I've already written a lot about this in the other thread, but as I said, many of the original articles had contributions and quotes backing the position NV's crashes were a result of higher market share in segments more susceptible to crashes at Vista launch:
From the original CRN article:
Nvidia (NSDQ:NVDA) appeared to be having shocking problems with Vista crashes as soon as the OS was released. Nvidia's own support forums were full of frantic users whose systems had turned from being rock solid on XP to flaky as soon as Vista was installed.
That argument has a supporter in Joe Toste, VP of marketing at Equus Computer Systems in Minneapolis, Minn. Toste's take on Microsoft (NSDQ:MSFT)'s crash numbers was that enthusiast systems with top-of-the-line GPUs were likely causing a disproportionate amount of driver crashes.
"If you look at Nvidia's dominance of the high-end, well that's where these bugs occur. The instability of those drivers is in the high-end dual SLI. And Nvidia has 90 percent of that market," Toste said.
There's tons of data to back these claims up, as "simple market share" data simply isn't getting it done. Even the quoted market share from Xbit paints a very different picture where it counts:
On the desktop Nvidia was the clear winner, claiming 43% against Intel?s 38.5%, while AMD had a modest gain to 23%. But in the mobile market Advanced Micro Devices saw a decline to 21%, Intel held its dominate position and grew slightly to 51.5%, with Nvidia number two at 27%.
A few other points to keep in mind about this crash data:
- 1) The MS Crash chart does not distinguish between video drivers and other product drivers. There are also situations where a crash will report a certain driver but the fix might be completely unrelated to the driver or part itself (memory/power problems were known to cause TDR errors). While I'm a big proponent of NV video products, I'm much less confident in their chipsets and core logic.
2) The MS Crash chart does not specify a time period, but clearly the authors are focusing on launch and the 4-6 months period after. They also do not specify what type of crashes were logged, whether they were appcrashes (CTD/TDR) or kernel crashes (BSOD). There's also no info about frequency of crashes or number of crashes from unique samples. This would obviously help identify whether the problem was endemic of all NV parts, in which case a simple market share analysis would work, or whether it was a small % of machines causing a disproportionate number of crashes (as hinted in the articles and market data).
3) ATI at this point had *NO* DX10 parts in the market in this time frame. They did not have any mainstream or enthusiast parts in the market until June, and as the Steam survey shows, even now, have very little market penetration in the gaming/enthusiast market with their latest generation of parts. The CRN article does note that users perceived an increase in ATI driver crashes in the June/July time frame.....coinciding with their DX10 part launches.
4) The crash chart does not specify DX10 parts as the main reason for excessive NV crashes, however common sense does. New machines sold with Vista would most likely be outfitted with newer DX10 parts. Gamers with the most up-to-date hardware would also be *most* likely to be early-adopters of Vista for some of its enhanced features over XP, most notably DX10. In both cases the only game in town for DX10 was NV. DailyTech, Ars, and CRN all hint at this with various quotes. While DX10 applications at the time were rare/insignificant, DX10 was and still is a significant check-box feature for forward-looking enthusiasts most likely to upgrade to new hardware and a new OS (this is substantiated in Steam Survey results).
5) Microsoft drastically changed the video stack in Vista, making numerous changes to the kernel most notably removing the driver from kernel space to reduce crashes. This had the desired effect of reducing BSODs at the expense of increasing TDRs and appcrashes. Overall I've found this to be a good change, however, MS released at least 2 major hot fixes to address problems with the video stack, most notably a virtual memory allocation hot fix and the reliability/compatibility update for TDRs. These hot fixes also coincide with the periods NV and ATI stabilized their drivers....sometime after June/July, or about 6 months after Vista's launch. While I'm sure NV didn't have perfect drivers at launch, clearly MS had a significant hand in the video driver problems, as their hot fixes largely undid the offending changes to the video stack in Vista.[/li]
In any case, I could write more about the topic but there's enough data out there to see the MS Crash data doesn't mean much at all by itself. If anything NV was a victim of their own good fortune, delivering a timely and wildly popular product that left the competition without an answer.
I did a lot of number crunching using the Steam Survey in the other thread to back up some of my claims, and while its not perfect, its 1.5 million samples that are just about the best market data you'll find for free besides the "simple market data" available around the web. You do need to keep in mind when viewing the Steam data its 1) machines used for gaming and 2) many of my conclusions are based on the belief that many Vista upgraders would be gaming machines and 3) applications most likely to cause video driver crashes are 3D games. In any case, if you're interested in video cards and gaming, you might want to take a look at the
Steam Survey even if you don't agree with anything I've written.
There's some
really great info in there that may help support/debunk some common beliefs/misconceptions out there. Like the high % of gamers out there with 2GB+ in their rigs. Gaming resolution still has a lot of 1280x960 users. Most gamers are running at least dual-core CPUs. AMD and Intel are very close in CPU share. Intel's crappy chipsets might be able to run Source, but no one
actually running Source on Intel's chips (2.61%).
Here's the breakdowns I posted in the other thread, with highlighted relevant info for easier viewing:
2:1 is for all systems polled, including XP which is still the overwhelming majority of all systems at
~82% XP compared to ~17% for Vista (32/64 for both).
9.20% of that 17% run DX10 parts or
54.1% of Vista machines run DX10 cards.
8.05% of that 9.20% are NV parts or
87.5% of those DX10 cards in Vista are NV parts.
If you apply the 60/30 split to the remaining non-DX10 Vista parts, you'll get 16.96 (All Vista users) - 9.20 (All Vista+DX10) = 7.76%(x.60) =~4.65 Non-DX10 NV parts.
Add that 4.65 to the 8.05 NV DX10 parts= 12.70 (All NV Vista) and divide by 16.96 (All Vista) =
74.9% of all Vista systems used for gaming are running NV cards. So yes, this definitely shows that if 75% of Vista Gaming Systems are running NV hardware, 30% failure rate is par for the course and is certainly better than any other FUD offered as evidence to the contrary.
59,907 out of 140,011 or
42.78% of Geforce 8800 parts are run in Vista. Now Compare that % to the total breakout again of XP and Vista of 82 and 17% respectively. Ed: This substantiates the claim high-end gamers are more likely to be early Vista adopers.
Keep in mind, this survey accounts for ATI DX10 parts sold to date where there were none available at Vista's launch and for most of the scrutinized article periods. Yes NV's total increased but their market share at the time would be much higher also (with no competition). If anyone has a Steam Survey link for the January-July 2007 time frame that'd also be interesting for comparison purposes.
Facebook
Twitter