• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD and NVIDIA butt heads over physics

Cyber-Mav said:
manufacturers own physics implementation will always be superior to an open source based approach when it comes to performance.
Click to expand...

Yeah like glide was better than OpenGL or Direct3D :)

Seriously, yes it was when there was nothing else, but once the hardware agnostic implementations pick up pace the market just won't accept a lock-in to a particular hardware, and once that happens the more "open" protocols quickly overtake the proprietary version in features, performance and convenience.
 
Last edited:
exactly, glide provided more performance than opengl and direct3d did at the time. shame the voodoo guys couldnt sort out thier new hardware otherwise id expect glide to still be used even today.
 
No it didn't it was all that was available. 3DFX switched from glide to OpenGL and Direct3D as soon as they realised it was a lost cause which was a long time before they went under. Same thing will happen to Physx and CUDA

What happened to glide

1. 3DFX release Voodoo1 using glide because there was nothing else, Direct3D was rubbish, and OpenGL was for CAD
2. OpenGL is used in Quake and 3DFX port the calls to glide with a special GLQuake version driver
3. Direct3D starts to take shape and the competition starts to hot up hardware wise with Nvidia and PowerVR so 3DFX realise they need proper Direct3D and OpenGL drivers
4. Glide useage dies out as Direct3D and OpenGL take over
5. 3DFX make some bad business decisions mainly down to inexperience and arrogance
6. 3DFX get absorbed by Nvidia
... etc
 
Last edited:
To be honest I think if you dig deep enough in Nvidia you'll probably find some of the same people responsible for Glide managing CUDA. The situation just seems so similar - they have some hardware ideal for breaking into a new market, so they try and make sure that ecosystem becomes rooted around their company as much as possible by attempting to create a de facto industry standard that is centred around their hardware.
 
errr no you got it all wrong mate. re read what i have written, i said glide was faster than open gl and direct3d which is true. since there were games out at the time that could run in different renderers and using glide on the voodoo cards always resulted in better performance than running a different api.
 
Heh I think you have your green tinted glasses on tonight Mav :) the way i wrote it is what happened. I'm not disagreeing with you, yes at an early point in history glide was the fastest until driver development caught up on the more open formats. 3DFX virtually abandoned glide by the time the Voodoo2 was superceded.
 
Biffa said:
Heh I think you have your green tinted glasses on tonight Mav :) the way i wrote it is what happened. I'm not disagreeing with you, yes at an early point in history glide was the fastest until driver development caught up on the more open formats. 3DFX virtually abandoned glide by the time the Voodoo2 was superceded.
Click to expand...

If it's fact, he can't disagree really.
 
what you have written is correct in the way glide died, but im not discussing that topic here, we were discussing how the proprietary api of the manufacturer, glide, in this example was faster than when given the choice running opengl or direct3d.
 
Only at an early stage of development of the market was glide faster, once the development of OpenGL drivers properly took off and Direct3D (at a bit slower rate) glide was redundant and useless and stopped being used.

Same thing will probably happen to CUDA and Physx hopefully. And Physx isn't even Nvidias own protocol!
 
Well of course, if you code for one target platform it will in most cases (provided the support is there, and the competing hardware is more or less equal in performance) be faster than similar code targeting multiple platforms. But then if you were to re-code every game for each graphics vendor, that would take large amounts of time, effort, and importantly for the companies writing this software, money.

If the majority of developers were to only code for one vendor, that vendor would gain a huge amount of the market share and the market would more than likely stagnate from lack of competition and so in the long run the performance gains from writing for that API would be mitigated from the slowed rate of development in the market (say everyone were writing for Glide in 2010, the equivalent 3dfx boards might only be as fast as say a 6800 GT because 3dfx would not have had to compete with any other companies as those companies would not have had the cash to develop and market anything to compete with them).
 
Mr Paul said:
IMHO ATI can't and shouldn't bitch about PhysX until they have their own working system out in the wild GPU accelerated physics engine.
Click to expand...

Or NV could open PhysX up and allow anyone who wants to implement it on their hardware thus making it de-facto standard
 
I actually think that NV owners are the biggest loosers from the way Physx is implemented. They're paying a premium for it but will never get the full potential from it in any decent games.
 
I'm slightly ignorant about the performance of one technology or another, but I thought that OpenCL would perform prety much exactly the same as CUDA does on an nVidia card. Is this not the case? Though nVidia had a hand in the design of OpenCL and it was very much like CUDA?

I think one of the reasons we see apps using CUDA is because nVidia is helping them with it. If they used OpenCL instead, nVidia wouldn't be interested. Just my opinion :D
 
No one knows yet, OpenCL is still waiting for decent drivers from both Ati and Nvidia. Same with DirectCompute from M$

Once the the investment in drivers are there I'll predict that the differences in performance will be small to non-existant.
 
I honestly don't understand Nvidia's decisions with Physx. I see that they wanted to control the market but I also see that no mainstream development is going to make use of it to the extent of shutting out 30-35% of the market (ATI users). I can't see how anyone at Nvidia saw a different outcome.

AMD/ATI took the obvious business decision of promoting an open standard. This gives them popular support, support in niche (but important) markets such as the scientific community and kudos from developers. If they had gone the proprietary route they would be a joke; that would have just pushed people to PhysX instead. This was nothing more than a shrewd business decision.

Nvidia could have allowed other implementations of PhysX, ensuring its future, directing the development (that ATI would have to implement) and improving the whole ecosystem of PC gaming. With this approach PhysX could have been ubiquitous already and Nvidia could have had a huge leg up on their competition. Why didn't they do this? The only reasons I can see are greed and incompetence. Happens to every company sooner or later, look at 3dfx.
 
Last edited:
Last edited:
You must log in or register to reply here.
Back
Top Bottom