NVIDIA On Their Role in PC Games Development
GamingHobo writes "Bit-Tech has posted an interview with NVIDIA's Roy Taylor, Senior Vice President of Content/Developer Relations, which discusses his team's role in the development of next-gen PC games. He also talks about DirectX 10 performance, Vista drivers and some of the upcoming games he is anticipating the most. From the article: 'Developers wishing to use DX10 have a number of choices to make ... But the biggest is whether to layer over a DX9 title some additional DX10 effects or to decide to design for DX10 from the ground up. Both take work but one is faster to get to market than the other. It's less a question of whether DX10 is working optimally on GeForce 8-series GPUs and more a case of how is DX10 being used. To use it well — and efficiently — requires development time.'"
As an early 8800GTX adopter, I'd like to tell NVIDIA where they can shove this $700 paperweight..
http://en.wikipedia.org/wiki/QXGA#WQUXGA
Apparently, the existing monitors at WQUXGA (worst. acronym. ever.) resolution run at 41hz, max. These days, top of the line game systems will pump out upwards of 100 frames/sec in some cases. A 41hz refresh rate is essentially caps you at 41 FPS, which is enough to turn off any gamer looking at blowing that much on a gaming rig.
"As the only manufacturer with DirectX 10 hardware, we had more work to do than any other hardware manufacturer because there were two drivers to develop (one for DX9 and one for DX10). In addition to that, we couldn't just stop developing XP drivers too, meaning that there were three development cycles in flight at the same time."
Didn't ATI kick out some DX10 hardware the other day? I'm sure the ATI x29xxx is DX10.
"Our research shows that PC gamers buy five or more games per year, and they're always looking for good games with great content.
Interesting, but makes me wonder what they lay in the definition PC gamer.
"Tony and David are right, there are API reductions, massive AA is 'almost free' with DX10. This is why we are able to offer CSAA [up to 16xAA] with new DX10 titles - the same thing with DX9 just isn't practical. Also interesting, but I'm skeptical. Turning on AA is just one API call, how does AA affect overhead?
"So yes we will see big performance jumps in DX10 and Vista as we improve drivers but to keep looking at that area is to really miss the point about DX10. It's not about - and it was never about - running older games at faster frame rates. Wait, rewind. Are he saying my DX7/8/9 games will run faster once Nivida gets their DX10 drivers together? Or is he saying games with DX9 level of graphics will run faster if ported to DX10?
"Five years from now, we want to be able to walk into a forest, set it on fire and for it to then rain (using a decent depth of field effect) and to then show the steam coming off the ashes when the fire is being put out."
No, I can do that in real life. A Pyromaniacs VS firefighters burn fest OTOH....
"Given how many copies of Vista are in use, a surprisingly small number of people came back to say they were not happy with our Vista drivers when we launched Vista Quality Assurance. Within a month the number of reported problems had been halved."
Customers are funny, if you ignore them long enough eventually they go away.
I don't want this to sound like the famous "640k should be enough for everyone", but...
WQUXGA, 3840x2400, or nine million pixels.
Sounds like overkill to me. I mean, I'm used to play my games @ 1280x1024 and i feel this resolution, maybe combined with a wee bit of AA, does the trick.
I'd rather see all that horsepower invested in more frames/sec or cool effects. I know, it's cool to have the capability, but it makes me wonder about what another user posted here regarding the 8800 being a 700$ paperweight 'cause of early adoption. You'll have a card capable of a gazillion pixels on a single frame, yet no monitor capable of showing it fully, and when finally the monitor comes out or achieves a good price/value relationship, your card is already obsolete. Null selling point there for moi.
Just my "par de" cents.
Speaking of Nvidia PC game development. Why the hell are all their new versions of their useful utilities like FX Composer 2 (betas I tried to test) now requiring Windows XP (with SP2) and no more Windows 2000 support? Win2k and WinXP have virtually zero differences in hardware support and driver system architecture. I should know since I've programmed a few drivers using Microsoft's driver development kit and according to the docs nothing has changed from Win2k to WinXP for drivers and majority of the APIs, just additional features.
The thing that pisses me off is that Nvidia seems to have done this for absolutely no reason at all and Windows 2000 is still a fine operating system for me. I have no reason at all to switch to Windows XP (and hell no to Vista), I especially don't care fot the activiation headaches (I like to switch around hardware from time to time to play around with new stuff and go back once I've gotten bored with it if I don't need it, such as borring a friends Dual-P4 motherboard).
Anyway, my point/question why must Nvidia feel the need to force their customers who use their hardware for developing games into later Windows operating systems like that? Anybody got any tips on how to 'lie' or disable the windows version check to force say FX Composer 2 to install on Windows 2000? It isn't like we're talking about Windows 98 here, Win2k is a fine OS and in my opinion actually the best one Microsoft has ever done.
My question would be how NVidia's helping the game developers write for and port to Linux. If popular cames were more compatible there, it'd be a lot easier to get converts; and I'd expect the game developers would be happy to see more of my software dollars go to their products rather to OS upgrades.
Yeah, but think of the points you could rack up in Scrabble.
*sigh* Where do people come up with this garbage? Look at some evidence already instead of making stuff up:
http://mckack.diinoweb.com/files/kimpix-video/
Actually, you know, it's sorta funny to hear people ranting and raving about how 32 bit killed 3dfx or lack of T&L killed 3dfx, without having even the faintest clue what actually happened to 3dfx.
In a nutshell:
1. 3dfx at one point decided to buy a graphics card manufacturer, just so, you know, they'd make more money by also manufacturing their own cards.
2. They missed a cycle, because whatever software they were using to design their chips had a brain-fart and produced a non-functional chip design. So they spent 6 months rearranging the Voodoo 5 by hand.
The Voodoo 5 wasn't supposed to go head to head with the GeForce 2. It was supposed to, at most, go head to head with the GF256 SDR, not even the DDR flavour. And it would have done well enough there, especially since at the time there was pretty much no software that did T&L anyway.
But a 6 month delay was fatal. For all that time they had nothing better than a Voodoo 3 to compete with the GF256, and, frankly, it was outdated at that time. With or without 32 bit, it was a card that was the same generation as the TNT, so it just couldn't keep up. Worse yet, by the time the Voodoo 5 finally came out, it had to go head to head with the GF2, and it sucked there. It wasn't just the lack of T&L, it could barely keep up in terms of fill rate and lacked some features too. E.g., it couldn't even do trilinear and FSAA at the same time.
Worse yet, see problem #1 I mentioned. The dip in sales meant they suddenly had a shitload of factory space that just sat idle and cost them money. And they just had no plan what to do with that capacity. They had no other cards they could manufacture there. (The tv tuner they tried to make, came too late and sold too little to save them.) Basically while poor sales alone would have just meant less money, this one actually bled them money hand over fist. And that was maybe the most important factor that sunk them.
Add to that such mis-haps like,
3. The Voodoo 5 screenshot fuck-up. While the final image did look nice and did have 22 bit precision at 16 bit speeds, each of the 4 samples that went into it was a dithered 16 bit mess. There was no final combined image as such, there were 4 component images and the screen refresh circuitry combined them on the fly. And taking a screenshot in any game would get you the first of the 4 component images, so it looked a lot worse than what you'd see on the screen.
Now it probably was a lot less important than #1 and #2 for sinking 3dfx, but it was a piece of bad press they could have done without. While the big review sites did soon figure out "wtf, there's something wrong with these screenshots", the fucked up images were already in the wild. And people who had never seen the original image were using them all over the place as final "proof" that 3dfx sucks and that 22 bit accuracy is a myth.
A polar bear is a cartesian bear after a coordinate transform.