Your New Year Wish

Reverend

Banned
So, taking into account how 3D technology that will be offered by the IHVs will more or less be "governed" by Vista, what are the things you wish for?

Besides more speed, of course.
 
That we'll be able to buy top-end DX10 cards from both IHV's no later than August.

That DX10 actually ships with Vista, no later than September.

That we have high-quality drivers (at least from a stability pov) from both IHV's for DX10 when Vista ships.

That DX9 apps run faster in Vista than they do on WinXP with the same hardware (maybe more system memory).

That the first gen of high-end DX10 cards are faster at DX9 in WinXP than the last generation of DX9 cards are at DX9 apps in WinXP. (I've been assured this will be true, but I'm covering my bases).

And since no wishlist would be complete without asking for what you know won't happen. . .that DX10 for WinXP, API and drivers, is available as a free download (and in high quality stability) at the same time DX10 ships for Vista.

Lastly, and most importantly of all, that we never ever ever see any threads about shimmering in the Vista 3D UI. :LOL:
 
Last edited by a moderator:
1. That Vista won't suck.
2. That the registry will go away.
3. That the registry will go away.
4. That the registry will go away.
5. That the current trend of ~$300 cards that offer pretty damn good performance continues.
6. That ATI will get back on its game and be competitive across the entire line instead of just at a few segments.
343. Stochastic.
 
Hopefully in 2006, there will be more games using OpenGL (and perhaps the OpenGL board will finally get something done, and contribute to something that DX10 won't initially support).

Also, perhaps MS finally decides not to "effectively kill" OpenGL in Vista (a longshot, but hopeful here nonetheless).
 
I hope to see more of Vista(Should have been named Longhorn btw). And to see how this 3:2(Nvidia) and 3:1(ATI) ratio irons out.

LVS
 
Might as well make this a (small) G80/R600 wishlist...
- Good granularity for dynamic branching in both PS and VS - not just either one (G70: VS, not PS; R520: PS, not VS).
- New "creative" rendertarget formats for G80/R600, ala INT8/8/16 for NAO32.
- A PPP that doesn't suck.

Uttar
EDIT: Yeah, and OpenGL, goddamnit! I want OpenGL ES to be extended to Desktops, with all new DX10 functionality but optional over DX9 functionality. It's high time it got fully competitive again! And properly accepted by Vista...
 
1. SM4 support in DX9.
2. That the G80 incorporates high-performance dynamic branching, multithreaded execution, MSAA on FP16 rendertargets, and is available late next year.
3. Good OpenGL support in Vista.
 
Uttar,

The GS in D3D10 is what's left of what could have been a highly potential PPP; I doubt IHVs will waste HW to go beyond the according basic requirements.

--------------------------------------------------------------------------------------
To the question:

Adaptive shader antialiasing. Now that we've finally got adaptive AA for transparencies, the 4-5 years D3D10 is aimed to last is enough to get a bit more greedy.
 
1.) Improvments in IQ, mainly texturing.
2.) GPU power usage to go down a bit (ahh, a pipe dream).
3.) That the $300 price point continues to be basically the same thing as the $500 but only a few percent slower.
4.) For HDR to go away because it completely sucks.
5.) CCC to improve performance, and Nvidia's drivers to stop giving me weird issues with every new game it seems.

So yeah, there we go.
 
Ailuros said:
The GS in D3D10 is what's left of what could have been a highly potential PPP; I doubt IHVs will waste HW to go beyond the according basic requirements.
Sure, but I'm curious how potent the basic requirements are for real-world situations. If the basic implementation is cool and efficient, I don't really care whether it's not very programmable at all.
Personally, I'd already be very happy with a system that would let me automatically put select a LOD that was created on the CPU (that is, with a fixed number of LODs being present on the installation disk, artist-approved) and lerp between the different possibilities properly.
Yes, this can be done on the VS, but there are multiple problems with it imo. Of course, such a special-purpose implementation won't happen I'm sure, bleh. Oh well.
Adaptive shader antialiasing. Now that we've finally got adaptive AA for transparencies, the 4-5 years D3D10 is aimed to last is enough to get a bit more greedy.
Yup, being able to select how much of the MSAA should become SSAA on a per-pixel basis would be cool, so as not to limit that feature only to alpha-testing-related-shaders. Of course, that would only be useful if the problem didn't come from textures, in which case a standard HLSL function to do that without having to rewrite the code a zillion times could be nifty.


Uttar
 
Uttar said:
Sure, but I'm curious how potent the basic requirements are for real-world situations. If the basic implementation is cool and efficient, I don't really care whether it's not very programmable at all.
Personally, I'd already be very happy with a system that would let me automatically put select a LOD that was created on the CPU (that is, with a fixed number of LODs being present on the installation disk, artist-approved) and lerp between the different possibilities properly.
Yes, this can be done on the VS, but there are multiple problems with it imo. Of course, such a special-purpose implementation won't happen I'm sure, bleh. Oh well.

Yup, being able to select how much of the MSAA should become SSAA on a per-pixel basis would be cool, so as not to limit that feature only to alpha-testing-related-shaders. Of course, that would only be useful if the problem didn't come from textures, in which case a standard HLSL function to do that without having to rewrite the code a zillion times could be nifty.

Past history tells me rather that I shouldn't expect too much from developers in finding sollutions for any sort of problems. I'd love to put on my personal wishlist that developers would stop treating antialiasing as some sort of luxury feature, but it would probably be in vain.
 
Ailuros said:
Past history tells me rather that I shouldn't expect too much from developers in finding sollutions for any sort of problems.
*cough* NAO32 *cough* - I know I'm insisting on that one, but it feels so humiliating for FP16 blending in the SM3.0. generation (read: would have been it totally useless had it been figured out earlier) that it completely disproves your claim: talented programmers can and will find astonishing solutions to key problems.
But another problem remains: most videogame programmers are either not very talented, or lack time to do proper algorithmic research due to obvious deadline constraints.

Uttar
 
Not to have to wait until 2008 before we see any SM4 games that truly take advantage of the tech.

Sigh.

Jawed
 
Back
Top