, before trying any of the advice above, download and run furmark. If you can run this for a half hour to an hour without it crashing, then the problem does not lie with your GPU. You can also try the same with prime95 with the CPU. if your system can survive these for an hour each, then you can count yourself clear of a hardware issue.

[Velhart]

Alright good news everyone. I have been working with a representative at Nvidia who has been working with folks who have been communicating with me attempting to find the issue. Through the benchmark, the group at Nvidia were able to create the error and now are investigating exactly what is causing it and fix the issue. I will keep posted in here the progress that happens.

Response By Email (Josh) (07/18/2017 04:11 PM)
Thanks ???,

Looks like we got a crash on the benchmark app for this game, so I think we are making progress on this now.

Josh
NVCC