>> "These comments (especially the third one!) seem to indicate that Linux
>> takes a back seat to driver development (to Microsoft Windows) at ATI."
>
> Well what do you expect? You want AMD to risk losing Windows certification
> for its Windows drivers, thereby hamstringing the single largest (by a
> huge margin) target audience of its consumer graphics hardware?
I didn't know that companies developing drivers for Linux can loose their Windows certification because of that! Why is that? Can you be more specific?
Posted Jan 6, 2009 11:06 UTC (Tue) by tialaramex (subscriber, #21167)
[Link]
The DRM (as in restrictions on using what you own, not the kernel DRI component) enforcement regime requires that the user be prevented from understanding how their own computer works.
I never looked into this very closely, but the present generation of DRM seems to expect that each hardware vendor (AMD in this case) will present the "trusted" DRM implementor, ie Microsoft, with secrets about their hardware which make it difficult to emulate correctly. The DRM implementor uses this to ensure that secret data is sent only via genuine hardware devices certified not to leak the user's data to the user, but rather ensure it can continue to be "protected" by DRM, ie extract monopoly rents indefinitely.
In theory if AMD reveals (even by accident) how their hardware works, Microsoft is obliged to provide a "bug fix" to Windows which disables the AMD hardware for viewing movies and other perfectly legal activities which need to be "protected" under DRM. This would be financially catastrophic for AMD of course. But in the present climate, even rumours that this would happen soon would probably be enough to send them into Chapter 11.
So they (and nVidia) are in the same poor state that IBM were in for a while, they've inadvertently created a situation where Microsoft gets to dictate how they run their business.
Not to mention that like most businesses, AMD will have made decisions when it was sure it was keeping this data secret that seem rash now that it wants to make it public. Sure, the hardware wasn't on /sale/ until 2007, but it was in development for 2-3 years prior. Suppose you're a senior AMD engineer and you know that one of the Z buffer speed-ups you used in the R7xx series is sort of broken in early silicon, its disabled in the proprietary drivers for that hardware and it has never been admitted to customers. Well, leaving an entry in the documentation "0x08 - enable Z-Foo - broken on 3xxx to r4.3" is a good way to cause customer disatisfaction and get yourself fired, so you're going to want to at least change it to "0x08 - reserved" or "0x08 - Z speedup, reserved" or something.
The most important reason is simply this: This isn't top priority (and since AMD is a for-profit business, it probably shouldn't be) and so a lot of the time the process was stalled waiting for this or that person to get a free hour to review the documentation.