[CRITICAL] memory data corruption with GM965 on Dell machines

[7oby]

current workaround (if no bugfix BIOS available):

use intel display driver 7.14.10.1272 (v15.2.6) dated 05/11/2007 or older

bugfix BIOS (yet only released for the following systems):

Dell Inspiron 1420 : Update BIOS to >= A09 (BIOS A09 here)
Dell Inspiron 1520 : Update BIOS to >= A09 (BIOS A09 here)
Dell Inspiron 1525 : Update BIOS to >= A09. Verify VBIOS >= 1566
Dell Inspiron 1720 : Update BIOS to >= A09 (BIOS A09 here)
Dell Latitude D530 : Update BIOS to >= A07 (BIOS A07 here)
Dell Latitude D630 : Update BIOS to >= A12 (BIOS A12 here)
Dell Latitude D830 : Update BIOS to >= A13 (BIOS A13 here)
Dell XPS M1330 : Update BIOS to >= A12 (BIOS A12 here)
Dell Vostro 1400 : Update BIOS to >= A09 (BIOS A09 here)
Dell OptiPlex 330 : Update BIOS to >= A05 (BIOS A05 here)

still waiting for fixes for at least these notebooks:

Dell Vostro 1500

I don't know whether these notebooks are affected as well, nor whether a BIOS update is being prepared:

Dell Vostro 1510, 1710, 1700
Dell Latitude ATG

List of notebooks NOT affected:

Dell Vostro 1310

Bottom Line:

I have sufficient reason to belive in some memory corruption bug with the X3100 drivers v15.8 + v15.7.3. I do not think it affects your hardware, but I have the problem. There is also the chance that my GM965 chip is faulty. In addition there is the chance that my bug is not related to the gfx drivers, but just emerges with those two particular gfx drivers and is caused by some other faulty driver or device.

I filed a bug report at intel, with the outcome that they haven't reproduced anything like that. Since I found people with bugs that sound very similar (see below), I want to raise awareness in this forum and help other people google it (if they encounter the same problem). I will updates this posting if I get more information or by some magic it's not the drivers fault.

My walkaround is to use the slightly outdated 7.14.10.1253 of the GM965 gfx drivers. It does not show up in this version.

Selected Details:

I've got a fancy Dell M1330 notebook with intel X3100 graphics integrated into the GM965 chipset. Dell ships this Core2Duo notebook with 7.14.10.1253 of the graphics driver version runnning Vista Business 32Bit @2GB memory. It runs perfectly except that sometimes it doesn't recognize HDMI attached video beamers and big monitors.

Thus I tried:
7.14.10.1437 (v15.8)
7.14.10.1409 (v15.7.3)

Both fix the external HDMI attached device bugs.

However I encounter a severe bug: Memory corruption!
Unfortunately it occurs very infrequent and my test setup to detect memory curruption in this case requires reading big files from harddrive in a loop and checking their md5sum. In general I have to read ~50 - 1000 GB of Data u ntil I encounter a single md5sum error.

The whole course of isolating this bug and accounting the gfx system is not that interesting here. Let me just mention that I threw all the memory tests (incl. the brand new GM965 supporting MemTest86+ V2.01), changed memory modules, CPU stre testing at every single P-State, Harddrive checking (incl. different drivers such as IASTOR.SYS 7.8.0.1012, 7.0.0.1020, MSAHCI.SYS 6.0.6000.20765). I do not encounter these bugs running ubuntu 7.10 or Mac OS X 10.5.1 on the same hardware reading the same files. It took me every evening the last week and some nights to do a thorough test. The bug seems not to be file/harddrive related - just my testing setup is built this way, since it gives me an easy way to validate some of the main memory content.

I am extremely careful if it comes to making hardware or software responsible for a given bug and try to think of every possible inference.

Certainly while switching back and forth with the drivers above I did not change anything else, after having isolated the bug. At that point I tried to find evidence that I'm not the only person. And I found:

Posting #161 (X3100 on MacBook):
"Our assessment of the error is that a bug in the graphics driver code is causing memory corruption within the data used by WoW under some circumstances, leading to a crash."
http://forums.worldofwarcraft.com/thread.html?topicId=3686740801&sid=1&pageNo=9

This is written by a WoW/Blizzard employee. Well, but it's 3D and it's Mac OS X!

There is code sharing amoung different intel drivers (Linux, Windows, Mac OS X). But the same crash also occured using WINDOWS by means of Boot Camp on the same hardware! posting #182:
"heres my windows crash report on the same macbook"
http://forums.worldofwarcraft.com/thread.html?topicId=3686740801&sid=1&pageNo=10

Regarding 3D: I'm using Aero, but the errors occurs with Aero being enabled and disabled.

[levicki]

taoshen1983:

I am surprised to see you, a software engineer, arguing that beta testing is not acceptable.

You seem to have misunderstood me on purpose in an attempt to imply that I am stupid. So, to clarify what I said for the sake of everyone else here:

I am not against beta testing as a method of improving software quality I am only against beta testing of hardware products while paying the full price tag. There has to be a line in the sand somewhere.

taoshen1983:

Instead of having a brain 1000 times the complexity, we spend time on the internet arguing that 4GB of ram is too much for a laptop.

Now you seem to be misquoting me on purpose as well. I never said that 4GB is too much for a laptop I said that 4GB may be too much for someone's budget, and that it is an overkill for the majority of laptop users.

taoshen1983:

So I am not sure industry should be reminded of the old days. They are trying as hard as they can.

They are only trying hard to make a profit, customer satisfaction seems to have fallen to the second place or it is perhaps even lower on their list of priorities.

As an example of what I am saying, a friend of mine bought X38 based mainboard in Q1 2008 which is now obsolete by two chipset generations (X48 and soon to be X58). He has been left with a buggy piece of hardware and he is already without updates and bugfixes because mainboard manufacturers are focused on churning out X58 based mainboards and are probably already designing X68 based ones. That isn't even beta testing anymore it is an eco-unfriendly resource wasting rip-off.

Lets end this offtopic discussion here if you want to argue further please start an appropriate thread in Watercooler section.

[psiworld]

Still waiting for a fix from Dell for the Vostro 1500.

[yuriylsh]

BTW, any dates for the next driver version (if it's gonna be released at all...)?

[7oby]

Regarding the Vostro 1500 and all other still affected notebooks: Please contact Dell and get to 2nd level support. Be polite, precise but confirmed and ask for a BIOS update. This helps more than I can help.

You probably recognized that v15.9.7/8/9 can't be used for GM965.

But v15.11 is a unified driver for X3100 and X4500 again. This one 7.15.10.1537 (v15.11.0.1537) works fine give the VBIOS is sufficient recent:
http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&ProductID=2377&DwnldID=12091&strOSs=164&OSFullName=Windows%20Vista*%2032〈=eng
It offers at least OpenGL 2.0 and one bug in wmv acceleration resolved.

[yuriylsh]

Thanks a lot for pointing the drivers out!
Do you know if it is all possible to see the changelog for the 15.11 version? Anything else besides OGL 2.0 and wmv bug? Performance, bugs, tuning, new features?

P.S. It's interesting that after installing 15.11 my vista graphics score went down from 3.4 (with 15.9 version) to 3.0.

[psiworld]

Notice there are new drivers posted on the intel website for the Intel 965 video card. Anyone tried them yet with Vista?

Version
15.11.2.1554

Date
9/8/2008

Link
http://downloadcenter.intel.com/T8Clearance.aspx?url=/16939/a08/winvista_15112.exe&agr=Y&ProductID=2800&DwnldID=16939〈=eng

[7oby]

psiworld:

Anyone tried them yet with Vista? Version 15.11.2.1554

These are two threads, where links to current drivers are posted and results are discussed:
http://software.intel.com/en-us/forums//topic/52983
http://forum.notebookreview.com/showthread.php?t=128777
But to give you an answer: Yes the drivers work fine.

[devicepedia]

I did google like an hour for Lenovo, HP devices containing GM965 chipset, but did not find any error reports yet.