ATI Radeon HD 4890 vs. NVIDIA GeForce GTX 275
by Anand Lal Shimpi & Derek Wilson on April 2, 2009 12:00 AM EST- Posted in
- GPUs
The Widespread Support Fallacy
NVIDIA acquired Ageia, they were the guys who wanted to sell you another card to put in your system to accelerate game physics - the PPU. That idea didn’t go over too well. For starters, no one wanted another *PU in their machine. And secondly, there were no compelling titles that required it. At best we saw mediocre games with mildly interesting physics support, or decent games with uninteresting physics enhancements.
Ageia’s true strength wasn’t in its PPU chip design, many companies could do that. What Ageia did that was quite smart was it acquired an up and coming game physics API, polished it up, and gave it away for free to developers. The physics engine was called PhysX.
Developers can use PhysX, for free, in their games. There are no strings attached, no licensing fees, nothing. Now if the developer wants support, there are fees of course but it’s a great way of cutting down development costs. The physics engine in a game is responsible for all modeling of newtonian forces within the game; the engine determines how objects collide, how gravity works, etc...
If developers wanted to, they could enable PPU accelerated physics in their games and do some cool effects. Very few developers wanted to because there was no real install base of Ageia cards and Ageia wasn’t large enough to convince the major players to do anything.
PhysX, being free, was of course widely adopted. When NVIDIA purchased Ageia what they really bought was the PhysX business.
NVIDIA continued offering PhysX for free, but it killed off the PPU business. Instead, NVIDIA worked to port PhysX to CUDA so that it could run on its GPUs. The same catch 22 from before existed: developers didn’t have to include GPU accelerated physics and most don’t because they don’t like alienating their non-NVIDIA users. It’s all about hitting the largest audience and not everyone can run GPU accelerated PhysX, so most developers don’t use that aspect of the engine.
Then we have NVIDIA publishing slides like this:
Indeed, PhysX is one of the world’s most popular physics APIs - but that does not mean that developers choose to accelerate PhysX on the GPU. Most don’t. The next slide paints a clearer picture:
These are the biggest titles NVIDIA has with GPU accelerated PhysX support today. That’s 12 titles, three of which are big ones, most of the rest, well, I won’t go there.
A free physics API is great, and all indicators point to PhysX being liked by developers.
The next several slides in NVIDIA’s presentation go into detail about how GPU accelerated PhysX is used in these titles and how poorly ATI performs when GPU accelerated PhysX is enabled (because ATI can’t run CUDA code on its GPUs, the GPU-friendly code must run on the CPU instead).
We normally hold manufacturers accountable to their performance claims, well it was about time we did something about these other claims - shall we?
Our goal was simple: we wanted to know if GPU accelerated PhysX effects in these titles was useful. And if it was, would it be enough to make us pick a NVIDIA GPU over an ATI one if the ATI GPU was faster.
To accomplish this I had to bring in an outsider. Someone who hadn’t been subjected to the same NVIDIA marketing that Derek and I had. I wanted someone impartial.
Meet Ben:
I met Ben in middle school and we’ve been friends ever since. He’s a gamer of the truest form. He generally just wants to come over to my office and game while I work. The relationship is rarely harmful; I have access to lots of hardware (both PC and console) and games, and he likes to play them. He plays while I work and isn't very distracting (except when he's hungry).
These past few weeks I’ve been far too busy for even Ben’s quiet gaming in the office. First there were SSDs, then GDC and then this article. But when I needed someone to play a bunch of games and tell me if he noticed GPU accelerated PhysX, Ben was the right guy for the job.
I grabbed a Dell Studio XPS I’d been working on for a while. It’s a good little system, the first sub-$1000 Core i7 machine in fact ($799 gets you a Core i7-920 and 3GB of memory). It performs similarly to my Core i7 testbeds so if you’re looking to jump on the i7 bandwagon but don’t feel like building a machine, the Dell is an alternative.
I also setup its bigger brother, the Studio XPS 435. Personally I prefer this machine, it’s larger than the regular Studio XPS, albeit more expensive. The larger chassis makes working inside the case and upgrading the graphics card a bit more pleasant.
My machine of choice, I couldn't let Ben have the faster computer.
Both of these systems shipped with ATI graphics, obviously that wasn’t going to work. I decided to pick midrange cards to work with: a GeForce GTS 250 and a GeForce GTX 260.
Facebook
Twitter
RSS
294 Comments
View All Comments
sbuckler - Thursday, April 2, 2009 - link
Big difference between Havok Physics and HavokFX physics. With physx you can just turn on hardware acceleration and it works, with havok this is not possible - unlike physx it was never developed to be run on the gpu. Hence havok have had to develop a new physics engine to do that.No game uses the HavokFX engine - it's not even available to developers yet let alone in shipped games. The ati demo was all we have seen of it for several years. It's not even clear HavokFX is even a fully accelerated hardware physics engine - i.e. the version showed in the past (before intel took over havok) was basically the havok engine with some hw acceleration for effects. i.e. hardware accel could only be used to make it prettier explosions and rippling cloth - it could not be used to do anything game changing.
Hence havok have a way to go before they can even claim to support what physX already does, let alone shipping it to developers and then seeing them use it in games. Like I said the moment that comes close to happening nvidia will just release an OpenCL version of physX and that will be that.
z3R0C00L - Thursday, April 2, 2009 - link
It's integrated in the same way. Many game developers are already familiar with coding for Havok effects.Not to mention that OpenCL has chosen HavokFX (which is simply using either a CPU or a GPU to render Physics effect as seen here: http://www.youtube.com/watch?v=MCaGb40Bz58">http://www.youtube.com/watch?v=MCaGb40Bz58.
Again... Physx is dead. OpenCL is HavokFX, it's what the consortium has chosen and it runs on any CPU or GPU including Intel's upcoming Larrabee.
Like I said before (you seem to not understand logic). Physx is dead.. it's proprietary and not as flexible as Havok. Many studios are also familiar with Havok's tools.
C'est Fini as they say in french.
erple2 - Friday, April 3, 2009 - link
I think you're mistaken - OpenCL is analogous to CUDA, not to PhysX. HavokFX is analogous to PhysX. OpenCL is the GPGPU compiler that runs on any GPU (and theoretically, it should run on any CPU too, I think). It's what Apple is now trying to push (curious, given that their laptop base is all nVidia now).However, if NVidia ports PhysX to OpenCL, that's a win for everyone. Sort of. Except for NVIdia that paid a lot of money for the PhysX IP. I think that the conclusions given are accurate - NVidia is banking on "everyone" (ie Game Developers) coding for PhysX (and by extension, CUDA) rather than HavokFX (and by extension, OpenCL). However, if Developers are smart, they'll go with the actually open format (OpenCL, not CUDA). That means that any physics processing they do will work on ANY GPU, (NVidia and ATI). I personally think that NVidia banked badly this time.
While I do believe that doing physics calculations on unused GPU cycles is a great thing (and the Razor's Edge demo shows some of the interesting things that can be done), I think that NVidia's pushing of PhysX (and therefore CUDA) is like what 3dfx did with pushing GLide. Everyone supported Direct3D and OpenGL, but only 3dfx supported Glide. While Glide was more efficient (it was catering to a single hardware vendor that made Glide, afterall), the fact that Game Developers could instead program for OpenGL (or Direct3D) and get all 3D accelerators supported meant that the days of Glide were ultimately numbered.
I wonder if NVidia is trying to pull the industry to adopting its CUDA as a "standard". I think it's ultimately going to fail, however, given that the industry recognizes now that OpenCL is available.
Is OpenCL as mature as CUDA is? Or are they still kind of finalizing it? Maybe that's the issue - OpenCL isn't complete yet, so NVidia is trying to snatch up support in the Developer community early?
sbuckler - Friday, April 3, 2009 - link
CUDA is in many ways a simplified version of OpenCL - in that CUDA knows what hardware it will run on so has set functions to access it, OpenCL is obviously much more generic as it has to run on any hardware so it's not quite as easy. That part of the reason why CUDA is initially at least more popular then OpenCL - it's easier to work with. That said they are very similar so to port from one to the other won't be hard - hence develop for CUDA now then just port to OpenCL when the market demands it.All in my opinion Ati want is their hardware to run with whatever physics standard is out there. Right now they are at a growing competitive disadvantage as hardware physics slowly takes off. Hence they demo HavokFX in the hope that either (a) it takes off or (b) nvidia are forced to port PhysX to openCL. I don't think they care which one wins - both products belong to a competitor.
Nvidia who have put a lot of money into PhysX want to maximise their investment so they will keep PhysX closed as long as possible to get people to buy their cards, but in the end I am sure they are fully aware they will have to open it up to everyone - it's just a matter of when. From our standpoint the sooner the better.
erple2 - Friday, April 3, 2009 - link
Sure, but my point was simply that HavokFX and PhysX are physics API's, whereas OpenCL and CUDA are "general" purpose computing languages designed to run on a GPU.Is CUDA easier to work with? I don't really know, as I've never programmed for either. Is OpenGL harder to program for than Glide was? Again, I don't know, I'm not a developer.
ATI's "CUDA" was "Stream" (I think). I recall ATI abandoning that for (or folding that into) OpenCL. That's a sound strategic decision, I think.
If PhysX is ported to OpenCL, then that's a major win for ATI, and a lesser one for NVidia - the PhysX SDK is already free for any developer that wants it (support costs money, of course). NVidia's position in that market is that PhysX currently only works on NVidia cards. Once it works elsewhere (via OpenCL or Stream), NVidia loses that "edge". However, that's a good thing...
SiliconDoc - Monday, April 6, 2009 - link
I guess you're forgetting that recently NVidia supported a rogue software coder that was porting PhysX to ATI drivers. Those drivers hit the web and the downloads went wild high - and ATI stepped in and slammed the door shut with a lawsuit and threats.Oh well, ATI didn't want you to enjoy PhysX effects. You got screwed, even as NVidia tried to help you.
So now all you and Derek and anand are left to do is sourpuss and whine PhysX sucks and means nothing.
Then anand tries Mirror's Edge ( because he HAS TO - cause dingo is gone - unavailable ) and falls in love with PhysX and the game. LOL
His conclusion ? He's an ati fabbyboi so cannot recommend it.
tamalero - Monday, April 20, 2009 - link
the ammount of fud you spit is staggeringz3R0C00L - Thursday, April 2, 2009 - link
On one hand you have OpenCL, Havok, ATi, AMD and Intel on the other you have nVIDIA.Seriously.
z3R0C00L - Thursday, April 2, 2009 - link
I'm an nVIDIA fan.. I'll admit. I like that you added CUDA and Physx.. but are we reading the same results?The Radeon HD 4890 is the clear winner here. I don't understand how it could be any different.
CrystalBay - Thursday, April 2, 2009 - link
I agree if nV wants to sell more cards they need to include the video software at no charge...