UPDATE 7/29/2007: Whatever. Stupid notebook. Stupid Heisenbugs. Sooooo on Friday at the conferrence I realized Safe Mode was probably working for me because of the low resolution it starts in. So I tried running in normal mode at low resolution. I pressed F8 during startup and noticed an option to boot Vista at 640x480 but without having to use safe mode. Wow - very cool - exactly what I need! I observed that at the CTRL-ALT-DEL screen I was indeed at that resolution - but then fairly soon after logging on - my native 1920x1200 resolution was restored (there is some nVidia stuff that runs at logon via rundll32.exe and I presume this is what changed the resolution back up to max res). Quickly before things could lock up - I changed it back to 640x480 and lo and behold - everything was fine for several minutes. So I decided to work my way up through the resolutions one at a time testing each one for several minutes and exercising the GPU by doing Flip 3D for about a minute at each resolution. To my utter amazement - I was able to work my way all the way back up to 1920x1200 resolution without any issues (i.e. the screen didn't go all black or all white etc. as it was doing previously). So what certainly seemed like a true hardware issue is now gone. I have since re-installed the latest Dell driver for my D820 (they have a driver from '1/2007' on their support.dell.com site for 64bit Vista - but the driver is actually from December 2006). I used that old driver on Friday / Saturday without issue at full res. Today I have installed the latest nVidia driver from www.laptopvideo2go.com and it is dated from 7/6/2007 which is MUCH newer - and again - I have no problems after rebooting and using it for about 2 hours. For those that are curious - this laptop has a GeForce Go 7400 based chipset and the driver version is reported as: 126.96.36.19922 and it was signed by the Verisign Class 3 code signing CA cert.
The only thing I can think of is that this all started when I plugged in an external display on Thursday night . . . Vista (well - the nVidia drivers I had installed at the time?) tried to do some sort of auto-detection because the instant the VGA cable was plugged in - the screen went black and that's when the badness started. We actually eventually got it working later that night (after migrating my VM's to a demo machine) by bringing up the mobility center *first* and then plugging in the external display and choosign the 'connect' option and specifying a refresh rate and resolution for that etc. So I'm wondering if what was happening was simply that the drivers were trying to pump out a resolution / refresh rate that wasn't supported by the LCD . . . but what I don't understand is why it would take about a minute after the machine started up and I had logged in for the display to get whacked out (thus making it seem like a hardware problem since I had already booted Vista and logged in and was using apps). Also - the machine would gradually grind to a halt before the video would black out or white out . . . IE would start running slowly, things would start displaying the 'not responding' thing and THEN the badness would happen - again making it seem very much like a hardware issue with something not working right.
UPDATE: Well the presentation went well - hope to get some good ratings / feedback. The notebook however - is dying. It has a very obvious hardware failure. The standard Vista VGA driver at max resolution displays a white screen with pretty diagnol lines through it after about a minute of use after I start Vista . . . 1024x768 and lower resolutions seem to work reliably but going above that causes severe pain for the notebook. I'll have Dell come and swap it out when I get back (hopefully). I think it must have shorted out last night at the 'tech check' when I plugged in the external VGA cable. I spent a few hours last night trying various drivers (the latest from Dell, the inbox VGA driver, etc.) and they all behave the same after high resolution. :( In 9 years of exclusive Dell hardware use - this is my first hardware related failure.
So I flew into Seattle Wednesday afternoon and I'm set to deliver my 'Targeted Attacks' presentation (yes, the same one I gave in Spain) at TR5 here in Seattle. Speakers get the chance to do a 30 minute 'tech check' before their speaking slot. This seems like a good idea - so I go. I plop my notebook on the desk, and the tech plugs in the video cable to the back of my notebook and the screen immediately goes black with only a mouse cursor and the HDD goes nuts. No biggie - so I engage in some idle chat while the drive cranks and wait for the screen to come back. It doesn't. No video on the LCD or the external display. I start to move the mouse and observe that it's very sluggish and choppy. Not good. I hit the FN+F8 key a few times to cycle through the display modes - nadda. So I hold the power button down for 5 seconds to power off the notebook. This notebook has been extremely reliable and generally well behaved but its track record at public events or outside of the daily norm has been un-good lately (bugcheck at FIRST, now some obviously bad kernel mode code going berserk in response to an external connection event). I power the machine on and login to Vista. The tech once again plugs in the video cable and once again the screen goes black and the notebook stops responding. I once again power cycle the notebook. Now I'm nervous. It's only now that I am told I can use their demo machine if I want. :) So I copy 20+GB of VM's over to their demo machine @ 100mbps . . . 30 minutes later it was done.
I told you all of that, to get to this. You see - the OEM's like Dell don't release the latest drivers from the chipset manufacturers the day, week or even month that they become available. I'm not quite sure of the reason but I'm sure it has to do with them wanting to test things and repackage them in their own particular formats / installers etc. What's worse - if you own a Dell D820 like me and you try to go to www.nvidia.com - you can download the latest drivers from their site - but good luck installing them - they won't install on your notebook. Why? Because the INF file used by Windows to figure out what drivers to install where - doesn't contain the right information and Windows thinks you don't have the right hardware for these drivers.
The solution? www.laptopvideo2go.com There is a whole community of folks devoted to solving this problem - they dilligently watch the nVidia web site and when new drivers post - they create mod'd INF files for them that allow you to install them on your machine. I'm a huge fan of always using the latest bits . . . so I'm about to install the latest drivers from their site to see if I can fix the suckage I encountered earlier.
Oh and for the record - the two previous hard resets were the result of yet another Heisenbug . . . after the 30 minutes worth of file copy - we tried one more time - and this time connecting the external VGA cable didn't kill the notebook - it worked (for the most part) as expected (I had some issues getting Vista to 'clone' the display onto the external - it kept wanting to extend it instead of clone it). I use multi-mon every day at work so I'm not sure why this paritcular setup was causing pain and suffering - hopefully the latest nVidia drivers I'm about to install will help. Fortunately I've got the demo machine to use for my presentation tomorrow. If you're at TechReady - it's in Willow/B at the Sheraton.
I have seen similar behaviour with nVidia cards (mine is a 7800) where I couldn't get both the projector and my laptop screen to work at the same time. Oddly after I'd given my talk facing the screen rather than the audience I tried again and it all sprang into life.
I suspect the Vista dialog box that asks if you're trying to do a presentation is getting in the way here.
One final thought is that I have also seen funny results when using dual monitor depending on if I've been using the VGA or DVI port on the same monitor (for VGA I have to use a DVI->VGA dongle on the laptop) when in DVI mode the secondary screen resolution was fixed incorrectly in Widescreen mode. I wonder if this dual use of the same port is confusing the world.