Read more.BSOD strikes again.
Read more.BSOD strikes again.
As you pointed out, bad drivers strike again. That's not Microsoft's fault. Well, they should have made sure not to use anything from nVIDIA for their demonstrations as nVIDIA drivers cause the most BSODs.
I never got the significance of the BSOD and why they are NOT within microsofts control. I'm playing devils advocate here, but surely they demonstrate that the OS does not handle the driver well enough. I mean, these are Microsoft certified drivers... Where's the value that MS is adding by certifying these?
Just a random thought that crossed my mind when reading this....
Just a note: kernel-mode drivers need full privileges to do their job. There is no good way to police them without slowing everything down. Certification has it's limits: you can't reasonably expect MSFT to check the source for every single driver out there looking for bugs.
In fairness there are projects like Singularity the micro-kernel.
The idea that you have to be in ring 0, in a low level language for a device driver is really rather antiquated!
throw new ArgumentException (String, String, Exception)
I think BSOD indicates a double processor fault in windows (in an x86 system a triple fault automatically resets the processor). Same thing for a kernel panic in linux, it is caused by a double fault. The used to use triple faults to switch a processor (286) out of protected mode into real mode iirc.
Under windows 7 I believe all drivers run as user mode (that the idea anyway) so a driver cant bring down the system. I am not sure if that theory translates into practice. It is one of the reasons all drivers had to be re-written for vista because XP drivers could run in ring 0 but vista ones couldnt.
You're missing my point.
the question I'm alluding to is WHY bother to certify them? This is microsoft branding them as safe and up to Microsoft standards. CLEARLY microsofts standards are not high enough if when using a certified driver you can experience the BSOD. If it's caused by a clash in resources due to two certified drivers then the method by which microsoft are using to test is somewhat floored. I don't accept that the most porfitable company on wall street (according to Mr Baulmer anyway) can't test for this, but that they DON'T. It's not like Nvidia don't have a standard chipset design or anything......
I'm going to play devils advocate (as usual).
Could this have been human error? The team had a well set up PC with the right drivers that had been tested and running just fine until someone decided to update the drivers with a version that hadn't been tested. Or perhaps they just got the wrong end of the stick in a discussion and thought they were supposed to do something and they weren't (or didn't do something they were supposed to).
Or perhaps this is a case of the kind of planning we have here most of the time. The team were only able to actually put it all together the night before for some reason (like imply not being given enough time to d their job) so the system was untested and obviously had major problems.
All these are simple human error. Any organisation can suffer from this.
Any thoughts on how many drivers there are? You'd have to run every single driver with every single other driver, in combination and alone, on every single piece of hardware, to be sure. It's a ridiculous idea. They could do a very basic testing with common (e.g pre-built systems with lots of sales) hardware and common drivers for those hardwares, but not really much more than that.
...So what you're saying is that the microsoft certification doesn't actually guarantee or mean anything.
Again, I repeat my statement WHY BOTHER. For me it's just a license for MS to print money, as I am sure that they charge for this certification and it's not a valid validation process.
Assuming that this isn't human error (heaven forbid! ) this is, if the first comments are to be true about this being due to an Nvidia driver problem, microsoft falling foul of their own lack of testing.
I don't subscribe to the driver comment particularly - how much variation is there REALLY out there. Certainly for GFX cards there are only 2 manufacturers AMD & Nvidia. Both companies sets of silicon are based on the same design, scaled up or down for the number of shader models etc. Likewise with northbridges and chipsets there are fundamentally only 2 of these producing modern mainstream kit - AMD and Intel. There's only so many permutations of these chips that exist. Granted historically there have been more manufacturers, and I know that others exist, but the actual models of chips out there and the permutations are actually quite finite for a large portion of the market, and if we look forward, it's actually getting fewer with the integration of northbridges and graphics silicon into the CPU.
Besides, and this only holds true if microsoft do actually charge for the validation checks, what value do the 3rd party manufacturers of gfx, network, sound etc cards get if they are submitting drivers to MS for validation/approval if they are not being unit/integration tested. That as a QA process from one of the worlds biggest most trusted brands, not to mention WEALTHIEST companies, sucks!
(Sorry if that's a pointless tirade/rant)
Off to go and hug a tree, then back on with testing Microsoft Dynamics NAV...
well they don't make money directly on the certification.
Its a lot more complicated that saying "oh you've got an invalid pointer there".
You often find that some hardware too is faulty, only a little bit mind, and all it does is create a state which the driver development team never anticipated. Then your in for a world of hurt!
throw new ArgumentException (String, String, Exception)
Tattysnuc (02-06-2010)
The driver certification programme is more or less the same as SSL root certificate authorities. Simply a system to verify that the host you're communicating with, is who it says it is. Or in this case, that the driver you're installing is from a genuine hardware vendor/driver provider.
Its a *little* bit more than that:
http://en.wikipedia.org/wiki/WHQL_Testing
bit of black boxing too.
throw new ArgumentException (String, String, Exception)
Tattysnuc (02-06-2010)
Meh, blackboxing holds mediocre effectiveness at testing code at the best of times. Still, point made, there is that, in that at least the driver wont make your windows installation blow into a thousand bits just by loading it, most of the time.
There are currently 1 users browsing this thread. (0 members and 1 guests)