Personally I've had problems with both, but in actuality nVidia for sure leads the camp with both releasing drivers that kill cards (remember the fan problems?)
And also releasing drivers which blow up the power circuit of their cards...
I think you'll find that the driver that "killed cards" was actually a single driver (196.75) that could in some instances bork the fan profile. The driver in question was removed within three days and superseded 2 days later by 196.78. The power circuit issue was the initial driver allied with incorrectly set max vCore in BIOS for a single card- the GTX 590- an issue that generally presented itself when the card was overclocked.
So, you've highlighted two drivers in how many Detonator and Forceware revisions to prove a point regarding reliability ? This differs from some Nv***** cherry picking and constructing an argument from a couple of AMD's faux pas how? IIRC, I think you'll also find that many people intentionally used the driver and/or claimed a dead card to trade up on their warranty. There were a great many forum posters looking for the driver once it was taken down from official download- the posts are easy enough to find
As a counterpoint, I'd bring your attention to two cards whose failure rates are known to be very high due to cooling issues of both GPU and VRM stages. Inadequate cooling and a bad design gave the HD 3870 X2 and HD 4870 X2 (as well as the reference HD 4890, and 9800GX2) an unwanted reputation for a short life span
It should also be noted that even with the high failure rates and very high local temperatures, none of the reference designs were altered during their respective production runs.
And yet still nVidia hardware & drivers are praised as the mother of all, when in actuality they are the company responsible for the death of millions of G8x & G9x GPU's in laptops & desktops "with inaduquate cooling"; which means if you don't watercool the things are gonna fail in large percentages.
I think you'll find that most enthusiasts (as opposed to randoms braying at the moon) base their opinion upon the hardware at hand, and on merit...or otherwise. Making a decision based on underfill failures in a series of cards that have been EOL'ed for 5+ years makes no more sense than boycotting AMD cards for the PowerPlay issue (Grey Screen of Death) that plagued the initial 7-8 months of the Evergreen series launch.
The GPU's affected by the high-lead solder underfill were based upon the G86/84/73/72/72M, and MCP67 / C51 onboard chipsets - I.e. single slot cards and mobile. Nvidia and TSMC transitioned to eutectic solder before all the G9x arrived excepting an early tranche of G92. The only part that could have been affected- 8800 GTS seems unaffected because of the effective dual slot cooler the card shipped with.
I'd also note that the issue of heating/cooling expansion and contraction cycles on solder underfill wasn't a greatly explored topic until the wholesale failures of the G84/G86 came to light (ATI's similar issue was happening concurrently)- and while the widespread failures bought a lot of attention to the Nvidia issue, it isn't by far the only one (ATI were only saved the same PC graphics issue because their high-lead solder issue occurred with their Xenos chip on the Xbox 360 causing the RRoD) although Nvidia and the OEMs in particular could have handled the issue much better. I don't think it was until the underfill issue and the
more widespread use of thermographs for monitoring graphics card local heat build up and dissipation that anyone really started taking serious notice - hence the advent of proprietary (and often, open shroud) cooling and the almost complete lack of single slot cooling cards at anything above entry level performance.
