GPU - driver stopped responding

[Mordregon]

Hello everyone

I'm new to this forum, have visited in the past reading up on things, but it's first time posting.

I'm wondering can you help me out with a problem. Recently I got my hands on 2 Sapphire 6970's, long story but it's a result of a replacement + friends favor, cost me hardly anything in the end. I've been using Nvidia for years, had 3 generations of SLI setups, this is my first ATI set since the days of Radeon 9800.

I got them last week, and since then all I get is "driver stopped responding" issue... managed to dig out some information on forums, finding out that it's caused by ATI not picking up it's clock fast enough from idle state, and essentially windows deeming the cards non-responsive because they didn't pick up in designated time frame, and attempting to reset the driver.

It's covered here:
Timeout Detection and Recovery of GPUs through WDDM

I also managed to find a work-around of sorts, editing manually profiles.xml file and lifting to bottom line clocks from default 250mhz on clock and 150mhz on memory, to 375mhz on clock and 400mhz on memory. This helped the issue a bit, it's crashing far less now, but it's still not perfect. Every time you boot it crashes, because profiles didn't get loaded yet, and once in a while it will crash when using simple things on desktop, just opening Internet Explorer, often while opening Windows Live Mail, sometimes for no reason just after perion of inactivity opening Start menu... mostly it recovers, sometimes it also hangs the system.

I've done everything from trying different drivers from older revisions, to cleaning system off all the drivers with sweeper, to manually deleting ati registrys, any reminder of nvidia is fully gone, removed old versions and updated to new ones almost everything, directx library, flash, C++ redist, heck even onboard intel controller and whatnot... problem still persists. Never experienced something like this with nvidia.

On AMD forums some say to just reinstall the system but that's not an option for me, too many things would have to be reinstalled, reprogrammed and setup all over again, far too much work for something that's supposed to be only a peripheral card working!...

Any ideas?

 

[ronaldbrijo]

Hi,

Had the same problem with my system not long ago. It is the ATI driver. Ive found that using the latest ATI driver tend to do this. I am using a older driver 10.5 and don't have it happening anymore.

Hope it helps.

R

 

[Mordregon]

Hey mate

Sorry for late reply, yea I've tried literally everything, every possible thing got removed, registry cleaned, updated to latest version, drivers, controllers, bios, you name it, checked ram sticks for integrity, together and separately with windows memory diagnostic and memtest, swapped card around, disconnected them, removed sound card... nothing.

Some folks on AMD forum claiming that it MUST be my system, so reluctantly backed up all my work related stuff and personal things, formatted the drive, reinstalled the whole damn system... on very first boot after installing catalyst drivers - driver crash...

Shocking.

How I solved it is I found some forums hinting that it's related to WDDM and TDR in windows Vista and 7, system deems the cards unresponsive and without waiting for them to speed up resets the driver, guess ATI doesn't fit in systems response time windows.

Clocking the cards bottom clocks in idle state seems to help it, almost removed it completely, still happens but few times only and mostly on initial boot.

In system, tick to show hidden files and folders.

Got to:
C:\Users\"your name"\AppData\Local\ATI\ACE and there you will find file called 'profiles.xml'

Edit it in notepad, and you should find this section:

<Groups>
<Group name="Overdrive5">
<Feature name="TimeUnlocked" />
<Feature name="OverclockEnabled">
<Property name="OverclockEnabledProperty" value="True" />
</Feature>
<Feature name="AutoTuneSupport" />
<Feature name="CoreClockTarget_0">
<Property name="Want_0" value="30000" />
<Property name="Want_1" value="60000" />
<Property name="Want_2" value="92500" />
</Feature>
<Feature name="PowerControl_0">
<Property name="Want" value="0" />
</Feature>
<Feature name="MemoryClockTarget_0">
<Property name="Want_0" value="50000" />
<Property name="Want_1" value="137500" />
<Property name="Want_2" value="140000" />
</Feature>
<Feature name="CoreVoltageTarget_0">
<Property name="Want_0" value="900" />
<Property name="Want_1" value="1000" />
<Property name="Want_2" value="1175" />
</Feature>
<Feature name="MemoryVoltageTarget_0">
<Property name="Want_0" value="0" />
<Property name="Want_1" value="0" />
<Property name="Want_2" value="0" />
</Feature>
<Feature name="FanSpeedProtocol_0">
<Property name="FanSpeedProtocolProperty" value="Percent" />
</Feature>
<Feature name="FanSpeedAlgorithm_0">
<Property name="FanSpeedAlgorithm" value="Automatic" />
</Feature>
<Feature name="FanSpeedRPMTarget_0">
<Property name="Want" value="1146" />
</Feature>
<Feature name="FanSpeedPercentTarget_0">
<Property name="Want" value="20" />
</Feature>
<Feature name="CoreClockTarget_1">
<Property name="Want_0" value="30000" />
<Property name="Want_1" value="60000" />
<Property name="Want_2" value="92500" />
</Feature>
<Feature name="PowerControl_1">
<Property name="Want" value="0" />
</Feature>
<Feature name="MemoryClockTarget_1">
<Property name="Want_0" value="50000" />
<Property name="Want_1" value="137500" />
<Property name="Want_2" value="140000" />
</Feature>
<Feature name="CoreVoltageTarget_1">
<Property name="Want_0" value="900" />
<Property name="Want_1" value="1000" />
<Property name="Want_2" value="1175" />
</Feature>
<Feature name="MemoryVoltageTarget_1">
<Property name="Want_0" value="0" />
<Property name="Want_1" value="0" />
<Property name="Want_2" value="0" />
</Feature>
<Feature name="FanSpeedProtocol_1">
<Property name="FanSpeedProtocolProperty" value="Percent" />
</Feature>
<Feature name="FanSpeedAlgorithm_1">
<Property name="FanSpeedAlgorithm" value="Automatic" />
</Feature>
<Feature name="FanSpeedRPMTarget_1">
<Property name="Want" value="1160" />
</Feature>
<Feature name="FanSpeedPercentTarget_1">
<Property name="Want" value="20" />


The values I put in bold are what I've changed, raised it up from default 250 on the clock and 150 on memory. Seems to alleviate the problem to larger extent. It's not a solution, just a workaround, but still, some people may find it helpful.