The AMD Mantle Thread

[blastingcap]

Gsync is theoretically the better solution, but you can have both so why not have both if it's possible?

Thing about Mantle is that even a +20% increase can't close the gap between 30 and 60 fps sometimes. Let's take an example. Say a game bounces between 40 and 70 fps and your monitor refresh rate is 60Hz.

With Vsync you will bounce between 30 and 60 fps. Without it you get screen tearing. With Gsync you get the best of both worlds: the full 40 to 60 fps (the limit of your monitor) without screen tearing.

With Mantle giving, say, +20%, you will bounce between 30 and 60 fps with Vsync or else get 48 and 60 fps with screen tearing.

Bouncing between 30 and 60 fps is unpleasant. Now, you could use triple buffer vsync to avoid that, but that is not always an in-game option and it also means more input lag, problems g-sync would not have.

Different how? According to the Anandtech preview the performance was the same above 60 fps, but on dropping below 60 fps was where Gsync looked a lot better.

http://www.anandtech.com/show/7436/nvidias-gsync-attempting-to-revolutionize-gaming-via-smoothness

AMD says this about Mantle - http://beyond3d.com/showpost.php?p=1807817&postcount=677

Seems like 2 different ways to get to the same result to me.

 

[Spidre]

Mantle and Gsync are both great, and I think both play to the strengths of their companies target market.

 

[3DVagabond]

Mantle will not change this issue at all:

2 completely different things.

At, or above, the refresh rate, GSync will do absolutely nothing. It doesn't increase performance, It doesn't improve visuals, it'll only make marginally playable games look better (To a point. below 30fps it'll do nothing as well.) Add to it that it's only going to be available on one model of monitor at first, and not any of the better visual quality IPS/PLS models. If you happen to already own that one model that supports GSync it'll cost you ~$100 to retrofit it and make it compatible. If you do need to buy that one model of monitor that supports it new, that will be $100 more expensive than the same model cost before GSync.

For what it does, I think GSync is an awesome development. It's just too bad nVidia came up with it first because they're going to try and make a fast buck off of us while they can. So hopefully the technology can be adapted fast so more people can enjoy and take advantage of it rather than the other way around. You would think that sooner or later people would bet tired of that kind of stuff?

Meanwhile Mantle will actually improve game play performance and IQ, allow the hardware company and the devs to add new features without having to wait for M$ to develop a new standard, get more out of existing hardware, will cost you and me nothing, and they're talking about it being implemented on other companies hardware.

You are right, not the same thing at all.

 

[stahlhart]

This is the Mantle thread, not the GSync thread, and not the "what's better, Mantle or GSync?" thread. Please stay on topic.
-- stahlhart

 

[smackababy]

If AMD came out with that demo you'd all just line up to shoot it down for not being an independent source or a real game.

I'm quite certain that when the BF4 patch launches with spectacular performance gains you'll say it was all because AMD spent $8 million and that the average improvement will be much lower.

If AMD came out with a demo, I'd take it as such: a demo showing the strengths of their API. Them just blindly saying "this is going to give X% of performance" means very little to anyone with some sense.

They could easily release sample code showing how similar it is to port certain things over to Mantle just as easily as writing something to port it over. A simple scene written in DX, Mantle, Xbox One, and PS4 APIs would suffice. They can even showcase how each performs at the same image quality.

However, such a simple and easy to produce test actually showcasing the Mantle advantage they have marketing slides about being so easy to produce AND not being put forth makes me think Mantle isn't this savior to PC gaming most of you are hyping it up to be.

Again, a lot of us remember the days of DOS dying and developers crying the Windows abstraction would kill gaming because it was so slow and inefficient. Well, we all know how that played out...

 

[VulgarDisplay]

If AMD came out with a demo, I'd take it as such: a demo showing the strengths of their API. Them just blindly saying "this is going to give X% of performance" means very little to anyone with some sense.

They could easily release sample code showing how similar it is to port certain things over to Mantle just as easily as writing something to port it over. A simple scene written in DX, Mantle, Xbox One, and PS4 APIs would suffice. They can even showcase how each performs at the same image quality.

However, such a simple and easy to produce test actually showcasing the Mantle advantage they have marketing slides about being so easy to produce AND not being put forth makes me think Mantle isn't this savior to PC gaming most of you are hyping it up to be.

Again, a lot of us remember the days of DOS dying and developers crying the Windows abstraction would kill gaming because it was so slow and inefficient. Well, we all know how that played out...

Mantle is ready to be shown yet. They aren't doing it to obscure facts and build hype. They just don't have a solid enough demonstration that they can release performance numbers.

BF4 at the end of December is going to be the first demo.

Also, many of the games that are going to be using it aren't close enough to release to be showing footage yet.

 

[smackababy]

Mantle is ready to be shown yet. They aren't doing it to obscure facts and build hype. They just don't have a solid enough demonstration that they can release performance numbers.

BF4 at the end of December is going to be the first demo.

Also, many of the games that are going to be using it aren't close enough to release to be showing footage yet.

If Mantle as an API is ready to be baked into BF4, it is ready enough to show a tech demo or even a rendered scene using Mantle calls and DX calls. They could easily show sample code (even if it might change in syntax some before the final Mantle version) and compare that to the same code written on the consoles' API. Thus, confirming how "similar" they are.

Yet, we get nothing.

I don't think they are trying to hide "the truth" about Mantle, it just think they would rather let the hype build and under deliver than dispel some of it and give a chance for bad publicity before it launched.

 

[3DVagabond]

If Mantle as an API is ready to be baked into BF4, it is ready enough to show a tech demo or even a rendered scene using Mantle calls and DX calls. They could easily show sample code (even if it might change in syntax some before the final Mantle version) and compare that to the same code written on the consoles' API. Thus, confirming how "similar" they are.

Yet, we get nothing.

I don't think they are trying to hide "the truth" about Mantle, it just think they would rather let the hype build and under deliver than dispel some of it and give a chance for bad publicity before it launched.

They did show a demo at APU13. It was geared towards devs not consumers since that's who they have to sell it to first.

 

[psoomah]

I don't think they are trying to hide "the truth" about Mantle, it just think they would rather let the hype build and under deliver than dispel some of it and give a chance for bad publicity before it launched.

My read is the scarcity of hard numbers are to keep people talking, but that the only stated 'official' numbers out there so far (15%, 20%) are to give the naysayers ammunition to keep expectations down so the actual much higher hard numbers when BF4 Mantle is released will have that much more impact.

Site after site cautiously quote only those few stated hard numbers and add caveats on top of them so they become the expectation.

I'm thinking it's a sandbag job.

 

[krumme]

My read is the scarcity of hard numbers are to keep people talking, but that the only stated 'official' numbers out there so far (15%, 20%) are to give the naysayers ammunition to keep expectations down so the actual much higher hard numbers when BF4 Mantle is released will have that much more impact.

Site after site cautiously quote only those few stated hard numbers and add caveats on top of them so they become the expectation.

I'm thinking it's a sandbag job.

No its not. Look at bf4 ps4 xbox and you will know you dont get 100% on average. Mantle is not furmark. Gpu utilization is not so bad. And its not like modern dx is not trying to solve eg the drawcalls it just takes some work. Ofcource mantle can be 1000% faster in some situations but they are not typical and mantle will just expose other bottlenecks. The difference in mantle is those bottlenecks is visible to the programmer and can be ajusted to.

But Mantle is much more than performance!
Its cheap porting, more games to pc, better quality ports, cheaper programming on performance game meaning more highend games for us, the liberation of devs from boring work, new games we dont know about yet.
It will take 5 years untill we see the potential of mantle unfold.

I think there is a striking difference from amd ppt presentation of mantle with its performance goal and then the q&a with the devs where they talk about everything else. Lol.

All the above nonperformance perspectives is far more important than if its 5 or 50% added to an old dx. We get that each half year/year anyway. The other perspectives is very seldom seen and is imho from a gamers perspective the most interesting.

 

[frozentundra123456]

No its not. Look at bf4 ps4 xbox and you will know you dont get 100% on average. Mantle is not furmark. Gpu utilization is not so bad. And its not like modern dx is not trying to solve eg the drawcalls it just takes some work. Ofcource mantle can be 1000% faster in some situations but they are not typical and mantle will just expose other bottlenecks. The difference in mantle is those bottlenecks is visible to the programmer and can be ajusted to.

But Mantle is much more than performance!
Its cheap porting, more games to pc, better quality ports, cheaper programming on performance game meaning more highend games for us, the liberation of devs from boring work, new games we dont know about yet.
It will take 5 years untill we see the potential of mantle unfold.

I think there is a striking difference from amd ppt presentation of mantle with its performance goal and then the q&a with the devs where they talk about everything else. Lol.

All the above nonperformance perspectives is far more important than if its 5 or 50% added to an old dx. We get that each half year/year anyway. The other perspectives is very seldom seen and is imho from a gamers perspective the most interesting.

So you think seriously that Mantle specific games will come out, and if so that is a good thing? Otherwise, I dont see how it will lead to more games being ported to PC because they will have to be ported to DX as well, actually increasing the work that has to be done.

 

[smackababy]

So you think seriously that Mantle specific games will come out, and if so that is a good thing? Otherwise, I dont see how it will lead to more games being ported to PC because they will have to be ported to DX as well, actually increasing the work that has to be done.

This argument has been brought up many times. The Mantle-ites will continue to claim it will take less work to port games and not cost any extra money.

 

[Tempered81]

Are there any mantle benchmarks yet?

 

[SiliconWars]

So you think seriously that Mantle specific games will come out, and if so that is a good thing? Otherwise, I dont see how it will lead to more games being ported to PC because they will have to be ported to DX as well, actually increasing the work that has to be done.

Do you seriously think AMD won't pay whatever $$$'s is required to ensure it?

As for it being a good thing, yes...I believe it would be.

 

[Erenhardt]

Btw, that 20% extra performance claim from Katsman in Thief? Read again. http://techreport.com/review/25683/delving-deeper-into-amd-mantle-api/3

...and then the statement -

So Katsman is saying that 20% extra performance is possible through using compute for optimizations alone. Just through the ability to run compute synchronized with normal GPU functions an extra 20% performance is possible - this is not including any other Mantle optimization, just the asynchronous compute.

I said it way back when... No one listened:

Form what I understand we can expect heat problems on GCN cards with second class coolers while under 125% :ninja: Mantle load. That depends if they are going to release the beast within.

AMD is trying to hit two birds with one stone. With Mantle, they try to improve their graphics performance enabling parallel computing and at the same time increase CPU performance, by going wide on cores and off-loading work from single thread.

It's hard to not be excited with such a bold move.

 

[3DVagabond]

I said it way back when... No one listened:


AMD is trying to hit two birds with one stone. With Mantle, they try to improve their graphics performance enabling parallel computing and at the same time increase CPU performance, by going wide on cores and off-loading work from single thread.

It's hard to not be excited with such a bold move.

Pretty brilliant. I'm surprised nobody saw this coming during the "console wars". I wonder if AMD was planning this all along? Sure seems like it.

Mantle is going to improve the performance of their entire product stack. Where they are already competitive it's going to give them an advantage and where they struggle (discrete CPU) it's going to make them much more competitive. At least where gaming is concerned.

 

[bystander36]

Gsync is all about the smoothness isn't it? Games done under Mantle probably give a similar effect. Slide 4 - http://i.imgur.com/nt5217p.png

Gsync syncs your display with your GPU. Without Gsync, your GPU has to sync to the display. Unless you are capped at your refresh rate, Gsync will always look smoother at playable FPS compared to v-sync. Without v-sync Gsync will be just as fast, but without tearing.

G-sync will only be on 120+hz monitor at first, so hitting your refresh rate is unlikely in the majority of games and will look smoother than v-sync.

More FPS is great and all, but unless it gets you to 120 FPS 100% of the time, G-sync will be very important still. And there are plans for Overlord IPS screens to get it.

 

[SiliconWars]

Pretty brilliant. I'm surprised nobody saw this coming during the "console wars". I wonder if AMD was planning this all along? Sure seems like it.

Mantle is going to improve the performance of their entire product stack. Where they are already competitive it's going to give them an advantage and where they struggle (discrete CPU) it's going to make them much more competitive. At least where gaming is concerned.

Yes it's 100% planned. You can't pull off all this by mistake. While it started much smaller there's no doubt that it just snowballed into what it is now.

Only AMD could have envisaged this on this level, because only AMD has all the pieces of the puzzle. Nvidia doesn't have the APU behind them to make asymmetric crossfire worthwhile. Recently they've been shying away from compute on all but their highest end cards while AMD goes further into compute through their entire stack.

Intel doesn't have the discrete GPU and has deliberately been holding back on their low-end because they thought they could get away with it due to AMD being bottlenecked by a single weak cpu core, and games being single-threaded. There is no way they saw this coming, they just don't have the expertise. They don't understand graphics like Nvidia and AMD does.

AMD has all the pieces from top to bottom. They have to see it through because this was a long term plan. This is what repays the ~$5 billion ATI aquistion and then some. Notice I didn't even mention consoles.

 

[3DVagabond]

Yes it's 100% planned. You can't pull off all this by mistake. While it started much smaller there's no doubt that it just snowballed into what it is now.

Only AMD could have envisaged this on this level, because only AMD has all the pieces of the puzzle. Nvidia doesn't have the APU behind them to make asymmetric crossfire worthwhile. Recently they've been shying away from compute on all but their highest end cards while AMD goes further into compute through their entire stack.

Intel doesn't have the discrete GPU and has deliberately been holding back on their low-end because they thought they could get away with it due to AMD being bottlenecked by a single weak cpu core, and games being single-threaded. There is no way they saw this coming, they just don't have the expertise.

AMD has all the pieces from top to bottom. They have to see it through because this was a long term plan. This is what repays the ~$5 billion ATI aquistion and then some. Notice I didn't even mention consoles.

Yeah, but without the consoles, they never could have pulled this off.

 

[SiliconWars]

Yeah, but without the consoles, they never could have pulled this off.

I'm not even sure of that any longer. The more I hear devs talking about Mantle the less consoles seem relevant. It's just a much better way of doing stuff on PC's.

If Nvidia adopted Mantle tomorrow DirectX would be dead overnight, that's how good Mantle is.

 

[frozentundra123456]

Yes it's 100% planned. You can't pull off all this by mistake. While it started much smaller there's no doubt that it just snowballed into what it is now.

Only AMD could have envisaged this on this level, because only AMD has all the pieces of the puzzle. Nvidia doesn't have the APU behind them to make asymmetric crossfire worthwhile. Recently they've been shying away from compute on all but their highest end cards while AMD goes further into compute through their entire stack.

Intel doesn't have the discrete GPU and has deliberately been holding back on their low-end because they thought they could get away with it due to AMD being bottlenecked by a single weak cpu core, and games being single-threaded. There is no way they saw this coming, they just don't have the expertise. They don't understand graphics like Nvidia and AMD does.

AMD has all the pieces from top to bottom. They have to see it through because this was a long term plan. This is what repays the ~$5 billion ATI aquistion and then some. Notice I didn't even mention consoles.

Pretty optimistic projections for something that is not even usable in a game yet and has no performance figures. Despite your unbridled enthusiasm and some of the others on here, only time will tell, and not just BF4 benchmark, but at least a couple of years to see how good it performs in a wide variety of games and how widespread the adoption is. You can quote all the vague developer quotes you want, and infer whatever makes you happy, but the fact is nobody knows.

 

[SiliconWars]

Pretty optimistic projections for something that is not even usable in a game yet and has no performance figures. Despite your unbridled enthusiasm and some of the others on here, only time will tell, and not just BF4 benchmark, but at least a couple of years to see how good it performs in a wide variety of games and how widespread the adoption is. You can quote all the vague developer quotes you want, and infer whatever makes you happy, but the fact is nobody knows.

So you're not going to make up your mind until 2 years are past? :|

Interesting to see you brought my latest prediction true however. Even when BF4 shows fantastic gains you'll still be refusing to believe it until...I dunno what you have to see tbh. 2 years of sustained gains? 6 months of being disMantled is all it'll take and Nvidia will be on the way out anyway.

Also, I wasn't projecting anything, that previous post was 100% fact.

 

[bystander36]

So you're not going to make up your mind until 2 years are past? :|

Interesting to see you brought my latest prediction true however. Even when BF4 shows fantastic gains you'll still be refusing to believe it until...I dunno what you have to see tbh.

You have to at least have something to base your opinion on. Even if one or two games are great, that doesn't necessarily mean the API was great. It could have just been an outlier, but at least it will give some evidence of it being good or not.

BF4 could just show that the dev's spend an exceptional amount of time to fine tune it with their new API. It could also show the new API is awesome. It could be that the more it is used, the better results people will get, as they learn how to take advantage of it.

Time will tell.

 

[SiliconWars]

BF4 could just show that the dev's spend an exceptional amount of time to fine tune it with their new API. It could also show the new API is awesome. It could be that the more it is used, the better results people will get, as they learn how to take advantage of it.

Time will tell.

I'm sure that better devs will get more out of it yes. But from my pov that just means that the gains will increase as time goes by.

 

[SiliconWars]

You have to say it's a real eye-opener when a dev says that 20% more performance is "not unreasonable", using something that has basically always been there but hardly ever used.

We know Mantle will give huge gains in CPU bottlenecked scenarios, but to read about the compute gains, which are basically free...that's in GPU bottlenecked scenarios. I mentioned compute shader lighting before and we might see that in BF4, or some kind of post process AA. I think they talk about other gains in the deep dive from BSN? 5% from the new memory model was one thing mentioned I believe.

There is just so much here, so many new things that devs can take advantage of. It's unbelievable that it's taken so long to realise it.