Hardware Farewell to DirectX?

At this point of time, the games are already looking good enough. Sadly, they all lack in game play variety and AI. Would a difference in API address that? I guess not. IMHO, the game devs should work more on the story execution and gameplay mechanics, rather than worrying about putting more impressive graphics. It'd really be sad if the games end up like the effects overloaded movie Avatar, shiny on the outside, fluffy on the inside.

 

QFT! Have some rep.

What I wouldn't give for developers to just focus on making good, engaging games with proper storylines and good mechanics rather than just focusing on graphics and DRM. We have already ascertained beyond reasonable doubt that you can polish a turd, but it's still going to be a turd at the end of it.

 

I dont like that the company that pays millions of dollar to keep some games from the PC (Microsoft), to sell more off its own hardware(Xbox) in charge over the preferd API, Microsoft dont want the PC plattform to outshine and embarras the XBox. so its clearly an conflict of interrest. and Microsoft is in charge bigtime.
The best scenario for Microsoft is if they could be in control over the visuals and performance. The tool: DX11 for PC and voala! even the latest High end PC only runs a little bit better with some minor visual tweaks.

 

Isn't a bigger issue that lead development has clearly shifted towards consoles from the PC.

If developers are struggling to provide decent ports of consoles games now (i.e. limited settings, poor UI, no AA), why would they develop 'close-to-the-metal' when it's technically more challenging and resource intensive. Developers aren't even taking advantage of PCs using DirectX APIs now.

The reason 10 times more power doesn't translate to 10 times more visuals in games is because PC games aren't designed to run at 720p and 30fps.

 

Why don't just AMD/Nvidia make their own OS that bypasses alot of the quirks and restrictions windows-pc's have? It doesn't have to be an OS just a bootable partition with good drivers and steam-like gui. AMD/nvidia should make HW that was built the same way only with different power/budget levels. So compabillity would not be a problem.

 

I suspect if it was as simple as that, they would.

 

They do both release demo's that must have a certain amount of low level programming, some of them only work on the cards they are designed for.

 

A lot of the demos released in the past were more impressive than the actual games the cards ended up playing. The 'Toyshop' and 'Whiteout' demos were particular favourites of mine.

 

I didn't read that as him .'backing away' ,he merely clarifies his view, that the, 'thicker layer' DX as we now know it, is proving unpopular with some studios and a rethink may be in order. It'd be a shame if people like Huddy can't speak out, debate is a good thing .

 

That makes much more sense.

 

As we travel away from the concept : "boot your Pc with the DVD game+minimal drivers, and play" the performance ruins more more and more, thus top-notch hardware ends being an SO events+API slave, downgrading efficiency.

And guess what? Things are getting worse , now they're pushing consumers to browser+HTML5+Js events!!

Which in terms of performance and efficient cycles is awesome stupid.
Try to imagine some periodic time based function in JS that triggers every n periods of time, ruled by the browser internal core, then releasing from time to time to the S.O the API (OpenGL/DX) calls, then the drivers releasing the effective drawcalls to the metal.... as I've said awesome stupid, we're going BACKWARDS. But some has to make profit on selling un-needed top-notch GPU's, hehehe. See the point?

 

He's realised no one else agrees with him and he's done a labour party sized back track.

Didn't really expect anything else from Elma.

 

He didn't backtrack, he re-affirmed the usefulness of DX in some circumstances while repeating his original idea of maybe seeing the direct to metal. Read the article Snips!

 

I think you're confusing two different issues.

If you're creating a game, you want to minimise the number of draw calls you make, because each draw call has an overhead related to it. (This is CPU overhead mostly. So it results in the CPU being a performance bottleneck).

If the reasons for the overhead have to do with the API, then bypassing the API would be faster, but only if you know what the hardware involved does.

There doesn't have to be a proprietary layer if you have sufficient documentation for the hardware itself, which is what console developers can do because there's only 1 hardware configuration.

A PC developer could do the same thing in theory. (and used to do so in the past), but they'd need to redo most of the graphics code for each individual graphics card.

The reason an API is slower is because it's abstract, and the game calls the API, which then calls a driver written to work with that API, which then sends instructions to the hardware.

If you bypass the API in the way they're suggesting, you are bypassing not just the API code, but the driver code as well, so that's 2 less layers between your game code, and the hardware.
But... Doing that is pretty complicated. And, as already noted, you'd have to redo it for each hardware variation.

(Currently it'd be impossible with Nvidia hardware too, since Nvidia doesn't release low-level documentation.)

To get back to the draw calls, when you're writing a game, the overhead of each draw call means you want your game engine to merge as much stuff as it can into a single draw call.
But that's quite difficult to do, and there are a lot of times when you are forced to split things into multiple calls. (For instance, if two bits of geometry have different shaders or textures, they need separate draw calls).
What this means is, if the way you access the hardware allows you to make more draw calls, you don't need to make as much effort to optimize each call, and you can also get away with more complex scenes by virtue of allowing more effects that cannot ever be done in a single call.

So, if a PC can only do 2-3000 draw calls, and a console can do 20,000 - That's going to make it a lot harder to get a PC to create complex graphical scenes, even if technically the PC's graphics hardware is 10 times more powerful.
(Because the limit on draw calls isn't the graphics hardware, but the CPU overhead associated with each graphics call.)

Basically, if you can re-write the code involved in making a graphics call, that's going to have a bigger effect than making efficient use of each graphics call.
And an abstract API like DirectX or OpenGL is inherently going to create a lot of overhead in something as fundamental as a graphics call compared to coding the hardware functions directly into your game engine is.

 

In your view, can you see the idea happening? Would it take massive work from developers? Would gfx card manufacturers be able to inject their knowledge of their products and make this happen in cooperation with developers? Is it too much work for everybody?

Is it too late to go back and do it all again properly at the root of the hardware? Has everyone got too complacent with DX doing all the hard work that everybody is now too reliant on it?

If a game can be created by utilising hardware directly, surely that will make porting to consoles easier, rather than (currently) porting console to PC (or rather, PC ending up as the poor cousin)?

Side note: MS turned down Machinarium (allegedly) because it had already been released on Mac and PC, and MS had wanted it as an exclusive for the 360. So now Sony has picked up the game for PSN. The game is an extremely clever point'n'click with beautiful artwork. MS rejecting the game out of spite added fuel to why MS should not be in control of how games are made ... my personal view. BUT, I do appreciate that DX has made life easier. Too easy for the major labels that should be more innovative?

 

I don't think the elimination of an API would make things any better. The need to optimize for every piece of hardware instead of a set standard would make things obscenely difficult as far as I can tell. Couple it with slowly dwindling PC funds, and you have a disaster.

I'm wondering when the current consoles will be replaced, it's been almost five years and I've heard nothing about a future console. I only want that post G80 graphics surge once again. Man those prices were great.

 

Going to be very interesting to see where this leads in the future. IMO the hordes of noobs and their xboxes are causing PC Gaming to suffer from severe consolitis, and bare-metal programming could pull it out of this rut. Death to GFWL!!!

 

A serious case of forgotten lessons of the past

The problem I see here is that you are assuming the devs will actually impliment such a system reliably in their products. In the past whenever this was how things were, companies would release support for a brand or chipset and good luck if you don't use that generation of that manufacturer's product it is more likely than not that it just won't work at all. This kind of thing is the reason the API was adopted in the first place.

Perhaps a "thinner" API with greater direct to metal capabilities provided to the devs would work, but as soon as you drop the API model entirely and go with letting the Devs code everything just how they want, you get into the kind of territory where your game wont run on anything but a ATI HD8XXX or Nvidia GT 5XX line card.

When we had all this back in the leadup to the 3Dfx debacle we had all these problems. Every computer out there is a little different and this results in a near impossibility of coding for all senarios. In the end, all these games are coded within a budget of both time and money. Games can be written for PC reliably because of the API. If Developers had to write their programs to explicitly support each piece of hardware out there, it just wouldn't happen by and large. yes it would be possible to do this, but it would greatly increase the budget necessary to get any given game running due to seriously extended QA trials and far longer developement times due to coding concerns. If you want to say that all they would have to do would be to come up with a common framework for their game to use when addressing any set of hardware, that is an API...

I believe it would be a mistake to do away with the API system for the reasons stated above. It just isn't a good idea to ignore the realities of game developement and human nature. To do away with the API because some devs want total control (not realizing that that means total liability and responsibility for universal suppport if the PC is to remain a workable gameing platform) is a major case of wishful thinking gone wrong. The claim that it doesn't have to be that way doesn't change that this is how it would play out in the real world.

 

Well one way would be to have Amd and Nivdia make there own api for there cards, Then the dev's at the gaming companies can add into the game install program a system that detects what we have in are gaming rigs (Amd or Nivdia) and install the right files.It might help the gaming companies optimise the game for a Amd card or Nivdia card as the know what each companies gpu could do. that my idea not a computer programmer. any one think it could work?