Forums | MacLife

From what I unerstand, racing games often get bogged down in license wrangles- exclusive contracts, unwillingness to show cars damaged (bad for sales), more. Regrettable, but nothing new. Gameplay, that's another issue.

What's going on here, I've no info about... but it shore should be prett3h.

That article seems to be a little light on anything but "ZOMG Look at us! We're delaying because we're SOOO cutting edge!" I will be interested to see what the DX11 vs. DX9/10 shots look like, though. With more GPGPU capabilities there is certainly a lot more flexibility in what you can do in terms of effects. Of course, the question is, how much is practical to do with a steady frame rate on a large monitor? Going more general purpose certainly isn't free: for example, though newer cards allow stuff like dynamic branching and looping, it has a large penalty and is a good idea to avoid whenever possible.

I have a feeling that this is largely a PR stunt, probably also sponsored by Microsoft and/or AMD/ATI. (they have to get pre-release versions for the DX11 SDK and hardware, after all) I have no doubt that they will be able to improve the visuals, but I doubt that the difference will be particularly jarring, at least not with something that can be implemented with affordable hardware in the near future. However, they may have ultra high quality settings that have much better visuals that can only be playable on multi GPU systems and/or future GPUs.

akb825 wrote:

That article seems to be a little light on anything but "ZOMG Look at us! We're delaying because we're SOOO cutting edge!"

How much detail did you expect in a press release? It wasn't GDC or Siggraph.

hardware tessellation, multi-threading and Shader Model 5.0 | Hardware tessellation will further improve the appearance of water and other surfaces as well as crowd animations  | The shader performance of Shader Model 5.0 (SM 5.0) will aid in [..] enhancing key image quality parameters such as depth of field, ambient occlusion and shadows

Tesselation, tho not this advanced, has been in the 360 all along. Shouldn't be too hard; ATi gave a demo at the recent hardware trade show, since they had working silicon. The rest is about as detailed as I'd expect in a PR, and has a few specifics.

I will be interested to see what the DX11 vs. DX9/10 shots look like, though.

Everyone should be.

I have a feeling that this is largely a PR stunt, probably also sponsored by Microsoft and/or AMD/ATI. (they have to get pre-release versions for the DX11 SDK and hardware, after all) I have no doubt that they will be able to improve the visuals, but I doubt that the difference will be particularly jarring, at least not with something that can be implemented with affordable hardware in the near future. However, they may have ultra high quality settings that have much better visuals that can only be playable on multi GPU systems and/or future GPUs.

Partly, no doubt. ATi has said they know of 5 titles coming in that timeframe. It's still remarkably fast takeup; TR:AOD and Halo PC weren't out until ~10 months after the DX9's release- this is more like 6 weeks. Naturally they won't come near to fully or even heavily exploiting the possibilities. Still, someone I know was opining on how long it'd be before any 11 games... :whistle:

NV won't have DX11 parts 'til next year, so it's AMD/ATi's show, and gain, for some time. Also, on multithreading, I'm sure this will be exploited on the API/driver level, not just multithreading the game. That's been looked forward to for some time- better thread handling, IIRC. (Should look that up..)

Their mention of

by giving developers tools to effectively utilize the multi-CPU, multi-GPU systems now available to players

...jibes with your

ultra high quality settings that have much better visuals that can only be playable on multi GPU systems and/or future GPUs

(Subconscious plagiarism as in music! Codemasters sue! )


Sure, they'll be enjoying as many multi-GPU sales as they can get, and a game that'll drive upgrades for a long time.

...if it's good enough. Ear has his doubts, I lack info.

Still, it's the first, and will have a place in game history. (And you didn't expect me to let this go by unmarked, didja? )

Oh, and on this score...

akb825 wrote:

(they have to get pre-release versions for the DX11 SDK and hardware, after all).

As I mentioned, working silicon is already around, and Win7 is set for a July RTM. It's close. At least early version[s] of DX11 must be workable for ATi to code a demo around it/them (reportedly heavy on tesselation).

~Coxy wrote:

akb825 wrote:

That article seems to be a little light on anything but "ZOMG Look at us! We're delaying because we're SOOO cutting edge!" I will be interested to see what the DX11 vs. DX9/10 shots look like, though. With more GPGPU capabilities there is certainly a lot more flexibility in what you can do in terms of effects. Of course, the question is, how much is practical to do with a steady frame rate on a large monitor? Going more general purpose certainly isn't free: for example, though newer cards allow stuff like dynamic branching and looping, it has a large penalty and is a good idea to avoid whenever possible.

That's a very good point. Reminds me of when DX10 first came out and in the DX9/10 comparison shots for Company of Heroes and Bioshock you could barely make out the minute differences.

Even then, and those subtle differences were just that --differences.  It wasn't better or worse; it was just ever so slightly different.  The big thing about DX11 seems to be hardware tesselation, which I believe is being blown waaayyy out of proportion.  Compared to software methods, the hardware approach likely requires less memory, and changes in detail are less expensive, but I just don't think this is going to be DX9:round 2.  On the other hand, there's too many crappy engines out there without any form of tesselation --Perhaps DX11's greatest practical benefit will be the boost it gives to these engines.

Tessellation is certainly interesting, but from what I understand it doesn't work by triangles, but by patches. Which makes sense: you can't suddenly make something look better just by looking at an existing triangle mesh, you need more information. In this case, it's done with a surface that uses control points to control the curve of the surface, from which you can tessellate to different levels of detail from the same source data. While that may sound easy enough, there are some issues that come up. For one, you need to model using these patches that can be used by the GPU instead of just triangles. However, once both the artists are trained and the tools support it, it's easy to convert to triangles for older renderers or for super low LODs where the cost of tessellating for so few triangles isn't worth it, so once it's mainstream that part is fine. The other issue I can see is skinning. It's very straight forward for triangles, but I can see it being much more difficult for the patches used for tessellation. For static environments, though, there is no such issue. I will be interesting to see both image comparisons and speed comparisons, though, since obviously it's not for free.

Bat: Obviously I wasn't looking for a SIGGRAPH paper, but they give no real information. The only thing they give is "enhancing key image quality parameters such as depth of field, ambient occlusion and shadows." All three of these effects are already used in games, and "enhancing" them is a little ambiguous. It can be as simple as "wow, we are now able to use 8x8 PCF on our shadows instead of 4x4!" which only means that they are able to do more texture samples. On the other hand, it can be as amazing as "we can now do physically accurate ambient occlusion instead of current image-based methods."

That said, though there are prototype DX11 chips around, these are hardly mainstream, and it's not going to be available to just any developer. According to the Direct3D 11 page on Wikipedia "Microsoft have stated that Direct3D 11 is scheduled to be released to manufacturing in July 2009, with the retail release coming in October '09." This means that us "normal people" will be able to reap all the rewards by October. If the developers of Dirt 2 are announcing the change now, that means that they have already gone through the technical work of making sure that they can handle the development change and made sure that they can get a hold of the necessary hardware and software. Not to mention the extra time that would have to be added for testing and QA. This has likely bee in the works months before it was announced, probably in tandem with Microsoft and AMD.

As for my comment about the availability of DX11 games, I obviously wasn't saying that there will be no games that use DX11 for 3 years. What I was saying was, if you look at DX10 currently, there are some games that have DX10 support, but almost no games rely on it completely, and even when used often don't have that much change in visual quality. Not to mention the fact that games have to run acceptably on 2-3 year old hardware in order to have a market to sell to. Developers aren't going to spend as much time working with a new edition of DirectX at the expense of allowing their game to run on 90% of their userbase's hardware. There are exceptions, though, who try to cater to the niche market of extreme gamers.

As always, improvements are evolutionary rather than revolutionary. Though there are some large improvements in terms of theoretical ability, such as the introduction of pixel sahders in DX9 or GPGPU features in DX11, the hardware isn't fast enough at the beginning to take full use of them. Each new revision of the shader model increases lower limits on aspects such as the number of instructions, which alone allows game developers to introduce more effects. The latest trend of making GPUs more general purpose does increase the set of tools available when writing shaders, but often times if you try to use all of those tools you will tank the framerate. I'd say that we are still a couple of years away both from DX11 becoming mainstream and having the true power of its new capabilities to be utilized in a way that allows previously impossible or prohibitively difficult effects to be implemented at playable framerates

Last edited by akb825 (2009-07-07 3:28 am)

Hope to get back to this soonish; a few things left to mention, not too far out of agreement anyway. Meanwhile, this seems worth tossing out on its own...

DirectX 11 Compute Shader: Three times faster than DX10.1 due to Local Data Share
AMD's Developer Relations boss Richard Huddy explains the mode of operation of Ambient Occlusion - the DirectX 11 Compute Shader is said to provide up to three times the performance possible with DirectX 10.1

Using a Compute Shader path in DirectX 11 could offer up to three times the performance possible with DirectX 10.1, mentions Richard Huddy, AMD Developer Relations, in an interview with PC Games Hardware. And even in DirectX 10.1 modern Radeon cards with their Fetch4 feature are able surpass pure DirectX 10 devices like the current Geforces.

But at the same time Huddy explains that Compute Shader is uncharted territory for developers and thus it is hard to integrate. Getting the optimum out of it is even harder currently.

In matters of possible performance benefit Huddy calls the Local Data Share (the memory shared between different threads), which is required by DirectX 11, a Killer-Feature. Especially in Screen Space based Ambient Occlusion it is possible to safe [save?] huge amounts of data traffic resulting in less workload for the bandwidth and texture units which are used to read the data.

PC Games Hardware has already reported about AMD's DirectX 11 graphics cards of the Evergreen family, which are said to be released in 2009. We have integrated the first video with Richard Huddy's interesting remarks about the expected performance into this article again.

http://www.pcgameshardware.com/aid,6899 … hare/News/  (more, links, video links)

Also, AMD/ATi plans on releasing DX11 parts before 7 launches. No doubt Codemasters has multiple units now, and it's not unprecedented; it happened with the R300 & DX9. What's majorly different is the time to first game release.

Last edited by Bat (2009-07-15 1:07 am)