DegustatoR
Legend
Can you compile a C++ program for the R5xx/6xx/7xx assembly? Can it be ran on a videocard with that GPU without a need for a driver?
Never. And your point is?
AMD: R8xx Speculation
- Thread starter Shtal
- Start date
Uhm, no. You're saying that it makes more sense for an ISV to implement a locked/exclusive/dying tech over the latest/greatest/most widely adopted standard?
I really disagree with you, no way I see PhysX being more dominant than DX11.
DegustatoR
Legend
Damn, that's not what i'm saying at all.
For ISV sense = money return. And money return = installed base. ISV don't really care how locked/exclusive something is if it's "most widely adopted standard". And _for now_ PhysX is clearly more widely adopted than DX11. And thus it's probably safer to bet on wider PhysX usage in the nearest future than on DX11 usage.
What don't you understand?
All i'm saying is that it's funny to see people praising DX11 and bashing PhysX at the same time. To me it all comes down to graphics in games I play. And for now I see more potential in PhysX then in DX11 for deliveling better graphics. _For now_.
DX11 is an API. PhysX is a middleware solution that can be ported to DX11 and run over DXCS. How's the first can be more dominant than the second? It's like saying that UE3 can't be more dominant than DX9 (while it is more dominant in practice thanks to gaming consoles).
Deg, if you'd indulge me, which language do you use/which programming language are you most proficient in?
DegustatoR
Legend
I'm not a programmer at all.
Still, your point?
But before making it you probably should ask the same questions to my oppenents in this thread. Just an advice.
See, your opponents might have some programming experience and are snickering at the thought of "Oh hey, my piece of CUDA code here can be recompiled for OCL when I press F9... CUDA is like a kind of C, right? C++? C#???? anyone?"
Silent_Buddha
Legend
Just a thought point, do you think ATi has put up money for developers of DX11 games? In the form of ATis form of TWIMTBP? I doubt many studios will choose to support DX11 at first due to the dollars involved in extra programming hours vs. how many boxes will get moved soley for the DX11 checkmark.
But never discount motivated programmers that just want to see sweet features in their games
-Plack
I doubt it. It's already been mentioned by a few devs that many of the performance oriented features of Dx11 (inherited from Dx10.1) are easily implemented into any Dx10 engine.
So you immediately gain performance increases for any Dx10 engine that wasn't already Dx10.1. Additionally, when Nvidia finally releases Dx11 hardware they'll finally get those Dx11/10.1 performance increases also.
The fact that Nvidia will HAVE to support Dx11 means more devs (TWIMTBP) will start incorporating Dx10.1 features that Nvidia have been discouraging them from using.
So right there you have the beginnings of some massive performance increases for everyone that supports 10.1 and 11.
Add to that, there's already devs showing Dx11 specific features in their upcoming games.
Battleforge has already been shown to the press using Dx11 Compute Shaders. Dirt 2 has already been delayed a couple months due to needing Dx11 for their Dx11 specific features. AvP has been shown with Tesselation. The next Stalker has also been shown with Dx11 specific features
And for most of these. Unlike PhysX, most devs don't have to take a massive performance hit to enable extra features over the base version.
Stalker added more features with Dx10.1 for example yet kept FPS the same. Most other devs used it to increase performance over the base Dx10.0 and 9.x rendering path.
Sacred 2 added some blowing leaves with PhysX and took a massive performance hit in doing so. Mirrors edge added far nicer PhysX effects but also took a large hit when enabling them on PhysX capable hardware. I have yet to see a single PhysX game that has used it AND gained FPS over the base version.
Regards,
SB
Not much more ridiculous than most of the other marketing bullcrap that surrounds GPGPU/stream compute/hybrid compute/call-it-what-you-will and the alleged impact it'll have on proven real-world codes.
DegustatoR
Legend
My opponents with some programming experience generally understand what i'm talking about.
How's your programming experience by the way?
I wouldn't be too sure about that.
"Oh what a wondrous gift 'twould be to see ourselves as others see!"
Never going to happen.
I wouldn't be so sure. I can see nvidia porting at least parts over to OpenCL, so that it'll run on next-gen consoles as well -- if they have non-nvidia hardware (there is a slight possibility that all these delays etc. we see from nvidia is due to working concurrently on next-gen console HW ...). After all, PhysX is popular among game developers (see for example http://www.bulletphysics.com/wordpress/?p=88), and they surely want to keep it that way. Of course, the "full" PhysX might still require CUDA for some parts (for instance, stuff that is not exposed in OpenCL), but I wouldn't be surprised if they port at least some parts over to OpenCL.
Funnily enough, Havok might do the same and go OpenCL/LRBni, so it will run sort of ok'ish on AMD/NVIDIA hardware, but for the full performance you'll need LRBni, which would leave AMD in a bad position.
DX11 will add support for things on all dx10+ class hardware. That market is nvidia g80+ and ati r600+ , running on vista+.
GPU physX is g80+ on xp+vista.
I suspect "g80+ on xp+vista+win7" vs. "g80+ & r600+ on vista+win7" ratio is gonna be skewing towards the second quite fast. But is not there yet.
Simplifying it, it's a ratio of Nvidias g80+ market share vs. Vista+Win7 market share. Guess which one wins.
Steam August HW survey:
28.95% are DX10 systems.
60% are DX10 capable (30% DX10 running on XP).
As good as *all* new system are DX10 capable (Vista/win7 force feeding and running out of DX9 cards in the market), expect windows 7 and "back to school" and xmas season to up this percentage, that's the next 3 months.
Nvidia market share (total systems): 65% (how many of those are dx9 class (no-physx) or so weak enabling physX will kill the few FPS you got?)
Quickly checking (im not sure about the % relative to each other): 65 minus
8600 5%
7600 2%
...
In summary, as a developer i would do the exact oposite of what you propose. I would concentrate on DX11 feature giving me *and my customers* an advantage on all current and *future* DX10+ systems (like deferred DX11 contexts).
PhysX... well, maybe if NV drops some moneys and parachutes in some monkeys to code it up for me, or if it's already in the engine.
Actually they will
brain_stew
Regular
Just a thought point, do you think ATi has put up money for developers of DX11 games? In the form of ATis form of TWIMTBP? I doubt many studios will choose to support DX11 at first due to the dollars involved in extra programming hours vs. how many boxes will get moved soley for the DX11 checkmark.
But never discount motivated programmers that just want to see sweet features in their games
-Plack
What significant dollar investment is that? Porting a DriectX 10 engine to DirectX 11 is a couple of days work at most by all accounts, and doing so will give advantages to everyone running DX 10/10.1/11 and that's a huge proportion of the market. I just don't see what the motivation (Apart from Nvidia pressure) is to stay with DX 10 at this point, when DX11 is a better solution for DX10 hardware than DX10.
Most modern games are expected and do indeed ship with a DX10 codepath these days, so why shouldn't any titles like this that were formally targeting DX11 not go ahead and upgrade to DX11? Sure these sort of situations aren't going to yield many titles with DX11 exclusive rendering modes, but if a couple of evening's coding is going to yield a significant perf. increase for a large part of your market and allow you to market your title as DX11 ready why wouldn't you go this route?
You do realise there's a version of DX11's compute shader that target's DX10 hardware, right? Targetting that'll give you access to a huge proportion of the market, much more than CUDA could ever hope for. Sure, it may be more limited than CUDA but then the version of compute shader that targets DX11 hardware almost certainly offers certain things that CUDA on current Nvidia hardware can't either, so the point seems moot
Last edited by a moderator:
Because all of that is bad business sense.
Any C++ program needs an operating system for a (normal) end-user to run it, surely?
True, I should revise to say 'won't be many DX11 titles for some time'. There is what, Dirt2 and the next Crytek game and a few others with DX11 patches? Both of which look sweet
I do hope DX11 gets here soon, but there are still a lot of games that don't even put much of DX10 to use.
-Plack
THIS year we count *four*; there was a LOT of reasons the devs ignored DX10 but not DX11 imo.
http://www.pcgameshardware.com/aid,...of-Pripyat-Dirt-2-and-Alien-vs-Predator/News/
NVIDIA's strategy is to give away a physics engine which is only GPU accelerated on one platform (and is heavily optimized for Cell). Yes the fact that it's only GPU accelerated on one platform hurts it's adoption, but that's not the point. They aren't in it for charity. For anything non NVIDIA they will only try to keep close enough to performance parity with the competition so it's adopted ... and no more.
As long as Havok/Bullet don't pull too many developers away by supporting supporting OpenCL or DX compute PhysX won't support them either. They might do some work already so they can release it near the same time, but that's it.
Similar threads
