• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

GDC - NVIDIA, AMD, Intel Explain How OpenGL Can Unlock 15x Performance Gains

TheD

The Detective
Really? Look at the specs. That the graphics on a 9 year old console can hold its own today is quite something.

For PS4, it's a new system. It will take some time before devs learn to optimize. So many users are so fast to jump on the graphics when it's not as good as it can be.

The old consoles can not hold a candle to current gaming PCs.
 

robin2

Member
Dont talk, show us some shit.
Open gl has been around for decades but i never saw something significant.
Wasn't Quake 3 engine OpenGL? I'd say it was quite a significant engine.

(Doom 3 engine too is OpenGL based, as far as I read on wiki).
 
So, any chances Valve are going to implement these in Source 2? I mean if someone wants to show us examples/benchmarks about how cool OGL is, it's probably them. Would be in their best interest.
 

EVIL

Member
Isn't Valve's support for Source really bad compared to other middleware solutions? I think the small developers have better alternatives.

Source.. is.. well.. can be frustrating. It doesnt have a super friendly UI and toolset like cryengine/unity/UE.

But Source2 will have a heavy focus on ease of use and ease of content creation.
Quote from gabe's reddit AMA said:
“The biggest improvements [out of Source Engine 2] will be in increasing productivity of content creation. That focus is driven by the importance we see user-generated content having going forward. A professional developer at Valve will put up with a lot of pain that won’t work if users themselves have to create content.”

I fully expect them to try to compete with the other engines in terms of feature set, and ease of use.
 
Who are they addressing here? Sounds odd sounding sounding like an commercial with hyperbolic figures if you are trying to reach out to people(developers) who in all likelihood can see through the BS easily. Sounds counteractive and just makes people more distrusting.

Down to earth facts backed up by facts would be ten times more impressive even if the numbers does not show *15 times* or whatever.

This just sounds like PR for the public.

*edit* somehow totally missed the slides. Thats way different from my first impression. *edit*
 

nasanu

Banned
PS3 uses OpenGL doesn't it? Will it benefit from this in any way at all?? Tech GAF required.

Doesnt need to. Console APIs are made specifically for the hardware, these 'gains' are built in. It is bizarre to me watching PC gamers get excited over stories like these while still posting about how their specs are x amount better than a console. Great compartmentalization of thought, I'll give them that.
 

Seanspeed

Banned
Doesnt need to. Console APIs are made specifically for the hardware, these 'gains' are built in. It is bizarre to me watching PC gamers get excited over stories like these while still posting about how their specs are x amount better than a console. Great compartmentalization of thought, I'll give them that.
You are the worst.
 

TheD

The Detective
It is bizarre to me watching PC gamers get excited over stories like these while still posting about how their specs are x amount better than a console. Great compartmentalization of thought, I'll give them that.

Those are not contradictions, PC gamers can like the idea of lower driver overhead (which helps in CPU limited cases) and still have systems that beat the consoles without it.
 

Wiktor

Member
For PS4, it's a new system. It will take some time before devs learn to optimize. So many users are so fast to jump on the graphics when it's not as good as it can be.

They will get better, but I wouldn't expect huge jumps. What you see at exclusives like Order or Killzone: SF is pretty close to full potential of the platform. Games will get better, but not to the level where they would make those two look bad in comparision.
 

SparkTR

Member
Doesnt need to. Console APIs are made specifically for the hardware, these 'gains' are built in. It is bizarre to me watching PC gamers get excited over stories like these while still posting about how their specs are x amount better than a console. Great compartmentalization of thought, I'll give them that.

These are more tailored to lower-end PCs and longevity of PC performance. The higher end stuff really is that much more powerful than consoles, and this is only the beginning of the generation.
 

Ty4on

Member
Haha, we'll see with Broadwell, I guess.

You mean Haswell that probably overclocks even less? :p

I'm jealous of 1366 owners. Overclocked they're still going strong five years later. The 2500k might be better with the price and copies that could hit 5Ghz.
 

Slavik81

Member
The 'driver overhead' stuff seems overhyped to me. What percentage of frame time could reasonably be reclaimed under realistic gaming workloads?

Dont talk, show us some shit.
This is a presentation for developers showing examples and explaining how. NVIDIA, AMD and Intel don't create games.
 
Who are they addressing here? Sounds odd sounding sounding like an commercial with hyperbolic figures if you are trying to reach out to people(developers) who in all likelihood can see through the BS easily. Sounds counteractive and just makes people more distrusting.

Down to earth facts backed up by facts would be ten times more impressive even if the numbers does not show *15 times* or whatever.

This just sounds like PR for the public.

*edit* somehow totally missed the slides. Thats way different from my first impression. *edit*

They're targeting people who are new to the industry or new to OpenGL and it's not BS

OpenGL's goal has been "zero driver" for probably about a decade now... and they've made significant advances

there has been Siggraph workshop/demo sessions showcasing these tricks on the latest OGL versions since the dawn of time itself

most of this stuff is just a refresher
 
Exactly. Buzzwords and claims aside, it has to raise a few eyebrows to see these three working together to oust DirectX, with Valve also pitching in. That's a lot of muscle.

That's a pretty wild assumption there.

Just before this presentation we had AMD, Intel and Nvidia taking a group photo on MS' stage...
 

Alcifer

Neo Member
From reading the slides what I get is that, using current best practices, OpenGL can outpace a naive D3D11 solution by 15x on synthetic benchmarks.

What I take away from this is that OpenGL is now so convoluted and poorly documented that most developers have been ignoring the benefits of various features it has supported for some time.
 

Nzyme32

Member
The only thing I'm worried about is that this stuff has been there all this time. Yet no one knew about it or was aware of how to use it? Now its out there, will it actually be used and matter at all?
 

M3d10n

Member
The only thing I'm worried about is that this stuff has been there all this time. Yet no one knew about it or was aware of how to use it? Now its out there, will it actually be used and matter at all?

1) Almost nobody is writing high-end games for OSX or Linux.

2) None of these features are supported on mobiles (which use OpenGL ES).

3) Consoles don't actually use OpenGL, they have their own APIs. The PS3 supports OpenGL and OpenGL ES, but it's only there for lazy porting and isn't as fast as Sony's own API.

4) Console's don't have much of "driver overhead" and developers have been able to batch stuff in ways PCs only dreamed of for over a decade.

5) You're guaranteed to get more compatibility problems by releasing an OpenGL game on Windows than a DirectX one. Microsoft doesn't provides OpenGL support in their default drivers, but they'll run DirectX 10 and 11 games out of the box. For OpenGL you must go to the vendor website and install a proper driver.

6) Since these features were barely touched, you're almost guaranteed to find lots of bugs and glitches when trying to use them (like most of the high-end rearely-used OpenGL features), specially on non-NVidia hardware and on Apple's shoddy OSX OpenGL drivers.

OpenGL's "zero driver" isn't correct: OpenGL defines only an API and all implementations (including the goddamn shader compiler) is the GPU vendor responsibility. This means some vendors can interpret things differently, specially in details that weren't specified in the documentation (this often includes error reporting, how to deal with invalid data, etc).
 
From reading the slides what I get is that, using current best practices, OpenGL can outpace a naive D3D11 solution by 15x on synthetic benchmarks.

What I take away from this is that OpenGL is now so convoluted and poorly documented that most developers have been ignoring the benefits of various features it has supported for some time.

The only thing I'm worried about is that this stuff has been there all this time. Yet no one knew about it or was aware of how to use it? Now its out there, will it actually be used and matter at all?

the thing is OpenGL in general hasn't been used much on PC's until recently when it started picking up again because it's getting harder to ignore OSX/Linux/iOS/Android nowadays and if you wan't to develop for any of those you're going to have to use OpenGL at which point it might make sense to just do OpenGL everywhere and have a smaller more managable code base (vs OpenGL + Direct X)

some of it ist just refresher for people are new/rusty and some of the stuff is new/clever ways of doing things people might not have realized/put together

graphics in generall generally just moves at such a fast pace it's almost impossible to remain productive and keep track of all the new specs/changes and tricks people are using
 

riflen

Member
The only thing I'm worried about is that this stuff has been there all this time. Yet no one knew about it or was aware of how to use it? Now its out there, will it actually be used and matter at all?

Of course some developers are familiar with OpenGL. But for game developers targeting Windows, DirectX was usually the better choice for a whole host of reasons, but mainly tools and support. Developers are going to go with whatever makes their lives easier. Microsoft have a vested interest in keeping DirectX the defacto API for PC games development. They are able to spend huge resources to make that happen and as a consequence, the tools for developing DirectX games are very good.

The key to greater OpenGL adoption is better tools. There are very promising advancements in that area happening right now with projects like VOGL recently being released. Almost all major game engines have an OpenGL renderer, all GPUs support OpenGL, many platforms support OpenGL (Windows, Mac OS, Linux, iOS, Android), meaning your game has the potential to reach a wider audience. There are many positives, but they were outweighed by the negatives. Now the negatives are being eroded, I expect to see more OpenGL games in the future,
 
PS4 already has an API with as much low level access as devs could want, I assume.

Sony will be constantly improving their PS4 software throughout the gen, they're just not trumpeting it every time they send a new SDK or whatever to developers.

All of this new API activity is mostly based around better performance for PC games, not consoles.
 

Skinpop

Member
I always thought that direct x was a more advanced api on a graphical level, even people like john carmack saying that open gl just couldn't keep up with direct x anymore. On PC, as far as I am aware, all the big graphics intensive (best looking) games are all direct x, making me think that you will still see direct x being dominant and relevant for quite some time.
that was the case years ago. like 7-8 years ago. ogl isn't behind like that anymore. Really no serious developer should be using DX anymore unless they specifically target xbox.
 
Well, I'd say this is probably Mantle's deathknell, but maybe it achieved what AMD was going for all along (forcing the other vendors to get together and improve what we already have).
 

atbigelow

Member
Just read through the slides. There's some pretty smart techniques in here. Feels like there's some pretty good momentum behind GL these days. Hopefully it can keep going. Mantle is bad for the PC world.
 
Well, I'd say this is probably Mantle's deathknell, but maybe it achieved what AMD was going for all along (forcing the other vendors to get together and improve what we already have).
Well, OpenGL already had these features. At best, it may have forced the OpenGL vendors to advertise the benefits of some calls from newer versions of the OpenGL Ali.
 

Durante

Member
I'd like to note that the most significant techniques (according to results) presented in the slides are exactly the features I talked about before in OpenGL discussions here :p

Who are they addressing here? Sounds odd sounding sounding like an commercial with hyperbolic figures if you are trying to reach out to people(developers) who in all likelihood can see through the BS easily. Sounds counteractive and just makes people more distrusting.
Did you miss the slides? Nothing BS about any of it.

*edit* somehow totally missed the slides. Thats way different from my first impression. *edit*
Ah.

The only thing I'm worried about is that this stuff has been there all this time. Yet no one knew about it or was aware of how to use it? Now its out there, will it actually be used and matter at all?
If nothing else, Mantle at the very least was a good advertisement for writing better graphics code on PC.
 

Alcifer

Neo Member
Of course some developers are familiar with OpenGL. But for game developers targeting Windows, DirectX was usually the better choice for a whole host of reasons, but mainly tools and support. Developers are going to go with whatever makes their lives easier. Microsoft have a vested interest in keeping DirectX the defacto API for PC games development. They are able to spend huge resources to make that happen and as a consequence, the tools for developing DirectX games are very good.

It might just be me but I find MSDN documentation and the structure of D3D versions to be much easier to follow than for OpenGL.

When I want to work with D3D 9 or D3D 11 I know where to find documentation explaining the core principles, the main pathways and the optional capabilities for some hardware.

With OpenGL and especially the extensions I find it a lot harder to determine what the optimal solution is to any particular problem, it depends heavily on which features are supported by the hardware and whether they've been superseded by newer features. It might just be I've not found the right place to look but all too often the documentation I find is out of date or doesn't refer to alternative features.
 

Durante

Member
It might just be me but I find MSDN documentation and the structure of D3D versions to be much easier to follow than for OpenGL.

When I want to work with D3D 9 or D3D 11 I know where to find documentation explaining the core principles, the main pathways and the optional capabilities for some hardware.

With OpenGL and especially the extensions I find it a lot harder to determine what the optimal solution is to any particular problem, it depends heavily on which features are supported by the hardware and whether they've been superseded by newer features. It might just be I've not found the right place to look but all too often the documentation I find is out of date or doesn't refer to alternative features.
Well, I think that's exactly why we should see more presentations like this.

This presentation answers the question "how can I do a lot of draw calls in OpenGL with minimum CPU overhead?", and it does so in the most up-to-date fashion.
 

atbigelow

Member
It might just be me but I find MSDN documentation and the structure of D3D versions to be much easier to follow than for OpenGL.

When I want to work with D3D 9 or D3D 11 I know where to find documentation explaining the core principles, the main pathways and the optional capabilities for some hardware.

With OpenGL and especially the extensions I find it a lot harder to determine what the optimal solution is to any particular problem, it depends heavily on which features are supported by the hardware and whether they've been superseded by newer features. It might just be I've not found the right place to look but all too often the documentation I find is out of date or doesn't refer to alternative features.

Doesn't help that there is so much legacy GL code and articles floating around the internet.
 

On Demand

Banned
They will get better, but I wouldn't expect huge jumps. What you see at exclusives like Order or Killzone: SF is pretty close to full potential of the platform. Games will get better, but not to the level where they would make those two look bad in comparision.

You really don't believe this do you? Improvements always happen over time. There will be games in the future that look much better than the two games you mentioned.

Are you a developer or have experience with consoles to be able to make a claim like that?
 
When discussing real numbers it is impossible to get more vague than 'up to 15% or more'

mathematically_annoying.png

Not sure where this is from, but it is amazing.
 

Wiktor

Member
You really don't believe this do you? Improvements always happen over time. There will be games in the future that look much better than the two games you mentioned.

The jump we've witnessed in PS3 was mostly because it was damn hard to make games from. This time it's anything but. PS4 uses architecture people have been using for decade or more. Immiedietely they are able to utilize it early on much more effectively than with PS3, where they needed to learn the ropes first.

So a jump from Uncharted 1 to 3 just won't happen. A jump from Uncharted 2 to 3 is more realistic.
 
Top Bottom