<!–#set var="article_header" value="Meltdown 2001 –
DX9, XP gaming, and more 3D graphics” –>
DirectX 9 – More 3D graphics
Let’s cut to the chase and get into the juicy happenings. First off, DirectX 9. Yeah, well, it’s true that we’re just getting the DirectX 8.1 Beta Interim 2, and it will finally ship only with XP so, why DirectX 9? I think it pretty much sets the tone for DirectX 8 as well. Most of you non-XP users will probably be upgrading to DirectX 8.1, and will eventually upgrade to DirectX 9. From a graphics standpoint, we shouldn’t expect too much from future versions of DirectX in new, cool terminology. The more programmability is added into DirectX the more the emphasis is going to be on reducing Microsoft’s direct role in the 3D pipeline.
What we do know is that DirectX 9 will be a completely new .dll file so, that’s one more upgrade to look forward to. It will introduce Scissor Planes, which is not really a gaming feature but more for the serious application, and will expect hardware support for a rectangle that can clip pixels. That’s what it means, you put a box around your view, and scissor out what’s not visible. Great for putting up multiple windows and clipping the viewport accordingly. Doesn’t get done very well without a certain level of percision and hardware support. It’s in OpenGL. OpenGL scissor and stencil functions are generally used to stop OpenGL writing to parts of the screen. Haven’t anything more to say about that.
DirectX 9 – More 3D graphics, Continued
If developers start to implement fog in DirectX shaders, we may see the feature rolled into the programmable shaders in future versions of DirectX. Presently fog is supported in the pipeline as a separate feature. Developers will probably end up doing better fog with their own shader routines then using existing fixed function support. That’s probably not a bad assumption.
There should be support for Displacement Mapping. More on that in my next article in the section on Matrox (they gave a pretty a in-depth presentation on displacement mapping). There’ll be a minimum upgrade to the Vertex Shader, V.s.2.0. Flow control should be introduced, adding jumps, loops and subroutine rules. This is just adding more flexibility to the programmability of the shader, and increasing the instruction count.
There’ll be an upgrade to the Pixel Shader, P.s.2.0 (Hmm, remarkable, reads PS2. Could be Freudian slip or Microsoft console-envy. You tell me.). There are two levels of support for Pixel Shader programming, one for higher level languages, and one for Assembly Code. In P.s.2.0 the Assembly Language should allow for even more direct control of hardware. This can be done through exposing register and instruction counts on the hardware. In addition, developers can embed their Assembly code within their higher language effects routines. Basically, P.s.2.0 will just open up the hardware, and that should unload Microsoft from the burden of accounting for every new texture register or blending operation that a hardware vendor throws up.
We should expect to see tighter integration between the Vertex and Pixel Shaders in DirectX, probably leading to an eventual unification. Future hardware units may even allow the two shaders to share the same math unit, which could be a benefit to low cost graphics vendors. Right now, DirectX 8 and in future, DirectX 9, are turning the graphics processor into a co-CPU, as much as anything else. So, maybe a GPU wasn’t such a bad marketing term to begin with. Maybe it is an accurate reflection of where we’re pushing the graphics subsystem. Be that as it may, not everyone is going to want to pay for a full GPU. There is still a market for the low-end.
The other interesting inclusion for DirectX 9 is going to be support for Gamma Correction. You can read a definition of Gamma Correction here. Gamma Correction on the PC has been a specialized function, but mainstream graphics hardware is getting precise enough to support it now. Color is used for appearance, atmosphere, and realism so, it’s use needs to be handled with care. Adding Gamma Correction gives developers one more tool to enhance the viewing experience, and will also allow for better lit texture effects, among other things. It’s been there in some hardware products, mostly at the very high-end. Now, it slips into the mainstream.
Windows XP – What’s in it for gamers
Courtesy of Microsoft
Obviously, Meltdown wouldn’t have been a Microsoft event without a whole lot of talk about Windows XP. The Microsoft presentation says, “The biggest OS release since Windows 95?” Well, I wouldn’t know. This isn’t really a forum for discussing the merits, or not, of Windows XP.
Looking at it from the viewpoint of DirectX, Windows XP Professional stood out more at Meltdown. XP Home, I’ll abbreviate from now on, doesn’t have multi-processor support, remote desktop (Terminal Services), encrypted file system, and a host of other features useful to developers and IT managers. So, Microsoft thinks that hardcore gamers are going to treat XP Professional as the preferable OS version. Like it looks on the roadmap above, XP Home is Windows ME with an NT kernel (shudder).
The premise of the consumer experience, nevertheless, is over 80% of PC consumers share a computer with other family members. So, XP gives you fast user switching. I could be running an application, Outlook or IE, and go get a cup of coffee, and my kid comes around, switches to his settings and starts playing Monster Truck Madness, and it’s all hunky dory. I could comment here about how horrific the idea of administrating a single home PC for multiple users is going to be, no matter how easy it is to set up and switch, but I won’t. Obviously, Microsoft’s XP marketing team didn’t do any research on the psychology of providing tech support for your spouse. Prediction, divorce rates are set to sky rocket in PC owning households.
On a more positive note, it’s what we’re stuck with, and for most tech savvy people, it’s probably a good thing. Knowledge is power and all that, and how many of us wish we had more power at home. Now, I guess, you could be IT manager of your own family. Go figure.
Windows XP – What’s in it for gamers, Continued
So, the issues facing developers are this:
- When it’s not your applications turn to run, minimize your use of resources. Go idle. Expect that there may be a couple of users running your game, and that each one may switch in and out of his or her session at any time. I think that’s a plausible scenario, but I am a little cynical about the premise of fast user switching. Looks good in demos, but real world applications escape me.
- Full screen titles will not allow fast user switching unless you switch tasks and then switch users. There’s ways around it, but at the end of the day, you could be playing a full screen game at work and some remote technician can come in and terminate the session. Not good. Legal, but not good.
- User installing a game needs to have Admin privileges. Where each users game files or personal folders, irrespective, are stored and handled is important. Each XP machine is really a multi-user location with varying degrees of access and privileges available to the users on that machine. Games, any applicaton in fact, has to account for the number of users, and protect them from having Access Denied. Doesn’t seem like XP makes life any easier for the consumer. Sounds fine for business, but at home? Enough to drive people to a Mac.
Not your mother’s Windows. Might be your brother’s, or grandma’s. If you are the Admin, you can nuke them all.
Having said all that, what little contact I’ve had with XP has been positive. It’s a good thing to get all of us to move onto the NT kernel. It’s a nice interface, and that’s okay after staring at the same Windows icons for a couple of years. I don’t really buy into the benefits coming from the interface, and the new multi-user features, but that’s just me.
OpenGL – Does it have a future
Ran into Neil Trevett, VP of Marketing for 3Dlabs. For those of you who don’t know, Neil was the original Mr. 3D. Back in the early 90s, he was the guy on the trade show and conference circuit putting up the slides about 3D graphics on the PC, and the 3D graphics pipeline, that ended up in the deluge of 3D pipeline articles that continue to appear on some sites. I think Neil was actually the first guy to have a PowerPoint slide with the 3D pipeline on it. Blame this man for the trend.
Peter Glaskowsky, Senior Analyst on 3D and Multimedia for MicroDesign Resources, and Neil Trevett of 3Dlabs. Peter was one of only three press passes at the event which is why he is smiling. Neil’s just ebullient.
We had a brief conversation, one that we will extend on these pages in the near future, about OpenGL. Neil is very active in the OpenGL community, and there has to be some consternation among OpenGL proponents that with DirectX 8 Microsoft has, for the first time, stepped ahead of that venerable API.
The argument has always been held that OpenGL was driven by committee and was therefore too slow to adapt to the changes in the PC market. It was Microsoft’s argument, actually. In fact, the problem with OpenGL was that it couldn’t accommodate Microsoft’s ego, and as a result, it has never been fully supported by Microsoft. Lip service is given to OpenGL at best, and now, we have to wonder what OpenGL is going to do next.
Does it matter what OpenGL does? I don’t know. It would be hard to see it as marginal, or to have it relegated to scientific and research applications. We don’t see anywhere near the same level of game developer commitment to supporting alternatives to DirectX as there was in the early days of its release, but then, it’s a better API these days.
There should be a new version of OpenGL announced at Siggraph. We’ll look into it in more detail closer to that time, but for now, we have to wonder if there’s anything to be even concerned about. Does anyone care about OpenGL that much? Have to look into it.
As for Neil, I don’t think there has ever been a better 3D evangelist at any company since so, it was good to see him again, and I am looking forward to some announcements coming from 3Dlabs in August, or September, that might help the company to get back into the mainstream graphics market. We’ll also be getting him to give us some insight into what is going on with OpenGL as well.
Trident says Blaze XP is a success
Le Trong Nguyen, Assitant VP of Graphics Marketing for Trident, was at Meltdown and he took umbrage with our review of the Blade XP from Computex.
Computex 2001 Exclusive: New Graphics Chips from SiS and Trident
As he told us, while pointing out the demos running at Meltdown, “This isn’t ho-hum graphics. We’re shipping two million chips before the end of this year.”
Le Trong Nguyen of Trident looks on adoringly at Blade XP demo. Actually, it wasn’t too bad. Good enough is good enough.
Mr. Nguyen raises a good point. We have to be careful sometimes to separate our need for speed from the realities of the industry. Trident is a public graphics chip company, and although not a stellar performer, the company has done well by focusing on the category of add-in cards that retail for under $40.
As Mr. Nguyen pointed out, Blade XP is the only DirectX 7 core logic graphics chipset, something that even Intel doesn’t have. He also pointed out that Trident’s business in the mobile segment is up.
At the end of the day, for a graphics board that costs under $40, and for the motherboard that requires a good enough 3D solution with low power consumption, Trident’s Blade XP does the job. Someone’s making money in this segment, and not everyone wants the latest and greatest technology. So, I’m glad we gave Mr. Nguyen his chance at a rebuttal.
STMicroelectronics – Kyro’s future
Another guy we got to spend some time chatting with was Joe Kreiner, Developer Relations Manager for the Kyro at STMicroelectronics. He was there with people from Hercules and Imagination Technologies.
Joe Kreiner of STMicroelectronics by the Kyro II booth. Overheard at the booth was a comment about how you could afford to buy three Kyro II boards for every GeForce3 board and why would you want to buy a GeForce3 board at that price when it would be out of date in six months. These game developers have no mercy, and rolling credit.
STMicroelectronics really does hope to be a player in the graphics game. They’ve tried with varying degrees of success in the past, and now, they’re hoping that the very complexity that DirectX is pushing in games will turn out to be exactly where the PowerVR architecture finds its advantage.
Mr. Kreiner said, “Talking to high-end game developers like Epic and Valve, they’re looking at larger numbers of polygons per frame. Depth complexity is going up. For instance, Quake III’s depth complexity is 3.3, which means that for every one pixel drawn 3.3 get chopped, and in the future that figure will probably be 4 or 5 and our advantage actually gets larger.”
Well, PowerVR’s architectural advantages and its arguments to that effect have remained consistent for the last five years. The reality is that PowerVR will probably have to look at incorporating programmable shader functions, probably in the form of a Vertex Shader of some form, in order to remain competitive. I wouldn’t jump to any conclusions right now because, putting a front-end Vertex Shader, or T&L engine, isn’t necessarily a slam dunk for PowerVR. If the graphics industry is moving more towards “exposing” its hardware, the advantages of PowerVR have to be sold to developers in a way that shows that the offbeat architecture does deliver value add.
That means either the best performance, which is unlikely considering the focus of the product on lower price bands, or great sales. However, I have a feeling that Kyro is going to be a retail product, with some exposure through Taiwanese OEM board vendors for some time to come. That means STMicro have to be in for the long haul. Favorable buzz among enthusiasts for the Kyro II hasn’t necessarily translated into stellar sales for STMicro.
However, if the PowerVR architecture can keep delivering simple price/performance advantages, STMicroelectronics can continue to mine the $90-149 retail board price range, and with a few strong OEM wins, stay in the race.
Summing Up
I am going to take a closer look at Nvidia’s presentations, and give you an update on ATI in the next installment on Meltdown. Also, there was an interesting presentation from Matrox which may hint at what that company has up its sleeve for a future 3D product.
In the meantime, it’s worth taking note that the graphics industry is opening up again. Slowly, and hesitantly, but it is opening up. First of all, we still don’t enough about the underlying performance of next generation hardware. Games haven’t begun to take advantage of their features, and as they do, we still have little idea of what the true performance advantages are going to be. It’s going to be highly dependent on how the developer works it out. In most of the Meltdown presentations, the thrust of each presentation was that there are infinite permutations of how you can program the 3D pipe, and the developer is going to have to experiment to figure out what works for his game.
So, it’s an open playing field. No fixed function features means that no graphics chip company is going to be looking to steal a march on the competition by forcing developers to go down a specific route that favors their hardware. In fact, no 3dfx approaches again. It’s going to come down to educating developers on the techniques, and providing drivers, and coding help to facilitate the use of effects that may not be otherwise considered. Take a look at our Hard News section today, we have some new screen shots from ATI. At best, ATI has a lot of work to do to get developers to use all of the features it promises on the R200, no matter how cool they look.
This is an opportunity for the graphics wannabes. Whether you are a game developer or an IHV. The doors kind of open for innovation. The boundaries have been removed. For IHV’s, particularly those chip companies looking in, the direction that DirectX sets is quite simple – you don’t have to build a specialized graphics team to get in the game.
Programmable graphics processing is nothing new. Check out the TMS34020 from TI, or the i860 from Intel. Try a search on Google on both chips and you’ll find a number of references. At one time, TI had the chance to rule the graphics world with the TMS34020. At one time, Intel’s i860 was light years ahead of anything else on the PC or workstation.
These were programmable graphics chips in their time. TIGA, TI’s graphics interface for the TMS34020, was your only API to the board. Intel actually licensed some 3D APIs from my old company, SPEA (the guys are now working as FireGL as a division of ATI), for the i860. DuPont, Evans and Sutherland (I think), a lot of research outfits, used the i860 and had their own 3D software libraries for it. So, in those days (wow, the early 90s) we didn’t have any way of standardizing on these architectures, and neither TI nor Intel were the best companies to develop APIs or help software developers. Microsoft is good at both.
So, maybe its time for TI to get back in the game with its DSP expertise. Maybe Intel can get off its high horse about CPUs and take all that knowledge of registers and ALUs, and make a really powerful graphics co-processor tied to its CPUs. Microsoft has set it up for them.
It may not happen this year, but I have to believe that exposing the core functions of the graphics processor is exposing the graphics industry to new opportunities. Maybe there’s a start-up brewing, or maybe there’s a TI, or Intel waiting to jump back in. We’re no way finished with 3D.