Interviews - Tessellation hugely exciting says AMD exec

[HEXUS]

Richard Huddy, AMD's European head of developer relations, has told HEXUS he feels Tessellation is the most interesting piece of technology currently still in development.

Read more.

[kalniel]

Great. So how about enabling the tesselation features that are present in earlier series (eg hd4000) as well?

[HW_90]

Great. So how about enabling the tesselation features that are present in earlier series (eg hd4000) as well?

because it's easier and cheaper to code for DX11 then to code for one specific device.

[scaryjim]

because it's easier and cheaper to code for DX11 then to code for one specific device.

Surely not that point. If the hardware capability exists to do it, it should be possible to enable the software path by either a driver or BIOS update... making earlier Radeon cards able to support DX11 tesselation...

[kalniel]

because it's easier and cheaper to code for DX11 then to code for one specific device.

Device and even game- specific coding is present in drivers already. And while the tesselating engine had changed a little for dx11 it shouldn't have changed so much that the brilliant software engineers couldn't come up with an elegant fall-back to ATI tesselation when developers call dx11 tesselation.

[Syllopsium]

Device and even game- specific coding is present in drivers already. And while the tesselating engine had changed a little for dx11 it shouldn't have changed so much that the brilliant software engineers couldn't come up with an elegant fall-back to ATI tesselation when developers call dx11 tesselation.

Where's the profit in them doing so? They can't even release half decent Linux drivers either; there's no resource free for coding card specific workarounds to create a partial DX11 when there's probably some hardware limitations anyway.

PK

[kalniel]

Where's the profit in them doing so? They can't even release half decent Linux drivers either; there's no resource free for coding card specific workarounds to create a partial DX11 when there's probably some hardware limitations anyway.

Existing owners of 4000 series cards aren't likely to move up just one generation of AMD card, but might be tempted to switch to nVidia's dx11 solutions if they offer something extra. Giving them a reason to stay with their 4000 series card until at least the next generation of AMD solutions would increase profit.

Hardware limitations are unlikely to be major - tesselation units have been present in ATI cards for some time, the limitation is more likely to be the way that tesselation is called from the software, which they had no reason to workaround in the drivers because no developers were using tessellation. Now that tesseletion is part of DX11 developers will use it, and the driver team therefore have a reason to add in the workaround, just like driver teams from both camps have added workarounds for things like AA with deferred rendering.

[Syllopsium]

Existing owners of 4000 series cards aren't likely to move up just one generation of AMD card, but might be tempted to switch to nVidia's dx11 solutions if they offer something extra. Giving them a reason to stay with their 4000 series card until at least the next generation of AMD solutions would increase profit.

Hardware limitations are unlikely to be major - tesselation units have been present in ATI cards for some time, the limitation is more likely to be the way that tesselation is called from the software, which they had no reason to workaround in the drivers because no developers were using tessellation. Now that tesseletion is part of DX11 developers will use it, and the driver team therefore have a reason to add in the workaround, just like driver teams from both camps have added workarounds for things like AA with deferred rendering.

Technically, switching to Nvidia has precisely the same impact on ATI's profit as not switching from a 4xxx series product..

The advantages Nvidia offer are PhysX, CUDA and 3D Vision. None of those are particularly widely used at the moment and ATI has upcoming offerings or existing third party drivers (for the 3D) in all those areas.

Nvidia isn't going to be releasing DX11 products until December if you're lucky, and that's probably only for the high end next generation. It's entirely possible that by the time Nvidia's next generation is out, ATI will have a product refresh too (at the very least the 5890 and X2 cards will be out).

I don't see why users of 4xxx cards are likely to switch to Nvidia, or why it's a good idea of ATI to schedule lots of expensive development and support resource just to enable a few people to use a very limited selection of games and have a few future proofed product warm fuzzies.

PK

[kalniel]

Technically, switching to Nvidia has precisely the same impact on ATI's profit as not switching from a 4xxx series product..

I'd argue not - AMD maintain a higher market share than otherwise, which in turn gives more leverage with developers and creates better exposure and investment. By not switching to nVidia's dx11 cards they're also more likely to be in the position to buy AMD's next generation cards ('6000' series).

I don't see why users of 4xxx cards are likely to switch to Nvidia

Because people will upgrade usually when they feel their current cards are slow, and when they do they will then jump to the best performance. When nVidia's next gen come in they are likely to beat AMDs cards for performance. AMD have won the now because there isn't any competition, but people with 4000 series are usually happy with the existing performance because there are no games out that use dx11. When tessellation becomes more mainstream this will change, and nVidia will have cards out by then.

or why it's a good idea of ATI to schedule lots of expensive development and support resource just to enable a few people to use a very limited selection of games and have a few future proofed product warm fuzzies.

They already have that development and support - the driver team is in place. We don't know what selection of games in the future will use tessellation, but as I said earlier, there are already card and single game specific optimisations in the drivers, so any change that benefits more than one game is already more widely beneficial than those that have been funded in the past.

[Jimbo75]

Nvidia's dx11, 6 months later and possibly done in software, that sure looks enticing yeah.