The AMD Mantle Thread

[Lonbjerg]

Giving you the benefit of the doubt, you have completely missed the point. The point isn't that you are draw call limited, it's that the game is draw call limited. By design they have to limit the draw calls the game runs so that it can run at an acceptably optimized level with DX. DX, not you is draw call limited.

So were does this hamper ARMA3?
You type a lot, but it's all beating around the bush.

 

[monstercameron]

i dont think dx overlays will work, doesnt bf4 have a builtin fps overlay?

 

[SiliconWars]

So were does this hamper ARMA3?
You type a lot, but it's all beating around the bush.

What happened at 19 seconds on your video Lonbjerg? I notice you still haven't answered that one.

 

[Lonbjerg]

What happened at 19 seconds on your video Lonbjerg? I notice you still haven't answered that one.

I'm recording on a HDD, not a SSD...my performance is EVEN better when not recording..so m guess is HDD I/O

 

[SiliconWars]

I'm recording on a HDD, not a SSD...my performance is EVEN better when not recording..so m guess is HDD I/O

Lol yeah JUST AT THE VERY MOMENT YOU LOOKED UP. THAT'S TOTALLY BELIEVABLE. :awe:

 

[Lonbjerg]

Lol yeah JUST AT THE VERY MOMENT YOU LOOKED UP. THAT'S TOTALLY BELIEVABLE. :awe:

It's more believable than your:
"R290X = SLI Titan's in BF4 with Mantle"
Or your
"No none releases in December"

Wouldn't you say?

 

[SiliconWars]

Deflecting Ljonberg? Who'd have believed it?

 

[Lonbjerg]

Deflecting Ljonberg? Who'd have believed it?

Not deflecting, told you what I think went on...but will make a "spin around, look up and spin some" video for you tomorrow.
Right now there is a nice blonde demaning my attention...and we both know you will never answer my question anyways...but it will make a great sig when (if) Mantle appears

 

[VulgarDisplay]

Is there a way to bulk report off topic posts? Thread is begging for it.

 

[SirPauly]

Or AMD is preparing for a time where their CPU performance will fall further and further behind Intel's..and Mantle is a stopgap attempt at remaining relevant.
Their last hope seeem to be "moar cores"...and Mantle seems to fit into that philosophy.

Makes sense based on improving efficiency and APU and SOC Designs! AMD can leverage their success with the console wins and bring this to the PC desktop/mobile and mobile platforms over-all. AMD possibly can really create AMD Gaming awareness for their brands!

 

[krumme]

I did not bring up the assertation that dx was disappearing. However, it was specifically stated or implied by several posters. If you want to accuse someone, accuse them. As a matter of fact, you stated that assertion in post 3665.

3665??, thats not my post.

But i think there is a good chance mantle will be the most important api in 6 years. But if it isnt its because its not worth it. It might be small and dx big. Or opposite. Peace with it. There will be new technology even then. As long as i get better games.

...and a better k/d ratio in bf4. The only thing saving me is sabot. Lol.

 

[krumme]

We have talked a lot about cpu perf in relation to mantle, but if you go and have a look at Johan presentation at apu13 at about 18 min there is more about gpu perf.
https://www.youtube.com/watch?v=N_6CAneoW-0&feature=youtube_gdata_player

Asynch dma transfers
Asynch compute together with graphics
Multiple compute kernels colaborating

And funky stuff like compute as frontend for gfx pipeline (compute runs ahead and prepares geometry for pipeline !!!)

Lots of stuff with great future potential

I am not a programmer and engine programming is probably the most complex stuff out there so Johan speak is difficult to understand. But i guess some of you could get more from it.

 

[BrightCandle]

Arma 3 is mostly dominated by the game world updates (what they call the SimuA process). The DX calls hover around 10ms most of the time, so there is certainly draw call overhead but from my tests the time is dependent on a variety of things. Mostly the problem with Arma 3 is its single threaded, it would achieve 90 fps or so if it wasn't for the simulation being both single threaded and on the same thread as everything else.

Mantle would help it but there are other more important things A3 needs to make it run than reducing the draw calls.

 

[frozentundra123456]

3665??, thats not my post.

But i think there is a good chance mantle will be the most important api in 6 years. But if it isnt its because its not worth it. It might be small and dx big. Or opposite. Peace with it. There will be new technology even then. As long as i get better games.

...and a better k/d ratio in bf4. The only thing saving me is sabot. Lol.

3671

 

[krumme]

Arma 3 is mostly dominated by the game world updates (what they call the SimuA process). The DX calls hover around 10ms most of the time, so there is certainly draw call overhead but from my tests the time is dependent on a variety of things. Mostly the problem with Arma 3 is its single threaded, it would achieve 90 fps or so if it wasn't for the simulation being both single threaded and on the same thread as everything else.

Mantle would help it but there are other more important things A3 needs to make it run than reducing the draw calls.

Yeaa obviously mt is the way forward and its old tech. What do you think in Johans presentation would help besides the draw calls? (btw draw calls are hardly mentioned in the 30min presentation)

 

[Noctifer616]

When Mantle actually does show up, will Fraps and other FPS readers even work with it at first?

https://twitter.com/repi/status/407169521753915392

PerfOverlay.DrawFps / DrawGraph will continue to work with Mantle.

 

[mikk]

What you said was 50% more on a 6c cpu in bf4 or it would be waste. And i asume it was not the phenom x6 you were referring to. It was just a pethetic attemp to create a strawman before mantle patch hits.

I said 6+ cores. Depending on core scaling with DX11 it might require a 8 core CPU in BF4 and when I say 6+ cores I don't limit it to Intel, this is a stupid assumption. I thought you are intelligent enough to understand that. That a 6 core IVB-E is already more than fast enough in this game is a different topic and shouldn't surprise anyone.

The only way a i7 6c will be cpu limited in bf4 is if the drawdistance can be improved.

Lower the res to 1280x720 or 960x540, disable AA and use a highend GPU like 780 Ti. I would expect a CPU limit. I guess you don't understand the purpose however. A serious tester who is testing Mantle has to eliminate GPU bottlenecks, otherwise no representative results for CPU bound test cases are possible.

 

[SiliconWars]

Mikk, the Oxide demo with 100k batches shows they can't get CPU bound with 8 cores at 2 GHz so there's not much chance of getting CPU bound in BF4 is there?

 

[Stewox]

We have talked a lot about cpu perf in relation to mantle, but if you go and have a look at Johan presentation at apu13 at about 18 min there is more about gpu perf.
https://www.youtube.com/watch?v=N_6CAneoW-0&feature=youtube_gdata_player

Asynch dma transfers
Asynch compute together with graphics
Multiple compute kernels colaborating

And funky stuff like compute as frontend for gfx pipeline (compute runs ahead and prepares geometry for pipeline !!!)

Lots of stuff with great future potential

I am not a programmer and engine programming is probably the most complex stuff out there so Johan speak is difficult to understand. But i guess some of you could get more from it.

Johan said something about DMA queues working as they should in Mantle ...

 

[BrightCandle]

Arma doesn't even offload to multiple threads let alone using multiple command queues or execution kernels. All of that requires that your game is highly parallel internally already and that you want to some amount of it on the GPU. So initially there is nothing about Mantle A3 is directly going to benefit from beyond a direct reduction in the draw call overhead. Simply put the engine does not scale to multiple threads in any way so the other aspects of Mantle and GPU kernels etc just wont matter to it.

 

[OCGuy]

Why are the same 2-3 people making the same type of Mantle posts over and over? This thread has devolved due to people who appear to have more than a casual interest in the reception of this announced product.

 

[Stewox]

Arma doesn't even offload to multiple threads let alone using multiple command queues or execution kernels. All of that requires that your game is highly parallel internally already and that you want to some amount of it on the GPU. So initially there is nothing about Mantle A3 is directly going to benefit from beyond a direct reduction in the draw call overhead. Simply put the engine does not scale to multiple threads in any way so the other aspects of Mantle and GPU kernels etc just wont matter to it.

Don't you think maybe they would took the time to implement multi-threading with ease when they would add Mantle support ?

Are you forgetting the whole other aspect of Mantle enabling easier and simple multi-threading that if attempted with DX would be way more time consuming and costlier for the resulting benefit.

Doesn't matter if engine scales or not, they can modify that code at any time. It's not even a big chunk of the engine, we're talking about parts here.

I've raised these before, some people think engines and engine code works like a washing machine or something, some casual rigid and fixed product, no, it doesn't mean they would have to rewrite the whole engine ...

 

[Silverforce11]

Why are the same 2-3 people making the same type of Mantle posts over and over? This thread has devolved due to people who appear to have more than a casual interest in the reception of this announced product.

Thousands of posts ago, I had said the same thing, camp negative have no evidence for negativity, just hyperbole and their own opinions. Likewise, camp positive also have very little evidence since we don't have a working game with Mantle yet to see what benefits it will bring, just presentations running alpha software.

It devolved a long time ago to poo flinging essentially because neither side has any evidence to the contrary.

We are still waiting for BF4's Mantle.

 

[Stewox]

Thousands of posts ago, I had said the same thing, camp negative have no evidence for negativity, just hyperbole and their own opinions. Likewise, camp positive also have very little evidence since we don't have a working game with Mantle yet to see what benefits it will bring, just presentations running alpha software.

It devolved a long time ago to poo flinging essentially because neither side has any evidence to the contrary.

We are still waiting for BF4's Mantle.

I've been discussing the hard stuff like since the 11 november conference as I simply used my brain to find all the slides out there but people didn't took it seriously until this demo came out so they had some context added to those slides and to tell you it does make a big difference with the talk, it's even better than I though, can't recall them now but there were many points there that made a lot more sense, it just validated my thoughts as I wasn't as eager to talk about it until I we got the video.


Oh, I may alleviate some of those negativity in the spirits ... very soon now, standby.

 

[3DVagabond]

So were does this hamper ARMA3?
You type a lot, but it's all beating around the bush.

It might seem like beating around the bush, but I'm not claiming to know things that I don't. I read/listen to what the devs are saying about Mantle. I have no reason to doubt them and figure it's exactly what they are saying it is. I assume that DX is as difficult to work with as they claim it is. I assume they are telling the truth. So, I post relating to what I've learned about Mantle so far from the Devs. There have been no hard claims about performance. I've read from 20% to 300% depending on the particular game and where in the game. I'm waiting just like everyone else to find out. You sure can't play a game in DX11 though and gleam any kind of idea one way or the other what improvement Mantle would make. You aren't going to get 65fps w/105K draw calls like in the Oxide demonstration from DX though. Despite what some are trying to make it out to be, I'm very impressed. When I make game models I have to keep a constant eye on the poly count with a hard ceiling on poly count. Constantly "*optimizing" the geometry and textures to try and get the most detail while using absolutely the least amt of resources. It's a real struggle artistically to make a design that I'm happy with artistically and not blow the polygon budget. Or have enough details in the textures. I'm hopeful that Mantle will give me the ability to add the detail I want to my models. Reread hopeful. Here's waiting just like everyone else (should be) but with a different interest.

*optimizing means trying to make something look good but still be able to be rendered at an effective frame rate. It's a constant compromise and balancing act.