Ubisoft: AMD's Mantle API is a double-edged sword

[ShintaiDK]

Shader model 3.0 really started to get the ball rolling for ATI and nVidia, imho!

We got a winner!

 

[GaiaHunter]

Unified shaders got nothing as such to do with STREAM capabilities as seen on the R580.

We are talking about Mantle being compatible with future AMD architectures.

You are telling me a non unified shader is compatible with a unified shader?

 

[Skurge]

You guys can't be for real.

 

[ShintaiDK]

You are telling me a non unified shader is compatible with a unified shader?

STREAM wise? Yes.

Unified is not about capabilities. Its about efficiency usage.

I dont even think the Xbox360 with unified shaders are STREAM capable.

 

[GaiaHunter]

STREAM wise? Yes.

Unified is not about capabilities. Its about efficiency usage.

I dont even think the Xbox360 with unified shaders are STREAM capable.

http://en.wikipedia.org/wiki/Stream_processing

Stream processing is a computer programming paradigm, related to SIMD (single instruction, multiple data), that allows some applications to more easily exploit a limited form of parallel processing. Such applications can use multiple computational units, such as the FPUs on a GPU or field programmable gate arrays (FPGAs),[1] without explicitly managing allocation, synchronization, or communication among those units.

http://en.wikipedia.org/wiki/Xenos_(graphics_chip)
On the chip, the shader units are organized in three SIMD groups with 16 processors per group, for a total of 48 processors. Each of these processors is composed of a 5-wide vector unit (total 5 FP32 ALUs) that can serially execute up to two instructions per cycle (a multiply and an addition). Thus each of the 48 processors can perform 10 floating-point ops per cycle. All processor in a SIMD group execute the same instruction, so in total up to three instruction threads can be simultaneously under execution.
-------------------------------------------------------------------
A Cuda core is a unified shader. Those haven't changed extensively for NVIDIA since the G80.

AMD changed their unified shader from VLIW5/4 to SIMD with GCN.

 

[OCGuy]

Take it to PM?

 

[24601]

Take it to PM?

Shaming is pointless unless it's public.

Credibility must be destroyed to prevent future incidents.

 

[Slomo4shO]

Shaming is pointless unless it's public.

Arguing over pointless semantics and using Wikipedia for references makes it obvious that both parties have no shame...

 

[dguy6789]

Double or triple performance. Dream on.

DirectX has massive overhead compared to low level coding. This isn't debatable, you're either aware of it or not. Simply looking at how consoles perform compared to a PC with similar number crunching power running DX or OpenGL it makes it overwhelmingly obvious that it's a pretty plausible best case prediction. A computer running Windows with double the CPU and GPU power of an Xbox 360 will run Skyrim or Bioshock Infinite at the same settings at a lower FPS than it does entirely because of DirectX inefficiencies.

We don't really know just how much control Mantle will give game programmers but if it is as much as they get on consoles then double performance is definitely possible if they put the time in.

 

[24601]

DirectX has massive overhead compared to low level coding. This isn't debatable, you're either aware of it or not. Simply looking at how consoles perform compared to a PC with similar number crunching power running DX or OpenGL it makes it overwhelmingly obvious that it's a pretty plausible best case prediction. A computer running Windows with double the CPU and GPU power of an Xbox 360 will run Skyrim or Bioshock Infinite at the same settings at a lower FPS than it does entirely because of DirectX inefficiencies.

We don't really know just how much control Mantle will give game programmers but if it is as much as they get on consoles then double performance is definitely possible if they put the time in.

Console games never ever ever have the same settings as PC games (excluding performance non-dependent games like say tetris)

The people who peddle these lies really need to stop. DirectX's inefficiencies are all CPU related. Any performance increases would be as a result of relieving any CPU bottlenecks. If there are none present in the specific test system, then the difference would be negligible.

The console advantage for developers is mostly that they can focus on not hitting any single bottleneck too soon, be it shader performance, pixel fillrate, texture fillrate, memory bandwidth etc. which would stop the designer short from their desired goal. The "solutions" they come up with are basically tailoring the entire game design to the target hardware, so as to use as few resources (manpower, manhours, capital, time) as possible to reach their goal.

Many ugly hacks are used to reach their desired goals on consoles that PC gamers do not tolerate.

 

[ShintaiDK]

DirectX has massive overhead compared to low level coding. This isn't debatable, you're either aware of it or not. Simply looking at how consoles perform compared to a PC with similar number crunching power running DX or OpenGL it makes it overwhelmingly obvious that it's a pretty plausible best case prediction. A computer running Windows with double the CPU and GPU power of an Xbox 360 will run Skyrim or Bioshock Infinite at the same settings at a lower FPS than it does entirely because of DirectX inefficiencies.

We don't really know just how much control Mantle will give game programmers but if it is as much as they get on consoles then double performance is definitely possible if they put the time in.

Consoles uses many features to boost performance. You are not comparing apples to apples. Just look at Crysis Ryse. 900p with dynamic downscaling from day 1. I wouldnt be surprised if the PS3/Xbox360 ran 360p in some parts of games. And API is about CPU performance mainly.

 

[OCGuy]

DirectX has massive overhead compared to low level coding. This isn't debatable, you're either aware of it or not. Simply looking at how consoles perform compared to a PC with similar number crunching power running DX or OpenGL it makes it overwhelmingly obvious that it's a pretty plausible best case prediction. A computer running Windows with double the CPU and GPU power of an Xbox 360 will run Skyrim or Bioshock Infinite at the same settings at a lower FPS than it does entirely because of DirectX inefficiencies.

We don't really know just how much control Mantle will give game programmers but if it is as much as they get on consoles then double performance is definitely possible if they put the time in.

This has to be a joke...console games use all sorts of tricks to help their performance.

Lower IQ, slower panning, closer view distances........

I have a feeling DX11.2 will bring as much of a performance increase to the PC as "mantle" does.

 

[Spjut]

Console games never ever ever have the same settings as PC games (excluding performance non-dependent games like say tetris)

The people who peddle these lies really need to stop. DirectX's inefficiencies are all CPU related. Any performance increases would be as a result of relieving any CPU bottlenecks. If there are none present in the specific test system, then the difference would be negligible.

Another advantage for the consoles is that their APIs can expose every feature the GPU supports.

For example, both the Xbox 360 and AMD's DX10 cards have support for tessellation. Viva Pinata, Banjo Kazooie and Halo Wars are known to use it on the 360.
But since it wasn't part of the DX9/DX10.1 spec, it was never used on the PC.

 

[24601]

Another advantage for the consoles is that their APIs can expose every feature the GPU supports.

For example, both the Xbox 360 and AMD's DX10 cards have support for tessellation. Viva Pinata, Banjo Kazooie and Halo Wars are known to use it on the 360.
But since it wasn't part of the DX9/DX10.1 spec, it was never used on the PC.

That's not a performance issue, that's merely a support issue.

Usually these things are hashed out between Nvidia, AMD/ATi, Intel, and Microsoft in developing DirectX. It is discussed between basically everyone who matters in the Khronos group in developing OpenGL.

 

[GaiaHunter]

DX gives no direct access to NVIDIA/AMD drivers on the PC side.

 

[Carfax83]

Except no one can tell a difference in gameplay from this "huge" lead in a strategy game where a last generation card is getting 70 fps. A strategy game is not like a racing or an FPS game where you can actually tell apart 90 vs. 138 fps. Where are these other games where NV badly beats AMD in a DX11 game in a similar price bracket as a result of multi-threaded DX11 driver?

Right, and you think the driver command list was a recent feature?

It's been around since Civ V first launched, before Kepler and GCN, and NVidia's GTX 580 on the 270xx driver was positively brutalizing AMD's 6000 series..

What difference does it make that AC3 uses driver command lists? 770 cannot beat R9 280X/7970GE in that title.

We can all pick and choose benchmarks. Driver command lists only help in CPU limited situations, and apparently, TPU didn't even know to find an area in the game that was CPU limited.

Assassin's Creed 3 benchmarks

With NV refusing to lower prices on the 770, the 4GB version has become a laughing joke, priced $150 more than R9 280X cards and offering no performance advantage in DX9-11 games.

I agree, the GTX 770 needs a price cut, but apparently people are still choosing the expensive GTX 770 over the 7970Ghz because it's still holding the same introductory price of 399 and 450.

Apparently, it's not all about performance per dollar. NVidia's extended feature set is superior to AMD's, and that's why people are willing to pay the money for it.

Right now a PC gamer can buy GTX770 level of performance for only $280 and set aside $100-170 for a next GPU upgrade. Considering 770 cannot win in Crysis 3, Metro LL or Tomb Raider, NV are absolutely mad if they think PC gamers are so brand-brainwashed/ignorant as to pay 42-61% more over HD7970Ghz for the same average gaming performance.

What? Crysis 3 runs faster on the GTX 770. Metro Last Light did as well, but I suppose AMD has tuned their latest drivers for it a bit more.

I'm sure there's a tessellation "optimization" in there somewhere to increase frame rates :sneaky:

 

[Carfax83]

Having 64 UAV slots is definitely the biggest gaming related feature in DX11.1. I wonder why they don't mention this.

I'm sure it's in there, probably under different terminology.

UAVOnlyRenderingForcedSampleCount and UAVs at all shader stages are also gaming related features.

According to NVidia, UAVOnlyRenderingForcedSampleCount is only used for 2D acceleration.

 

[Carfax83]

Uhm, your image is for a different resolution... So not really an apples to apples comparison there. Although its obvious driver changes have been made on both sides.

Yeah, like there's going to be a huge difference between 1080p and 1200p

 

[Spjut]

Got any source about AC3 using driver command lists?

All I know is that Dice repeatedly has said it's not working for them.
Far Cry 3 had it initially but it was disabled in a new patch because it was reported to cause stability issues(people have been able to activate it manually in the config file though)

 

[Erenhardt]

snip

No, really, 770 SLI was the best purchase decision you could ever make. Now, stop this OT.

Back on track. I think it is double edged sword. If PCs can get part of console optimization, PC gaming will be limited by "slow" consoles" even earlier. If suddenly lower end and integrated GCN can play ps4 games without any problems, no need for better cards then. Helps move low-end parts by the cost of higher-end ones.

 

[sontin]

Having 64 UAV slots is definitely the biggest gaming related feature in DX11.1. I wonder why they don't mention this.
UAVOnlyRenderingForcedSampleCount and UAVs at all shader stages are also gaming related features.

Kepler (+Fermi) supports UAV:
http://nvidia.custhelp.com/app/answ...sion/L3RpbWUvMTM4MTk1Mjc5MC9zaWQvSE95NktZQ2w=

 

[Carfax83]

Got any source about AC3 using driver command lists?

Here's a PDF about Deferred context which mentions AC3 using it

All I know is that Dice repeatedly has said it's not working for them.
Far Cry 3 had it initially but it was disabled in a new patch because it was reported to cause stability issues(people have been able to activate it manually in the config file though)

DICE said they couldn't get it working properly in BF3, but for BF4, who knows?

I don't think Far Cry 3 had it. Far Cry 3 had a DX 11 multithreading option, which isn't the same thing as far as I know. I tried enabling it and it didn't make any difference.

*Edit* I did some checking, and apparently, it is the same thing. But Ubisoft must have botched the implementation in Far Cry 3. They did a better job in AC3, as it does make a difference. AC4 will likely have it as well, and so will Watch Dogs.

 

[24601]

Here's a PDF about Deferred context which mentions AC3 using it



DICE said they couldn't get it working properly in BF3, but for BF4, who knows?

I don't think Far Cry 3 had it. Far Cry 3 had a DX 11 multithreading option, which isn't the same thing as far as I know. I tried enabling it and it didn't make any difference.

Thanks for that PDF

It's a good read.

 

[Carfax83]

Thanks for that PDF

It's a good read.

NP..

It really makes you wonder though why AMD didn't implement it in their drivers, as it seems to make a huge difference when done correctly. The PDF says that Civ 5 would make more than 10K draw calls in a frame at certain times, and I remember when NVidia first implemented this feature in their 270xx drivers; the GTX 580 really demolished AMD's 6000 series as a result.

I think a further refinement of this technology in DX12 could make Mantle redundant, as Mantle comes with strict limitations and disadvantages of it's own.

 

[cytg111]

"So, it’s really a double-edged sword: we will have to analyse the pros and cons before taking a decision. "

- Translates into : AMD might have to put in a dime here to make it worth our time. And Im sure AMD will.