The AMD Mantle Thread

[Gloomy]

Just realized as I was writing up a reply that I play at 2560x1440, with an overclock on my 290, so my personal experience doesn't apply here.

Don't mind me.

 

[blackened23]

Even at GPU limited settings (eg. 2560x1600/1440, ultra, with a strong processor like an extreme edition), improvements should be 10-15% with a card like a 290 or 290x.

You know this how? Citation? Source? Experience? Direct playing of BF4 with Mantle? What and how? So you're stating that AMD 290 users should expect 10-15% after AMD touted 45%?

So if we put the average 290 framerate at 60 at 2560x1600, that means the average FPS jumps to 66-68 with Mantle being 10-15%? Am I following? I would assume most are going to be GPU limited, at least with Haswell or IB-E CPUs.

6 more frames doesn't seem like "ridicule" territory to me. SHRUG.

 

[ASM-coder]

It's a pity Mantle is destined to fail

Unbiased opinion, or fact?

 

[Grooveriding]

And I bet they are all in the single player campaign, as AMD is better able to compete because it's far less CPU intensive than the multiplayer..

The reason why I did that was to show the CPU dependency that I was referring to and how crippled AMD's current drivers are in that regard. That's the whole purpose of creating Mantle, to make exploiting multicore processors easier and more effective.

It's a pity Mantle is destined to fail, because two APIs cannot co-exist indefinitely. One will eventually be eaten, and I can guarantee you it won't be Direct3D, as Direct3D despite it's flaws, has two massive advantages; cross architectural and backward compatibility. Enhanced multithreading capability can be added over time, as they did with deferred contexts.

You're not showing any CPU dependencies. You showed a GPU limited benchmark and I posted another. Your theory about AMD's vs Nvidia's drivers and CPU usage is just your theory. Those benchmarks certainly didn't give the theory any substance, but you put forth your opinion on it as fact.

You're speaking in a lot of absolutes without any actual inside knowledge, but then again, that is 90% of this thread at this point.

Ultra 2X MSAA 2560x1600. I see nothing unplayable here on the GTX 780 vanilla. Fact is, carfax83 is correct: nvidia substantially increased BF4 performance through drivers. I personally remember how it performed at launch, and it DEFINITELY performs way way better now.

So ? A completely unrelated derail point about driver improvements doesn't mean the points I was actually addressing, which you paid no mind, are also correct. He's passing off a theory about drivers and CPU usage that has nothing to do with anything I was actually initially saying, not even realizing the settings of the benchmarks he's posting.



Thread is rife with logic impediment.

 

[chimaxi83]

Unbiased opinion, or fact?

A quick look at his posting history will answer that for you.

 

[3DVagabond]

This has already been explained. Just because it is free to the user doesn't mean it doesn't come without a price. Additional development time (no matter how small) costs money.

It's called ROI. You need to spend money to make money.

 

[VulgarDisplay]

The other thing I'm interested in is IQ improvements with mantle. They can use that loss of overhead to do something great like push draw distances.

I would love an unfair advantage in bf4.

 

[3DVagabond]

Wow NV does show strong now in BF4.....that wouldn't be good if the 780ti was king in a AMD optimized game that is also using a AMD proprietary API.

I see this sentiment often that it's not good for AMD if nVidia is faster in games they sponsor. It should be obvious by now that AMD doesn't concentrate on making games perform worse on nVidia cards. It's good for gamers. That's all any of us should care about. All AMD gets is early optimization advantage. Rather than seeing it as a negative for AMD you should simply see it as a positive. Especially if you own nVidia hardware.

 

[desprado]

You know this how? Citation? Source? Experience? Direct playing of BF4 with Mantle? What and how? So you're stating that AMD 290 users should expect 10-15% after AMD touted 45%?

So if we put the average 290 framerate at 60 at 2560x1600, that means the average FPS jumps to 66-68 with Mantle being 10-15%? Am I following? I would assume most are going to be GPU limited, at least with Haswell or IB-E CPUs.

6 more frames doesn't seem like "ridicule" territory to me. SHRUG.

U are right AMD has this kind of habit to hype but at end it is not was that people hoped for.

Nvidia has advantage on multi thread games and AMD will try to finish that Gap with Mantle or much more.

 

[el etro]

My bet with Mantle on BF4 multiplayer(1920x1080 res):


Mantle + FX6300 plus 7990@1,2Ghz* on BF4-64Players Multi Player: equals to i7 4960x + 7990Ghz@1,2Ghz without Mantle on same 64 Players BF4 MP Map;


Mantle + FX9590@5Ghz plus xfire r9 290x@1,2Ghz** on Bf4-64Players Multi Player: faster than i7 4960x@5Ghz + Sli GTX 780Ti@1,2Ghz without Mantle on same 64 Players BF4 MP Map;



*7990@1,2Ghz: Equals/Outperform to stock GTX 780 Sli performance;

**290x@1,2Ghz Xfire: Equals/Outperform GTX 680 4-way SLI performance.

 

[Carfax83]

Your theory about AMD's vs Nvidia's drivers and CPU usage is just your theory. Those benchmarks certainly didn't give the theory any substance, but you put forth your opinion on it as fact.

Yes, and all theories have good supporting evidence. Whilst the BF4 benchmark I posted doesn't admittedly constitute good evidence (I basically posted that to show you that you were wrong in saying that the 290x was faster than the Titan in BF4), the benchmarks at PClab.pl do:

Benchmarks

These benchmarks show beyond a doubt that what I'm saying is true.

You're speaking in a lot of absolutes without any actual inside knowledge, but then again, that is 90% of this thread at this point.

What does it matter? Do you want AMD to come out and admit that, "Yeah, our drivers really suck at exploiting multicore processors which is why we came up with Mantle in the first place!"

The evidence is there for any free thinking individual willing to look for themselves.

 

[Grooveriding]

Looking at those pclab benchmarks they in no way 'show beyond a doubt that what I'm saying is true.'

Priceless stuff.

So much FUD and personal opinion being tossed out as fact. FUD spreaders from both sides just wait on the benchmarks then gloat or cry all you want when there is something real to make claims on.

 

[itsmydamnation]

First level implies Singleplayer.

Singleplayer is hardly CPU bound

im pretty sure you will find repi uses eyefinity

https://twitter.com/repi/status/16942728402


edit: i've seen him in video as well working on a 3x screen setup. ( god know where.....lol)

 

[el etro]

I bet Mantle stronghold will be a heavy decrease in CPU performance needed to drive powerful cards without bottleneck. With Mantle we will see ridiculously weak processors driving top-class GPUs with relatively low bottleneck.
And then most of us will no more need to spend money on expensive processors like i7s if the most CPU-intensive task we do is web browsing or made little works in word or excel.

 

[Gloomy]

im pretty sure you will find repi uses eyefinity

https://twitter.com/repi/status/16942728402


edit: i've seen him in video as well working on a 3x screen setup. ( god know where.....lol)

Uhh, that picture has like 8 monitors, a laptop, and an HDTV lol. They could have been playing on anything.

 

[itsmydamnation]

not the link the comment

Trying Eyefinity 3x1 portrait with 3x 24" (3844x1920), great for gamedev! Thank you AMD for the screen upgrade

 

[Carfax83]

Looking at those pclab benchmarks they in no way 'show beyond a doubt that what I'm saying is true.'

Priceless stuff.

And they show absolutely nothing then is that it?

Ah screw it. There's no point in debating this issue with you, as your mind is already made up that AMD's drivers are just as good at using multicore processors as NVidia's..

Anything to the contrary is purely FUD..

 

[DiogoDX]

What benchmarks are those? The thing about benchmarks, is that they are subject to patches and driver updates.

The Radeons had a good lead on the NVidia cards when BF4 first became available, but after patch and driver updates, NVidia has all but closed the gap in SP, and actually has a large lead in multiplayer thanks to the aggressive multithreading implementation in their drivers:

NVidia drivers seem to be much better at leveraging multicore CPUs for performance. AMD is way behind NVidia in that front, and that's probably why they need something like Mantle, which is more geared towards multithreading at the foundation.

This test is a joke. 30s empty server bench.:thumbsdown:

http://www.youtube.com/watch?v=oCrjWuNJl94

 

[Carfax83]

This test is a joke. 30s empty server bench.:thumbsdown:

http://www.youtube.com/watch?v=oCrjWuNJl94

If you're testing GPU performance, what difference does it make whether the server is empty or full?

CPU performance on the other hand relies on a populated server, because the more players you have the more CPU power is needed...

GameGPU knows this, which is why their CPU test used a populated server.

 

[krumme]

Gpu frame to frame pace tend to be more consistent than cpu frame to framepace.
We have consoles using jaguar cores working fine giving smooth gameplay.

I expect far better pacing of the cpu frame to frame with mantle giving smooter experience.
Add to that the vital min will probably be greatly reduced.

Measured the above could give a low average fps increase but the actual experience gaming could be far better non the less.

When i get the mantle update i dont give a damn about the fps. I will just simply game game game as usual. Its very simple to judge by yourself. Framepacing charts could help and 1% min assessments fps but my subjective feeling will decide.

 

[krumme]

Yeah, we should make a petition to remove those wicked coffee machines in their offices.

Lmao and spot on. Its about time those wicked coofee machines have to go.

 

[zlatan]

I agree with this. There's a reason they used an APU instead of an 8 core FX processor for the claim.

Maybe they implemented a complex multi-GPU solution, so the iGPU in the Kaveri APU is also gives an extra performance boost compared to the D3D path.

 

[Gloomy]

Maybe they implemented a complex multi-GPU solution, so the iGPU in the Kaveri APU is also gives an extra performance boost compared to the D3D path.

way too optimistic

 

[desprado]

Gpu frame to frame pace tend to be more consistent than cpu frame to framepace.
We have consoles using jaguar cores working fine giving smooth gameplay.

I expect far better pacing of the cpu frame to frame with mantle giving smooter experience.
Add to that the vital min will probably be greatly reduced.

Measured the above could give a low average fps increase but the actual experience gaming could be far better non the less.

When i get the mantle update i dont give a damn about the fps. I will just simply game game game as usual. Its very simple to judge by yourself. Framepacing charts could help and 1% min assessments fps but my subjective feeling will decide.

U want to say that at 30fps with Mid to high settings and at barely 1080p than yes.

 

[3DVagabond]

Maybe they implemented a complex multi-GPU solution, so the iGPU in the Kaveri APU is also gives an extra performance boost compared to the D3D path.

While that's possible and it was an APU demo, I still think they were also trying to show a best case scenario. FWIU (and this is 2nd hand info) the "up to 45% in BF4" isn't meant as up to 45% in one spot of one map, it's up to 45% boost in BF4 overall/avg. depending on hardware. Remember this is just one game, so it can't be extrapolated as meaning all games will be up to 45%, just BF4. Other games could be more or less. This is an early implementation and not on an ideal game genre (fps as opposed to rts, for example), so it should get even better as more mantle optimizations are explored.