AMD Releasing Public Mantle SDK This Year, Encourages Nvidia and Intel to Use it... For Free.

[Faa]

you keep bringing up the same super optimized fud games. who knows what is really happening under the hood to gain such performance.

It's your 3rd trollish nonsense post.

[holytoledo]

Some of the people defending Nvidia are sounding like Fanboys and I don't think those arguing for AMD are fanboys. They sound very level headed and their arguments make sense.

 

Higher memory bandwidth permits the 290x to perform better at 4k. Mantle lifts CPU bottlenecks for games.

 

These are all good things that we should be pushing for regardless of what card we have.

 

I recently bought a 970.

[monstercameron]

It's your 3rd trollish nonsense post.

ok alatar, what ever you say.

[Ashaira]

Still, Mantle is only useful for games that are CPU bound where the CPU is the bottleneck, or any game where the CPU is simply too weak to keep up with the GPU. Other than that, the gains are pretty small. If I remember correctly, Mantle caused a performance drop in some titles. It may have been rectified since then though, or may have never occurred.

 

For gods sake people that's what mantle was supposed to do. Remove CPU bottlenecks from the api. Look at ACU while it may be poorly optimized a big reason why it is so hard to run it is because it uses a lot of cpu for the npc (useless though they may be). In civ beyond earth mantle has tangible benefits in large maps with lots of units. the turn and ai proccessing is much faster due to more cpu being available.

[Humbug]

Still, Mantle is only useful for games that are CPU bound where the CPU is the bottleneck, or any game where the CPU is simply too weak to keep up with the GPU. Other than that, the gains are pretty small.

an API which can render the same scene with less CPU time is clearly an example of superior engineering. It's clearly doing something right. If DX12 and OpenGL Next can do the same it is good for everyone. This is why it is so wrong when people say it doesn't matter to me because I have a good CPU. Yes it does matter to you! Whether you have powerful hardware or not we should all strive for good optimization. Good software means achieving a job in an efficient and non-bloated manner. e.g. if the min GPU for AC:Unity is a geforce 680 then that's not good, and even people who have Titan Blacks and R9 295x2s should consider it not good, it means that the dev has failed to properly utilize his resources and it means that they are unable to get decent performance which means they have no headroom to push graphical boundaries.

 

The fact that for the moment in some games you can still brute force it with more powerful hardware is besides the point, better optimization is good for the industry and will benefit us all in the long-run.

-Widespread adoption of more efficient APIs mean that devs can free up the CPU to do other things which make the game world more immersive. e.g. more advanced AI routines and more going on on-screen.

-GPUs are getting faster and faster so rapidly. CPUs are not. This means that CPU bound scenarios will become more and more common if we keep doing things the way we are atm. Already today we have loads of CPU bound games, especially multiplayer. This is going to get worse in future unless we address it.

[Godlygamer23]

an API which can render the same scene with less CPU time is clearly an example of superior engineering. It's clearly doing something right. If DX12 and OpenGL Next can do the same it is good for everyone. This is why it is so wrong when people say it doesn't matter to me because I have a good CPU. Yes it does matter to you! Whether you have powerful hardware or not we should all strive for good optimization. Good software means achieving a job in an efficient and non-bloated manner. e.g. if the min GPU for AC:Unity is a geforce 680 then that's not good, and even people who have Titan Blacks and R9 295x2s should consider it not good, it means that the dev has failed to properly utilize his resources and it means that they are unable to get decent performance which means they have no headroom to push graphical boundaries.

The fact that for the moment in some games you can still brute force it with more powerful hardware is besides the point, better optimization is good for the industry and will benefit us all in the long-run.

-Widespread adoption of more efficient APIs mean that devs can free up the CPU to do other things which make the game world more immersive. e.g. more advanced AI routines and more going on on-screen.

-GPUs are getting faster and faster so rapidly. CPUs are not. This means that CPU bound scenarios will become more and more common if we keep doing things the way we are atm. Already today we have loads of CPU bound games, especially multiplayer. This is going to get worse in future unless we address it.

I didn't say anything about not needing it because I have a good CPU. In no way did I say that. Games that are not inherently CPU bound, like Borderlands, will see next to no benefit to Mantle. Games like Star Citizen will because it will be CPU intensive which will benefit people with decent CPUs and people with horrid CPUs. In addition, I fully support optimizations in titles.

[Godlygamer23]

For gods sake people that's what mantle was supposed to do. Remove CPU bottlenecks from the api. Look at ACU while it may be poorly optimized a big reason why it is so hard to run it is because it uses a lot of cpu for the npc (useless though they may be). In civ beyond earth mantle has tangible benefits in large maps with lots of units. the turn and ai proccessing is much faster due to more cpu being available.

I wasn't complaining about Mantle.

[Humbug]

I didn't say anything about not needing it because I have a good CPU. In no way did I say that. Games that are not inherently CPU bound

Cool.

I was making a general point. If you can do the same job with less CPU time then that is the way to go regardless of the fact that other bottlenecks exist for now in some games.

[Shahnewaz]

Said it multiple times it's not always there, learn to read perhaps.

Then why are you trying to prove something bad about a thing that's volatile? Today you whine that AMD drivers cause "overhead", tomorrow there might be none!

 

What job? I provided everything already with official response from Nvidia themselves. Also if that is the case, then it's done with Mantle already with DX12 coming up.

LOL haven't seen anything like that. Are you an Nvidia PR? I meant when I asked you to be transparent, is to provide references to what points and arguments you are bringing. You are showing performance benefits of Nvidia cards over AMD cards when I had no idea Nvidia made a recent update to their driver. Now obviously that makes a world of a difference. I don't keep up with Nvidia drivers because I don't have an Nvidia GPU.

 

You've seen it scaling up to 100%, just like Mantle did. Don't come here ask us how much fps it basically adds when you've seen it.

LOLZ!!! Even Nvidia themselves don't claim 100% performance improvement in ANY game with 337.50 drivers. Please read this time: http://www.nvidia.com/download/driverResults.aspx/74636/en-us

Oh, and also for a bit of fun, please read this: http://www.extremetech.com/gaming/180088-nvidias-questionable-geforce-337-50-driver-or-why-you-shouldnt-trust-manufacturer-provided-numbers

 

6 fps difference between the 290x & 780 Ti which is with both at full load 99%, even if the 290x suffers from driver overhead that's no longer valid anymore since its running at its full potential. You still do not want to understand that the amount of CPU bottlenecks you remove has nothing GPU bound scenario's. Read my bottleneck explanation further before you're arguing a point nobody here is arguing except you guys who are misunderstanding the meaning of what CPU overhead caused by a driver means.

How do you know that the driver is causing a CPU overhead? Do you know there is something called performance profiles? Game specific performance improvements? I can't trust an explanation from some like you! If there was actually a "driver overhead bottlenecking a cpu" situation then I would have heard that from far more reputable and trusted sources like Tom's Hardware, AnandTech, SemiAccurate etc. You assuming that won't change anything.

 

Mantle has done nothing Directx couldn't do. Copying here salestalk doesn't help you proving it is really happening, Linus is just copypasting what the PR guys said.

You're linking 4K benchmarks with a single GPU, the GPU will be max'ed out all the time which comes down to which GPU is more powerful. Like I said 10 times we aren't discussing this. We all know the 290X is better for 4K because of its wider bus.

5960x with 970 150 fps

That's a 80% difference.

Why don't you care so much about CF/SLI? It increases the CPU load.

 

Are you trying to run away from my point? I think I've clearly established ground in my argument. And benchmarks I previously posted from AnandTech and SemiAccurate clearly shows just that.

I quote from the SA article: http://semiaccurate.com/2014/10/28/look-civilization-beyond-earth/

We also tested this game on a PC build on Intel’s Core i7-5960X on Gigabyte’s X99-UD7 motherboard with RAM and water-cooling from Corsair and a reference Radeon R9 290X send to us by AMD.

We tested Beyond Earth on the Ultra present with at 1080P with 8x MSAA and V-sync disabled.

Seems like YOU need to learn to read.

And again, just because no one has to look for my previous post, the results:

And just because you don't get confused:

More from the SA article:

We were expecting to see a small performance bump from using Mantle based on the graphs that AMD supplied in its reviewers guide. Much to our surprise though using Mantle doubled and nearly doubled minimum and average frame rates respectively. It’s important to note that frame rates were measured using the lowest performance slice of Beyond Earth’s built-in late game benchmark. Thus these numbers are as close to a pathological worst case scenario as you’re going to get and as always your mileage may vary. With all of that said it appears that Mantle really unlocks the rendering bottleneck present in DirectX 11. We probably shouldn’t be surprised though as Civ games are almost always CPU hogs and Mantle is designed to free up resources on just that. Subjectively and with Vsync off, there was less stuttering and tearing using the Mantle version of Beyond Earth.

 

The system used in the AnandTech benchmark: http://www.anandtech.com/show/8640/benchmarked-civilization-beyond-earth

CPU Intel Core i7-4770K (4x 3.5-3.9GHz, 8MB L3)

Overclocked to 4.1GHz Motherboard Gigabyte G1.Sniper M5 Z87 Memory 2x8GB Corsair Vengeance Pro DDR3-1866 CL9 GPUs Sapphire Radeon R9 280

Sapphire Radeon R9 280X

Gigabyte Radeon R9 290X

EVGA GeForce GTX 770

EVGA GeForce GTX 780

Zotac GeForce GTX 970

Reference GeForce GTX 980

Laptops:

GeForce GTX 980M (MSI GT72 Dominator Pro)

GeForce GTX 880M (MSI GT70 Dominator Pro)

GeForce GTX 870M (MSI GS60 Ghost 3K Pro)

GeForce GTX 860M (MSI GE60 Apache Pro) Storage Corsair Neutron GTX 480GB Power Supply Rosewill Capstone 1000M Case Corsair Obsidian 350D Operating System Windows 7 64-bit

 

Result @ 1080p Ultra:

 

Minimums:

 

I don't know where you're getting your graphs from but I'll trust AnandTech more than whatever you're showing. If there was a bottleneck then AnandTech's benchmark should cap at 82.9 fps too.

Lastly I don't care about SLI/Crossfire benchmarks because their performance is heavily dependent on Crossfire/SLI profiles. They can cause a night and day difference in performance.

[Faa]

Then why are you trying to prove something bad about a thing that's volatile? Today you whine that AMD drivers cause "overhead", tomorrow there might be none!

It's been for almost a year like this.

 

LOL haven't seen anything like that. Are you an Nvidia PR? I meant when I asked you to be transparent, is to provide references to what points and arguments you are bringing. You are showing performance benefits of Nvidia cards over AMD cards when I had no idea Nvidia made a recent update to their driver. Now obviously that makes a world of a difference. I don't keep up with Nvidia drivers because I don't have an Nvidia GPU.

We're not comparing cards at all.

 

LOLZ!!! Even Nvidia themselves don't claim 100% performance improvement in ANY game with 337.50 drivers. Please read this time: http://www.nvidia.com/download/driverResults.aspx/74636/en-us

Oh, and also for a bit of fun, please read this: http://www.extremetech.com/gaming/180088-nvidias-questionable-geforce-337-50-driver-or-why-you-shouldnt-trust-manufacturer-provided-numbers

Seriously fuck off. You're linking fucking GPU limited benchmarks all the freaking time when we are talking about how much CPU bottlenecking Nvidia's newer drivers & Mantle removes. 

GPU loads were around 40/40%-50/50%

Thats nearly a 100% gain, seeing how that GPU load cranked up from 50% load to 99% explains enough nvidia managed to remove a lot of CPU bottlenecking. 337.50 didn't make any difference over the older drivers there, that spot is my reference for benchmarking/testing etc.

 

How do you know that the driver is causing a CPU overhead? Do you know there is something called performance profiles? Game specific performance improvements? I can't trust an explanation from some like you! If there was actually a "driver overhead bottlenecking a cpu" situation then I would have heard that from far more reputable and trusted sources like Tom's Hardware, AnandTech, SemiAccurate etc. You assuming that won't change anything.

Performance profiles, you're really flapping any bs you can. This is the 2nd time I'm posting this, just points out you're not reading my posts at all.

PCPER: Why do we see performance improvements in some sections of games but not in other sections (Skyrim for example)?

NVIDIA: For any given set of in-game settings, sections of the game may differ in terms of CPU versus GPU bottlenecks. Driver overhead reduction (one of the key improvements in 337.50) doesn’t help much if the area of the game is GPU limited. The CPU-limited segments of the game will benefit however.

We can also only reduce overhead in the driver. If the app is CPU-limited in the game itself, then it won’t see much improvement from an optimized driver.

PCPER: Why would SLI benefit more often and in higher numbers than single GPU configurations?

NVIDIA: SLI effectively doubles the available GPU horsepower, shifting the bottleneck from the GPU to the CPU. The net result is a more CPU-limited load. Consequently, our driver overhead optimizations are more likely to improve performance when SLI is enabled.

PCPER: Why do some games see more scaling with higher end CPUs than with lower end CPUs (this seems kind of counter-intuitive)?

NVIDIA: If a game (or portions of a game) are heavily CPU-bound in the game itself, rather than the driver code running on the CPU (which is the target of our 337.50 driver optimizations), then it’s possible for higher-end CPUs to show better scaling overall. But if the CPU-boundness is more driver-related, the 337.50 optimizations will show more scaling benefit on lower-end CPUs than higher-end CPUs. DX11 gives you the opportunity to tailor your optimizations to the issue you’re trying to solve. If you are dealing with a CPU bottleneck, your focus is on reducing overhead. However, with a scene that is facing a GPU bottleneck, your goal is to get the most GPU performance possible. Different games will require different strategies, and the effectiveness of those strategies will vary with different hardware configurations.

Source: http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-33750-Driver-Analysis-Single-GPU-and-SLI-Tested/Single-GPU-Testing-and

 

Are you trying to run away from my point? I think I've clearly established ground in my argument. And benchmarks I previously posted from AnandTech and SemiAccurate clearly shows just that.

I quote from the SA article: http://semiaccurate.com/2014/10/28/look-civilization-beyond-earth/

Seems like YOU need to learn to read.

And again, just because no one has to look for my previous post, the results:

Anandtech is using 4x MSAA which will increase the GPU load which would obviously hit 99% load. When are you going to understand that there's no CPU bottlenecking when the GPU is at full load? When? Even if the drivers cause a bunch of overhead, if you have a powerful CPU you can mitigate this to a point that you still can reach 99% load.

What part aren't you understanding of drivers causing CPU overhead? What part? Is it so difficult to understand that it makes the CPU bottleneck worse or just creates a CPU bottleneck? A 290x at 99% load vs a 780 Ti at 99% load is the typical 5-10% difference we've seen in all reviews. I'm not looking at such benchmarks, I'm posting benchmarks where the CPU is bottlenecking both cards. The 15time I'm explaining this.

Pclab was clearly not having the GPU max'ed out on the 290x due to driver overhead which was confirmed by at least 4 different sources and I'm posting another source that confirms my points again.

50% difference between a 780 Ti and 290x in Watch Dogs which Pclab even confirmed as well;

i5 4460 has 30 fps with a 290x and 45 fps with a 780 Ti. Again a 50% difference. That's not coincidence anymore.

 

 

I don't know where you're getting your graphs from but I'll trust AnandTech more than whatever you're showing. If there was a bottleneck then AnandTech's benchmark should cap at 82.9 fps too.

Lastly I don't care about SLI/Crossfire benchmarks because their performance is heavily dependent on Crossfire/SLI profiles. They can cause a night and day difference in performance.

Techreport, Pclab, gamegpu, pcgameshardware.de, PCPER all confirmed massive CPU overheads in DX11 on AMD. 

Copypasting results from the link you gave me;

== Hardware Configuration =================================

GPU: NVIDIA GeForce GTX 970

CPU: GenuineIntel

      Intel® Core i7-3930K CPU @ 3.20GHz

Physical Cores: 6

Logical Cores: 12

Physical Memory:  17114951680

Allocatable Memory: 140737488224256

===========================================================

 

 

== Configuration ==========================================

API: DirectX

Scenario: ScenarioFollow.csv

User Input: Disabled

Resolution: 1920x1080

Fullscreen: True

GameCore Update: 16.6 ms

Bloom Quality: High

PointLight Quality: High

ToneCurve Quality: High

Glare Overdraw: 16

Shading Samples:  64

Shade Quality: Mid

Deferred Contexts (D3D11): Disabled

Small Batch Optimized (Mantle): Enabled

Temporal AA Duration: 16

Temporal AA Time Slice: 2

Detailed Frame Info: Off

===========================================================

 

 

== Results ================================================

Test Duration: 360 Seconds

Total Frames: 20189

 

Average FPS: 56.08

Average Unit Count: 4441

Maximum Unit Count: 5406

Average Batches/MS: 1111.09

Maximum Batches/MS: 3980.34

Average Batch Count: 20360

Maximum Batch Count: 170891

===========================================================

Thats nearly twice as much.

== Hardware Configuration =================================

GPU: AMD Radeon R9 200 Series

CPU: GenuineIntel

Intel® Core i7-3770K CPU @ 3.50GHz

Physical Cores: 4

Logical Cores: 8

Physical Memory: 17117880320

Allocatable Memory: 8796092891136

===========================================================

== Configuration ==========================================

API: Mantle

Scenario: ScenarioAttract.csv

User Input: Disabled

Resolution: 1920x1080

Fullscreen: True

GameCore Update: 16.6 ms

Bloom Quality: High

PointLight Quality: High

ToneCurve Quality: High

Glare Overdraw: 16

Shading Samples: 64

Shade Quality: Mid

Deferred Contexts: Disabled

Temporal AA Duration: 16

Temporal AA Time Slice: 2

Detailed Frame Info: Off

===========================================================

== Results ================================================

Test Duration: 360 Seconds

Total Frames: 18980

Average FPS: 52.72

Average Unit Count: 4390

Maximum Unit Count: 5593

Average Batches/MS: 1280.15

Maximum Batches/MS: 3224.82

Average Batch Count: 28892

Maximum Batch Count: 205885

===========================================================

And we outperformed it. Here's the video before you act like a fanboy again; youtube.com/watch?v=eLrRg3rVxcY

 

Lastly I don't care about SLI/Crossfire benchmarks because their performance is heavily dependent on Crossfire/SLI profiles. They can cause a night and day difference in performance.

 

No, SLI/CF performance is heavily CPU dependent. A moderator on the geforce forum wrote a nice long article about SLI/CF Scaling, have a nice read; https://forums.geforce.com/default/topic/532913/sli/geforce-sli-technology-an-introductory-guide/post/3749687/#3749687

Crossfire/SLI profiles are nearly damn perfect these days. http://nl.hardware.info/reviews/5622/3/nvidia-geforce-gtx-980-sli--3-way-sli--4-way-sli-review-benchmarks-battlefield-4

Noticing in 1080p there was no difference at all from a 3rd or even a 4th card in 1080p, when they moved to 4K which shifted the CPU bottleneck towards the GPU the scaling was lineair. 

1 -> base

2 -> 100% more than 1

3 -> 50% more than 2

4 -> 33% more than 3.

Why wouldn't you care about SLI/CF benchmarks? NVIDIA: SLI effectively doubles the available GPU horsepower, shifting the bottleneck from the GPU to the CPU. The net result is a more CPU-limited load. Consequently, our driver overhead optimizations are more likely to improve performance when SLI is enabled.

We were arguing how much overhead they were causing, nvidia's DX11 vs mantle, so why not argue something that taxes the CPU harder? PCPER showed a 50% difference between 980 SLI vs 290x CF with Mantle.

Don't know why you're still arguing when I linked a source where a nvidia guy is pretty much saying what I'm saying. You're just being a fanboy, that's all. 

== Hardware Configuration =================================

GPU: AMD Radeon R9 200 Series

CPU: GenuineIntel

Intel® Core i7-3770K CPU @ 4.60GHz

Physical Cores: 4

Logical Cores: 8

Physical Memory: 17117880320

Allocatable Memory: 8796092891136

===========================================================

== Configuration ==========================================

API: DirectX

Scenario: ScenarioAttract.csv

User Input: Disabled

Resolution: 1920x1080

Fullscreen: True

GameCore Update: 16.6 ms

Bloom Quality: High

PointLight Quality: High

ToneCurve Quality: High

Glare Overdraw: 16

Shading Samples: 64

Shade Quality: Mid

Motion Blur Frame Time: 16

Motion Blur InterFrame Time: 2

Detailed Frame Info: Off

===========================================================

== Results ================================================

Test Duration: 120 Seconds

Total Frames: 4013

Average FPS: 33.44

Average Unit Count: 3868

Maximum Unit Count: 5385

Average Batches/MS: 847.03

Maximum Batches/MS: 1610.35

Average Batch Count: 27747

Maximum Batch Count: 219345

===========================================================

[Diceair]

Jus because Nvidia released a driver to make it better than AMD mantle the proof should be by adding mantle to their own cards. I would love to see mantle in every card cause when they make mantle for linux it should open a world of possibility. You can then run most future games on SteamOS