Nvidia GT300 core: Speculation

[Sxotty]

It's not unethical, it just doesn't engender competition ... which in the end is always going to make things worse for us.

Not really.

If PhysX was hugely popular then some open standard would have way more impetus to get adopted. Once there was an alternative Nvidia would be the one behind if they did not support it well.

No gpu physics isn't better than nvidia PhsyX if those are the options. Until there is another viable option that AMD actually has support for it is just complaining.

 

[MfA]

If PhysX was hugely popular

If it was competitors would be dead in the water and we would be fucked.

 

[KimB]

I agree that the simplest (and hence most likely) explanation is that this is just NVidia FUD. And I know that being optimistic about NVidia is a tad passe right now, but let's for the moment consider that they might have something.

How big would G300 have to be to warrant the kind of statement that's been quoted? If (big if) G300 ships in December, how much faster/better does it have to be to warrant passing on the ATI part? 3TFlops? more? Complete rearchitect of G80 that out-larrabee's larrabee? I dunno, I haven't heard any talk like that....

FWIW, the gents at nvidia seem to be burning the midnight oil of late (I still pass their campus, but now on the way to and from diving, not gymnastics ).

-Dave

Actually, by definition it's still FUD, whether or not they have something.

 

[DegustatoR]

It only goes to show AMD doesn't have a well greased fud machine for picking that up.

Of course they have. Saying that Bullet Physics is better than PhysX is a nice example of AMD's FUD.

 

[Sxotty]

If it was competitors would be dead in the water and we would be fucked.

No then AMD would already have physics running on their card and in games. That is what would happen.

Regardless looking at the rage3d interview linked in the other thread it seems AMD might finally be serious about getting GPU physics going. I am certainly happy about that. I would probably buy a 5850 if they had such support already, but waiting a bit won't hurt.

 

[Richard]

Saying that X is better than Y is a nice example of Z's FUD.

FUD isn't saying x is better than y. It's saying that y sucks and is going down the drain for <whatever_obscure_reason> that isn't readily verifiable. Substitute the variables by your favourite/hated IHV and features.

 

[MfA]

No then AMD would already have physics running on their card and in games. That is what would happen.

Which would either also be proprietary and section of the games market into warring incompatible camps in which case we're still fucked, just slightly differently ... or which would not be proprietary and run equally well on NVIDIA hardware and thus not do a damn thing about the competitive advantage NVIDIA would hold. If PhysX was wildly popular it would be bad for us in the long run ... proprietary APIs exist for the exact purpose of being anti-competitive, there is nothing unethical about it but it's not good for us.

 

[LordEC911]

This remark was pretty sad though:
"Personally I think that the Vantage team should have never implemented a technology that can boost your score, but only be used by one side."

Back when Futuremark implemented PhysX, it was still an Ageia product, and their PPU could be used in combination with any GPU. Futuremark realistically had no way of knowing that nVidia or any other GPU manufacturer would buy out Ageia and make it into something that can only be used by one side.

Wait... what?
Vantage was released in Feb '08 and that is also when Nvidia bought Ageia...

Sure, they already had coded all the PhsyX stuff into the benchmark but why couldn't they disabled it for certain runs or maybe made it an optional/additional score that doesn't add to the total...

Lots of different ways to not have it skew the results.

 

[trinibwoy]

Hmmm strange. The first pic is from Hardware.fr's 4870 review, the second is from the 5870 review. So GT200 8xAA Z performance improved dramatically but there's still a much bigger hit compared to AMD's stuff so something else is to blame.

 

[Sxotty]

Which would either also be proprietary and section of the games market into warring incompatible camps in which case we're still fucked, just slightly differently ... or which would not be proprietary and run equally well on NVIDIA hardware and thus not do a damn thing about the competitive advantage NVIDIA would hold. If PhysX was wildly popular it would be bad for us in the long run ... proprietary APIs exist for the exact purpose of being anti-competitive, there is nothing unethical about it but it's not good for us.

You can get upset, but I still disagree. Glide went away. Things that are popular can cease to be. Walkman, discman where art thou?

 

[Vincent]

According to my sources they're pushing for a december release. GT212 was resurrected for a while, but supposedly canned indefinitely because it couldn't keep up with Cypress. So they're putting all their effort into GT300 and partners should be getting designkits next month. If this is true, more info should be out in the next couple of weeks.

G300 is up and running. ( Beta Driver ).

 

[Richard]

You can get upset, but I still disagree. Glide went away. Things that are popular can cease to be. Walkman, discman where art thou?

New games no longer use GLIDE. Games that already used it will continue to. That's the reason why I can't play EF2000 with hardware acceleration despite the fact I have a computer probably 37 times faster than the PC I had when the game was released, or the GLIDE patch for it at any rate.

I kept a VooDoo 2 card for exactly this reason where games either only had GLIDE or also a crappy-lesser-quality D3D/OGL mode. Unfortunately the old PC that used to house it stopped posting one day. Not that it would be convenient having a decrepit piece of computer history in my living room just to play a handful of games that use a God forsaken API that used to be the bee's knees of a particular brand.

It's bad enough when supposedly backwards compatible APIs don't live up to the promise (D3D and Virtua Fighter PC I'm looking at you) but the longer a proprietary API lives on, the greater the number of games you won't be able to play in 10 years time when that API is forgotten.

PhysX isn't quite in the same league as GLIDE obviously; not with the running on the CPU bit. In ten years time we may even be able to play a PhysX-game on the CPU and get double digit fps too!

 

[KimB]

At the time that Glide was common, there weren't a whole lot of other options, though. By the time that DirectX and OpenGL became viable, most everybody made use of them.

And bear in mind that the one popular game of that era that used an open API, Quake, had IHV's release mini GL drivers specifically for the game. That's just not a feasible path for most devs.

So as far as I'm concerned, it might be a bit annoying that there are old Glide-only games out there, but there also weren't a whole lot of options at the time. If ATI were pushing hard for an open physics API right now, they might have room to talk. But in the mean time, there just aren't a whole lot of other options than nVidia's proprietary API.

 

[Richard]

<snip>

So as far as I'm concerned, it might be a bit annoying that there are old Glide-only games out there, but there also weren't a whole lot of options at the time. If ATI were pushing hard for an open physics API right now, they might have room to talk. But in the mean time, there just aren't a whole lot of other options than nVidia's proprietary API.

I agree completely. I'm merely pointing out to Sxotty that just because something disappears doesn't mean its (ill) effects will stop as well.

 

[KimB]

I agree completely. I'm merely pointing out to Sxotty that just because something disappears doesn't mean its (ill) effects will stop as well.

Yeah. I'd also like to point out, though, that we do have Glide wrappers available. If GPU-accelerated physics catches on, we'll probably have PhysX wrappers, too.

 

[neliz]

Yeah. I'd also like to point out, though, that we do have Glide wrappers available. If GPU-accelerated physics catches on, we'll probably have PhysX wrappers, too.

You think nVidia will start writing those after everyone abandoned ship? or would they be so nice to have them available for OCL this year?

 

[KimB]

You think nVidia will start writing those after everyone abandoned ship? or would they be so nice to have them available for OCL this year?

Why would nVidia write one? 3dfx never wrote a Glide wrapper that I am aware of.

 

[XMAN26]

Why would nVidia write one? 3dfx never wrote a Glide wrapper that I am aware of.

No, but creative wrote one for Glide for teh GeForce256 cards when they launched. Awesome app that worked perfectly. The caught hell from 3DFX for it tho.

 

[Scali]

Wait... what?
Vantage was released in Feb '08 and that is also when Nvidia bought Ageia...

So their product was basically finished by the time nVidia got involved. My point exactly. It had probably been in development for 2 years or more.
And PhysX acceleration IS optional. You have the option to disable it when you run the benchmark.

 

[neliz]

So their product was basically finished by the time nVidia got involved. My point exactly. It had probably been in development for 2 years or more.
And PhysX acceleration IS optional. You have the option to disable it when you run the benchmark.

You'd have to wonder why reviewers still enable PhysX when testing Graphics cards as the PhysX benchmark in Vantage is meant to show the difference between a CPU and CPU+PPU combo. Not for benching one graphics card to the other.