Develop Conference 2006 - D3D10

[Rys]

<a href="http://www.beyond3d.com/articles/develop-d3d10/"><img border="1" src="http://www.beyond3d.com/includes/images.php?id=224" align="right" width="120" height="75"></a>A couple of weeks ago the <a href="http://www.developconference.com/" rel="external">Develop Conference and Expo</a> took place. Develop is a UK-based multi-day conference designed to get the entire games industry -- from ISV to IHV and everyone in between -- together in the same place to talk tech and development.

On the last of the Expo's days, ATI held a Developer Day in conjunction with Microsoft and Intel, in order to talk about D3D10, Windows Vista and multi-theaded games programming, both on the PC and Xbox 360. Beyond3D made the trip down to sunny Brighton on the UK's south coast, and we've got a short developer-focussed summary of the day's D3D10 presentations.

<a href="http://www.beyond3d.com/articles/develop-d3d10/" rel="external">Read on for chatter about geometry shading performance, state change overhead, DX10.1 and more....</a>

[Demirug]

Nice sum up

Richard mentioned that the first D3D10 drivers under Vista won't be heavily optimised
I don’t like such things. Maybe it is true but I could mean that the hardware is not that good with D3D10, too. Make people hope for better drivers is a bad game.

[Rys]

I don’t like such things. Maybe it is true but I could mean that the hardware is not that good with D3D10, too. Make people hope for better drivers is a bad game.
Hopefully it just means the first few months of drivers won't be too optimal, and that soon after they'll get better. It all depends on how long the driver team will have been programming for the real hardware, before Vista launch, and is it reasonable to expect a brand new architecture to have great drivers bare months after silicon came back?

[AlStrong]

In the short-to-midterm it might not even be a huge deal. I certainly don't expect millions of people to suddenly upgrade to Vista. On the other hand, new systems from Dell/HP etc, those would be the center of troubles, but then again, even if the drivers are not optimized for gaming, they should be stable enough to do business apps, no?

[DegustatoR]

Sam let the aseembled developers know that in addition to having double the per-pass rendertarget count compared to D3D10
D3D9 i believe :)

[Rys]

Of course, fixed :grin:

[Geo]

Some bits that caught the eye:

Think multi-GPU, but not just SLI, Crossfire or Multichrome.

What be that about?

Sam finished off his presentation by saying DirectX 10.1 would come fairly soon after Vista's launch

If I threatened to take away your broadband for two weeks if you didn't color that a bit more, what kind of time frame do you think he had in mind there? 3 months after release? 6 months? Something less than a year after release anyway?

He said that those pre-raster caches would get bigger in upcoming hardware,

Probably that's not a surprise, but always nice to see.

With R580+ launching in late August and RV560 and RV570 a wee while later,

Also nice to see with a little more cred than L'Inq.

Lastly, he urged developers not to use the SDK samples as performance indicators on the new hardware, when it arrives. It's common for developers to use SDKs almost as cut-and-paste repos, for certain effects, but they shouldn't do that this time around (at least with the current SDK). Use the samples as inspiration, not as the solution.

Yes, I can see that causing some problems. Any hints at when the SDK revs will get more in line with reality?

[Demirug]

If I threatened to take away your broadband for two weeks if you didn't color that a bit more, what kind of time frame do you think he had in mind there? 3 months after release? 6 months? Something less than a year after release anyway?

Don’t expect it too soon as D3D10.1 requires WDDM 2.1 hardware. This is not a minor change.

Yes, I can see that causing some problems. Any hints at when the SDK revs will get more in line with reality?

“We don’t talk about unannounced products.“

[Geo]

“We don’t talk about unannounced products.“

I'm not asking them to talk about unannounced products! I'm asking them to talk about unannounced SDKs! :lol:

[Demirug]

I'm not asking them to talk about unannounced products! I'm asking them to talk about unannounced SDKs! :lol:

There is a new DirectX SDK every two months. Next update should be only days away. But as long as Microsoft has no hardware they could not change the SDK samples.

[AlStrong]

a new DEVICE_REMOVED enum exists to let your app know the GPU you were executing on is no longer available. Think multi-GPU, but not just SLI, Crossfire or Multichrome.

What be that about?


One possibility is if there is an integrated GPU that gets disabled. Maybe? Or how about a third GPU used for physics processing?

[Demirug]

One possibility is if there is an integrated GPU that gets disabled. Maybe? Or how about a third GPU used for physics processing?

Think about a notebook that is removed from it's docking station. Vista will support hot plugin PCIe.

[Geo]

Think about a notebook that is removed from it's docking station. Vista will support hot plugin PCIe.

Whoooooa. So a base unit with the Big Daddy gpu in it, scaling down to the onboard when you disconnect and walk away, and without rebooting?

[trinibwoy]

What is this 17th shader unit? Is that the geometry shader? Because when I surmised that it would be separate from the unified bits I was told that all shaders would be GS capable :?:

[Geo]

it was a unified part in hardware, he asked the assembled crowd to think of it as a 17 shader unit part, so that he wouldn't have to give the game away before it was announced officially. Surely he understands we know already that it's 19? Regardless, he forged on.

I'm pretty sure it was a joke on Huddy's part to avoid feeding any rumor fires on whether it's 48, 64, 80, 96. . . whatever unified cores in R600. Were Huddy, even for example purposes, to use a plausible looking number then some folks (like us!) might go all :shock: and run with it. 17 is safely low and weird enough that he could be sure no one would take it seriously.

[Demirug]

Whoooooa. So a base unit with the Big Daddy gpu in it, scaling down to the onboard when you disconnect and walk away, and without rebooting?

Yes, something like this would be possible.

[AlStrong]

"whoa whoa whoa....DOOOOD but who would actually do that? :shock: "

One Answer: QA testing would love it, I'm sure. Do I get a cookie? :) But still, I'd rather shut everything off and disconnect all power just to be absolutely sure. :???:

[Bastion]

Now you handle DEVICE_HUNG on your 3D device inside your app, to detect hung cases
Sorry but I couldn't help but smile reading that :)

Big changes include the driver and hardware working on Set/Draw calls without the CPU, and Set calls mapping much more closely to actual register changes on the hardware.
This is certainly very nice. Just thought I'd state the obvious.

Moving on to D3D10, Richard mentioned that the first D3D10 drivers under Vista won't be heavily optimised
This is kinda expected but did he also say what sort of specific improvements/optimizations can be expected with subsequent driver releases?

I enjoyed reading the article which was very precise. The intro page stated that you had chats with game developers but I didn't see any pages dedicated to this. It would have been interesting to see what the developers had to say.

[Rys]

There's additional content to come from Develop, that first piece was just for the D3D10 bits. Stay tuned for the developer chat-related bits, but I won't promise when you'll see it, or in what form :wink:

As for the rest of the thread, Demi's case for DEVICE_REMOVED is the poster child they use just now, when giving an example, but you can also think of devices that don't draw any pixels on your screen, too.

And it's not that the first D3D10 driver will be horrible, it's just that all optimisation opportunities won't have been explored by release, so expect incremental perf improvements as time goes by.

[Geo]

Just a wee post to get Last Post link working. . .