Some months ago I e-mailed ATi asking them to participate in an interview and they were positive but after some confusion and misunderstanding this interview has been delayed for some time. Thus some of these questions are a bit outdated and some we had to remove completely since they since long are common knowledge etc.
But we still got some interesting information that we which to present you with here…
Some months ago I e-mailed ATi asking them to participate in an interview and they were positive but after some confusion and misunderstanding this interview has been delayed for some time. Thus some of these questions are a bit outdated and some we had to remove completely since they since long are common knowledge etc.
But we still got some interesting information that we which to present you with here.
Note: As a few readers have noticed some of these questions are identical to the ones DriverHeaven.net has published: there’s a simple reason for that I, Anton Nilsson, sent these questions to ATi many months ago. As mentioned there were delays. In the meantime Driverheaven asked their users to post their questions for ATi: and I posted some of the questions I sent to ATi there. Thus some of the questions are almost letter for letter identical.
My username is ”Uriel_is_UndaC” in the DriverHeaven forums so this is easily verifiable for the skeptics.
I hope this clarifies any misunderstandings.
If so when will this be realized in software?
[ATi] We are not sure what you mean by ”10bit precision per color”. Is it 10bit render target or 10bit lookup table? 10bit render target is supported, under DirectX 9. [NH] It has been noted that ATi products have a problem with stuttering. A bunch of enthusiasts seems to have nailed it down to a READ/WRITE problem in DirectX as explained in this discussion: http://www.beyond3d.com/forum/viewtopic.php?t=3053
How does ATi respond to this obvious problem with their products?
(The problem has been widely discussed at Fan sites such as rage3d.com and is well commented around the net, still there hasn’t been any word from ATi yet.)
[ATi] Every bug reported to us, regardless of the reporting origin, is assigned to a software engineer in our organization and is being addressed at a high level of priority. We have seen the recent reports on ”stuttering” and we are currently investigating. We are confident that we will be able to release a fix to this problem very soon.
(ATi recently released an updated version of Catalyst 3.0 which fixed this issue.) [NH]The R300 currently doesn’t support the usage of FSAA when using a 16 bit framebuffer.
When will this be fixed and how?
[ATi] Demand for the support of FSAA on 16bit frame buffers remains very low. ATi currently has no plans to add this to its list of supported features.
(A leaked driver from ATi supports FSAA when using a 16 bit framebuffer so it certainly seems as if it will indeed be supported.)
What type of improvements can we expect, as far as performance is concerned, from R300 drivers?
[ATi] ATi is committed to optimize the performance of its products to the highest possible level. A precise forecast for the level of gain in performance for a new driver set is however not possible.
[NH] When, if ever, will we see a Super Sampled Anti Aliasing implementation on the Radeon 9700-series?
Early statements from ATi indicated that it was possible and that it would be up to 6 times faster than
previous generations. I know personally that I would love to see it implemented in the drivers and
there’s a lot of people that agree with me. Will you grant us our wishes?
[ATi] ATi has investigated this possibility, however we believe that sample-for-sample, the RADEON 9700 & 9500 series multisampling implementation, provides better performance and image quality than any supersampling method. So we currently have no plans to support supersample anti-aliasing on these products.
[ATi] Given the current price level of DDR II RAM it seems unlikely that a graphics board equipped with it would be affordable soon. ATi however certainly leaves it to it certified Partners to decide on the additional features of their graphics adapters.
(There certainly seems to be horsepower to do 8x FSAA)
[ATi] The RADEON 9700/9500 series cannot support multisample anti-aliasing with more than 6 samples per pixel, nor can it support anti-aliased resolutions above 1600×1200 due to buffer size limitations. Supporting 8x FSAA would require a multisampling/supersampling hybrid, which would drastically impact performance and in many cases reduce image quality (by blurring fine text and other detail). We believe that implementing ”higher” FSAA modes that detract from image quality is confusing and unnecessary. [NH] How do you feel about a 256 MB Radeon 9700 aimed at consumer level. Is it worth while?
What benefits would it bring? (Look at the previous question for an example)
[ATi] Apart from the ”All In Wonder” product line , ATI does not build complete graphics boards. Most graphics boards are designed and manufactured through certified partners and hence ATi leaves it to them to define features like the memory size. [NH] Is there any chance of a multichip consumer R300 solution?
What benefits could such a solutions potentially bring?
[ATi] ATi does not have plans for a multichip solution. [NH]Will ATi ever bring us users true Stereo Drivers for 3D-glasses?
[ATi]This issue is being raised from time to time. We always try to be attentive to the market needs and respond accordingly. At the moment, we do not see a great deal of market demand for this feature. Of course we will react if we see market conditions change. [NH] At the 9700 launch there were rumours about ATi having a fix for the absence of AA on alpha textures when using MSAA.
Will this fix ever be implemented?
[ATi] ATi does not have Information on such a fix.
(There is indeed a way to solve this in OpenGl with an alpha coverage mask call but it has a couple of limitations.)
Our thanks goes out to Rubeena Hussein of ATi Europe and her colleagues.
Leave a Reply