The AMD Mantle Thread

[Lonbjerg]

Buy another 2 screens and tell us how it goes with surround.

The ever moving goalposts eh?
I am going 3 x G-syncmonitor sometime next year fyi

But how about you start answering some of the questions you run and hide from?

Like if AMD never has released hardware in December?
Like what your excuse is for my lastest AMRA3 video with no HHD I/O?


It says a lot about a poster than runs away from posts when his arguments gets debunked.

But nice to know that Matle is now relegated to tripple-monitor setup before it shows any benefits...

 

[SiliconWars]

The ever moving goalposts eh?
I am going 3 x G-syncmonitor sometime next year fyi

But how about you start answering some of the questions you run and hide from?

Like if AMD never has released hardware in December?

Yes of course they have released hardware in December. It's not the normal thing to do however, and is generally only done so they can say that they made it in certain year. The real launches happen at CES, every time.

Like what your excuse is for my lastest AMRA3 video with no HHD I/O?

I'm not really impressed by videos of a guy walking around a world with almost nothing in it at barely over 40 fps.

It says a lot about a poster than runs away from posts when his arguments gets debunked.

But nice to know that Matle is now relegated to tripple-monitor setup before it shows any benefits...

You said you maxed it out. You aren't even close. I've been running higher resolution than you have for 3 years with much less powerful graphics cards and Mantle is going to make that even easier. You don't know what you're missing mr "maxed out".

 

[Lonbjerg]

Yes of course they have released hardware in December. It's not the normal thing to do however, and is generally only done so they can say that they made it in certain year. The real launches happen at CES, every time.

So you where wrong, thanks for clearing that up.

I'm not really impressed by videos of a guy walking around a world with almost nothing in it at barely over 40 fps.

So you avoid amitting that the HDD I/O at 19 sec in the first video was exactly that...HDD I/O...and you try and beat around the bush and move the goalposts, instead of admitting you were wrong.

You said you maxed it out. You aren't even close. I've been running higher resolution that you have for 3 years with much less powerful graphics cards and Mantle is going to make that even easier. You don't know what you're missing mr "maxed out".

Please tell me what settigns I need to run at before you consider ARMA3 maxed out...be SPECIFIC please, so you cannot move the goalposts again, thank you very much.

 

[SiliconWars]

So you avoid amitting that the HDD I/O at 19 sec in the first video was exactly that...HDD I/O...and you try and beat around the bush and move the goalposts, instead of admitting you were wrong.

No I have seen 2 videos and I choose to believe the one that you weren't trying to avoid stutters in is the usual case. It's easy to change settings or set up an easier load that proves your point. What counts is how it works when you weren't trying to prove the point, and that stutter at 19 seconds in your original video tells far more of a story.

Please tell me what settigns I need to run at before you consider ARMA3 maxed out...be SPECIFIC please, so you cannot move the goalposts again, thank you very much.

Who's moving the goalposts? I can switch off both my side screens and "max out" a ton of games at 1680x1050 on what is now basically a low end graphics card.

When the 290X crushes Titan or anything Nvidia has under Mantle at Eyefinity/4K you'll know what "maxing out" is then, because you'll be further away than ever.

 

[Lonbjerg]

No I have seen 2 videos and I choose to believe the one that you weren't trying to avoid stutters in is the usual case. It's easy to change settings or set up an easier load that proves your point. What counts is how it works when you weren't trying to prove the point, and that 19 second stutter in your original video tells far more of a story.

So you accuse me of being dishonest now, because I recorded on a SSD and not a HDD now and there are no more HHD I/O?

Who's moving the goalposts? I can switch off both my side screens and "max out" a ton of games at 1680x1050 on what is now basically a low end graphics card.

We are not talking a ton of games now...we are talking ARMA3.
So please tell me what settings are needed to be maxed out in ARMA3 thanks

When the 290X crushes Titan or anything Nvidia has under Mantle at Eyefinity/4K you'll know what "maxing out" is then, because you'll be further away than ever.

So you are still sticking to your claim aboutR290X = SLI Titan's in BF4.
Nice to know...I will hold that in mind and hold you to your claim...when Mantle comes out.

 

[SiliconWars]

First of all Lonbjerg, I couldn't give a flying monkeys about ARMA 3 as it's a game I have no intention of playing.

I'll play this game though, just for you. What screen resolution are you currently playing at? And what resolution are these 3x G-Sync monitors you're going to be buying?

 

[krumme]

What game could possibly use that many draw calls? You'd overtax the GPU and have so much clutter on the screen you wouldn't even be able to see anything.

It's great that Mantle offers such high performance and reduced overhead, but quoting big numbers is a bit of a spurious argument in my opinion, because the majority of games are not really draw call limited.

DX11 multithreading can probably do a maximum of 20K draw calls on a well designed multithreaded engine like Firaxis's Lore Engine. I have no idea how many draw calls 3D engines like Frostbite 3 and CryEngine 3 as they use immediate context via manual threading.

And PCs also have something called instancing, which allows you to batch draw calls for objects with similar properties. This can be very effective at reducing the amount of draw calls, as the CPU doesn't care about the size, only the amount.

At any rate, IF we ever have games that issue that many draw calls, by the time they roll around, we'll have DX14 or something which will have a reduced operating overhead similar to Mantle, but not as thin as to retain a reasonable degree of abstraction and backward compatibility.

Its easy for a rts to use 100k. If you take the oxide engine demo of the spaceships its not like you need 100k spaceships but what you want is to eg build your spaceships with eg 15 configurable turrets, and eg have a lot of shooting. There is much more objects to it than the major spaceship you see.

Secondly having lots of headroom make it far cheaper to program for. You dont eg ask the artist to put more elements in a texture for no logical reason.

 

[Carfax83]

I guarantee we can all agree on something. The performance we get from direct X is awful. We know how much "horsepower" our graphics cards have (ESPECIALLY when compared to consiles) , but how they are squandered away and we end up with games that don't look that great, and don't run that great.

AMD is finally trying to solve this problem.

While the performance of DirectX definitely needs improvement and refinement, I would never call it awful.

Native DX11 typically look great, and run great. Crysis 3, BF3, Civilization V, Assassin's Creed IV, and yes, even BF4. Despite it's problems (mostly related to online play), BF4 also looks and performs very well.

Next year we'll get Watch Dogs, Dragon Age 3 and the Witcher 3, which will also be shining examples of native DX11 games. The latter in particular will be fairly groundbreaking, and will be the first game to my knowledge to have a completely open massive World with no loading screens and no chapters or artificial transitions.

Feast your eyes on the uncompressed Witcher 3 VGX trailer (1080p)

A game like the Witcher 3 could not be done in DX9 (or 32 bit systems) without some serious sacrifices.

 

[Carfax83]

Its easy for a rts to use 100k. If you take the oxide engine demo of the spaceships its not like you need 100k spaceships but what you want is to eg build your spaceships with eg 15 configurable turrets, and eg have a lot of shooting. There is much more objects to it than the major spaceship you see.

Secondly having lots of headroom make it far cheaper to program for. You dont eg ask the artist to put more elements in a texture for no logical reason.

I'd like to see an actual GAME that could use 100,000 draw calls.

As for your example of 15 turrets and lots of shooting, remember you can use instancing that can use a single draw call for multiple objects of the same type, drastically lowering the amount of draw calls needed.

 

[SiliconWars]

I'd like to see an actual GAME that could use 100,000 draw calls.

10 years ago 300-1000 batches was about as good as it got and people probably said similar stuff about 10K back then. Are we now settled for 10-15k (in good examples), forever being held back by the API?

GPU's continue to get more powerful and continue to be held back. We aren't going to get to VR or holodecks with 10k batches.

 

[krumme]

I'd like to see an actual GAME that could use 100,000 draw calls.

As for your example of 15 turrets and lots of shooting, remember you can use instancing that can use a single draw call for multiple objects of the same type, drastically lowering the amount of draw calls needed.

We have all been playing RTS seeing how they are slowed down. Its not like a factor of 20 is that different. Come on. Show me a GAME that can use 15k? say civ5 - and its not like it would have any trouble using 20 times as much.

And it doesnt even count the game we can not imagine.

Secondly its a programming hell you descripe.

Here it goes on dx to reduce the drawcalls.
Ask the artist to put more elements in the texture
Bind the objects and use a single call to call for multiple objects
Use complex MT programming

And so on.

But what is reality.

You have a certain budget. Getting the last performance out of this half black box of dx is extremely expensive, and you get half the way not matter what you do.

In a budget restrained reality, dx simply dont deliver high performance and here many drawcalls, as is evident from games on the market today.

In a new world where a handfull of engines drives most games, and not 1000 different studios, dx is simply a bad solution. It doesnt make sense not to let people like Johans programmers not take control. They have the specialist competence.

Some keep argumenting like this is 2001 when the situation on the market is clearly different. Dx is improving as it should. Its fine, but the context is simply radically different.

Arguing for DX is like arguing the gaming industry should go back to 2001 structure.

 

[Lonbjerg]

First of all Lonbjerg, I couldn't give a flying monkeys about ARMA 3 as it's a game I have no intention of playing.

That wasn't the point...the point was HDD I/O...and you try and run and hide from that now....your choice.

I'll play this game though, just for you. What screen resolution are you currently playing at? And what resolution are these 3x G-Sync monitors you're going to be buying?

I use CRT's (due to flatcreens currently being an I.Q downgrade), and don't know what screens I will get, except it will be Samsung's...waiting for monitors to hit the market.

 

[Lonbjerg]

Arguing for DX is like arguing the gaming industry should go back to 2001 structure.

No arguing for DX is simply not wanting to go back to the API wars that ran before DX...you are putting the cart before the horse and making a untrue argument now.

 

[dacostafilipe]

No arguing for DX is simply not wanting to go back to the API wars that ran before DX...

So, you then want to leave it as it is, and don't improve what can be improved?

Why? There's no reason NOT to try something new. That's how new things get invented in the first place.

This sounds like the "good old days syndrome".

 

[ShintaiDK]

So, you then want to leave it as it is, and don't improve what can be improved?

Why? There's no reason NOT to try something new. That's how new things get invented in the first place.

This sounds like the "good old days syndrome".

Fragmentating API is not an improvement.

 

[Final8ty]

Should of been posted here and not the other thread..

Guys

To take advantage off Mantle, the game developer needs to release a patch for the game to use the API, BF4 shall be the first and very soon. The improvements in FPS are very good and worthwhile.

So it is just game patches, if more and more game developers start patching and releasing games with Mantle support then AMD will be doing very well indeed.

http://forums.overclockers.co.uk/showpost.php?p=25567161&postcount=1346

 

[Paul98]

Fragmentating API is not an improvement.

Then why use DX at all why not just use OpenGL?

 

[cmdrdredd]

Then why use DX at all why not just use OpenGL?

Cause DX is at the point where it is superior to opengl at the present time. Even John Carmack said do but he refuses to switch for his own usage because he is heavily invested in opengl and just continues with using that.

 

[Paul98]

I'd like to see an actual GAME that could use 100,000 draw calls.

As for your example of 15 turrets and lots of shooting, remember you can use instancing that can use a single draw call for multiple objects of the same type, drastically lowering the amount of draw calls needed.

Lots of games could use that, it allows you to increase the verity of objects, increase the draw distance, increase the number of different objects, split up what currently is a single object into different objects to make it look more realistic, more dynamic changes to the objects. If you have an object that you could blow up, you could keep the rubble it turns into.

There are lots of interesting things you can do, or the artists can do if you raise the ceiling by a lot.

 

[Paul98]

Cause DX is at the point where it is superior to opengl at the present time. Even John Carmack said do but he refuses to switch for his own usage because he is heavily invested in opengl and just continues with using that.

This is the point that people have been trying to make, compatibility is not the only reason for use.

 

[dguy6789]

Can you go back for yourself and find the post where I linked to OpengL and 100.000 draw calls...above the 65.000 in the Mantle Tech demos...or do I need to make a repost?

I am confused. Are you trying to suggest that it's possible for a high level API like DX or OpenGL to approach the performance offered by extremely low level coding?

 

[Lonbjerg]

I am confused. Are you trying to suggest that it's possible for a high level API like DX or OpenGL to approach the performance offered by extremely low level coding?

No, why make a fallacy?
Fact is OpenGL can do 100.000 drawcalls...today.
(Actually it can do millions if you just focus on drawcalls)
Sorry if that burst Mantle's PR campaign...but facts are facts.

 

[OCGuy]

Any word if Mantle is going to show a tangible FPS improvement in BF4?

 

[frozentundra123456]

So, you then want to leave it as it is, and don't improve what can be improved?

Why? There's no reason NOT to try something new. That's how new things get invented in the first place.

This sounds like the "good old days syndrome".

Well, despite its higher overhead, Dx has a proven track record making gaming possible across multiple architectures. Obviously one is always looking for improvement. But there is rarely a "free lunch" so to speak. Improvement in one area comes at a sacrifice in others. For mantle, obviously the improvement in efficiency comes at the cost of cross platform compatibility. IMO, it will take considerable time to see how much improvement mantle gives in a wide variety of games and how (or if) it can be adopted to a wide variety of situations.

 

[Paul98]

No, why make a fallacy?
Fact is OpenGL can do 100.000 drawcalls...today.
(Actually it can do millions if you just focus on drawcalls)
Sorry if that burst Mantle's PR campaign...but facts are facts.

Nowhere have you shown millions of draw calls, you have shown a single draw call doing lots of draws. This is the sort of thing that programmers have been doing for years to reduce the number of draw calls.