Quote:Original post by Yann L
It's more about DX10.1 (or will be). Means pre-resolve access to individual multisamples, cubemap texture arrays, etc. But these will probably be quickly added as extensions. On NV, that is...

Except they won't because NV doesn't have a DX10.1 part, they only have is a massive DX10.0 part where as AMD are on their 2nd generation of DX10.1 parts (the HD3xx0 being the first generation); and right now, for the majority of people AMD's HD4xx0 series is looking like a better bet. The HD4850 is a very good card for it's price, performing as well or better than NV new GT2600 part. While I doubt the HD4870 will take any top performance crowns if it is properly priced and clocked it could end up being a good high-mid card indeed.

What has this got todo with anything? Well, ATI were never a massive OpenGL driver supporter and while under AMD they have made some improvements there is still (afaik) a lack of DX10.0 support via extensions.. hell, they haven't even exposed their tessalator yet! So, if this continues and AMD claws back market share in the midrange again (which imo they are well placed to do with this series) you can kiss goodbye to 'DX10 on XP via OpenGL'.

Hell, I'm not even thinking about the idea of the Mt. Evans update to OpenGL3.0; if they continue their current level of uselessness it won't be upon us until NEXT SIGGRAPH in 2009, at which point Windows 7 is lurking around the corner...

Quote:Original post by Yann L

Quote:Original post by phantom
Except they won't because NV doesn't have a DX10.1 part, they only have is a massive DX10.0 part where as AMD are on their 2nd generation of DX10.1 parts (the HD3xx0 being the first generation); and right now, for the majority of people AMD's HD4xx0 series is looking like a better bet. The HD4850 is a very good card for it's price, performing as well or better than NV new GT2600 part. While I doubt the HD4870 will take any top performance crowns if it is properly priced and clocked it could end up being a good high-mid card indeed.

As long as NV doesn't support DX10.1, nobody is really going to use it anyway for anything production level. NV is a much too powerful global player on the market.

Ditto, 10.1 isn't going to be common place until Nvidia has a 10.1 card out. Game developers aren't going to code their game for the highest common denominator, they will target DX10 hardware for now even with DX10.1 class hardware being available, due to the sheer volume of DX10 cards is far greater for now. And Dx10 is a far cry from being main stream as DX9 is still king of the hill. Why, consoles are stuck in DX9 land for years to come still, and game companies are targeting the PS3/360 PC gamers are going to get screwed for GFX tech until the next generation of consoles come out.

[Edited by - MARS_999 on June 21, 2008 1:07:38 PM]

Quote:Original post by Krohm
Not conceptually, but may actually contain ring-0 instructions in some operating systems, thus making it kernel in real cases at the end of the time. If it has a single ring-0 instruction (even in unreachable code) then I won't consider it userspace. The notion of OS evolves with time, you can pretend to still apply defintions from the previous century but would you - for example - consider 'safe' a car without ABS? A car nowadays must have ABS, although nobody thinks it's strictly necessary.

Considering a few modern operating systems work this way then yes I still consider it a valid description. Do you not make a distinction between services run by the system and the intrinsic system itself? For instance I wouldn't consider Cron to be part of the operating system, yet it is launched by the operating system at boot time. Subtle difference.