Topic Title: 7970 DX11 Hard Lock
Topic Summary: Is this problem related the GPU or something else?
Created On: 11/18/2012 06:20 PM
Status: Post and Reply
Linear : Threading : Single : Branch
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
 11/18/2012 06:20 PM
User is offline View Users Profile Print this message

Author Icon
Geowil
Peon

Posts: 12
Joined: 04/20/2012

So I have been grappling with my 7970 from Sapphire over the past sever or so months I have had the card.  During certain DX11 enabled games my system hard locks and I have to use the reset button to restart my computer (note that pressing my restart button actually restarts my computer so the computre itself is not hard locked but the video and sound outpput and keyboard input is).

I have tried about a billion things to fix the problem and have come to the conclusion that there is just something wrong with the way the card interacts with my system.

For things that it is not:

Not a incomplete driver uninstallation problem

Not a heat problem

Not a PSU problem

Not an overclocking problem

 

I can rule these out because everything runs fine in most games even DX11 ones.  Only specific games are causing an issue but even then it is intermittent.

First my system specs though:

CPU: Intel I7-3930K @ 4.6Ghz

RAM: G. Skill Quad-channel 1600Mhz 16GB

MB: Asus Sabertooth X79

System hard drive: OCZ Agility 3 120GB SSD (only contains Windows 7 Install)

OS Windows 7 Ultimate 64-bit

Gfx: Sapphire Radeon HD 7970 OC Dual-X

PSU: Ultra X4 1200W Modular

 

Here is what I have tried:

Updatring to the latest MB bios (did that last night)

Uninstalling drivers, using Driver Sweeper to get all of the files off, and then reinstalling

Full reformat of my system

Changing to BIOs profile 2 on the card

Down clocking from the factory overclock

Increasing millivoltage for the GPU from 1112 to 1128 and 1131

A different 7970 (borrowed from a firend)

A different PCI-E slot

Changing the PCI-E slots to Gen 2 instead of Gen 3

Updating my Realtek drivers

Updating chipset drivers

Using different 8/6 pin cables and oplaced in different slots on the PSU

Trying my card in a different system

Ran Memtest, my RAM is fine

PC can run on Prime95 for 10 hours with 6 threads with no problems

Running some hugely resource intensive PS2 games via PCSX2 did not cause my system to hard lock

 

None of these either fixed the problem or indicated that the card was entirely at fault.  I tried two other 7970 cards besides the one mentioned above and all of them caused this same thing to happen.

So far my computer only hard locks in four DX11 enabled games and only when playing these games:

Total War: Shogun 2- hard locks between 5 and 20 minutes of playing every time.

Black Ops 2: If it is going to hard lock it does it in about 30 minutes.  However I have been able to play the game for 5 hours straight at times and not have a hard lock.

Secret World: Hard lock within 15 minutes

WoW: Hard lock within 25 minutes

I am going to try the 12.11 beta drivers today but so far from 12.3 all the way to 12.10, none of them have fixed this problem.

These are the only DX11 games that cause any problems.  Crysis 2 with DX11 turned on, the tesselation pack and HD texture pack installed runs smoothly and without any problems.  Metro 2033 runs just fine and does not cause any hard locks.  Neither does Saints Row the Third.

The card temp does not go above 60C usually and average during gameplay on games that do and games that do not hard lock is about 56C.

Right now my card is using 900 gpu clock and a 1375 memory clock.  The factory settings for the card are 1050 gpu and 1500 memory.

So I now ask again, is this the GPU saying it is a lemon or is this something else?

 



-------------------------

"I thought what I'd do was, I would pretend I was one of those Deaf-Mutes...Or should I?" -- Laughing Man
_____________________________________

Laughing Man Productions² Gaming and Entertainment Network



Edited: 11/18/2012 at 10:07 PM by Geowil
 11/18/2012 10:29 PM
User is offline View Users Profile Print this message

Author Icon
Thanny
Alpha Geek

Posts: 1458
Joined: 07/13/2009

Originally posted by: Geowil I tried two other 7970 cards besides the one mentioned above and all of them caused this same thing to happen.

 

So the short answer is, it's not your 7970 card.

That leaves CPU, PSU, and memory as plausible causes for a lockup.

For the CPU, the two standard tests are Prime95 and Linpack (two common front ends of which are IntelBurnTest and OCCT).  If you have hyperthreading enabled, you should be using 12 threads for Prime95, not 6.  I find that Linpack is better for finding unstable overclocks.

For memory, the two common software-based tests are Memtest86 and HCI MemTest.  The latter is usually credited with being better able to find memory errors.  Keep in mind, though, that both Prime95 and Linpack can fail due to faulty memory as well.

As for the PSU, a quick search for reviews suggests it's a mixed bag at best.  It's supposed to be 1200W, but only supplies 85A on the 12V rail.  While that's probably plenty, it does call into question the overall quality of the unit (a typical high quality 1200W PSU will supply about 100A on the 12V rail).

Your first step should probably be putting everything to stock speeds and seeing if the problem continues to happen.

 

 11/19/2012 01:21 AM
User is offline View Users Profile Print this message

Author Icon
Geowil
Peon

Posts: 12
Joined: 04/20/2012

Originally posted by: Thanny
Originally posted by: Geowil I tried two other 7970 cards besides the one mentioned above and all of them caused this same thing to happen.

Thanks for the reply.

 

For the CPU, the two standard tests are Prime95 and Linpack (two common front ends of which are IntelBurnTest and OCCT).  If you have hyperthreading enabled, you should be using 12 threads for Prime95, not 6.  I find that Linpack is better for finding unstable overclocks.

I have hyper threading turned off, turned it off when I first setup the bios after building it.  As I mentioned running Prime95 full boar on 6 threads resulted in stable operation for the 10 hours I ran it for.  I will try OCCT tonight as well though for sake of double covering.

 

 

For memory, the two common software-based tests are Memtest86 and HCI MemTest.  The latter is usually credited with being better able to find memory errors.  Keep in mind, though, that both Prime95 and Linpack can fail due to faulty memory as well.

I will try running HCI either later tonight before OCCT or tomorrow and see what turns up.  I sohuld note that I have had 99% of my ram being used before by a memory leak in After Effects and no crashing took place but that might be comparing apples to oranges when it comes to what I have been having going on.

 

As for the PSU, a quick search for reviews suggests it's a mixed bag at best.  It's supposed to be 1200W, but only supplies 85A on the 12V rail.  While that's probably plenty, it does call into question the overall quality of the unit (a typical high quality 1200W PSU will supply about 100A on the 12V rail).

 

Your first step should probably be putting everything to stock speeds and seeing if the problem continues to happen.

 

This has been happening since day 1.  It took me two weeks to do an overclock because I wanted to make sure my system was stable.  This crashing was happening during that time as well and I have already tried returning everything to their normal clocks tonight after I thought about it.

Still ended up crashing in the same way.  I want to note that I tried a GTX240 and a Visiontek Radeon HD 6870 (borrowed both from an enthusiest friend of mine that supplied most of the other cards I have tested with) and the result was no crashes.

I gave him my 7970 to test on his rig (near same configuration as mine except with the 3960X and 32GB of ram and a RIVE [Rampage IV Extreme, an Asus board], and Coolmax ZPS-1600B PSU) and he started having the same crashes as I have been getting.  His normal setup is with quad GTX250's and he has never had the crashes he got when using my card before (this system has never had an AMD card in it).  He also tried the 7970's that he had lent me and the same crashing happened with them as well.

He was stumped and tried many of the things I have and more including heavily stressing all of our 7970s in OCCT for four hours each.  None of them crashed during that time.  Only when he got into the games I listed above did he get those hard locks.

The 7970's were part of an earlier rig he had been using for bench marking with a first generation I7 processor.  It was not effected by these types of hard locks.  His current system is a LGA2011 same as mine.

I will post back my findings from running the two alternitive programs you suggested sometime tomorrow.  Thanks again.



-------------------------

"I thought what I'd do was, I would pretend I was one of those Deaf-Mutes...Or should I?" -- Laughing Man
_____________________________________

Laughing Man Productions² Gaming and Entertainment Network



Edited: 11/19/2012 at 02:18 AM by Geowil
 11/19/2012 02:17 AM
User is offline View Users Profile Print this message

Author Icon
Geowil
Peon

Posts: 12
Joined: 04/20/2012

All right so I got OCCT, changed a few BIOs settings back to default (not the over clock but some voltage settings I had upped one or two pegs) to be on the safe side and ram the PSU test for 30 minutes at 1280x720 @ full screen.

Apart from seeing some high temps (cpu cores topped out at 76C gpu got up to 61C) the test finished with out any problems.  I also ran linpack (@ 90% of available ram) for 15 minutes with no issues either.

There was one problem, the OCCT core 0 temp kept going way out of bounds (127C) but that reading was proven irrelevent by CoreTemp which read the temp each of the times OCCT read 127 as being consistent with the other cores.

Disabled the warning for that core and stopped getting an over heating warning and was able to finish the tests.  This is with my 4.6Ghz overclock on the CPU.



-------------------------

"I thought what I'd do was, I would pretend I was one of those Deaf-Mutes...Or should I?" -- Laughing Man
_____________________________________

Laughing Man Productions² Gaming and Entertainment Network

 11/20/2012 12:27 AM
User is offline View Users Profile Print this message

Author Icon
Geowil
Peon

Posts: 12
Joined: 04/20/2012

I just ran HCI for the 20 minutes it stated to on their website.  Had to run 8 instances in orer to test all my ram but no issues detected.



-------------------------

"I thought what I'd do was, I would pretend I was one of those Deaf-Mutes...Or should I?" -- Laughing Man
_____________________________________

Laughing Man Productions² Gaming and Entertainment Network

 11/22/2012 02:23 PM
User is offline View Users Profile Print this message

Author Icon
Geowil
Peon

Posts: 12
Joined: 04/20/2012

I have been talking with AMD support about this issue and they believe it is being caused by driver issues when all other possbilities are ruled out.  Ther are "expecting" (he said hoping but that sounds way too apathetic to be the official position) that either the 12.11 drives or a future update fixes the hard locking issues being experienced.  Here is the email I recieved:

Your service request : SR #{ticketno:[8200518990]} has been reviewed and updated.

Response and Service Request History:

Thank you kindly for your extensive efforts in providing details to this issue. I can tell you that this issue is actively being investigated by AMD at this time.  I do not (and cannot) provide any details on ETA to fixes or any specifics on our own investigation of the issue. 

You may want to pay attention to Catalyst Creator's twitter feed, as he will often post updates on big issues.  I am hopeful that in the coming driver releases this issue will be resolved.

http://twitter.com/CatalystCreator

In order to update this service request, please respond, leaving the service request reference intact.  I believe that the information provided above will help to resolve your issue.  If we do not hear otherwise, we will assume that the issue has been resolved, and the service request will be permanently closed after 10 days.  If more time is needed to evaluate my suggestion(s) please let me know, and I will ensure this service request remains open for you.

Best regards,

Marc
AMD Global Customer Care



-------------------------

"I thought what I'd do was, I would pretend I was one of those Deaf-Mutes...Or should I?" -- Laughing Man
_____________________________________

Laughing Man Productions² Gaming and Entertainment Network

Statistics
86392 users are registered to the AMD Support and Game forum.
There are currently 2 users logged in.

FuseTalk Hosting Executive Plan v3.2 - © 1999-2014 FuseTalk Inc. All rights reserved.