DirectX 9 December 2005 SDK is now available...

Oh WOOT! It's downloading atm but please PLEASE say they've worked on MDX 2.0. In my team, we are using .NET 2.0 and hopefully generics and DX is going to give a performence boost. *Crosses fingers* Oh yea... and directx 10... but thats all...unmanaged and icky!

Thanks for this update jollyjeffers :)

One other thing... on the subject of Direct3D 10... I've been a bit busy, so I'm a little behind, but I've got a couple of articles written up for you guys.

I just want to check them on this latest version and I'll see about making them available [smile]

Jack

Wow, it's a lot bigger than the October one. Lucky that I'm downloading at 800 KB/s [grin]

It's nice to see that there's finally MDX for .NET 2.0. I assume they did FX highlighting for VS8 as well?

Quote:

Original post by Promit Wow, it's a lot bigger than the October one. Lucky that I'm downloading at 800 KB/s [grin]

I guess it's bigger because it's DX9 and DX10...

I've just finished downloading all the components, so I'm gonna go play [smile]

Jack

Quote:

Original post by jollyjeffers One other thing... on the subject of Direct3D 10... I've been a bit busy, so I'm a little behind, but I've got a couple of articles written up for you guys.

I also have some samples I would like to share, mainly centered around Geometry Shaders.

Quote:

However, note (emphasis mine) that you'll be needing the latest Windows Vista CTP to experience it... which might well be a sticking point for some.

Indeed...if you don't have an MSDN subscription, then you pretty much can't run it. You can look at code and samples for a few months though, until Vista B2 comes out. Yea, I know it sucks [crying]

Quote:

Original post by Promit Lucky that I'm downloading at 800 KB/s [grin]

Pretty much the one and only thing Resnet is good for hehe

Just marvelous. Thank you to David Weller and Mr. Jack Hoxley for making a post about this. Let's get some development going and we are looking forward to those articles by the DirectX MVP's Mr. Franklin and Mr. Hoxley.

Awesome.

Wait for some articles from DirectX MVP Mr. Suni also, concerning implementing networked AI using next-generation graphics hardware [wink]

(as soon as I can dig myself up from my huge pile of work).

Quote:

Original post by Nik02 Wait for some articles from DirectX MVP Mr. Suni also, concerning implementing networked AI using next-generation graphics hardware [wink] (as soon as I can dig myself up from my huge pile of work).

This is one I'm definitely looking forward to [smile]

Quote:

One other thing... on the subject of Direct3D 10... I've been a bit busy, so I'm a little behind, but I've got a couple of articles written up for you guys. I just want to check them on this latest version and I'll see about making them available [smile] Jack

As promised: [D3D10] A simple comparison of D3D9 and D3D10

Jack

Quote:

As promised: [D3D10] A simple comparison of D3D9 and D3D10 Jack

Man, I am really looking forward to reading it. Can you post a link that works, please? ;)

EDIT: Wow, that was speedy. Thank you!!!

Quote:

Original post by j stagner Quote: As promised: [D3D10] A simple comparison of D3D9 and D3D10 Jack Man, I am really looking forward to reading it. Can you post a link that works, please? ;) EDIT: Wow, that was speedy. Thank you!!!

Haha, yeah - fixed it now. Not entirely sure how that path ended up on my clipboard instead of the real one though [oh]

Jack

Here is one of the Geometry Shaders that I have been working on:

Tesselator.fx



Basically, it just takes a triangle, and breaks it down into smaller triangles. It's pretty simply, and doesn't really work all that well (it doesn't do it smoothly like you would want). For me, it was just an exercise in emitting lots of vertices.

I investigated further into the practicality of using GS for tesselation purposes, and it turns out that it is a very bad idea. First-generation hardware will simply not have the performance for that kind of mass output. You may be able to do 1 level of tesselation, but nothing big. I had looked into doing arbitrary n-patch tesselation (ie bezeir curved surfaces), but stopped when I figured out that it will be waaaaay too slow.

I had also looked at doing what I called "adaptive tesselation", where the GS would tesselate the primitive more or less, based on the camera's distance and viewing angle. Essentialy, this would let the parts of the model that you are looking at have finer detail. However, I stopped working on that too.

Also, I didn't test this with the new SDK, because the new Vista CTP isn't out yet. Hopefully soon....

Quote:

Original post by circlesoft I investigated further into the practicality of using GS for tesselation purposes, and it turns out that it is a very bad idea. First-generation hardware will simply not have the performance for that kind of mass output. You may be able to do 1 level of tesselation, but nothing big. I had looked into doing arbitrary n-patch tesselation (ie bezeir curved surfaces), but stopped when I figured out that it will be waaaaay too slow.

Because the GS has visibility to whole primitives, it should be very powerful when it comes to organising internal rendering. I'm expecting this to be (at least initially) the main focus.

If you think about it - apart from a few odd hacks, the VS and PS we currently have doesn't really know anything about where it belongs - what triangle the vertex is part of etc...

With the IA's counters (VertexID, PrimitiveID and InstanceID) along with the GS's topological visibility it means that the code we write for the GPU has a lot more knowledge of what it's dealing with.

As well as single-pass cubemap rendering, I'm expecting some funky GPU-based material systems to appear. Anything to offload the complexity of a CPU-based material sorting/management system would be nice [smile]

Jack

Quote:

Original post by jollyjeffers As well as single-pass cubemap rendering, I'm expecting some funky GPU-based material systems to appear. Anything to offload the complexity of a CPU-based material sorting/management system would be nice [smile]

I think you are going to see some very, very high-quality particle systems being developed. For those of you who haven't really looked into GS, it is great for particle systems, because you can essentially take a point (represented by 1 vertex) and output an entire primitive representing that point. This is great, because:

(a) You don't have to use so much memory in buffers for the particle systems, since you are only storing 1 vertex per point now

(b) You can animate the points on the GPU using stream-out. This means that you don't have to lock the buffer to update your particle system anymore.

In my opinion, we are going to see an explosion of natual effects. Of course, this is linked with the above particle system technique, since many natural effects (ie rain, snow, ect...) use them. Also, things like massive amounts of grass and fur can really be rendered well now.

Does this mean that windows XP cannot run dx10, *la cry.

I'm wondering if theres much scope to output multiple particles per input vertex...

It probably wouldn't be physically accurate as such, but just to clone a particle and offset it a number of times might well be an easy way to make "dense" particle systems...

Quote:

Does this mean that windows XP cannot run dx10, *la cry.

DirectX10, from the outset, was a Vista-generation technology. It wasn't intended to run on XP.

XP isn't going to suddenly disappear overnight, such that DirectX9 will still be a viable candidate for a while yet - especially with the "DirectX9.L" version appearing on Vista.


Jack

Or to learn some new technologies to get an edge on future markets :p

Anyway, so DX10 doesn't come with a Managed wrapper yet, right? Just making sure, since I'm almost considering gettting myself a MSDN subscription because of all the things I currently seem to be missing out on...

Isn't there some kind of discount on MSDN subscriptions for Microsoft Volume Licensees? Need to convince management ;)

Quote:

Original post by Anonymous Poster there will be no dx10 hw for a while there will be no dx10 hw in the massmarket for very long time vista has not one exceeding must-have-feature over XP so why would someone really need all this new useless stuff!? I bet, only some tech-geek's :-P

You're more than welcome to dislike Microsoft products (etc..) but this is a development forum for those that are interested in using these technologies - such that sweeping blows are neither constructive nor interesting.

By all means, if you have some facts and logical reasonining to back up any concerns or critiscms you have of DX10 then by all means post back. I'm sure people will be happy to discuss... Just do so in a professional manner.

We had enough trouble for the October '05 thread - it got boring pretty quickly and I'd prefer not to see a repeat of it [smile]

Quote:

Original post by Anonymous Poster "XP isn't going to suddenly disappear overnight" I don't see XP disappear in a lot of years vista will play only a minor role in the next years

It's worth realising that Microsoft tend to have a 5-year support cycle for their software. Windows 2000 expired on July 30^th this year. Windows XP's official life cycle ends on December 31^st 2006 (see here). This means that the big corporate customers won't be able to get official support for their platform anymore - which will drive them to upgrade to a supported platform.

Sure it's never that simple, nor does that necessarily reflect on the home-user market, but I wouldn't go much further than saying that XP will be around for as long as it takes for people to transition/migrate across..

hth
Jack

Quote:

Original post by Anonymous Poster but here we are on an gamedev forum, so the only important market is the home-user market

What do you mean home dev only? There are plenty of professionals that come here. Gamedev has long since evolved from hobbyist-exclusive environment.

Quote:

and I dont see a strong reason why they should/could move the vista in the next years

What is your reasoning behind this? Personally, I still favor using Windows XP at this point, since the Vista beta builds don't really offer that much. However, they are only betas and there is still 1+ years left of dev time. If you look at what the goals for Vista are, they are quite ambitious, so really, I don't know what you mean.

Quote:

I dont think it's a good move when developer's force the user to buy the newest microsoft system, when there are no really is a very good reason for it for me the vista exclusive support of dx10 or better of the nextgen hw is only a marketing decision to force the sells of vista

Well, the reason that D3D10 requires new hardware goes down to the fundamentals of D3D10 - mainly a new driver model, unified caps, and shared resources. Each of these is a large step foward from what we currently have, in terms of stability, portability across hardware, and of course performance.

Quote:

maybe you say there are this and that technical reasons it only works on vista, but I believe it would also possible on XP

Perhaps, if you made a completely new display driver model for XP (ie isn't going to happen). Plus, like Jack said, XP is already 4 years old, been through 2 service packs, and is on its way out. I like XP too, but it is getting to be a little aged.

We've all been discussing the new features exposed through the D3D10 API over in the #graphicsdev IRC channel. Feel free to come Join Us.