AMD: "[Developers use PhysX only] because they’re paid to do it"
- Thread starter Richard
- Start date
the batman stuff is true at the very least. You could hack the driver id file and change it to a geforce and you'd get fsaa on radeons. There is an old thread about it here.
Its why i wont buy batman 2 until i know all features work on my 5850
In fact the "GOTY" version of Batman apparently has support for AA on Radeons out-of-the-box. Normal version never got that patch, though.
Does a six-month head start really give you an advantage for a game releasing in 18 months? Seems like lots of time for Nvidia to exert their considerable "influence" ($$$$).
Did he really say that after trying to promote the laughable tessellation usage in AvP and Dirt2? Wow. Yes Richard, let's only apply tessellation where your cards can handle it and it makes absolutely no impact on IQ.
If Nvidia's message is tessellation, what's AMD's message?
I think there's a clear difference between wasting 75% of your tesselated work (what nV is promoting) and applying tesselation everywhere with sense.
And tesselation definately has impact on IQ in both cases of AvP and DiRT2, though I personally don't see the point in tesselating audience in DiRT2 for example, the water differences are huge, and flags.. well, they're so-so
So I take it you're refuting this?:
Thats their response, do you care to state how that is wrong in any specific way? Im not being snarky, im just asking.
Bouncing Zabaglione Bros.
Legend
I think Huddy's point is that Nvida's only advantage is tessellation, so that's what they want to fill every game with. It makes Nvidia products look better than they are, and makes AMD products look worse than they are. An ideal world for Nvidia is where every game is Stone Giant and they can crow about how much faster they are than anyone else.
It's exactly the same as what Nvidia did with PhysX - spend their own time and money filling games with the one thing they can do better than anyone else because otherwise they can't compete.
The truth is that tessellation is nice when it's used well, but a game has to have a lot more to it than tessellation.
It's exactly the same as what Nvidia did with PhysX - spend their own time and money filling games with the one thing they can do better than anyone else because otherwise they can't compete.
The truth is that tessellation is nice when it's used well, but a game has to have a lot more to it than tessellation.
Where did I refute that? But it's a synthetic benchmark and lots of synthetic benchmarks have unrealistic workloads - bilinear texturing for example. The only reason they're complaining is because they can't handle those levels of tessellation.
So their "message" is to complain about Nvidia? That's not a message....
@Kaotik: you probably played a different version of AvP than everyone else. Even in zoomed in marketing screenshots there's no difference.
Yes, obviously they will push their advantage. They didn't spend all those transistors on "Polymorph units" and "GPC's" for nothing. AMD was advocating trivial uses of tessellation before Fermi hit the market. So what's changed? Of course we don't want tessellation applied in a wasteful or unnecessary way. But we definitely want it at a level higher than what Evergreen supports. My question still stands though - what is AMD's message?
So their "message" is to complain about Nvidia? That's not a message....
@Kaotik: you probably played a different version of AvP than everyone else. Even in zoomed in marketing screenshots there's no difference.
Yes, obviously they will push their advantage. They didn't spend all those transistors on "Polymorph units" and "GPC's" for nothing. AMD was advocating trivial uses of tessellation before Fermi hit the market. So what's changed? Of course we don't want tessellation applied in a wasteful or unnecessary way. But we definitely want it at a level higher than what Evergreen supports. My question still stands though - what is AMD's message?
I don't know if NVIDIA really has that much "influence" to spread around these days…
Bouncing Zabaglione Bros.
Legend
That games are more than just tessellation.
Does anyone think that Nvidia's extreme tessellation performance is balanced, or that games are going to need/use that level of tessellation in the next couple of years? Is it useful in any game you are playing now?
ATI used to get criticised for too many forward looking features, where Nvidia was always "works great in current games, you'll upgrade in a couple of years when you need more". Are we now saying the situation is reversed and Nvidia are touting a forward looking feature that's not going to be useful before the product gets superseded?
You're calling this no difference?
vs
(and no, it's not zoomed in)
http://www.pcgameshardware.com/aid,...-benchmarks-and-graphics-comparison/Practice/
I think its more than six months. Sept 2009 to June 2010 right ?
Thas about 9 months and don't forget that the ati parts were avalible in much higher quanitys Ati could have flooded developers with the very cheap radeon 57x0 products in which nvidia still has no answer to .
So not only multiple months advantage but having much more avaliblity inside dev houses will help. Having one or two geforce 4x0 products vs having dozens of ati 5x00 products will mean ati cards are tested and used more in development.
Yes I think that the tessellation performance is balanced, and use I think games should need/use that level of tessellation. Why don't you? It is pretty much awesome and has been for awhile. I always was happy when ATI pushed it in the past and hoped it would catch on in a big way. The stone giant thing was weird though. Something always seemed off about it.
Bouncing Zabaglione Bros.
Legend
But they don't. And it doesn't look like they will for some time. Yes, it's nice to have, it's nice to encourage future tech by getting it out into the market, but what good is it to me if no one needs such massive tessellation performance above everything else because there's no software that needs such extreme tessellation performance?
It depends on what the cost trade-off is in terms of transistor budget. If other things have to suffer, then some of those other things might be more relevant to what I want to play for the next couple of years.
For instance, is the Nvidia tessellator useful enough for me to put up with all the other disadvantages of that particular product that brought me that tessellator to begin with? Does it have great performance when playing games, or is it simply dedicating so much of it's processing power to it's one-trick-pony of tessellation for demo benchmarks?
I suspect that Nvidia went overboard on tessellation because it was easy and cheap in terms of transistor budget, but do I want to pay for it now, for games coming down the pipe in a couple of years? Is it important enough (either in current performance or current usage) to live with the negatives of Fermi? For me the answer is no, I don't need that extreme tessellator performance if I'm losing out elsewhere.
It can have advantages too. Nvidias geometry aproach is faster without tesselation too. If your pixels are light on performance(and you have lot of it) than wasting 75% is not a big problem. (quadro cards for example)
Edit: But yes in games until tesselation on and off doesnt show any usefull difference than both aproaches are waste.
Is there any particular aspect of rendering that Nvidia has neglected with Fermi? Nvidia is pushing tessellation and compute which is a big "duh" given those are the big new features in DX11. In what respect is AMD pushing the envelope?
I wouldn't really call it forward looking, Quadros seem to be benefiting greatly from the new geometry setup. Also, the only bottleneck is the content authoring pipeline - there are no technical barriers to better use of tessellation. I wouldn't be surprised at all if games launching in the next 6-12 mths increase the geometry workload and make better use of all those triangles.
Yep.
Initial development yes but Fermi has been around a long time. Long enough for it to be included in the optimization process for any game coming out 6 months from now.
If it's so cheap and easy why did it now make it to DirectX and why didn't AMD also beef up tessellation performance?
Bouncing Zabaglione Bros.
Legend
Perf/watt, perf/$ as you well know. Nvidia doesn't reach required performance without massive overclocking and associated heat/noise/power. Without that, we'd be looking at every aspect of Fermi's rendering and saying the performance is under par.
Quadros are not mainstream nor gaming cards. It's irrelevant to mention them and is just a tangent on your part.
Different architectures of course. I could ask why if it was so easy for AMD to make cards that don't burn like the sun, why doesn't Nvidia.
ATI tried to push it into DX years ago with Truform, but Nvidia resisted as they had no capable hardware and no plans for it. Now Nvidia has tessellation as it's one trick pony, we're supposed to want and need every game to be nothing but tessellation?
It's just another single message Nvidia are going to use to try to leverage their hardware at the expense of everyone who doesn't have a top of the range Nvidia card. Not just AMD owners, but everyone who has a lower Nvidia card where the tessellation performance drops off a cliff will suffer for Nvidia's attempt to bend the market to it's own will and make everything all about tessellation, instead of about the best results for all gamers.
Richthofen
Newcomer
I thinks Nvidia's advantage is not only tessellation. It's tesselation, AF, 3D vision, PhysX, Cuda and the whole software environment around is. Nothing stops AMD from inventing by themselfs. To be honest Huddies message sounds like a whining child. Stop whining and do something more than just standard food in terms of graphics.
As a consumer I want everything to work across platforms. Why should buying an intel igp or an amd gpu stop me from using Physx ?
Why should Nvidia be able to buy basic features like AA ?