Link Copied
AnybodyM:
My current theories are:
- defect Ram, even though no standalone memtest can detect it (i'll get new ram tomorrow and try swapping that)
- defect HDD (perhaps HDD's 8MB cache defect?), which corrupts data especially while swapping
- weird driver problem.
I will investigate this further this weekend to confirm 7oby's theory of the Intel GFX drivers being the cause.
I have carried on my investigations this weekend, and i've come to the following conclusions:
- It's definitely NOT the Ram. I've swapped the Ram from 2x1GB Samsung to 2x2GB OCZ and the problem does still exist
- It's present in both Vista32 and Vista64
- It's not affected by the operational mode of the Intel SATA Controller (Compatibility Mode vs. Native AHCI)
Even though i've not yet changed the harddrive (theory "defect HDD"), i'm 99,95% confident that it's the Intel GM965 graphics driver which is corrupting my data and causing all this trouble. Why? I'm now using the REALLY OLD 7.14.10.1253 which is the latest one available at DELL, and even though I've tried hard I was totally unable to reproduce the bug with those (old) drivers installed.
==> Intel, please fix this ASAP (and perhaps withdraw affected drivers in the meantime) ! Data corruption is no trivial matter!
Even though i've not yet changed the harddrive (theory "defect HDD"), i'm 99,95% confident that it's the Intel GM965 graphics driver which is corrupting my data and causing all this trouble.
Disabling pagefile in Windows should rule out faulty HDD. Memtest won't swap then.
I haven't disabled pagefile yet, since I rouled out fault HDD earlier already. Errors also occur reading data from DVD-Rom (which is a different controller and driver).
I'm now using the REALLY OLD 7.14.10.1253 which is the latest one available at DELL, and even though I've tried hard I was totally unable to reproduce the bug with those (old) drivers installed.
Just to be precise: Even if the corruption does not occur with these old drivers ever, this just means it doesn't trigger the bug. For example you cannot conclude directly that the HDD or DVD-Rom driver is fine. If one component is faulty there are also always at least two ways to fix it: a) fix faulty component b) workaround in all other components talking to the faulty component to make sure bug is not trigged. Both types of fixes a) and b) are applied frequently and common.
I think that Dell realized that something is fishy because they withdrawed all the more recent graphic drivers from their support site.
I can not confirm this one. Here's v.7.14.10.1409 for Lattitude D630, which is also affected:
http://support.euro.dell.com/support/downloads/download.aspx?c=de&cs=dedhs1&l=de&s=dhs&releaseid=R179345&SystemID=LATITUDE%20D630&servicetag=&os=WLH&osl=GE&deviceid=12618&devlib=0&typecnt=0&vercnt=3&catid=-1&impid=-1&formatcnt=1&libid=6&fileid=244176
Disabling pagefile in Windows should rule out faulty HDD. Memtest won't swap then.
I don't know how Vista handles operation without a page file, but in Win2K and WinXP this was absolutely NOT recommendable, the OS totally freaks out without page file :-)
I haven't disabled pagefile yet, since I rouled out fault HDD earlier already. Errors also occur reading data from DVD-Rom (which is a different controller and driver).
So many faulty HDDs aren't that likely. But it could have been an incompatibility - perhaps Dell is always using similar drives. Both Dells in which i've so far been able to reproduce this bug (XPS M1330 + Vostro 1400) had a "WD1200BEVS-75UST0" Drive, so... just to be sure, i today acquired a Seagate Momentus 5400.2 SATA drive to replace it with something completely different.
Did my 12th reinstall of Vista, this time absolutely barebone, no SP1, only Ethernet Driver, nothing else.
Prime95 worked perfectly fine. Installed the very latest v15.9 drivers from Intel, rebooted, started Prime and who-would-have-guessed-it: Once again a Hardware fault detected shortly after starting Prime95-Blend-1500MB.
So if it's an incompatibility, it's now beyond any hardware I can change for myself.
I've now installed the latest bugfree driver (7.14.10.1272 (v15.2.6) dated 05/11/2007) according to your list on both of those Dell Laptops (M1330+Vostro 1400) and the problem is gone on both. Not really a solution though, because I too will need the HDMI port soon and am dependent on good HDMI behavior.
By the way: Do we have ANY specific report of someone who can reproduce exactly THIS (and not some hard to pinpoint World of Warcraft problem) on a machine that is NOT a Dell ? This seems to be pretty Dell specific from the posts i've read so far ?
I don't know how Vista handles operation without a page file, but in Win2K and WinXP this was absolutely NOT recommendable, the OS totally freaks out without page file :-)
As I said earlier: Disabling the pagefile is to rule out any swapping (and in general all harddisc i/o) related issues. Memtest doesn't involve hard disc i/o at all - except swapping, which is outside of the control of Memtest. Thus disabling swapfile. However other tests with the CD-Rom revealed to me that the hard disc or SATA controller is not the source of the problems (still they might trigger errors e.g. due to high interrupt traffic).
Both Dells in which i've so far been able to reproduce this bug (XPS M1330 + Vostro 1400) had a "WD1200BEVS-75UST0" Drive
Mine is different already: FUJITSU MHY2120BH.
I did much more to verify it's not hard disc related: E.g. I read linux driver source code to find those hard disc drives with NCQ defects.
So if it's an incompatibility, it's now beyond any hardware I can change for myself.
I would like to know if somebody got his mainboard replaced by dell and the bug vanished! I'd also like to know if there's somebody (one!) with one of the affected notebooks: Dell XPS 1330, Latitude 630, 830, Vostro 1400 etc. and does not suffer from this bug.
Not really a solution though, because I too will need the HDMI port soon and am dependent on good HDMI behavior.
I'm using HDMI on different monitors and video beamers. If you get to the point that the external monitor disappears in Windows and you can't select this output device, those two things help:
a) disable pivot services in Windows and/or driver
b) In case you still have problems: Boot by forcing booting splash screens to external monitor (Fn + F8).
Do we have ANY specific report of someone who can reproduce exactly THIS (and not some hard to pinpoint World of Warcraft problem) on a machine that is NOT a Dell ?
I did google like an hour for Lenovo, HP devices containing GM965 chipset, but did not find any error reports yet.
Everyone, do you have the latest BIOS update in your laptop?
I tried various BIOS updates for my XPS M1330 up to and including A10 released 05/06/2008.
If the answer to the above question is "yes", I would stronly suggest you to return the laptop and ask for a working replacement.
At least there is a workaround (see above: installing old driver). I'd be curious if such a "working" replacement exists. I have not yet found anybody suffering NOT from this bug with this hardware (still that doesn't mean such person does not exist; maybe this guy). Please somebody raise his hand, to whom this doesn't apply!
Finally got past 1st, 2nd level support. Dell is investigating this issue. If they can reproduce it and fix it in software, they save a tremendous amount of money for swapping hardware components (e.g. motherboards) meaninglessly. Most people will only discover BSOD, DVD burning and verification errors, CRC checksum unpacking errors and hardly suspect the gfx driver for it.
I have a Dell Inspiron 1525 and Windows XP SP2 with latest gfx driver from Intel's website and ICH8M in AHCI mode with MSM 7.8. I do not have memory corruption or crashes.
Unfortunately this bug is currently Windows Vista only related. Still thanks and maybe once you switched to Vista this thread turns out useful for you.
I am suspecting a hardware problem (perhaps a batch of partially defective or failing chips?)
You are exploring every possible source for the fault. That's nice. I had the same thought in this thread. Unfortunately the spread of manufactoring time of bug affected notebooks is quite big 07/2007 - 03/2008 which makes the theory of a faulty GM965 batch quiet unlikely.
Besides that people got mainboards holding the GM965 replaced (click or another one) and memory and hard drives exchanged (click). And ... nothing helped.
What if the bug is due to Vista itself then?
Microsoft Windows XP and Windows Vista are both incredible stable operating systems. It is well known that most instability issues are related to driver issues.
In 2007 all online reported Vista crashes reported to Microsoft have their source in components of the following vendors:
nVidia 28.81%
Microsoft 17,97%
Unknown 17,07%
AMD-ATI 9,30%
Intel 9,83%
...
source p.47 in this document or see here.
There is no such thing as an unstable Vista. This is well known also to Microsoft and they develop certification programs for drivers and extremely powerful tools to detect errors in drivers as soon as possible:
http://www.microsoft.com/whdc/devtools/tools/SDV.mspx
Therefore Vista itself is not the cause.
What version of Prime95 is being used?
Only AnybodyM used Prime95 to detect this kind of error. I trust him that the version he used is Vista compatible. But it doesn't actually matter, since other methods (memory test within vista and md5 checksuming) also expose this bug.
Please use Memtest 3.6 which is Vista compatible and detects errors within a few seconds (less 30 secs):
http://hcidesign.com/memtest/
1. Does the issue occur in AERO Basic?
I do have the M1330 using Vista Business 32 Bit and here the bug occurs regardless whether Aero is enabled or disabled.
2. Does the issue occur with 1GB memory?
On my M1330 the error also occurs with just one memory module of 1GB installed - however maybe less frequent.
On D830 there is one report by MikeAP that his error disappears with just one module installed (click).
3 What are the DVMT setting in BIOS?
Dell BIOS neither displays nor allows to change DVMT. Dell BIOS only displays the size of the linear framebuffer, which is mandatory since VESA 2.0 and is of 8MB size. Actually the allocation of DVMT is under the control of the operating system:
"The amount of video memory allocated depends upon the amount requested by the operating system."
http://www.intel.com/support/graphics/sb/CS-010488.htm
Using the older working drivers I can query from the intel tool:
Minimum Graphics Memory: 8 MB
Maximum Graphics Memory: 358 MB
Graphics Memory in Use: 177 MB
The actual used memory (here 177MB) changes during operation. I see 165 - 177MB within a single session. Those values do not seem to be different when installed one of the "bad" drivers above.
However DVMT is also something that came to my mind: Just consider what happens if the driver dynamically reduces the video memory and forgets to alter the Page Attribute Table / MTRR processor registers to switch from WriteCombining Caching strategy as used by framebuffers to WriteBack/WriteThrough. Strong ordering of read/write requests is no longer given => memory corruption.
