AMD GPU14 Tech Day Event - Sept 25'th

[Sunday]

http://www.gamesindustry.biz/articl...ew-consoles-a-generation-ahead-of-the-best-pc

"Our benchmarks on just the video and audio performance are 8-10 times superior to the current gen,"

Well this sounds like a benefit of Mantle/Tru Audio... and since he didn't specified is it Xb1 or PS4, it's safe to assume both! So in the foreseeable future (say 5 years) Mantle could dismantle both - OGL and DX, when it comes to game development!

 

[Sunday]

It instantly reminded me of this interview from early 2011:

Huddy now works for Intel. Perhaps someone should ask him will Intel support Mantle

 

[Davros]

A generation on a pc is less than a year

 

[pjbliverpool]

http://www.gamesindustry.biz/articl...ew-consoles-a-generation-ahead-of-the-best-pc

"Our benchmarks on just the video and audio performance are 8-10 times superior to the current gen,"

Well this sounds like a benefit of Mantle/Tru Audio... and since he didn't specified is it Xb1 or PS4, it's safe to assume both! So in the foreseeable future (say 5 years) Mantle could dismantle both - OGL and DX, when it comes to game development!

He's clearly talking about the performance of the new consoles vs the current consoles which would have nothing to do with either truaudio or Mantle. As for the title of that article and the claim within wrt PC performance...just lol. Its clear where EA's going to be focussing it's PR at least

 

[Lightman]

So far everyone here was exited about Mantle and TruAudio from high end PC (R9 GPU's) POV where I think benefits will be visible but not ground-shaking.

I'm on the other hand curious if upcoming Kaveri will have TruAudio block included as this would make for much quicker adoption of this new technology.
Mantle should shine even more on next gen APU's and admittedly slower AMD CPU cores as its main advantage from CPU point of view is significantly lower overhead and therefore achieving peak performance at a lower CPU speed.
This could mean perfectly playable BF4 with decent detail settings on desktop Kaveri APU and good to great performance from straight console ports, which currently is not possible.

 

[3dilettante]

Mantle doesn't have much to say about the bandwidth an APU with no discrete card has to play with.
The consoles would have a difference there.

 

[Lightman]

Mantle doesn't have much to say about the bandwidth an APU with no discrete card has to play with.
The consoles would have a difference there.

This is obvious, but I'm naively still hoping for 256bit memory interface for Kaveri at some point in it's life cycle. If not, then by 2015 we should get successor improving on available memory bandwidth and still based around GCN(nextTM) and therefore good candidate to offer close to console performance thanks to Mantle and TrueAudio.

 

[artist]

I dont get why people like DSC and sheepdog arent banned yet.

Moving to the topic, why are people assuming 260X is Bonaire? AFAIK AMD didnt mention it during Bonaire's deep-dive. So if you assume 260X is Bonaire, then it would be the equivalent of "secret sauce".

The rest of the stack looks like they could have re-used current SKUs or there are some minor spins (?) on existing SKUs.

 

[LordEC911]

Moving to the topic, why are people assuming 260X is Bonaire? AFAIK AMD didnt mention it during Bonaire's deep-dive. So if you assume 260X is Bonaire, then it would be the equivalent of "secret sauce".

The rest of the stack looks like they could have re-used current SKUs or there are some minor spins (?) on existing SKUs.

Because it's Firestrike score lines up pretty well with 7790 and seems to be out of reach of a Cape Verde based SKU.

 

[artist]

Because it's Firestrike score lines up pretty well with 7790 and seems to be out of reach of a Cape Verde based SKU.

Couldnt it be a new GPU?

 

[liolio]

Couldnt it be a new GPU?

To be honest if AMD were launching a whole line up of new GPU not integrating true audio in every single one of them would not be a bright move.

 

[LeStoffer]

Gotta say, that I’m pretty torn on Mantle.

On one hand people are paying so much for a quality GPU, that it is close to ridiculous when the performance are being capped by an inefficient API. I have always loved the thought of efficient coding and getting that latest features exposed and used to good effect. So in this sense, I actually love the idea of Mantle to the extent, that I’m grateful that somebody is taking the leap.

On the other hand I’m certain that nVidia will not support Mantle but will eventually create its own API. And what about Intel? This will almost certainly lead to some game engines running commandingly faster on neither AMD or nVidia hardware - which is obviously going to be beyond frustrating for gamers.

Having said that I do think something good will eventually come out of it. Things will settle into either a much more efficient DirectX or a couple of strong APIs like Mantle, nVcore(?) etc where at least a part of the higher layers of the APIs might be shared enough to make things a little easier on those small game studios not using one of the big engines.

If all of the big game engines embrace Mantle strongly, it will force nVidias hand, and it will be turbulent times ahead, but in a sense we need them.

 

[LeStoffer]

Wade Brainerd, Principal Technical Director at Activision:

We'll have to support #Mantle now; wish it was just #OpenGL extensions though! Maybe someone will lightly wrap D3D or GL around it for us?

https://twitter.com/wadetb

 

[Gipsel]

If you write a wrapper for D3D, the performance improvement will be lost. The added API/driver overhead isn't there for nothing. One either has to loosen some restrictions of DirectX or provide more information to the API to let the driver/hardware handle things more efficiently. Otherwise AMD/nV would have provided faster drivers already long ago.

 

[Jay]

If you write a wrapper for D3D, the performance improvement will be lost. The added API/driver overhead isn't there for nothing. One either has to loosen some restrictions of DirectX or provide more information to the API to let the driver/hardware handle things more efficiently. Otherwise AMD/nV would have provided faster drivers already long ago.

You probably know a lot more about it than me but it depends where the inefficiencies are.

You wouldn't need to deal with all the overhead that DX has to handle compatibility.
Could be a much more streamlined DX driver as it where, bypassing the DX stack.

We don't know what compromises mantle has made if any, the DX wrapper would be making the same compromises.

 

[LeStoffer]

If you write a wrapper for D3D, the performance improvement will be lost. The added API/driver overhead isn't there for nothing.

Yeah, I'm not sure what Wade Brainerd is asking for with a "lightly wrap D3D" either.

 

[LeStoffer]

Someone over at the hardOCP forum found this post over at NeoGAF:

i've just made draw call performance test in CryEngine SDK, quite surprising results
Video - http://youtu.be/GrSpm2AZWVU (draw calls are listed as DP 3rd row from the top)

Results, not from video, but a little more precise testing:
300 draw calls - 105 fps
2100 draw calls - 104 fps
3000 draw calls - 103 fps
4000 draw calls - 101 fps
5000 draw calls - 91 fps
6000 draw calls - 83 fps
7000 draw calls - 75 fps
9000 draw calls - 65 fps
13000 draw calls - 49 fps
17000 draw calls - 41 fps
20000 draw calls - 37 fps
on stock i5 2500k and GTX 560. Dunno if my GPU has anything to do with it, but recording with FRAPS havent affected fps in any meaningful way.

http://www.neogaf.com/forum/showpost.php?p=83775657&postcount=397

I don't know how much this applies to real world performance in games, but it is interesting nonetheless.

 

[silent_guy]

On current games, wouldn't it be true that a straight mantle ports won't give a major performance benefit because those games have learned the hard way not to use a lot of draw calls in the first place?

If this is true, the real potential of Mantle is for rendering techniques that can only be done with many draw call that are currently avoided (and that will continue to be avoided for the DX version.)

This could result in games having a special image quality setting that's only enabled for Mantle. (A bit like GPU accelerated PhysX.)

I foresee fun benchmarking complications around that one.

 

[silent_guy]

Yeah, I'm not sure what Wade Brainerd is asking for with a "lightly wrap D3D" either.

If the biggest advantage of Mantle is in fast draw calls by bypassing a whole bunch of Windows generic driver overhead (error checking etc.), then there's probably still quite a bit of value in making it look like a DX call while still bypassing the weight of the real DX layer?

 

[gkar1]

Someone over at the hardOCP forum found this post over at NeoGAF:



http://www.neogaf.com/forum/showpost.php?p=83775657&postcount=397

I don't know how much this applies to real world performance in games, but it is interesting nonetheless.

The way I read that is we'd be able to go from 2100 draw calls to 17000 and still maintain 104 fps. Correct me if I'm doing it wrong.