The AMD Mantle Thread

[blackened23]

3 pages of CIV 5. Jeez. Who cares? This has to be well into off topic territory, it is neither interesting or informative.

 

[SiliconWars]

3 pages of CIV 5. Jeez. Who cares? This has to be well into off topic territory, it is neither interesting or informative.

Lol yeah, it's no longer interesting now that I've given links of SKYMTL and the 7970 review destroying the entire argument. :whiste:

 

[mikk]

Lol yeah, it's no longer interesting now that I've given links of SKYMTL and the 7970 review destroying the entire argument. :whiste:

You didn't understand that command list support for DX11 mutltihreading improves CPU bound cases, your link clearly is GPU bound.

 

[f1sherman]

Lol yeah, it's no longer interesting now that I've given links of SKYMTL and the 7970 review destroying the entire argument. :whiste:

You mean wining against GTX 580 by 13%?
That means my 290 could be faster than 680!

Well allow me to hoist my red flag high above this epic Civ battlefield :biggrin:

Anyway... BF4 mantel patch; when can we expect it?

 

[blackened23]

Lol yeah, it's no longer interesting now that I've given links of SKYMTL and the 7970 review destroying the entire argument. :whiste:

I don't think AMD's drivers support multi threading. But I also don't think this has any context relating to Mantle, although I could be wrong. So i'm seeing three pages of CIV 5 talk and wondering why anyone should care. Maybe people cared 2 years ago. I have no idea.

I would not be surprised if AMD's driver did indeed NOT support MT rendering, given their super slow cadence for fixes. I mean, AMD hasn't fixed CF+EF frame pacing or DX9 frame pacing yet for 79xx cards after 2 years. Whatever though. Either way it isn't on topic, so i'm not sure why i'm reading about CIV 5 for 3 pages.

 

[SiliconWars]

You didn't understand that command list support for DX11 mutltihreading improves CPU bound cases, your link clearly is GPU bound.

Find me any other game where the 580 is 45% faster than the 6970 at 1600p, and the 570 is 26% faster at 1600p, then i'll agree it's GPU bound.

 

[SiliconWars]

You mean wining against GTX 580 by 13%?
That means my 290 could be faster than 680!

Well allow me to hoist my red flag high above this epic Civ battlefield :biggrin:

Agreed it's not much of a win but it proves the multi-threaded point is false.

Anyway... BF4 mantel patch; when can we expect it?

No idea but I doubt it'll be this year now.

 

[mikk]

Find me any other game where the 580 is 45% faster than the 6970 at 1600p, and the 570 is 26% faster at 1600p, then i'll agree it's GPU bound.

Here.

http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/15/
http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/19/
http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/21/


You can be sure Civ 5 in 1600p +4xMSAA on a 3960X @ 4.3GHz tested on old Gen GPUs is GPU bound.

 

[SiliconWars]

Here.

http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/15/

Starcraft 2 - 6970 is 10% faster than the 580 GTX

http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/19/

Battleforge - 580 is 13% faster than the 6970

http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/21/

LP2 (really?) 580 is 39% faster than the 6970

Even a heavy TWIMTBP game can't reach the 45% difference shown at 1600p in Civ 5 on Anandtech. The 6970's deficit in LP2 is almost certainly down to drivers. This whole dx11 multi-threading point is a complete nonsense. It doesn't exist except in Civ 5's "late game" canned benchmark.

 

[mikk]

Starcraft 2 - 6970 is 10% faster than the 580 GTX

Try again 1600p 4xMSAA and find your mistake. I help you.

Battleforge - 580 is 13% faster than the 6970

 

[Falafil]

Try again 1600p 4xMSAA and find your mistake. I help you.

This isn't 4Chan.
-- stahlhart

 

[krumme]

Do you have a source? Never heard this. DX11MT implementation seems complicated.

In the oxide mantle demo there is no lead thread for their engine.
Did someone actually grasp the potential of that?

Who cares about some pathetic ms patch using the word mutithreadding giving some minor improvement in a few games. And at the same time complicated and expensive to implement.

No lead thread is a revolution. Who would have thought it was even possible?

Implemented in the engines it can transform games.

 

[ShintaiDK]

In the oxide mantle demo there is no lead thread for their engine.
Did someone actually grasp the potential of that?

Who cares about some pathetic ms patch using the word mutithreadding giving some minor improvement in a few games. And at the same time complicated and expensive to implement.

No lead thread is a revolution. Who would have thought it was even possible?

Implemented in the engines it can transform games.

You seem to copy every single PR statement as the truth and hype it to no end. I thought history by now would have taught otherwise. Just see how it went for BF4. And where is the mantle patch for that?

Until we see a real world mantle game/application we can compare by 3rd party to DX/OpenGL. Then its utterly pointless to talk further about it.

 

[Falafil]

You seem to copy every single PR statement as the truth and hype it to no end. I thought history by now would have taught otherwise. Just see how it went for BF4. And where is the mantle patch for that?

Until we see a real world mantle game/application we can compare by 3rd party to DX/OpenGL. Then its utterly pointless to talk further about it.

Why would they lie about Mantle's properties?

 

[BrightCandle]

Until we see a real world mantle game/application we can compare by 3rd party to DX/OpenGL. Then its utterly pointless to talk further about it.

Even then its going to be really hard. If you take a game mostly developed on DirectX or perhaps openGL and you port it to Mantle then it could perform better or worse on Mantle. That wouldn't say much about Mantle really, it would simply tell us that particular game did or didn't benefit from the lower draw call and the multiple command lists and other features. Another game with a different engine doing the same thing could see the opposite effect.

However a game developed for Mantle and then ported to DirectX could also perform better or worse, but that equally wouldn't say much about DirectX. Its going to take a lot more than one data point before anything definitive can be said about its true performance impact and potential and its going to depend hugely on the architecture of the game and its engine, its just a lot more complicated than "Mantle will be faster".

 

[blastingcap]

Mantle is promising in that if it decreases debug times that may be reason enough for it even if performance doesn't improve. But if it's cheaper and easier to slap an ARM core onto a GPU and call it a day, I'm all for that, too.

 

[ShintaiDK]

Why would they lie about Mantle's properties?

Maybe you should check their history, then answer me why they wouldnt.

Marketing is marketing, they are gonna try spin the most positive cases as possible. Even going above and beyond what it can actually do.

And just look at Mantle so far.

Its in consoles! Ups, its not in consoles.
Its open source! Ups, its not open source.
Its super quick and cheap to implement! Ups, its not so easy or cheap after all.
nVidia/Intel can support mantle! Ups, they cant support mantle.

Conflicting statements.

Plus as demonstrated with Tomb Raider and the API overhead for that game in terms of ingame vs benchmark FPS. There is simply not so much to gain. Or a quick look at how badly consoles perform.

But again, Mantle can run from its biggest issue. It only works on GCN and it cost to add.

 

[ShintaiDK]

Mantle is promising in that if it decreases debug times that may be reason enough for it even if performance doesn't improve. But if it's cheaper and easier to slap an ARM core onto a GPU and call it a day, I'm all for that, too.

The lower level, the higher cost and extra debug time usually.

 

[Gloomy]

The lower level, the higher cost and extra debug time usually.

an exception being coding against black boxes, no?

 

[ShintaiDK]

an exception being coding against black boxes, no?

Its no different than C++ being much harder and complicated to code for than JAVA or VB for example.

 

[SiliconWars]

Maybe you should check their history, then answer me why they wouldnt.

Marketing is marketing, they are gonna try spin the most positive cases as possible. Even going above and beyond what it can actually do.

And just look at Mantle so far.

Its in consoles! Ups, its not in consoles.
Its open source! Ups, its not open source.
Its super quick and cheap to implement! Ups, its not so easy or cheap after all.
nVidia/Intel can support mantle! Ups, they cant support mantle.

Conflicting statements.

Most of which never existed. AMD never said it was in consoles, never said it was open source and never said Nvidia or Intel can support Mantle. All they said out of all that was that it was quick and cheap to implement - something that has been backed up by 2 or 3 different independent sources so far.

But again, Mantle can run from its biggest issue. It only works on GCN and it cost to add.

And yet both DICE and Oxide said it was easily worth it.

 

[ShintaiDK]

And yet both DICE and Oxide said it was easily worth it.

How is it working out for DICE again? When is the patch coming, if ever?

 

[Genx87]

Any word on when the BF4 patch is due out? I thought they said Dec?

 

[SiliconWars]

How is it working out for DICE again? When is the patch coming, if ever?

Who knows? Clearly BF4 is a mess but that's not the fault of Mantle. The problem DICE has just now is that if they release the Mantle patch they'll just get accused of putting the DX game aside. It's a no-win situation.

 

[Skurge]

How is it working out for DICE again? When is the patch coming, if ever?

Soon.