Crysis 2 being re-designed for GTX580

[Hell Hound]

[DilTech]

If you prefer, we can look at it from the other side...

A company offers your development studio cash, a FULL test lab, on-site assistance, AND extra developers to include features...

Do you take it, or do you say no?

Both companies release features from time to time that's only on their cards, just NVidia puts out the work to make sure their features get used while AMD sit back and hope someone gives it a spin.

[Vardant]

There's no mention of anything Cuda specific, right? So there's no reason, why it shouldn't work on AMD hardware to. It just might not run well, for now.

[Manicdan]

If you prefer, we can look at it from the other side...

A company offers your development studio cash, a FULL test lab, on-site assistance, AND extra developers to include features...

Do you take it, or do you say no?

i think the line needs to be drawn at optimization.
this is starting to sound more like intel vs amd with disabling instructions based on vendor id. and chances are it will only get worse.

cuda is great, i love the idea, i will not disrespect nvidia for being involved in the software side of things. but once a brand gets to reach into the cookie jar, whos stopping them from crushing all the cookies they didnt take?

the HAWX2 tessellation thing is close to this. if nvidia is average at rendering polygons in all methods, and amd is average in one, and sub par in the other, then maybe nvidia could choose to help them optimize code in the way thats sub par for amd, while offer a negligible difference for themselves. at that point they in theory are just helping, while actually sabotaging, and whos going to prevent it?

the other big "if" question, is what would have happened if nvidia didnt buy physx? would we see an implementation in games thats more optimized? since under the old brand they might have cared more about looking incredible vs havok, instead of requiring brute force to do anything decent to sell higher end cards

[Final8ty]

Like I said before, if this sort of polarization continues to the extreme we may actually end up with two types of software for two types of hardware.

Indeed but there will always be people who can not see that until its far too late.
At the moment all they are seeing is the one sided exclusive extras they are getting now with not having to go without anything like the people who don't own the hardware.

But if the same was to happen on the other side as well then both sides will be missing out on things & then the reality really sets in as people just cant put there self in other peoples shoes until there upto there neck in it first hand.

[VoodooProphetII]

the HAWX2 tessellation thing is close to this. if nvidia is average at rendering polygons in all methods, and amd is average in one, and sub par in the other, then maybe nvidia could choose to help them optimize code in the way thats sub par for amd, while offer a negligible difference for themselves. at that point they in theory are just helping, while actually sabotaging, and whos going to prevent it?

And why would AMD be only average in one and not all of them? Who are at fault there?

[Manicdan]

And why would AMD be only average in one and not all of them? Who are at fault there?

i only recently learned theres multiple ways to do tessellation (i think thats just stupid until i understand why theres different ways). and the reason might be something silly, like the other method is too much of a space and tdp hog for the same performance.

and like i said, "negligible difference" could still mean worse. even if its just 5% worse, would still mean its dumb to do it that way, unless their real goal was sabotage.

please keep in mind i use the words IF and COULD alot, meaning its all hypothetical. so whos at fault, right now i wonder if its the dx11 standards for poor implementation and leaving only recommendations

[VoodooProphetII]

i only recently learned theres multiple ways to do tessellation (i think thats just stupid until i understand why theres different ways). and the reason might be something silly, like the other method is too much of a space and tdp hog for the same performance.

and like i said, "negligible difference" could still mean worse. even if its just 5% worse, would still mean its dumb to do it that way, unless their real goal was sabotage.

please keep in mind i use the words IF and COULD alot, meaning its all hypothetical. so whos at fault, right now i wonder if its the dx11 standards for poor implementation and leaving only recommendations

No worries Manicdan, I'm just asking questions as I know that I'll be able to learn more by listening to what other people have to say.

[[XC] Oj101]

Hardware should have no influence over the software development.

{
if
nvidia offer moneys exploit their ass;
//This is my project punk
else
keep developing as intended ;
}

I'm not sure what language that is, but there seems to be no function for if...

//this is a comment

It should rather be
if (nvidia offers moneys (to?) exploit their ass)
{
print_r("This is my project punk");
}
else
{
keep developing as intended;
}




(Yes, I'm bringing humour into an otherwise waaay too serious thread)

[Hell Hound]

It was intended as a joke but I was called immature.

[570091D]

I'm not sure what language that is, but there seems to be no function for if...

//this is a comment

It should rather be
if (nvidia offers moneys (to?) exploit their ass)
{
print_r("This is my project punk");
}
else
{
keep developing as intended;
}




(Yes, I'm bringing humour into an otherwise waaay too serious thread)

very nice man, i think we needed a laugh!

[STEvil]

If you were a company, and you were investing all of this money into making things happen, would YOU let your competition take advantage of those things?

That's a serious question to anyone out there who still says NVidia should let AMD use all of their features freely.




The issue with AMD isn't ideas... I think we all know that. The problem is they lack the bankroll to help developers USE those ideas. I mean, we didn't see tessellation used even once until DX11 came around which added it to the API, while ATi have had it on their cards since what...2007?

This is why every time a new feature is added on an AMD card I don't pay it any mind unless it's a set standard...because they never get used. As many years as ATi had truform, I can only recall it being used once, and that was in a wolfenstein game.

Google "truform". Serious Sam had it, GLDSRC had it, Unreal had it...

[570091D]

Google "truform". Serious Sam had it, GLDSRC had it, Unreal had it...

truform was the first itteration of tessellation, and while boasting an impressive list of games, didn't appear consistantly on ati gpus. it debuted on the 8500, then they stopped including it with x1000. tessellation finally came back with r600, but even that form of tessellation isn't supported with dx11.

i think eyefinity is ati's first success at pushing at pushing gaming to the next level. it still has it's problems (only the 5970 can render three screens with playable framerates) but it's been embraced by players, devs, and monitor manufacturers. i'm hoping ati uses this as a model for promoting future innovations.

the main reason i support nvidia is their consistancy on this, once they bring a feature into their hardware, you can count on continuing development and support.

[railer]

If you were a company, and you were investing all of this money into making things happen, would YOU let your competition take advantage of those things?

That's a serious question to anyone out there who still says NVidia should let AMD use all of their features freely.




The issue with AMD isn't ideas... I think we all know that. The problem is they lack the bankroll to help developers USE those ideas. I mean, we didn't see tessellation used even once until DX11 came around which added it to the API, while ATi have had it on their cards since what...2007?

This is why every time a new feature is added on an AMD card I don't pay it any mind unless it's a set standard...because they never get used. As many years as ATi had truform, I can only recall it being used once, and that was in a wolfenstein game.

That's exactly what I'm talking about. I personally wish AMD would get more involved with game developers so their features would get used more and not 5 years later. Like i said they are building fine GPU's but really lack involvement with game studios. Sure there were few titles ATI had its hand in but compare that with Nvida and it looks like poultry.

Personally I'm not so found that Nvidia cuts ATI from the tech it develops and in perfect world everything would be open to anyone but lets get real here. They are running a business first and if they want to offer exclusive features to get customers can you really blame them? And most of the stuff that gets talked about on this forum was developed by them and did not exist prior. How can anyone expect them to hand it out to competition is beyond me. Or you guys forget that AMD is a competitor and Nvidia is truing to have an edge over them with exclusive stuff. From what i see no one forbids AMD to do the same thing and if they wanted to there is plenty of game studios that would like some support.

Like I said before, if this sort of polarization continues to the extreme we may actually end up with two types of software for two types of hardware.

Would never happen with PC gaming. Look both GPU company's have the same DX API's. It's a standard and all game developers have to do is follow it and it should work on both with no problems. Stuff like proprietary physics and features that were developed exclusively by ether GPU company will probably always stay that... proprietary. To crucify ether for it is just silly.

[kuroikenshi]

I'm not sure what language that is, but there seems to be no function for if...

//this is a comment

It should rather be
if (nvidia offers moneys (to?) exploit their ass)
{
print_r("This is my project punk");
}
else
{
keep developing as intended;
}




(Yes, I'm bringing humour into an otherwise waaay too serious thread)

hummm can we have that on visual basic v90??? shows my age

[STEvil]

truform was the first itteration of tessellation, and while boasting an impressive list of games, didn't appear consistantly on ati gpus. it debuted on the 8500, then they stopped including it with x1000. tessellation finally came back with r600, but even that form of tessellation isn't supported with dx11.

i think eyefinity is ati's first success at pushing at pushing gaming to the next level. it still has it's problems (only the 5970 can render three screens with playable framerates) but it's been embraced by players, devs, and monitor manufacturers. i'm hoping ati uses this as a model for promoting future innovations.

the main reason i support nvidia is their consistancy on this, once they bring a feature into their hardware, you can count on continuing development and support.

Go back a page: "render to vertex"

[Aerwidh]

hummm can we have that on visual basic v90??? shows my age

How about in Comal?

[kuroikenshi]

that's it, from now on we'll do everything on assembler

[STEvil]

that's it, from now on we'll do everything on assembler

i'm all for it!

[[XC] Oj101]

I'll whip out my ZX Spectrum and relearn BASIC

[Hell Hound]

01001110011010010110001101100101

[trinibwoy]

right now i wonder if its the dx11 standards for poor implementation and leaving only recommendations

DirectX doesn't have any implementations. It's an API. It just defines data structures, order, inputs and outputs. Implementations are done by IHVs and ISVs. Getting upset because there is more than one way to implement something seems a bit ummm....misguided.

[Lanek]

DirectX doesn't have any implementations. It's an API. It just defines data structures, order, inputs and outputs. Implementations are done by IHVs and ISVs. Getting upset because there is more than one way to implement something seems a bit ummm....misguided.

You are right... but there's data structures and code optimisations who are defined and offer to be used... there's 2 really important optimisation of the codes who are apply by ALL developper who use Tesselation in a game ( who benefit ALL gpu's ) .... one of this codes is designed for don't apply tesselation on hidden object, and distant object ( LOD apply now on textures, surfaces and "geometry" ( tesselation ) in DX11 ), the second is used for don't apply Tesselation on plane surface ( a "plan" don't need to be tesselated as he don't have any relief ) ..... if you reach a certain amount of tesselation applied on polygon you need thoses optimisations ( but you can and you want use them even if you don't reach this level ) or you will need to reduce the amount of polygon who will be tesselated.. All of this have not be used on Hawx2 . and for reduce the impact, they have just decrease the number of polygon used on certain object ( buildings, AI planes, etc etc )... ( it's a common practice on console games developpement to choose to reduce the numbers of polygon on certain object less visible for optimise the overall performance, not needed in our case ).. For Hawx2 looking on planes and environnement object, they have really need to go far far higher of what they had expected, we looking at AI planes who seems coming from a 1996 PS1 games...


anyway ..
Reading this thread, this was make me think to Hawx2 ... ( techreport have start a really interessant overview of it in his GTX580 review, i hope some will investigate a little bit further )

A GTX480 in Hawx2 have 105fps, a HD5870 2gb, 45fps ...
GTX460=75fps, GT450 = 60fps and a GT430 is a bit higher of a HD5870.....
(2560x1600 )

I m really curious to know if it was AMD who have put some money on this game, and the HD5870 was run at 105fps and the Nvidia GTX480 at 44fps, and if a HD5650 was faster of the GTX480 ( and surely the GTX580 ) what will say Nvidia users about it.


Now with AMD and his Gaming Evolved progamm, i just hope we will not enter in a new war of the gaming benchmarks... where both companies put a tons of money ( games productors will be happy ), just for try make shine their cards on hardware reviews... We are speaking about Crysis2 .... what will happend if AMD start to do same and you will find Battlefield 3 running at 50fps on a HD5770 and 35fps on a GTX580 ? i believe all of this will finish in a nice disaster for all the gpu's and games industry....

[Iconyu]

@Lanek

Not really a realistic scenario. AMD cards carry no special features that cripple nvidia cards when optimised towards. The only time building towards ATI crippled nvidia was when nvidia decided to be clever during DX9 era and tried to get developers to use either FP16 or FP32 while ATI stuck with the standard FP24.

FP16 looked ugly but was quick while NV30 really sucked at FP32 but didn't offer anything over FP24 (until SM3.0).

Since then nvidia has made sure not to self own again (they basically offered the standard + something else), but by putting such emphasis on tessellation, I think they've left a door open. It's not really something AMD can't push themselves and nvidia don't have the dx11 market share to force the issue.

But that's basically the difference between them, ATI stays standard while nvidia tries to force new standards and pays developers to follow them. That's why AMD don't bother, they won't get the same fps gains nvidia get from their TWIMTBP. Most optimisations from a AMD program would work on both companies cards and according to the way the game is designed could STILL end up faster on a nvidia card for a game that AMD put money into i.e. Dirt 2. Tessellation has probably been the fairest tactic nvidia have used in years.

[kuroikenshi]

I'll whip out my ZX Spectrum and relearn BASIC

right'o let get together and setup camp (lots of brew required) .... those tapes are gonna take quite some time to lad and save