The AMD Mantle Thread

[cytg111]

I realize multithreading isn't trivial, but it IS something that can be done with high level API and give immediate results. And it is something that would be better than having Mantle and DX.

There is one of those ultimate truths again. I wouldnt even know where to start going down the reasoning chain to arrive at that. Is any mantle like approach wrong? Avoid fragmentation at ALL cost? I dont know? It seems to me, to arrive at that, that any benefit an mantle-api can provide can never outweigh the damage its going to do? I suggest to begin putting some weight on those scales instead of just assuming either way.

 

[bystander36]

This is probably going to be a totally ignorant question, but from what I was reading here. DX is working from the upper end and Mantle from the lower end, could their effects be combined for a superior out come by working together?

It's not like that at all. High level just means each line of code does more things, and they are generally designed to work with a wide range of hardware, so there a lot of overhead. Low level code is code that does much less per line, but gives you more control over the details, allowing you to avoid a lot of overhead, but it requires the programmer to account for lots of different hardware, or exclude lots of different hardware.

 

[bystander36]

There is one of those ultimate truths again. I wouldnt even know where to start going down the reasoning chain to arrive at that. Is any mantle like approach wrong? Avoid fragmentation at ALL cost? I dont know? It seems to me, to arrive at that, that any benefit an mantle-api can provide can never outweigh the damage its going to do? I suggest to begin putting some weight on those scales instead of just assuming either way.

I never said it was better at all costs. Where did you read that? Are you just reading what you want?

I never said that there are no benefits to Mantle now. If you read both sentences, I clearly made the statement that if DX was improved and said in previous sentences that if it was made to be close enough to Mantle in performance, it would be better than having both Mantle and DX.

 

[frozentundra123456]

There is one of those ultimate truths again. I wouldnt even know where to start going down the reasoning chain to arrive at that. Is any mantle like approach wrong? Avoid fragmentation at ALL cost? I dont know? It seems to me, to arrive at that, that any benefit an mantle-api can provide can never outweigh the damage its going to do? I suggest to begin putting some weight on those scales instead of just assuming either way.

Everything is a trade off. Despite all the criticism of DX, justified or not, it has allowed PC gaming to be what it is today. It allows basically any game of the last what, 8 to 10 years, to be played on any hardware, whether it be indie games developed on a budget, or AAA big budget titles. It has a proven track record. It may be inefficient, but I cant think of any recent game that I was playing where I hated the game because of the graphics.

Mantle, OTOH, is just the opposite. It is more efficient, but can play no past games, only one current game, granted, perhaps a few more by the end of the year, but certainly not a majority of games coming out in 2014. In the few games that support it, in cpu limited scenarios, and with the right hardware, it gives a nice boost. I just dont see how it can overcome the lack of wide compatibility to become more than a nice additional feature on AMD cards, much like Physix for nVidia, but more useful.

 

[cytg111]

I never said it was better at all costs. Where did you read that? Are you just reading what you want?

I never said that there are no benefits to Mantle now. If you read both sentences, I clearly made the statement that if DX was improved and said in previous sentences that if it was made to be close enough to Mantle in performance, it would be better than having both Mantle and DX.

Nope, just this.

"I realize multithreading isn't trivial, but it IS something that can be done with high level API and give immediate results. And it is something that would be better than having Mantle and DX."

If thats not what you meant then my reply has little relevance, please do ignore it.

 

[cytg111]

Everything is a trade off. Despite all the criticism of DX, justified or not, it has allowed PC gaming to be what it is today. It allows basically any game of the last what, 8 to 10 years, to be played on any hardware, whether it be indie games developed on a budget, or AAA big budget titles. It has a proven track record. It may be inefficient, but I cant think of any recent game that I was playing where I hated the game because of the graphics.

Mantle, OTOH, is just the opposite. It is more efficient, but can play no past games, only one current game, granted, perhaps a few more by the end of the year, but certainly not a majority of games coming out in 2014. In the few games that support it, in cpu limited scenarios, and with the right hardware, it gives a nice boost. I just dont see how it can overcome the lack of wide compatibility to become more than a nice additional feature on AMD cards, much like Physix for nVidia, but more useful.

Yes and I think 10 years is a pretty good run for such an API but the mere existence of mantle tells its own story IMO.
When is the last time that an API had masswide adapoptation with the release of beta1? It is just going to have its run and I dont see anyone with the crystal balls(!) to see where its actually going to land. DirectMantle? We could hope! - (Or should we? Keep PC gaming locked in on the windows platform?) The factors that will drive mantle(or the concept of) forward is too volatile for my taste to make any kind of prediction.
But adoptation takes time for any new API.. and that gives the market a chance to counter. Interresting times ahead.

 

[Despoiler]

1st off, I didn't say what MS is doing. I said what they can do. I said IF they do this or that and you are the only one saying that they will or will not do.

I realize multithreading isn't trivial, but it IS something that can be done with high level API and give immediate results. And it is something that would be better than having Mantle and DX.

Your premise is the MS has more tricks up their sleeves for multi-threading performance that can somehow close the gap on Mantle. What I am saying you are completely ignoring or just don't understand the state of DX. What they CAN do they HAVE done. DX11 mutithreading is their solution and it doesn't even come close to Mantle's beta performance.

 

[3DVagabond]

All those measuring Mantle by it's early gains in BF4 only need to realize that BF4 was written to run with DX and all of it's limitations. So, the gains so far are modest in situations that don't really need Mantle. The 2 biggest gripes I took from the Devs about DX is that 1) They have to limit their artistic content to deal with the limitations of DX and 2) DX doesn't give them fine enough control. Let's assume they can double or even triple the draw calls with MT in DX. That allows for the removal of bottlenecks in current games. Once they up the draw calls to the max that they can reasonably run with DX MT we'll be back in the same boat we are now and Mantle still won't be bottlenecked. We're still waiting to see what gains can be had by reduced GPU overhead, as well.

There is no such thing as enough resources.

 

[bystander36]

Nope, just this.

"I realize multithreading isn't trivial, but it IS something that can be done with high level API and give immediate results. And it is something that would be better than having Mantle and DX."

If thats not what you meant then my reply has little relevance, please do ignore it.

How is this quote different than my reply to you?

Improving DX to a state that lifts the CPU bottleneck problems with multithreading should be enough to make it better than using two different API's for the same platform.

Yes, I think I'll ignore you.

 

[bystander36]

Your premise is the MS has more tricks up their sleeves for multi-threading performance that can somehow close the gap on Mantle. What I am saying you are completely ignoring or just don't understand the state of DX. What they CAN do they HAVE done. DX11 mutithreading is their solution and it doesn't even come close to Mantle's beta performance.

Why do you think multi-threading is so impossible with DX? They've added massive changes in several DX versions. Why do you believe nothing can be improved again?

 

[bystander36]

All those measuring Mantle by it's early gains in BF4 only need to realize that BF4 was written to run with DX and all of it's limitations. So, the gains so far are modest in situations that don't really need Mantle. The 2 biggest gripes I took from the Devs about DX is that 1) They have to limit their artistic content to deal with the limitations of DX and 2) DX doesn't give them fine enough control. Let's assume they can double or even triple the draw calls with MT in DX. That allows for the removal of bottlenecks in current games. Once they up the draw calls to the max that they can reasonably run with DX MT we'll be back in the same boat we are now and Mantle still won't be bottlenecked. We're still waiting to see what gains can be had by reduced GPU overhead, as well.

There is no such thing as enough resources.

This is the problem with 2 API's. Are they going to design 2 different games, one for DX and one for Mantle? I have my doubts that we'll ever see games take full advantage of Mantle as long as they also have to make the game run for DX. Mantle has great potential, but I find it hard to believe they'll design a game for Mantle only. There is going to be compromises so they can make the game work for both API's. If they try to do more for Mantle, it'll probably only be with additional cosmetic things, similar to GPU PhysX.

 

[Carfax83]

A lot of ifs in that statement.
Windows 9 is planned for mid-late 2015 depending on the rumours you believe. Even then there are no guarantees that DX12 will be part of that release.

Last rumor I heard was that Windows 9 was scheduled to be released this year..

I find it amazing that the vast majority of Nvidia fans are happy to predict Mantle's demise before it has been given a chance. Instead of waiting to see if it can bring benefits to PC gaming they are much quicker to declare it pointless. Even after many previews/reviews show that there are definite benefits for even enthusiast level PCs many Nvidia fans declare it "meh".

Mantle is bringing benefits to only a select few PC gamers. As long as Mantle cannot replace DirectX, it's doomed to failure.

Do you honestly think the PC gaming market wants to support two APIs indefinitely? What's the point?

Like bystander36 said, it makes much more sense to stick with the evolving DirectX.

It's in alpha at the moment and will continue to improve. Hopefully to the point where other devs and even Nvidia take notice.

Mantle has supposedly been in development for how many years now, and it's still in alpha stage? That doesn't bode well for AMD.

Something tells me they actually would prefer if the PC gaming industry stagnated rather than accept AMD tech in their Nvidia cards. I wonder how they feel about Eyefinity or GDDR3/4/5.

Sorry, but that's just an inane statement. In fact, that's probably the dumbest thing I've read in this entire thread.. :thumbsdown:

 

[bystander36]

Now, just to be clear, as all the attacks of on my posts seem to be clouding my words.

Mantle does great things for CPU bottlenecks, something we do face a lot today. I think this is awesome. Something I face a lot more than most, as I need high FPS to avoid nausea issues.

I do, however, think it would be even more awesome if that could be done with a single API for everyone, even if it isn't as quite efficient, because games can be polished around a single API, instead of 2+. All your dreams of awesome games with mega draw calls will not happen as long as they have to design them around the lowest common denominator. Fo the same reason why so many complain about how last gens consoles held back tech, DX will hold back Mantle.

 

[wand3r3r]

Last rumor I heard was that Windows 9 was scheduled to be released this year..

Mantle is bringing benefits to only a select few PC gamers. As long as Mantle cannot replace DirectX, it's doomed to failure.

Do you honestly think the PC gaming market wants to support two APIs indefinitely? What's the point?

Like bystander36 said, it makes much more sense to stick with the evolving DirectX.

Mantle has supposedly been in development for how many years now, and it's still in alpha stage? That doesn't bode well for AMD.

Sorry, but that's just an inane statement. In fact, that's probably the dumbest thing I've read in this entire thread.. :thumbsdown:

Replace mantle and API with physx, now where do you stand?

Why is it the opponents of Mantle are proponents of Physx? I thought exclusive "value" features are an icon of superiority? I should go review the physx threads just for humor.

At least AMD "might" let NV use the mantle API, I wouldn't see the reverse happening.

 

[BrightCandle]

If Microsoft made a significant change with DX12 that dramatically improved multithreading then the API would likely have to change somewhat to accommodate that. Even if all that happened was the contract of the methods changed so that now some of them could be run in parallel it would still have changed and it wouldn't just work without effort on the part of developers. Adding multithreading below the API itself might help but that is basically what Nvidia does already with its MT driver, MS is already trying to make the API as thin as possible. You might save some porting effort from a DX11 engine verses a whole new engine but its not zero. DX is backwards compatible by having all versions of DX available (well DX9 upwards anyway) its not actually backwards or forward compatible in the general sense.

I still haven't seen any documentation on Mantle, no dev kit or anything. I don't really consider it released until I can write some software against it and determine what is and isn't really GCN specific in their API. Kind of disappointed all we got was an alpha driver (its not functionally complete hence not a beta despite what they might call it) and no developer documentation at all. They are keeping it very close to them for something they want companies to adopt.

 

[Carfax83]

DX11 mutithreading can, but it shunts the responsibility to the driver, which is basically like building a mutithreaded OS at the GPU level. Not even remotely trivial.

It's definitely not trivial there's no doubt about it, but it's doable. It took NVidia about 2 years to do it, and it's still a work in progress and being constantly improved.

AMD is too incompetent to pull it off though apparently, because they still haven't enabled it in their drivers after almost 5 years.

All that aside, DX11 mutithreading implementation only doubles your object/draw-call throughput. Mantle is so far beyond double the draw calls.

Taking a page from Richard Huddy am I right?

Now the PC software architecture – DirectX – has been kind of bent into shape to try to accommodate more and more of the batch calls in a sneaky kind of way. There are the multi-threaded display lists, which come up in DirectX 11 – that helps, but unsurprisingly it only gives you a factor of two at the very best, from what we've seen. And we also support instancing, which means that if you're going to draw a crate, you can actually draw ten crates just as fast as far as DirectX is concerned.

But it's still very hard to throw tremendous variety into a PC game. If you want each of your draw calls to be a bit different, then you can't get over about 2-3,000 draw calls typically - and certainly a maximum amount of 5,000. Games developers definitely have a need for that. Console games often use 10-20,000 draw calls per frame, and that's an easier way to let the artist's vision shine through.'

Source

DX11 multithreading may "only" double draw call throughput based on AMD's research, but based on the Firaxis Lore presentation by Dan Baker, Civilization 5 managed to achieve 15,000 plus batches at 60 FPS and were GPU limited, so they could have gotten more.

Combined with instancing, draw calls should never be an issue unless you're purposely trying to make it one *cough* Star Swarm benchmark! :sneaky:

Also that was with NVidia's first DCL driver years ago, so I'm sure it's much better now. The point is, that DX11 multithreading does work and it CAN more than double draw call output.

The technology is not as effective as Mantle perhaps, but considering it's not tied to GCN and is part of the DirectX 11 spec, that's not really a drawback.

Project C.A.R.S uses it, and by all appearances, it's allowed a GTX 660 to achieve a 55% advantage in frame rate over a HD 7850 on a low end Core i5 760 CPU:

GTX 660 vs HD 7850 in Project C.A.R.S

 

[Pottuvoi]

It certainly would be great if we would see DX12 within a year.
Especially if it would have backward compatibility path for DX11 class cards. (Shader Model 5)

 

[Carfax83]

Replace mantle and API with physx, now where do you stand?

Why is it the opponents of Mantle are proponents of Physx? I thought exclusive "value" features are an icon of superiority? I should go review the physx threads just for humor.

At least AMD "might" let NV use the mantle API, I wouldn't see the reverse happening.

Mantle has absolutely no correlation with PhysX. PhysX runs on practically all platforms, from x86 to A.R.M to CUDA to DirectCompute. The only reason it hasn't run on AMD GPUs is due to artificial limitations driven by corporate politics.

Mantle on the other hand is a low level rendering API specifically designed to exploit and tap into GCN architecture.

Could it work on NVidia or Intel GPUs? Yes I'm sure, but not without some serious changes.

 

[ICDP]

Last rumor I heard was that Windows 9 was scheduled to be released this year.

Nope

Mantle is bringing benefits to only a select few PC gamers. As long as Mantle cannot replace DirectX, it's doomed to failure.

The point I was making is that right right now it is only an alpha that works for limited hardware, that could and will change. Here let me remind you.

"Instead of waiting to see if it can bring benefits to PC gaming they are much quicker to declare it pointless."

Do you honestly think the PC gaming market wants to support two APIs indefinitely? What's the point?

It's not what I think, some developers obviously want Mantle or none of them would adopt it. Stop thinking as an enthusiast gamer with an expensive SLI rig and think like a developer. Mantle will lower the minimum requirements dramatically for PC games. This opens up the possibility to increase sales by a significant amount for the devs. They could massively increase income for a few months work.

Like bystander36 said, it makes much more sense to stick with the evolving DirectX.

If DirectX was evolving the way developers liked they wouldn't have approached AMD to write a new API. DirectX bring significant bottlenecks for lower end hardware, mantle doesn't. Obviously it benefits PC game developers or mantle wouldn't exist in the first place. Of course AMD also have a stake in this but if there was no possibility of Mantle being adopted, AMD would not have wasted all the time/money/resources developing it.

Mantle has supposedly been in development for how many years now, and it's still in alpha stage? That doesn't bode well for AMD.

How long was it between major DX versions, DX9 to DX10, or DX10 to DX11? These are just an update to an existing API, not written from the ground up. Yet there are years between each major release.

Here let me save you a Google search.
DX8 Nov 2000
DX9 Dec 2002
DX10 Nov 2006
DX11 Oct 2009

Two - four years between major releases! Not a wonder some developers approached AMD and are enthusiastic about Mantle. In the time it took MS to release the next major variant of DX AMD managed to write a whole new API from the ground up. If this doesn't bode well for anyone it MS and DX.

Sorry, but that's just an inane statement. In fact, that's probably the dumbest thing I've read in this entire thread.. :thumbsdown:

The perception I get from you and many other Nvidia fans is that Mantle is pointless because DX is great. Rather than wait to see how it progresses over the coming months you have declared it pointless already. You even post the following drivel.

Originally Posted by Carfax83 View Post
If I were NVidia's CEO, I wouldn't support Mantle. I would play to my strengths, and try to expose AMD's weaknesses with Direct3D.

Yep, I stand by my statement that to many Nvidia Fans anything AMD = fail.

 

[ICDP]

It's definitely not trivial there's no doubt about it, but it's doable. It took NVidia about 2 years to do it, and it's still a work in progress and being constantly improved.

AMD is too incompetent to pull it off though apparently, because they still haven't enabled it in their drivers after almost 5 years.

LOL, you truly are the epitome of a blinkered Nvidia fan aren't you.

Nvidia = great for taking two years to add a feature to DX and AMD = fail for taking the same time to write a complete API from scratch. I've read it all now.

 

[3DVagabond]

This is the problem with 2 API's. Are they going to design 2 different games, one for DX and one for Mantle? I have my doubts that we'll ever see games take full advantage of Mantle as long as they also have to make the game run for DX. Mantle has great potential, but I find it hard to believe they'll design a game for Mantle only. There is going to be compromises so they can make the game work for both API's. If they try to do more for Mantle, it'll probably only be with additional cosmetic things, similar to GPU PhysX.

Never said anything about making Mantle only games. So, your reply doesn't really address my post at all???

I'll try and say it a bit differently. They will design games just like they do now. They will have to run on DX, just like games do now. They will push the games to the limit of DX w/ MT and they will become CPU limited. When this happens Mantle still won't be CPU limited. This very well could happen with the first console ports since they will have been designed with 8 cores and direct control over the thread scheduling, memory allocation, etc. which Mantle offers and DX doesn't. It might not happen right away, I understand (just saying it could), but it will happen, and when it does Mantle still won't be CPU limited running games that are cpu limited with DX.

Mantle has absolutely no correlation with PhysX. PhysX runs on practically all platforms, from x86 to A.R.M to CUDA to DirectCompute. The only reason it hasn't run on AMD GPUs is due to artificial limitations driven by corporate politics.

Mantle on the other hand is a low level rendering API specifically designed to exploit and tap into GCN architecture.

Could it work on NVidia or Intel GPUs? Yes I'm sure, but not without some serious changes.

You seem to misunderstand so much? Mantle is not tied to a specific platform or arch. It simply has a minimum feature set. Can you say what changes there would have to be to Mantle to have it run on nVidia and Intel hardware? Those who have actually worked with Mantle specifically say it's not GCN specific.

 

[3DVagabond]

Replace mantle and API with physx, now where do you stand?

Why is it the opponents of Mantle are proponents of Physx? I thought exclusive "value" features are an icon of superiority? I should go review the physx threads just for humor.

At least AMD "might" let NV use the mantle API, I wouldn't see the reverse happening.

Going by AMD's past record there's no reason at all to think they won't allow Mantle to be open. That's just the way they push to move the industry in the direction they want to go. Look at how AMD operates. They would love for everyone to pick up the Mantle ball and run with it. Freeing games from Windows and allowing any OS that wants to incorporate it run Mantle would be the best thing that could happen to gaming, IMO. Because it has to be the be all Windows is very bloated. Because it has to be so diversified hardware wise and requires backwards compatibility DX is bloated. Sometimes a Swiss army knife is the best tool for the job. Sometimes though, a stiletto is.

 

[Carfax83]

Nope

Windows 9 may release in October 2014

Apparently the hacker that claims this has been accurate in the past concerning Microsoft's plans.

The point I was making is that right right now it is only an alpha that works for limited hardware, that could and will change. Here let me remind you.

There is no indication that AMD has any plans to implement Mantle for other architectures. That's something that needs to be planned from inception, with the input of other IHVs. And even if they did, it doesn't appear that Intel and NVidia want in on it for aforementioned reasons..

But even if they did, would that not defeat the purpose of Mantle? How can Mantle claim to be low level, when it has to add more abstraction layers for other architectures?

It's not what I think, some developers obviously want Mantle or none of them would adopt it. Stop thinking as an enthusiast gamer with an expensive SLI rig and think like a developer. Mantle will lower the minimum requirements dramatically for PC games. This opens up the possibility to increase sales by a significant amount for the devs. They could massively increase income for a few months work.

I'm all for increasing accessibility to PC gaming for lower spec machines, but making a low level architecture specific API is hardly the way to go about doing it.

If DirectX was evolving the way developers liked they wouldn't have approached AMD to write a new API. DirectX bring significant bottlenecks for lower end hardware, mantle doesn't. Obviously it benefits PC game developers or mantle wouldn't exist in the first place. Of course AMD also have a stake in this but if there was no possibility of Mantle being adopted, AMD would not have wasted all the time/money/resources developing it.

DX11 multithreading is DirectX solution for CPU bottlenecks, and it works well. If AMD would get a clue and implement it in their drivers, then Mantle would never have been necessary.

Two - four years between major releases! Not a wonder some developers approached AMD and are enthusiastic about Mantle. In the time it took MS to release the next major variant of DX AMD managed to write a whole new API from the ground up. If this doesn't bode well for anyone it MS and DX.

You forget that DirectX has much greater functionality, cross architectural and backward capability which would necessarily be more complicated than Mantle which is as of now, tied to GCN architecture only.

So no, Mantle cannot be compared to DirectX.

The perception I get from you and many other Nvidia fans is that Mantle is pointless because DX is great. Rather than wait to see how it progresses over the coming months you have declared it pointless already. You even post the following drivel.

The perception I get from you and many other AMD fans is that DirectX is long in the tooth and no longer necessary and Mantle is the greatest thing since sliced bread and is the savior of PC gaming.

Yep, I stand by my statement that to many Nvidia Fans anything AMD = fail.

Why should NVidia help AMD? AMD is the one that is having problems with being CPU limited, not NVidia.

 

[Carfax83]

LOL, you truly are the epitome of a blinkered Nvidia fan aren't you.

Nvidia = great for taking two years to add a feature to DX and AMD = fail for taking the same time to write a complete API from scratch. I've read it all now.

NVidia took 2 years to add a feature that basically put the burden almost squarely on them for implementing multithreading, which likely involved massive changes to their driver code.. But after 2 years, it WORKED and gave massive performance increases.

AMD takes 3 years to develop Mantle, makes a big hullaballoo about it, delays it 3 times, and when it's available it's buggy and only works on certain GCN cards..

Job well done! :thumbsup:

 

[Carfax83]

You seem to misunderstand so much? Mantle is not tied to a specific platform or arch. It simply has a minimum feature set. Can you say what changes there would have to be to Mantle to have it run on nVidia and Intel hardware? Those who have actually worked with Mantle specifically say it's not GCN specific.

No unbiased or objective sources have verified this, as the spec hasn't even been published.

All we know is what AMD wants us to know at this point.