D3D, Down clocking Nvidia Factory overclocked card's...

StrongSad · 22 May 2015 at 11:14

OK

Before anybody Jumps down my throat, this seems to affect a small number of users and there are of course a million things that could also contribute to driver instability.

However......

I had an issue with Elite Dangerous where I would jump to hyperspace and the game would CTD with a D3D error. This started happening after a patch

In my testing I did the usual stuff and then started down clocking my card in increments and tetsing. When I got to below 1200 MHz base clock, the issue went away... fully away... If I overclocked/returned it to my Card's clock speeds 1233 MHz... it crashed... any sort of Overclock from my factory speeds and it would just crash....

I passed all this info onto support along with my testing results.. they eventually replied saying it was reported by other users and my work around was being past on.

Come GTA V

I have started getting the dreaded ERR_GFX_D3D_INIT CTD after the last patch

Again loads of messing around, raised a ticket with R* then found a steam discussion.

It would seem that again this issue might be resolved for some (me included) by down clocking their cards from Factory set clocks speeds to near reference clocks....

I have mine running at a base clock of 1199 and it has not crashed after 4 hours of game play.... put it back to stock clocks for my card and it crashes, overclock it and it crashes even quicker.....

As I said, not all are having this issue, not all people are getting it resolved....

My point.

Is this an issue with Nvidia, Direct 3D or our cards being factory overclocked?????

As I said, It's gone away for me and I can now play without crashing....

Silent_Scone · 22 May 2015 at 11:17

I think it's AIB partners getting caught with their pants down at how sensitive Maxwell is to overclocking. If it's not crashing at factory base clocks, it's the overclock.

Is it worth getting up in arms about? Possibly, frankly I'd just flash to factory base for 24/7 and find an overclock that works for you.

P.B · 22 May 2015 at 11:18

I get this with gta of my titan X I thought it was because I was pushing the card to hard ?

andybird123 · 22 May 2015 at 11:19

the card is faulty, it should run at it's advertised clocks and if it doesn't its faulty and should be replaced

benjii · 22 May 2015 at 11:28

andybird123 said:
the card is faulty, it should run at it's advertised clocks and if it doesn't its faulty and should be replaced

This. The whole point of factory overclocks and the reason you pay more for them is because you expect the AIB to have done quality control and be able to guarantee those clocks.

StrongSad · 22 May 2015 at 11:29

P.B said:
I get this with gta of my titan X I thought it was because I was pushing the card to hard ?

There is Titan X user who has had it fixed by dropping the clocks.... so worth a try.

andybird123 said:
the card is faulty, it should run at it's advertised clocks and if it doesn't its faulty and should be replaced

it did run at stated clock speeds for 150 hours of game play on GTA V and 80 hours on Elite dangerous, Both were patched and both started having the issue.

I have run a 1500 clock on sky rim with out issue and all other games without issue... also I don't thing I would be able to RMA a card that is running fine... My bench marks are great on Heaven and Firestrike... really great

PieEater · 22 May 2015 at 11:59

If you don't want to go down the RMA route you could try running the fans at 100% ingame to see if the instability is heat related. If this helps you could create a more aggressive fan profile. If that has no effect try adding a touch more voltage to see if that makes the overclock more stable ingame.

StrongSad · 22 May 2015 at 12:11

PieEater said:
If you don't want to go down the RMA route you could try running the fans at 100% ingame to see if the instability is heat related. If this helps you could create a more aggressive fan profile. If that has no effect try adding a touch more voltage to see if that makes the overclock more stable ingame.

My card runs at 60 C in GTA V, less when down clocked.

The voltage thing might be way forward, but my card does does not take voltage increases via AB as I think its locked in the bios.

There was some user analysis on the Elite forum that was along the lines of a sudden drop in voltage when the hyperspace cut scene/effects happened that could cause a problem. However that is way overmy head.

I will test in a min with GPU Z running and see what the results are.... I can make it crash pretty much at will online....

Also this was in the 350 patch note's

Improved stability and fixed several crashes and hangs while transitioning between Jobs in GTA Online.

That is where it mostly happens....

Nasher · 22 May 2015 at 12:56

I had this problem on a factory OCed 670, which would only be 100% stable when downclocking it. I switched it with a non-OCed card and no more crashes :/

It's a nightmare to RMA because it won't do it in all games or benchmarks. It only seems to happen when the card is in a certain state. I don't buy OCed cards any more because of this.

I suspect many people getting crashes and BSODs in games are having this problem and blaming the game. Not knowing it's actually a bad factory OC on their GPU.

PieEater · 22 May 2015 at 13:06

Strudders said:
My card runs at 60 C in GTA V, less when down clocked

TBH it was more the VRMs that I was thinking about than the GPU. Running the fans at 100% should provide sufficient additional air movement around the card to provide additional cooling to the rest of its components. I would give it a go just to fully rule temperatures out of the equation.

Aretak · 22 May 2015 at 13:09

I had to return my Asus R9 280X DirectCU II TOP because it wasn't stable at its factory overclock and would artifact in some games (WoW especially didn't like it at all for some reason). A lot of people seemed to have the same problem with that particular card, possibly due to the unusually high factory memory overclock that Asus decided to go with (400MHz over reference).

StrongSad · 22 May 2015 at 13:29

PieEater said:
TBH it was more the VRMs that I was thinking about than the GPU. Running the fans at 100% should provide sufficient additional air movement around the card to provide additional cooling to the rest of its components. I would give it a go just to fully rule temperatures out of the equation.

I will give it ago, but my Inno3d has a vram heat sync and a separate fan for PCB circulation LOL

It is not that pretty... more Dr Mephesto....

EDIT:

OK that did not help unfortunately... Kept the GPU at 52 c though. LOL

and if I ever want any sound effects of a RB211 spooling up... I am sorted...

Edit Edit:

From the latest Nvidia drivers release notes.... 352.86

known issues:

Grand Theft Auto V randomly crashes giving error: ERR_GFX_D3D_INIT

Whilst I am still on 350.12 I guess it is a driver issue.... Fair few on Reddit that have fixed via down clocking....

StrongSad · 27 May 2015 at 08:56

This from the ED team relating to the issue of down locking to resolve the problem with ED

Currently have a load of stuff to try from R*

Please note, the following is by no means a statement of absolute fact, it is our current understanding of the situation.

The issue seems to be that these cards are optimised for constant high load. Elite however tends to cause phases of maximum load interspersed with lower usage. Especially events like exiting from hyperspace near a star for example cause the load to rise from very low to very high. This seems to cause the cards in question to fail to adjust power levels quickly enough, their processor in turn briefly suffers from low voltage and behaves undefined, causing a crash.

Down-clocking the card in question lowers its power draw and alleviates the problem. It would be conceivable that small variations in production quality make a percentage of the affected card models more resistant to the issue, that would explain why having a replacement solved the problem for some people while it did not for others.

I am afraid as of now lowering the clock rate is the only solution at this point, we will continue to look into fixing the problem of course. I am sorry that I cannot be of greater assistance in this matter.
i

DeanT77 · 31 May 2015 at 11:12

Feels like we are in the same boat!
I have a gigabyte 780ti Ghz edition which has a factory overclock.
Experiencing the same D3D error in GTA5. Downclocking the gpu can help but rather than crash 5 mins into the game it will still crash but can be anywhere between 20mins and an hour or so.

From doing some searching it seems that a few people have issue with this card too, extremely disappointed in it, just over a year old and feels like £500 that could have been better spent!

smogsy · 31 May 2015 at 11:30

im getting a few more driver crashes lately at stock Nvidia reference 980s. but nothing like you are

StrongSad · 31 May 2015 at 11:37

I have a load of stuff to test that R* have suggested.

As we examined your dxdiag which you have provided the drivers are not updated. If you already updated your drivers, please provide us an updated Dxdiag. (Please refer the link to update your drivers

Disable your firewall for some time and play the game.

Ensure you have administrator rights on your computer. Run the Social Club
Installation application as an administrator by right clicking the application and choosing "Run as Administrator."

Temporarily disable antivirus/security, then uninstall if disabling does not work as a test. Be sure to reinstall and re-enable your antivirus after the test.

However

Nothing has made a difference other than down clocking, which solves it completely.

DeanT77 · 31 May 2015 at 11:47

I'm going to wait for new drivers, in a last ditch attempt and then try a 280x I have.

What is a good stress test app to use for the gpu? I don't want to cook it, just push it a bit under controlled circumstances!

fenchu · 31 May 2015 at 11:51

Well i wondered if anyone else had noticed this , you see for me my overclocks were 100% stable upto driver version 347.88 then with the release of the gta V driver 350.12 & 352.86 my card became unstable with d3d errors. I tryed numerous fixes however the only solution i found was to drop my clock speed on memory and gpu roughly 100mhz to become stable again.

Essentialy i think the newer drivers are pushing cards harder and as a result cards that would have passed a factory overclock say 6 months ago simply will not now and are showing they can be quite fragile in referance to clock speeds. Im fairly sure if these oc cards were tested with current drivers they would not pass there factory oc and would be binned / sold accordingly.

PieEater · 31 May 2015 at 11:54

PieEater said:
......try adding a touch more voltage to see if that makes the overclock more stable ingame.

So if it's a power draw issue has anyone tried the above suggestion to see if it helps?

Jono8 · 31 May 2015 at 12:23

I just think modern games and graphics techniques are showing up weak overclocks - some of which are factory ones. I had my 980 boosting to 1500mhz with stock volts and this would loop Heaven all day long. Then GTA V came a long and bamn, Driver crashes. Dropped the Core by 50MHZ to 1450mhz and then no more crashes...in GTA V.

I then started playing Witcher 3 and have started to get the odd driver crash every now and then so i have bumped the core down another 25mhz to see if that stops it - only time will tell.

Clearly i am still to find my 24/7 OC on stock volts!