Nvidia drivers named as lead Vista crash cause in 2007

Nvidia, not Microsoft, is apparently responsible for the drivers that caused the most Windows Vista crashes last year. That's if documents posted as part of a lawsuit probing PCs' ability to run the new OS are anything to go by. The files in question are a batch of Microsoft emails, available to read in a single PDF here. …

[James]

Trying to do too much too fast..

I'd suspect that these driver errors were caused by the developers trying to do too much, too fast with an overly complicated operating system. Possibly with inadequate documentation and information.

Maybe it'd be better if the various developers took time to fully and properly document/describe/develop/test everything before it was released. It may be a much longer development cycle - but it may actually produce a much more reliable operating platform and better "user experience". Which is what everyone wants, right?! (Ching ching, go the cash registers.......)

[A J Stiles]

Long term solution

There is only one long-term solution to the problem of driver instability. That is for governments worldwide to insist that Source Code must be supplied along with all driver software -- or the hardware will be banned from sale.

For one thing, AMD and nVidia are already reverse-engineering one another's products; so any competitive advantage that might be lost by having to supply Source Code is tiny. Also, the move won't be unilateral, so all parties will suffer equally. For another, it's suspected that manufacturers are being somewhat disingenuous (and, in some cases, stopping not far short of Creationism) with their hardware specifications, hiding behind the unavailability of driver Source Code to conceal their lies.

A corporation's "right" to keep secrets from the very people who keep it in business must be subordinate to a consumer's right to use goods they have lawfully purchased for their rightful purpose.

[Craig]

I can believe this

Nvidia's drivers last year were awful really, which is the main reason my media center is still using MCE. On my desktop, major features that are in the XP control panel were missing for most of last year in Vista. Even now, every time I update my nvidia drivers, it crashes my desktop PC without fail as part of the install.

I'm used to paying over the odds for new stuff, but paying for a 8800gtx and experiencing the borked drivers was the first time i truly felt gypped by a company.

[Anonymous Coward]

careful of statistics

They are almost always misleading.

An important reason nVidia has more crashes because they sell more stuff.

[Gerry]

@james

I don't think I understand your reasoning.

Surely the problem is the lack of modularity in the system, with unintended interactions giving rise to unforeseen consequences?

I'm sure things get fixed eventually, but...

Whatever the "this and that" of *nix - graphics card drivers don't get anywhere near the kernel as they interact with the communication layer. Things can be developed separately and put together successfully.

[Peter Kay]

Sounds accurate to me

Nvidia's Vista x64 drivers are a) shit and b) lacking in functionality as compared to XP. They don't handle either non plug and play monitors, or suspend/resume, and let's not even get onto SLI and various broken transparency APIs (you'd think they could get that right).

The other surprise is that Creative aren't in the list. It's taken a while for the Audigy and X-Fi drivers to mature, and they're still not perfect.

[Shagbert]

who's /= whose

Bad Register, no cookie for you!

[Andy Turner]

@AC

"An important reason nVidia has more crashes because they sell more stuff". Good point if you're comparing them against ATI, but *every* user of those Vista drivers is also using Vista too and in those cases, the figures are awful for nVidia, especially when you bear in mind what a small amount of the running binaries the video driver represents, compared to that which Microsoft has provided.

[Gordon Fecyk]

Isn't this what MS HWQL is for?

Here's a clue: Making drivers open source doesn't stop people from writing crappy drivers.

It's also nice to see someone blaming everyone else this time. I might have not touched everyone's driver code, but I haven't seen a crash I could attribute to a Microsoft driver since the release of Windows 2000.

But what really bugs me, is the MS HWQL. Isn't the Hardware Quality Lab supposed to test these things? Or doesn't anyone want to bother paying Microsoft to test and sign their code, knowing full well they'll find something wrong with it? Or maybe there's a "HWQL Lite" service, that only covers a cursory check of the device and doesn't perform serious "regression" testing, or whatever it's called.

And no, putting an infinite number of open source monkeys on a piece of driver code isn't the same as testing.

[Tez McArt]

It's not just me then...

I've had a Alienware laptop that has two SLI cards laptop running Vista for the last 11 months. Not only have Nvidia/Alienware been unable to produce a Vista Driver that has can use the SLI functionality, but the driver periodically crashes just as the desktop loads, and 100% crashes if you try and step down the power options.

I finally had enough few weeks back and started to look at how to slipstream raid drivers into an XP installation to dual boot the damn thing.

[Daniel Wilkie]

Statistics

Nvidia sell more cards than ATI perhaps, I'll give you that one though I haven't looked at it recently... But they don't have more cards installed on Vista than microsoft have copies of Vista that are crashing. I think the key is that, perhaps for once, its not 100% fair to blame all the crashes on MS and Vista, if the majority of reported crashes are caused by a 3rd party?

Mine's the one made from Nomex, I think I'll be needing it...

[Cameron Colley]

I'm not surprised either.

nVidia drivers seem to be a cause of crashes on my system -- and I use kmythbuntu!

[Anonymous Coward]

WHQL vs Reference Drivers

I wonder how many of the driver problems were due to people using beta or reference drivers. I for one try to get the most bleeding edge drivers for my video cards but I know full well what I'm getting myself into. If I wanted more stable drivers, but with fewer bells and whistles, I'd use the ones MS feeds me as part of Auto Updates. Those are typically all MS WHQL certified.

[Highlander]

Not that I'm sticking up for Nvidia....

But, I do remember reading several reports on the topic of Nvidia and Vista that suggested that Microsoft's changing goal posts effectively yanked the carpet right out from under Nvidia. The result was that the drivers that were apparently working well(ish) had to be redone, quickly, and much unreliability followed.

Some of the documents that have already surfaces in the Vista capable lawsuit paint a picture that sheds some light on the problems that Nvidia and others have had with their drivers. According to internal presentations from Dell "Late OS code changes broke drivers and applications, forcing key commodities to miss launch or limp out with issues".

http://www.nytimes.com/idg/IDG_002570DE00740E18002573FE006B7266.html?ref=technology

So Microsoft made changes to the driver model just before launch and then blamed everyone else for the resulting failures? Smart.

[Paul Crawford]

Why different from XP?

The real question I don't see discussed is why there are such problems with Vista, compared to the wide (and presumably moderately reliable) drivers for XP?

Is this anything to do with monkeying around to satisfy the DRM requirements?

http://www.cs.auckland.ac.nz/~pgut001/pubs/vista_cost.html

Is it really the case that the video card manufacturers can't write a proper kernel mode driver, or is just that Vista is too complex and/or badly documented?

[Patrick O'Reilly]

Busy doing other things.

They must be spending all their time making their drivers work like a dream on linux boxes.

[Edward Barrow]

@ a j stiles

You are absolutely right, but it shouldn't just apply to drivers. Copyright in all code should be conditional on publishing the source code. You don't get a patent without publishing how the invention works. There is in all intellectual property an inherent bargain: publish, for the good of society, and in return society gives you a restricted monopoly. Patent law makes it explicit; the fact that we've lost it in software copyright for a whole host of historical reasons has resulted in lots of monopoly abuses, broken drivers and worse.

[Don Mitchell]

Parity Errors, Modularity

There was also a study some years ago that showed that memory parity errors were a bit cause of system crashes -- one reason servers use ECC. Is that hidden in the "other" and "unknown" slices?

@gerry, Windows has always been modular. Microsoft was using binary replacable DLL's and COM interfaces long before Unix. The Windows device driver interface (DDI) was more or less copied by Linux.

[Syren Baran]

@Andy Turner

Argueing about the amount of binaries delivered by Microsoft and Nvidia is´nt a good measure either. Probably only code in ring 0 and 1 can cause fatal crashes, thats kernel code and drivers. Given that graphics drivers are the biggest and most complicated drivers i can well understand that a big portion of the crashes gets attributed to NVidia. On the other hand, if Microsofts portion of the crashes were even bigger that would really awfull, after all they wrote it.

[Peter Kay]

WHQL is worthless

They may well test basic functionality (and I do mean *basic*), but complex or edge cases are obviously not tested, because they still fail (I take the old fashioned view, that a quality certified product should *never fail*)

About the only thing I'd trust WHQL to do, is word processing, on a system with one graphics card and one monitor.

[kain preacher]

SHocked

No windows bashing

[Henry Wertz]

Driver errors..

"I'd suspect that these driver errors were caused by the developers trying to do too much, too fast with an overly complicated operating system. Possibly with inadequate documentation and information."

From what I've read, the REAL reason ATI and NVidia have so many crashes is Microsoft refused to ship them a Vista version without video-related rights restrictions. It's very hard to write a video driver, and it's just that much harder when there's hostile code in place actively preventing running a debugger.. looking for unusual behavior and killing the driver preemptively ("tilt bits", designed to try to keep the user from playing a rights-restricted video, then just copying the video back out of video RAM).. and so on. I'm surprised they got it to work at all.

@Gordon Fecyk

"Here's a clue: Making drivers open source doesn't stop people from writing crappy drivers."

Nope but, in the case of some VERY poor Linux drivers that have been released, driver geeks do work the poor drivers over. I've seen several drivers that were quite poor but open sourced get worked over into fully functional and well behaved within literally months.

However, I don't think they should have to release source. I do urge them to release card specifications though. (Which, now that AMD bought ATI, they are releasing the specs. Here's looking at you, nvidia.)

[Anonymous Coward]

@A J Stiles - are you serious?

I think that had to be one of the most out-of-touch-with-reality posts I've ever seen on The Register - and to be fair, the competition is pretty stiff.

I'm not sure if you *really* believe that the combined governments of the world are going to put aside their differences, temporarily ignore some of their other pressing issues (like war, AIDs and famine) and somehow bring in a Utopian new world of open-source driver software.

Perhaps it was meant to be ironic, in which case I wholeheartedly apologise. If not, I can only say that you have the political nous of Michael Jackson or a Miss World contestant.

[Colin Wilson]

Microsoft certified drivers

After making manufacturers pay to get their drivers certified, it's a shame Microsoft can't be held responsible to replace hardware for which faulty certified drivers exist - the code is essentially being released in their name (given that the manufacturer had to pay a fee for certification / permission to install on Fister in the first place).

Thinking further afield, since Microsoft or other media copyright "holders" have the facility to switch off hardware that might be breached via faulty drivers and (god forbid) allow a movie to be watched or copied without DRM - shouldn't Microsoft be liable to replace this hardware since it's their certification process that failed ?

[Anonymous Bastard]

Percentages

"Pretty much everyone who writes driver software is present on the list but after Intel, the percentages very rapidly tail off to below one percentage point."

1% of all vista crashes is still more than I would like to count.

Were the network/copying slowdowns logged as crashes? This is one of the most widely reported bugs, entirely microsoft's fault and likely is excluded from those stats.

[kain preacher]

@ac

"Were the network/copying slowdowns logged as crashes? This is one of the most widely reported bugs, entirely microsoft's fault and likely is excluded from those stats."

A slow down is not a crash. SO why would they be on that list. IF the list was about stuff that impacts the OS usibilty , fine then count slow downs.

Please explain how a slow down is the same as a crash.

[C Caulfield]

Zune , anyone ?

Surely a large part of this driver problem is that hardware manufacturers finally had their trust in Microsoft destroyed by the Zune debacle .

The MS emails reveal that few manufacturers believed that MS would ship an unfinished , untested OS . But then they're trying to ship an unfinished , untested file format "standard" at the moment .

In the case of Vista , MS has paid the price of it's grossly uncooperative behaviour (hidden hooks and deliberately incomplete documentation) .

In the case of Oxmeal we're all paying the price of it's grossly uncooperative behaviour (hidden hooks and deliberately incomplete documentation) .

Unfortunately they've also wrecked the world's standard setting mechanism at the same time .

As others have pointed out , blaming others having taken their money and given them certficates is pretty low behaviour .

[Paul]

They f&*cked the XP drivers as well

I wouldn't care how bad Vista support is if the fuckwits at Nvidia hadn't backported all the stupid restrictions into the XP drivers. Video overlay support Vista apparently insists on restricting (ATI didn't seem to have a problem) has now spread restrictions into XP drivers - upgrading XP drivers removes existing features!

The working XP style control panel is no more, replacing by the Fisher Price Vista incarnation, along the way losing half its control options and what's left usually doesn't work. Want to save a colour setup, get some pen and paper, the drivers won't remember it for you.

I've been hit hard by the incredible decision that monitors will always be autodetected as the driver starts. Autodetected even if I try to override detection (the option just doesn't work). Even better, after failing to find my second monitor hiding behind its monitor switch all related options are removed from the fscking panel so I can't manually enable it.

Braindead engineers following braindead DRM orders from braindead Redmond monopolists. Is it any surprise they're responsible for so many crashes?

(And maybe some decade soon NVidia will admit their NIC hardware doesn't work and do what they did with all other mboard drivers - disable the hardware acceleration and fallback to software).

[yeah, right.]

wowsers

Imagine that, Microsoft blaming everyone but themselves for creating an "operating system" that isn't robust enough to make sure the drivers do what they're supposed to do, and no more.

Maybe if they spent less time trying to buy votes for their OOXML debacle (or creating code that deliberately stops the machine from working if you try to run DR-DOS, or maliciously screwing Netscape, or any one of the hundreds if not thousands of shenanigans they've tried to pull off over the years) and more time actually designing something that was a bit more secure they wouldn't have nearly as many problems.

So, is it really Nvidia's fault? Or is it Microsoft that can't provide reasonable and accurate documentation for their O/S, because they're too busy trying to sodomize all potential partners as they desperately try to maintain their fraudulently and even illegally obtained world domination?

[Christopher A Light]

M$ - insecure certificate checking? Or simply grabbing the cash?

What I'd really like to know is: how many of the drivers responsible for the crashes in question were ones that DIDN'T throw up the 'This software isn't signed by Microsoft - continue (Y/N)?' dialogue on install?

Because my recent unhappy experiences with both Vista and XP64 lead me to believe that there are a lot of drivers (latest versions, downloaded from the manufacturers websites) which install happily, with no indication that the driver ISN'T M$ Certificated, which either don't work (Huh? But I've installed the RAID driver - how come no RAID?), or render system in question unstable - or better still, slow it unacceptably?

So, if the software isn't certified, why doesn't a warning appear?

And if it IS certificated, why doesn't it function correctly?

What exactly is going on here? Are manufacturers managing to disable the 'unsigned' warning built into the OS? Or are M$ simply taking the money for 'testing' the software and rubber-stamping the certification?

[DS]

Yes, well, was it not..

a shock to find that Vista, which supposedly changes the GFX driver model wholly to move in the direction of stable operation, hence the drop in performance in Vista Gaming for example, in fact has not resulted in the end product.

You could forgive Vista if GFX operations really became bombproof, but they have not. Worse, it seems to have made writing a driver a very tricky job, the framerates and performance are lower.

Another huge black mark/failure when getting down to discussing wether Vista really is a step forward.

Microsoft, again, what was the point of Vista....

[Mark Honman]

@ Don Mitchell

1) It's interesting that replacing NT with Win2K caused a significant decrease in memory parity errors.

2) Windows is so modular that graphics drivers have to run in Ring 0?

That registry corruption results in an unbootable system?

That you can't swap out a motherboard (to different chipset) and it will figure out the right drivers for itself?

That you can't run it without IE?

That until last year you couldn't run it without a GUI?

3) In 1990 Unix had not one but two RPC standards - ONC-RPC (originated by Sun) and DCE-RPC (originated by Sun's then-big competitor Apollo). At that time the best Microsoft could do was Windows 2.11 for '386, which certainly caused a lot of memory parity errors.

FWIW I consider both Windows and Linux to be second-tier operating systems and I'm waiting for the day that they are as stable as the HP and Apollo Unixes were in the early 90s.Sorry I can't offer comment on the other leading Unixes - SunOS (and later Solaris), IRIX, and AIX as I had no practical experience with those.

[A J Stiles]

@ Edward Barrow

Making the legal protection of copyright contingent on publication of Source Code is an excellent idea; I hadn't thought of it exactly like that, but you're spot-on.

I'd like to see it made law eventually that *all* software be published in Source Code form, even if users are not permitted to pass around copies. (That's how it always used to be; and anyway, concealing the Source Code to Windows never prevented anyone from making unauthorised copies .....) I said driver code *first* because it's been demonstrated clearly that having driver Source Code available is beneficial, but I see no reason to stop there.

[mike]

im glad nvidia is finally catching some hell over their drivers

I've had quite a few issues over the past year with Vista Business and an 8800GTS card. Lockups, reduced driver functionality(where'd my old and powerful nvidia control panel go?-even in xp), along with the expected 'new game crashes' that require a driver patched up for the game(Crysis, COD:4, ...).

[Stan]

linux drivers rock solid here

Don't know whats going on in vista but the nvidia drivers (64bit linux and a 8600gt) are working very well for me here, and I don't hear many XP users complaining either.

In contrary to a post above, all (including linux) video drivers do have direct access to the kernel. "X% of crashes due to driver Y" doesn't say a whole lot, is the kernel is issuing a request that the driver doesn't understand or is the driver asking for something the kernel doesn't understand? And in both those cases is the blame being put on the driver? f..ing statistics.

cheers

[Tom]

A little spin needed here.

Whilst I have so far managed to avoid Vista, I have been using Nvidia cards on XP for years (not a conscious decision, just the way things worked out).

I have never had a problem (at least, that I am aware of) with Nvidia drivers.

Therefore it seems to me that, rather than Nvidia drivers breaking Vista, it is the other way around.

[Dan Paul]

It is "Windows Offender" that is the greater problem!

MS provided WHQL drivers SUCK since they RARELY get updated and there never was a problem with XP and Nvidia reference drivers UNLESS you are talking about a laptop not a desktop. Get your laptop hardware drivers ONLY from the laptop manufacturer!!!!!!!!!!!

The real answer is that Windows Defender disables drivers and programs that it doesn't like (Or those that conflict with MS profit motives).

Here is my scenario: Let's say the new driver gets installed (after incessantly pressing "Yes" I want to frikkin continue" but "Offender" disables it but does not roll back to the old driver completely. There you have it, your driver is frakked and so are you.

SLI is another issue entirely as there are NO INTEL chipsets that actually support SLI on Vista only ATI crossfire and not well at that.

nForce Nvidia chipsets will support Nvidia SLI just fine but there is a problem with Vista's "Offender" preventing MANY chipset drivers from properly installing especially the newer ones.

MS needs to re-think Vista, strip out the worst "Nanny" bits and much of this problem would likely go away.

Then they need to HIRE people from Nvidia and ATI to work with MS developers to fix the frikking graphics driver problems.

Hey if you can't even change or update a driver without crashing or requiring a re-installation, what is the point of "playing" with your computer as many of us Reg-Heads do?

MS, here is a clue... I will do what I want with MY computer, when I want to and you have no right to mess with the programs or drivers I want to install or test.

Do you even understand this, Mr. Bill? I know "Ball-More" doesn't but YOU should.

[Tam Lin]

132% of Microsoft's statistics are wrong

This is really suspect. After encountering half a dozen servers and PCs with boot-up BSOD's attributed to non-identifiable <cough> driver bugs </cough> in the last couple days, I seriously doubt anyone at Microsoft has the skill and work ethic run a trace to find out which DL^Hriver was really at fault. It must be hard, as none of their 10,000 programmers can write trace code[1]. I'd guess it was some of the 20,000 Microsoft marketeers that pegged nVidia for business reasons. To what end I know not.

[1] I know the code's design is so bad that this is nearly true, but Microsoft has taken 4 or 5 opportunities to fix the driver model(s) - and blown every single chance.

[A J Stiles]

nVidia on Linux

For the record, I have been using the open-source "nv" driver with any nVidia cards that I have been unfortunate enough not to be able to avoid using, and this works fine.

Oh, and Evil Graham, I'm not suggesting multilateral action -- I've seen enough of life to know that as you put more people in one place, the number of disagreements increases quadratically. Rather, I'm suggesting several independent unilateral actions.

At first, there will simply be a substantial grey import market, from countries without such a restriction into countries where manufacturers take their ball and go home. But as binary-driver-only bans increase and customs enforcement improves, it will become harder and harder for manufacturers to sell a product without driver Source Code, and they will eventually be forced to comply in order to maintain a market.

There is also still the possibility that some third party could successfully develop a genuine alternative driver: this would place them in the situation of being able -legally- to supply hardware and drivers together to countries where driver Source Code was required by law. Without the risk of confiscation, such an operation is bound to be more profitable than a simple illegal importation operation. The only way for manufacturers to avoid people riding on their backs in this way would be to ship their own drivers in Source Code form.

[Dick Emery]

Aha! An opportunity!

Since elreg decided to ignore my email about Nvidia's 8 series lacking video mirror which is in all previous versions of their cards/drivers I will post it here instead.

Yes. The 8 series Nvidia cards are unable to perform video mirroring. That is. The ability to output FULL-SCREEN video from a window (Minimized or otherwise) to a TV output using either Composite, S-Video or Component.

This is an essential ability for those that using video editing tools like Adobe Premiere or DJ's who wish to display FULL-SCREEN video on a projector/TV and still be able to work with the video windowed on the desktop.

Nvidia have stated that 'this is by DESIGN'. So they are freely admitting that it is disabled on the 8 series drivers.

Now why could this be?

Let us look now. Could it be because of Vista? If so then why does it not work on XP? Could it be because the 8 series cards are not capable of it or they made a f*ck up during manufacturing of their chipset? Perhaps. Or more likely could it be that they do not wish for people to be able to output HD material over the 'UNPROTECTED' analog Component output? Hmm. My bet is on the last one.

Elreg. Why are you not reporting on this item? Why has Nvidia been able to hide this information from the public which has been known about for some months now?

Answers on the back of a postcard.

[Anonymous Coward]

@A J Stiles - really?

I probably wasn't specific enough. Not only do I think your suggestion is unworkable because of the zero chance of getting worldwide cooperation, I don't even agree with it as a concept.

You need to explain what is so special about software that requires companies to give up their right to keep secrets. The GPL (or whatever licence you choose) is great, provided that the licence is really your free choice and not imposed on you by the new Stilesian World Order. Real freedom means that if I want to sell some hardware and keep the driver code closed, then I can. Same as you can choose whether or not to buy it.

To extrapolate your argument logically, as soon as the driver code was opened you'd have to also open all the OS code as well, because (as anyone who has debugged anything will tell you) you only have half the story if you have an open driver and a closed OS. Then you'd also have to open all the apps, because how do you know what they are doing to your driver / OS combination? Obviously you need all the hardware specs for the card, because you can't write software for it unless you have that. Keep going with it and you'll have to have the source code to the engine management system of the car you drove to work in, plus the secret recipe for the Coke you bought to keep you going through your debugging session.

Even if you got your wish and all the source code was magically made available to all the drivers in the world, who's going to do all this debugging? When I get home from PC World with my latest Nvidia card and a copy of Quake 9, how am I going to fix a system hang 3 hours into an online multiplayer frag fest? An Nvidia G8800 GTX has 680 million transistors, how many million lines of code do you think it takes to keep it going in Vista? How many million lines are in Vista, talking back to it? Who's going to pay your armies of coders while they debug this stuff? Who's going to make the next Nvidia card, now that they've gone bust due to cheap Chinese copies (we opened all the specs, remember)?

There is a whole bunch of practical things you *could* do to improve the situation - consumer law could protect you better from MS and Nvidia's mess for a start - but I'm afraid your "plan" sounds like the sort of manifesto more normally seen around sixth form common rooms.

[Charles Manning]

Why I still blame MS

The new Vista display driver model (http://msdn2.microsoft.com/en-us/library/aa480220.aspx) is far more complex than anything before and was stabilised far too late. It's hardly suprising that these drivers were rushed out the door with inadequate testing. MS should have stabilised the display drivers far earlier in the development (a point that Dell raised IIRC).

There is also another missing data point. "Driver crash" does not necessarily mean that the driver was broken. Perhaps the graphics part of the OS (ie. MS's contribution) was calling functions with parameters that did not agree with what was documented. That sort of thing happens quite a bit.

[Aubry Thonon]

"Crash Cause"?

Hmmm... I wonder how they decide what *caused* the crash. Do they simply look at which program stopped running at the time and go "here's your problem"?

If so, I'd like to introduce them to a little concept known as "segmentation violation". It can be a pain to fix: That's when program A (illegally) shits all over program B's memory, causing Program B to crash when trying to access its stores. But the cause of the crash is not poor program B, it's program A.

So, depending on how scrupulous MS is about finding the *cause* of the problem, it might just be that NVidia drivers are more sensitive (perhaps because of size issues) to some sort of "behind the scene" corruption by an as-yet-unknown source. Or maybe (for example) they more often access MS functions which leak memory like sieves.

When find the fault for a crash, be very careful of the arrangement of horses and carts.

[jolly]

@Don Mitchell

Couldn't agree more. A while ago I built my own [WinXP] PC and it crashed with BSOD on a regular basis. BSOD always named the same DLL (a network adapter driver if i remember correctly) but the cause was actually faulty memory. Caused a few data corruptions too :-(

Prior to building the above PC I bought a Dell precision workstation (about 9 years ago) for my main business PC and this was fitted with ECC RAM. It was expensive but I can't remember ever having the OS on this machine crash since the day it arrived [NT4, then Win2K, then WinXP]. Don't think I'll ever buy a non-ECC machine for professional use after seeing what can result.

[Tony Paulazzo]

Unknown caused by sunspots?

Surely unknown should come under MS as well which would give them 34% of the crashes. When I got my laptop with Vista preinstalled (+ Nvidia 6150), I had error reporting on which turned out to be a colossal waste of time as each and every problem came back as unknown - except for a few or so graphic driver crashes which never resurfaced after I switched off transparency (never turned it back on as I tend to only work in maxed windows anyway), and stopped using the sleep function (tho with that I think it was more down to readyboost and sleep not working well together - leastways the laptop seemed to successfully wake when readyboost was disabled.

Plus, you could put the fastest graphics card in the world in a system, but if the other components aren't up to the task, poor PSU, low RAM, crap CPU; then load AVI backgrounds, transparent windows, 3D task switchin - and finally mix in DRM, beta drivers, brand new OS... and, well I'm no scientist, but I'm fairly sure the result is BANG!

And finally, aren't the numbers suspect anyway? Intel, AMD, ATI, Nvidia, Others and unknown, didn't they crash Vista, or cause Vista to crash? and shouldn't Vista stop the crash from occurring? being as it's the operating system.

Tony F Paulazzo.

[Rhiakath Flanders]

In defense of nvidia

As my title says, i dont think it's nvidia's fault. Basically, as was already said, Microsoft delayed the Vista launch a bunch of times, always rewriting something. Second, they were constantly changing it's driver model, which gives nvidia and other hardware vendors a reallly bad time.

Third.. Who made the statistics? Who's got the data? Microsoft! Who has the OS that connects all the drivers and runs them? Who is responsible for managing everything? Microsoft! They changed their architecture radically (or so they say), and they want everyone else to just code THEIR way? Come on!

If a windows error popup dialog comes out, and says that nvidia drivers borked the system, why should it actually be that way? can't it be that the OS doesn't know how to handle the situation, and then just says that the drivers performed an unknown or unauthorized action?

I've seen it happen a gazillion times. Besides. I use Linux, Vista, and XP. Love Linux, like XP a lot, but damn, i HATE vista.

I don't think Nvidia or the rest of the companies are doing a good job. They're working with what Microsoft gave them. A super buggy OS, and they've never even once acknowledged what most of us know... That Vista is a resource hogger with more holes in it that a fishing net.

Vista, for me, is a death trap. Half the stuff i had working on XP doesn't work there. Most of them even work in Linux, under wine, or something similar...

[Rhiakath Flanders]

And...

One more thing... How many of you have had a nvidia driver crash on you? I figure, not many. If you found a bug in these, you would certainly submit it, as it is a piece of software which you're used to work correctly, and want it to keep that way...

Now, how many of you have had a windows crash, or bug? Everyone! How many times did you bother to send the error report? A lot! Did it come back solved on the next Service Pack? .... I doubt it! No-one ever sends the error reports anymore, i guess. We're all used to windows crashing. It's not a bug, it's a WELL-documented feature.

I do use NVidia, and will keep using it. Nothing against ATI, as i have also an ATI card on one of my laptops, and i used those for a VERY long time, early in time. (Mach64, anyone? or RagePro 2C)

I do know that Windows is a freaking dinosaur compared to a mosquito which would be the drivers, in what comes to code size. It's a full OS. But damn, with an unstable base as that, It's a miracle if the drivers work.

[Mark]

weighted?

Are the results weighted against how many Nvidia vs Ati machines there are out there?

If not, then any stats are worthless....

I seem to recall on SteamPowered.com, their hardware survey showed a much larger number of NVidia based machines out there. (I can't get to the site, as it's websensed...). Perhaps someone way want to share with us the corrected skew...

[heystoopid]

Hmm

Hmmm , given the crap cheap and nasty buggy low rent drivers from Creative Labs for their various indifferent sound cards not performing well under Vista , I would have thought they would have been on the top of the list !

[Kurt]

NVIDIA Driver Still crashes

The latest driver for my Nvidia card still crashes. I will never buy another Nvidia card again.

[Scott Earle]

The other 17% 'unknown'

A couple of people have mentioned Creative, and I have not seen a good driver escape their codemills for a good few years (Maybe Audigy was the last?). I hereby nominate them as a strong candidate for the title 'unknown', and am surprised at the Ars for not hinting at it themselves. 'heystoopid' said they should be 'top of the list' ... looks like they are a strong contender for sitting in fourth place :)

Can anyone remember the last *good* Creative product? I had an AWE64 Gold that was pretty damn good in its day. SoundBlaster Live! was pretty good, too.

[Aaron]

Think Different 

With so much Windows love going around, it might be time to get a new girl friend and give OSX a try. I've always marveled at how fast Steve gets his crew to turn software around. Why are software requirement specs for OSX always half of XP (not to mention Vista)? Must be the differences in OS overhead.

Give it a shot, guys... nothings perfect, but my productivity and happiness has gone way up since I dumped Microsoft (and my wallet's thicker too).

There, I said it... that was easy!

[Anonymous Coward]

@ Mark Honman

On your UNIX comments.... completely agree... however, you forgot to mention

 OSX 

After all, they are UNIX 03 certified. C'mon, give credit where do.

[Brian]

creative..

To all of you wondering why creative isn't on the list.. the audio stack was moved out of the kernel. An audio driver can't crash the system now.

[Peter Kay]

Nvidia/Vista crashes

I've had several - the driver *really* doesn't like you not using plug and play monitors, and Vista will report the driver errored and had to be restarted. I'm fairly certain the suspend/restore BSODs have been due to Vista, too.

Forget all this 'driver model wasn't stabilised' bollocks - Nvidia and others have had *well* over a year and a half to create a working driver. The real problem is that they've chosen to spend all the time getting an extra 2fps out of OblivCrysisLifeProBlaster v3, instead of really solid 2D, colour management, refresh rate control, multi monitor support and so on.

[Andy Tyzack]

statistics

are purely shit.

[Anonymous Coward]

@ Who's /= Whose

Who's = Who is

"Whose" is correct in the demonstrated context.

Nice try though.