Memory corruption with Intel Graphics Driver

Hi all, I have a big problem with the Intel GMA Driver for my Dell Latitude D630: it is causing memory corruption

I used MemTest and it would continuously report read and write errors with 15.7 and 15.7.3 series drivers for my GM965. However, when the Intel graphics drivers are not installed, it detects no errors. You can download it to see for yourself at http://hcidesign.com/memtest

I could even start MemTest after a fresh install, and it would work. At the same time I would update the Intel GMA drivers, and suddenly MemTest starts detecting errors. When I uninstall the GMA drivers, with MemTest still running, the errors suddenly go away again.

Funny thing, Dell has a custom build 1253 driver on their support site, and this one does not cause the errors. A later version does cause the memory corruption.

Now MemTest is one thing, but at the same time something else goes awfully wrong: SFC /SCANNOW is finding corrupted files, when the Intel driver is active. There are no problems after a fresh install, without GMA driver. After installing the driver and rebooting, running SFC /SCANNOW from an admin prompt will find a couple of corrupted files. These appear to be random files, different after every reboot. The log states that these integrity violations can't be fixed because both the file, and the backup of it fail a hash test. I believe memory corruption is causing some files to fail a hash check.

Like with MemTest the problem is only there when Intel GMA drivers are installed, uninstalling them and rebooting will cause SFC /SCANNOW to function normally, without integrity violations. Like before, the custom Dell driver (build 1253) does not produce the erratic behaviour, a more recent custom dell driver does produce both memory errors and integrity violations.

My system is a Dell Latitude D630, with 2 x 1GB memory, a GM965 chipset with X3100 graphics, and a core 2 duo T7700. A colleague of mine has a D830 with similar specs (except for CPU and HD) and it has the same problem. Remarkably, the other system does not have the problem with only 1GB of ram instead of 2x 1GB of RAM... puzzling.

I now my memory is not defective or damaged:
- Preboot Memory Diagnostics find no errors after 30+ minutes of testing.
- MemTest does not find any errors with the original "1253" driver build
- MemTest does not find any errors when running XP with latest drivers.
- I've used memory from another computer and it has the same problem.
- Chance of a hardware error in two different machines is too small

I know the harddrive is not at fault:
- I have tested it with two different harddrives (a seagate and a fujitsu)
- Chance of a hardware error in two different machines is too small

If you have GM965 or maybe even G965 or G35 please test these two things on your machine and report back! Thank you.

 

Just a question. How do you know it's an actual error? Does it cause programs to crash? Data loss? Anything else?
Having a tool say temporary files are corrupted doesn't really matter in itself.
As for the memory part, how do you know it's not Memtest screwing up?

In short, are there any symptoms of this actually causing problems?

 

SFC is not a tool saying temporary files are screwed up. It is part of Windows Vista, and it compares system files in their actual locations (C:\windows\system32) with their signatures, and restores the ones in the store (C:\Windows\WinSxS\...) if necessary. It's called System File Integrity Checker or System File Protection and has existed since XP. If a file is hacked, Windows can restore a signed version of it.

This wouldn't bother me, but it can be reproduced so accurately. Other than that, I've had "a service stopped working" quite often, and a few downloads that were corrupted on disk. It doesn't happen all the time, but it shouldn't happen at all on a system that has nothing but a fresh install (from original media) and drivers, right?

It is possible that MemTest is screwing up, but SFC integrity violations can not be tolerated after a fresh install (Vista RTM and SP1, both in 32bit and 64bit, and it happens on a Seagate and on a Fujitsu harddrive, as long as soon as I have Intel GMA drivers 15.7.0 or 15.7.3 installed)

If someone with a similar system (preferably a D630 as well) can tell me he doesn't have this problem, then I know I have a weird hardware defect and need to return the machine. If anyone has the same problem, then it's the probably the driver malfuncioning. Please help me figure this out.

 

I ran this on my D630 for about an hour or two (walked away and forgot it for a while LOL) anyway, it found no errors. Interesting...

 

I'm clueless and need enlightening here, how does a video driver lead to RAM/memory corruption? I though memory corruption was a hardware problem which couldn't be caused by software issues?

 

@booboo12: thanks for running the test, but can you confirm that you are using Windows Vista with 15.7 series driver, because I don't have the problem with Windows XP or the older drivers. I've also contacted Dell about this, I want a working D630 like yours.

@harleyquin07: software can stress hardware to the point where it becomes unstable. that memory test is a pretty intensive test, pushing the CPU up to 70% and doing nothing but memory access.

At some point support for Hardware T&L was added to those drivers, it is possible my chipset has a defect that only causes problems when Hardware T&L is used. As I said I also only have the problem when I use 2x 1GB in symmetric dual channel, but either 1GB stick will work on its own in either location without the problem.

 

Hi.

have you tried "RightMark Memory Analyzer" runs in windows

from here

http://cpu.rightmark.org/download/rmma372bin.rar

regards

John.

 

True, of course. Your RAM either works or it doesn't. No software can change that.

However, there is still a lot of software involved in accessing memory. There's firmware/BIOS, there's drivers and so on, in addition to the OS itself. If any of these screw up, applications will "see" corrupted memory, even if the physical memory is fine.

And of course Intel's integrated graphics means that the GPU is using system memory in the first place (which means it has to work with the memory controller and set up the memory system appropriately), so there's plenty of opportunity for it to screw up incoming memory requests.

You might want to submit a bug report to Intel, if you can find an email address for that purpose. (Unless it's a known issue, of course)

Other than that, wait for a driver that fixes it, I guess. It does sound like it's only a software issue.

 

I have submitted a bug to Intel and launched a support call with Dell about it.

 

What you're saying then is that applications and integrated graphics "mistake" otherwise perfectly normal RAM for corrupted memory because of bad requests?

 

@tinderbox: I've run the RightMark memory test, and it's the same as MemTest, a single failure in every test run with the new drivers, not a single failure in any test run with the older driver or without graphics driver.

I'll try two new memory DIMM's and after that I'm calling it a motherboard defect. Intel has no known issues like this.

UPDATE: I've tried two new DIMM's and Dell provided me with a new mainboard, but the problem persists. Intel if you are reading this... you have a serious driver problem.

 

Hi. I have a D630 and I´m having Blue screens once in a while with XP SP2. I don´t know if it is the graphics driver, but the erros are 8E and EA. The EA error is related to graphics driver and the two erros occurs at the same time. I´m using the latest one at Intel website. If this continues I´ll try to switch back to Original Dell drivers. Where can I find older drivers at Dell website?

 

BUG CONFIRMED

I can confirm having the same bug on slightly different hardware (Dell XPS M1330, GM965, Vista Business 32-Bit). Some of my bugreport is available here:

http://softwarecommunity.intel.com/isn/Community/en-US/forums/thread/30250940.aspx

Bug present in
7.14.10.1437 (v15.8)
7.14.10.1409 (v15.7.3)

but NOT present in
Dell 7.14.10.1253

I ruled out AHCI, NCQ and any harddisc related stuff. Also Vista SP1. To further narrow down this bug:
. What OS are you using?
. I only tried with AERO enabled. Can anybody confirm having AERO disabled?

A more elaborate bug report I send to intel, but only received:

Thank you so much for your bug report! This one will get nailed!

I found one more person in a german XPS forum suffering from this bug:
http://www.xps-forum.de/thread.php?postid=74541#post74541

 

The curse of the Intel cards strikes again, plaguing with me incompatibly issue's on games for years, finally getting a Nvidia 8800 GTX in 2 weeks.

 

Edit: Prelim tests with the 7.14.10.1329 x64 driver suggest the it's free of the "memory corruption" feature. Won't be sure until I reinstall as I'm pretty sure that the combination of driver verifier + a twitchy driver has hosed a few system files at this point. Crossing fingers.

--

Having the same problem in vista64 with my vostro 1500 (spent the last two days slogging through memory dumps trying to figure this out, actually).

It started with a random reboots (no bsod, no error, just cut to black and back to post) and I've been progressively troubleshooting the issue. I'm very glad I found this thread: I was on the brink of sending the laptop back to dell to have the motherboard swapped out...all the memory dumps showed some kind of memory corruption but the memory itself kept test.1329ing out fine.

I'd be happy if I could find an old version of the 64bit drivers that don't cause this, but haven't, as yet.

 

My current finding as posted in intel forum (see link above), ist that

any driver beginning with
7.14.10.1277 (v15.4.2) dated 05/24/2007
contains this bug.

any driver including these and below are bugfree:
7.14.10.1272 (v15.2.6) dated 05/11/2007
7.14.10.1268 (v15.4.1) dated 05/03/2007

Therefore I suggest you try the 64-Bit version of
v15.2.3 64-Bit build 1255, which you can get here:
http://www.opendrivers.com/driver/2....64.1255-windows-vista-x64-free-download.html

Other versions are available here:
http://www.opendrivers.com/category...d-video-intel-free-driver-download-page2.html

Please report back, if you succeed. The fastest way (usually whith seconds or one minute) to detect this kind of memory corruption is by means of this program
http://hcidesign.com/memtest

I suggest you start testing with a buggy driver. If memtest then fails, it's likely you really suffer from the same bug. Then try v15.2.3 64-Bit driver above.

 

Ok, my head hurts.

Further testing suggests that whats really going on with my machine is memory corruption somehow triggered by the lpower state shifts on merom / stanta rosa. It's hardware at this point: something's not feeding somthing enough voltage somewhere and the memory's getting corrupted. I locked the voltages and disabled the chipsets cstates with rmclock and the problem went away (or at least became much more difficult to detect, not sure which at this point). Vista's inbuilt power management seemed to have no impact on the situation.

Interesting that it presented in much the way you described, though. The windows memory tester was the only thing that could catch the pool corruption between its onset and a system crash.

At one point I was actually watching as the md5s on some files i had being loop verified kept changing. Thought I was losing my mind.

 

I tried to do the same and watch memory corruption:
. using RMClock 2.35.0 I fixed the multiplier to 6x @0,9V
. C-states are idle states and P-states are operational states. I do not understand how manage to disable the C-states? In my eyes you can adjust some C-states, but you can not entirely disable them.
. In "Advanced CPU settings" / Processor you may disable the Deeper Sleep C2E, C4E states (which I also did for testing purpose)
. In "Advanced CPU settings" / Chipset you may disable "Enable C5 and C6 modes". But as soon as I disable this one, may machine hangs.

So I fixed multiplier and voltage and disabled "some" C-States, since e.g. I can't disable C0 state at all. Anyway: Still I do have memory corruption with the mentioned drivers.

In my eyes you conclude too fast. If you disable P-State transistions and disable some of the Deeper Sleep modes, you also reduce latency of the entire system. An incoming interrupt gets served faster, since waking up doesn't take that long or chipset can't be in the middle of adjusting P-State Voltage (= this takes some time).

Before I used Windows memtest I was also looping over large files and calculating md5 sums.

Did you / would you test v15.2.3 64-Bit build 1255 driver?

 

re: p- vs. c-states. Sorry. I may have spoken inaccurately. I was operating on nearly no sleep at the time. However, I believe you can disable c4 and un-enable c5 and c6 under the advanced chipset options in rmclock. To clarify: I've locked the multiplier at 10x and the voltage at 1.175 (I'm using a 2ghz merom with 2mb cache; don't recall the exact model). I've also disabled everything related to power management that rmclock gives me access to, including c 4-6, popup and popdown.

Have you tried locking the voltages higher than that? .9 is the lowest voltage mine supports.


I also get crashes during heavy io use, especially via usb or firewire, making me suspect something in the southbridge. But a lot of this is very educated guesswork.

re: latency. Good point, but wouldn't that still point to hardware in my case? One presumes that the production divers have been tested with the longer latency in place. I tend to think this still means something low level is functioning out of spec (if I've understood you correctly) even if it's latency rather than voltage related. I lean toward voltage because of the nature of some of the crashes I've had. In addition to blue screens, I've gotten a lot of blank reboots, with no warning and nothing but an unexpected shutdown in the event log.

I'll try testing with the drivers you mention over the weekend. I may have to reinstall with x64, as I'm not sure I have a clean backup (system stability would go down the tubes over time as the drivers got more and more corrupted. Currently have a 32 bit version of vista installed. I'll also try to test with the suspected good 32 bit drivers.

I've started to wonder whether the video driver is just more susceptible to memory corruption rather than the cause (at least in my case). It has certainly "casued" more cashes than other drivers, but it wasn't alone.

Dell is going to be swapping out my motherboard early next week, so that should put me in a better place to determine if we're talking about the same issue or not. It's certainly possible that I'm having the issue you describe in addition to some power related wonky-ness.

 

I am having the same problem with my D630. Windows Vista 32. Few weeks ago I installed the updated intel drivers and my laptop started doing strange things, black screen, blue screen, jumpy screen. I could not figure it out. I stumble across a webpage that told me to test my memory and it would fail when I ran memtest. I went today and picked up 2 new chips.

Installed them and booted up the machine. things seemed fine but then my micorosft outlook kept hanging and restarting. Then I recieved the flicker.

I came across this forum article and reverted to the 1253 video drivers. The memory test now runs fine and no errors with outlook, explorer or blue/black/flicker screens.

I decided to put the old memory which I thought was bad in and it does not error out on these drivers either.

something is definitly playing games with those newer intel drivers on a D630!

 

Well, turns out the comment re: me jumping to conclusions qualified as foreshadowing. When I'm wrong, I'm wrong and this is me admitting it.

In short, events whet like this:

1. The mother board replacement greatly improved the situation. To the point where, after cursory testing, I wrote it off as cured.

2. I continued to have stray blue screens now and again. The dumps didn't show a recognizable pattern: i concluded i had another misbehaving driver. Over time, the severity of the blue screens increased somewhat (I believe 15.9 made matters worse).

3. I finally got aggravated enough to start systematically testing the drivers. Signs pointed back to the intel display driver. I loaded the 15.2.3 1255 driver and the problem (which had been highly reproducible) seems gone. Doing more thorough testing now.

4. Currently kicking myself and hoping i've learned my lesson. 7oby: thanks for your great work on this issue. I read in another thread that a fix is in the works at dell.

Edit: has anyone tested to see if the situation is the same on server 08? My guess would be "yes" but I'd be curious to see...

 

Any updates to this issue for the vostro 1500? What seems to be the hold up in issuing a bios update for the 1500? The latest bios for the vostro 1500 is still A06 issued almost a year ago.

The inspiron 1520 received a update that resolves this issue with the latest Intel video drivers long ago.

I can reproduce this problem repeatedly using memtest.exe or Orthos blend test on my Vostro 1500. Memtest 86 or vista memery tests come up clean.

If I revert to the old drivers posted in this thread the problems disappears. But then multi monitor support suffers.

A lot of Vostro 1500 owners must be having this problem with the integrated x3100. I can't be the only one.

We need more vostro owners with this problem to put the pressure on Dell to update the bios with a fix for this video memory corruption problem with the Vostro 1500.