The AMD Mantle Thread

[Spjut]

Yes, not every engine is going to benefit from multithreaded rendering. In fact, only in CPU limited situations does it seem to confer benefits, and most games tend to be more GPU than CPU limited. Civilization 5 had huge benefits with it, as it's a massive strategy game and the Lore engine was designed with DX11 multithreaded rendering in mind..

If the technology is further improved and refined, then you could achieve much higher draw call submissions compared with the present implementation.

Also, Project CARS uses driver command lists as well. BTW, where was it confirmed that Assassin's Creed IV uses it? I know it was suspected, but not necessarily confirmed..

One big thing preventing usage of DX11 multithreading is the fact that it's not supported by AMD.

But look at the benchmarks for Civ5 today and you'll see there's not a big difference between Nvidia and AMD, not even for crossfire and sli systems. Seeing how much the HD 7000 series improved the Civ5 performance for AMD (and this series has had lots of additional driver improvements since launch) it's very likely that Civ5 was highly dependant on compute performance. And the 400/500 series were also known to have better compute performance than the 5000 and 6000 series.

If it were so easily fixable, hadn't Microsoft fixed it already in DX11.1 and 11.2?

Now that you mention it, I haven't read specifically that AC4 uses it, but since it's using the same engine as AC3, it's likely

I think you're oversimplifying it alot saying it's not used alot because of AMD not supporting it.
Big players like Dice complained about multithreaded rendering not working already when they developed BF3, and Dice/EA is also one of the big players endorsing Mantle.

Creating Mantle most likely wasn't cheap for AMD, and incorporating it into the engines obviously isn't free for the studios either.
If the draw calls issue had been solved by having support for multithreaded rendering, don't you think AMD would have opted for that solution instead? Not to mention Dice, Oxide and the other studios, since they could have had everything under the same API and have it working on all compatible hardware

 

[VulgarDisplay]

Why are you guys so vehement that draw calls are not a problem? Will it cause you great personal harm if mantle allows more draw calls than dx? It shouldn't since you obviously think multithreaded rendering is the fix for the problem.

In fact I do not at all understand why you are all against mantle? Even if you are jealous that your hardware vendor of choice currently does not support it there is a good chance it will cause Microsoft to break DX stagnation which benefits you directly.

 

[blightymate]

There is only so much you can fit on a screen though without absolute clutter. Tell me, what use would it be to have one million disparate objects on screen at the same time?

Wearing my 8K 3D VR headgear one million disparate objects on screen at the same time would be glorious.

My judgment isn't more valid than theirs, but I do raise valid points.

Backward compatibility and cross architectural compatibility are MAJOR concerns; particularly the latter.

Looking at the Steam hardware survey, GCN users only account for roughly 5% of users. That's a lot of work for the industry to go through for such a small percentage of users.

If AMD is serious about Mantle, then make it an open standard and bring in Intel, NVidia etcetera.. Without their support, it WILL fail.

All the developers are aware both Nvidia and Microsoft turned down Mr. Andersson's request to help create such an API. It was AMD that stepped up to the plate and created an API that will vastly expand the future potentials of PC gaming and the future profit potentials of developers and publishers alike.

There are very solid rea$on$ EA is fully committing to Mantle. Those reasons will apply to the other AAA developers and publishers.

Backward compatibility and cross architectural compatibility are obviously not a sufficiently MAJOR concern to deter EA and a growing number of AAA developers from implementing Mantle.

Intel has no reason not to get on board with Mantle when it is made available, they have no stable of proprietary graphics technologies or JHH class ego to protect. I just don't see the developers and publishers allowing JHH's petulance, should it arise, to spoil their golden chance for a streamlined, high performance write once, easily port everywhere future and the vastly increased profits that will bring.

Nvidia either gets with the program or it will be side lined by the developers and publishers and then the rush to buy AMD graphics will be that more intense and Nvidia's AIB demise that much quicker.

 

[DamnedLife]

You are the one making assertions without any evidence, but I'll indulge you.

http://www.youtube.com/watch?v=fFkLvopNKRk#t=238

EVE Online runs well at max with ~100 ships and however many drones flying about, and that's on DX9. They just recently added DX11 support, so draw calls are even less of a bottleneck now.

I watched that video and all ı want to say is Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!

repating 12436478426 times really dents someone's brains in man

 

[DamnedLife]

Quoted from a different forum post which I totally agree:
Guess we'll never know. Shame too as Mr. Andersson said he approached Nvidia with the same proposition that AMD ended up accepting and apparently JHH wasn't interested for some reason, possibly to do with already having PhysX, CUDA and NAVI in play (and maybe an ARM assisted Maxwell GPU in mind). Extremely short sighted as it turns out which, along with his decision to not compete for the consoles, put AMD in charge and all but paved a path of success for Mantle. With the hardware tie-in EA might have insisted on Nvidia collaborating with an also ready to play AMD even If Nvidia had initially signed on to such a project, but at least then Nvidia would have been at near par when such an API was released instead of now helplessly waiting for the performance delta axe to fall and able to join the Mantle party solely at AMD's discretion. You may not care if there's a 60-80% performance boost in BF4, I doubt that can be said for JHH and the employees of Nvidia. The bigger the performance delta, the less incentive AMD will have to open standard Mantle.

It would have been far better for gamers and the industry as a whole if Nvidia had stepped up to the plate and at least joined in if not taken the lead on such a project, that's for sure. It would have ensured a Mantle style API gaining instant wide acceptance across developers and publishers and ALL gamers with modern hardware able to immediately enjoy the performance gains of such an API on next gen games. That this isn't the case is TOTALLY on JHH's obdurate short sighted head.

 

[blightymate]

Quoted from a different forum post which I totally agree:
Guess we'll never know. Shame too as Mr. Andersson said he approached Nvidia with the same proposition that AMD ended up accepting and apparently JHH wasn't interested for some reason, possibly to do with already having PhysX, CUDA and NAVI in play (and maybe an ARM assisted Maxwell GPU in mind). Extremely short sighted as it turns out which, along with his decision to not compete for the consoles, put AMD in charge and all but paved a path of success for Mantle. With the hardware tie-in EA might have insisted on Nvidia collaborating with an also ready to play AMD even If Nvidia had initially signed on to such a project, but at least then Nvidia would have been at near par when such an API was released instead of now helplessly waiting for the performance delta axe to fall and able to join the Mantle party solely at AMD's discretion. You may not care if there's a 60-80% performance boost in BF4, I doubt that can be said for JHH and the employees of Nvidia. The bigger the performance delta, the less incentive AMD will have to open standard Mantle.

It would have been far better for gamers and the industry as a whole if Nvidia had stepped up to the plate and at least joined in if not taken the lead on such a project, that's for sure. It would have ensured a Mantle style API gaining instant wide acceptance across developers and publishers and ALL gamers with modern hardware able to immediately enjoy the performance gains of such an API on next gen games. That this isn't the case is TOTALLY on JHH's obdurate short sighted head.

Brilliant! Couldn't have said it better myself.

 

[JeansenVaars]

Sooooooooo....

I have been promised to have a battlefield mantle patch for free by now,

Only thing I have for free so far is left for dead 2...

:$

 

[jj109]

I watched that video and all ı want to say is Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!Mantle rocks!

repating 12436478426 times really dents someone's brains in man

The 'FC' in EVE is really just a glorified WoW raid leader. Some like to repeat the same thing over and over again.

 

[144HzGamer]

There is only so much you can fit on a screen though without absolute clutter. Tell me, what use would it be to have one million disparate objects on screen at the same time?

My judgment isn't more valid than theirs, but I do raise valid points.

Backward compatibility and cross architectural compatibility are MAJOR concerns; particularly the latter.

Looking at the Steam hardware survey, GCN users only account for roughly 5% of users. That's a lot of work for the industry to go through for such a small percentage of users.

If AMD is serious about Mantle, then make it an open standard and bring in Intel, NVidia etcetera.. Without their support, it WILL fail.

They are sure on the right path IMO.

Both AMD and DICE have confirmed that they are indeed adding the necessary abstraction for compatibility with other vendors, but it will definitely come down to politics at the end .

Btw on the whole OpenGL vs Mantle discussion... (not related to the post quoted).

We have to make one thing crystal clear here, DICE has always been on the cutting edge when it comes to API adoption/usage, they were one of the firsts to release a Direct3D 11 game, the absolute first to release a Direct3D 11.1 game and the firsts to add support for Mantle (it everything goes according to plan).

So don't think for one second that if OpenGL really solved all the issues they are having with Direct3D (and if it's indeed better than Mantle), that they wouldn't have used OpenGL for Battlefield 4/ Future Frostbite games, they are definitely not afraid of pushing the boundaries when it comes to API usage.

After all, this was something that was pitched and pushed by Developers themselves, so if OpenGL was indeed the answer for them, they wouldn't have gone through with the trouble of creating a completely brand new API.

I really liked what Johan and the Oxide guys said during a Q&A session with the media during the APU13 conference, it went something along these lines:

"Scratch everything, think of what a modern API should look like, and there you have Mantle".

 

[desprado]

There is only so much you can fit on a screen though without absolute clutter. Tell me, what use would it be to have one million disparate objects on screen at the same time?

My judgment isn't more valid than theirs, but I do raise valid points.

Backward compatibility and cross architectural compatibility are MAJOR concerns; particularly the latter.

Looking at the Steam hardware survey, GCN users only account for roughly 5% of users. That's a lot of work for the industry to go through for such a small percentage of users.

If AMD is serious about Mantle, then make it an open standard and bring in Intel, NVidia etcetera.. Without their support, it WILL fail.

Agree.I dont get AMD statement that dev asked for API but just for 5% user so it can be lie or Bs statement.Fact is Intel and Nvidia both have majority example Intel in CPU and Nvidia in GPU.

 

[Carfax83]

But look at the benchmarks for Civ5 today and you'll see there's not a big difference between Nvidia and AMD, not even for crossfire and sli systems. Seeing how much the HD 7000 series improved the Civ5 performance for AMD (and this series has had lots of additional driver improvements since launch) it's very likely that Civ5 was highly dependant on compute performance. And the 400/500 series were also known to have better compute performance than the 5000 and 6000 series.

I don't think the gap has necessarily narrowed as you much as you believe. Look at these benchmarks from the GTX 770 review, the last benchmarks of Civilization 5 on Anandtech:

In these benches, the GTX 780 is leading the HD 7970 GHz by almost 50%. Yet, the GTX 780 is on average only 16 to 20% faster than the HD 7970 GHz..

The GTX 770 on the other hand has a much slimmer lead, but that's because it's more GPU limited in this test. Remember, DX11 multithreading only helps in CPU bound situations, which explains why the GTX 780 has such a humungous lead, as it's the most CPU limited card due to it's greater raw speed.

If it were so easily fixable, hadn't Microsoft fixed it already in DX11.1 and 11.2?

I don't think it was a priority for Microsoft to truly fix it. But with Mantle bringing attention to the DirectX inefficiencies, I'm sure Microsoft has renewed focus on reducing the API overhead.

More likely, they are going to roll out the fixes with the next version of Direct3D, which may be based on the Xbox One's API..

Now that you mention it, I haven't read specifically that AC4 uses it, but since it's using the same engine as AC3, it's likely

It could be that they are using it, but we'll likely never know unless they tell us. The game engine is limited to 62 FPS internally, so we won't see the effects of DX11 multithreading in benchmarks.

Big players like Dice complained about multithreaded rendering not working already when they developed BF3, and Dice/EA is also one of the big players endorsing Mantle.

When DICE was creating BF3, NVidia hadn't released multithreaded drivers. So that probably had a lot to do with why they couldn't implement it properly. Also, it was quite a difficult thing for NVidia to do. Since then, the drivers have improved significantly in their multithreaded capabilities to be sure.

As for BF4, I think DICE could have done it if they wanted to, but since it was a G.E game, they didn't.

If the draw calls issue had been solved by having support for multithreaded rendering, don't you think AMD would have opted for that solution instead?

Not necessarily. After all this time, AMD still hasn't added support for multithreaded rendering in their drivers. Like I said before, it's a difficult undertaking to do so, and to be frank, I think their driver team lacks either the resources or competence to do so.

This recent PClab.pl review has shown that NVidia's drivers are FAR more optimized for multithreading than AMD's are. It's something that has helped NVidia big time as their hardware has a significant advantage on game engines that are multithreaded; which the next generation of game engines will all be.

Even BF4 now performs better on NVidia after patches and driver updates, especially in multiplayer mode.. AMD NEEDS Mantle to compete with NVidia on the next generation game engines, which will all support multicore processors.

 

[Carfax83]

Why are you guys so vehement that draw calls are not a problem? Will it cause you great personal harm if mantle allows more draw calls than dx? It shouldn't since you obviously think multithreaded rendering is the fix for the problem.

Draw calls are a problem, but I don't believe it's as big a problem as what we've been led to believe; certainly not enough to warrant an architecture exclusive API. Most games after all, are GPU limited rather than CPU limited.

In fact I do not at all understand why you are all against mantle? Even if you are jealous that your hardware vendor of choice currently does not support it there is a good chance it will cause Microsoft to break DX stagnation which benefits you directly.

I said several pages ago that if Mantle accomplishes anything, it will have been to give Microsoft a wake up call..

 

[Carfax83]

There are very solid rea$on$ EA is fully committing to Mantle. Those reasons will apply to the other AAA developers and publishers.

Well if rumors are to be believed, AMD paid a nice sum of money to EA to implement Mantle in their Frostbite 3 engine.

Backward compatibility and cross architectural compatibility are obviously not a sufficiently MAJOR concern to deter EA and a growing number of AAA developers from implementing Mantle.

You make it sound as though EA is doing this out of genuine concern for PC gaming and the health of the market..

EA would NEVER have done this without the cash incentive that AMD gave them.. EA loses nothing by implementing Mantle, as it's already been paid for.

But AMD cannot continue to hand out cash incentives to developers for implementing Mantle. Unless there is wider IHV adoption, there is no incentive for developers to implement Mantle..

None!

Intel has no reason not to get on board with Mantle when it is made available, they have no stable of proprietary graphics technologies or JHH class ego to protect.

Actually you're wrong. Over the years, Intel's superior IPC has given them a massive advantage over AMD in gaming. With Mantle, IPC may take a backseat to core count in importance, giving AMD a distinct advantage.

If you're Intel, why would you want to help your competitor equalize the battlefield when you're already kicking the crap out of them in IPC?

I just don't see the developers and publishers allowing JHH's petulance, should it arise, to spoil their golden chance for a streamlined, high performance write once, easily port everywhere future and the vastly increased profits that will bring.

Portability only increases if other IHVs adopt Mantle, and that remains to be seen.

Nvidia either gets with the program or it will be side lined by the developers and publishers and then the rush to buy AMD graphics will be that more intense and Nvidia's AIB demise that much quicker.

Or just as likely, Intel and NVidia refuse to support Mantle, and instead Microsoft delivers a lighter, more console like Direct3D API that has lower runtime overhead and can achieve a higher draw call count than current DX11 (thought not as high as Mantle) whilst not sacrificing important aspects like cross architectural and backward compatibility..

Which would effectively make AMD's investment in Mantle a waste..

 

[Carfax83]

Both AMD and DICE have confirmed that they are indeed adding the necessary abstraction for compatibility with other vendors, but it will definitely come down to politics at the end .

When was this? I've heard AMD and DICE say that Mantle can be expanded to support other IHVs, but I don't recall them saying they were actually doing it yet..

So don't think for one second that if OpenGL really solved all the issues they are having with Direct3D (and if it's indeed better than Mantle), that they wouldn't have used OpenGL for Battlefield 4/ Future Frostbite games, they are definitely not afraid of pushing the boundaries when it comes to API usage.

What issues are they having with Direct3D? BF4 performs brilliantly on any mid to high machine, achieving close to, or over 100 FPS, even at ultra settings..

The DX11.1 renderer in BF4 isn't holding back performance imo. How could it when we are achieving such high frame rates?

 

[Genx87]

Quoted from a different forum post which I totally agree:
Guess we'll never know. Shame too as Mr. Andersson said he approached Nvidia with the same proposition that AMD ended up accepting and apparently JHH wasn't interested for some reason, possibly to do with already having PhysX, CUDA and NAVI in play (and maybe an ARM assisted Maxwell GPU in mind). Extremely short sighted as it turns out which, along with his decision to not compete for the consoles, put AMD in charge and all but paved a path of success for Mantle. With the hardware tie-in EA might have insisted on Nvidia collaborating with an also ready to play AMD even If Nvidia had initially signed on to such a project, but at least then Nvidia would have been at near par when such an API was released instead of now helplessly waiting for the performance delta axe to fall and able to join the Mantle party solely at AMD's discretion. You may not care if there's a 60-80% performance boost in BF4, I doubt that can be said for JHH and the employees of Nvidia. The bigger the performance delta, the less incentive AMD will have to open standard Mantle.

It would have been far better for gamers and the industry as a whole if Nvidia had stepped up to the plate and at least joined in if not taken the lead on such a project, that's for sure. It would have ensured a Mantle style API gaining instant wide acceptance across developers and publishers and ALL gamers with modern hardware able to immediately enjoy the performance gains of such an API on next gen games. That this isn't the case is TOTALLY on JHH's obdurate short sighted head.

There are already two API's that do this. Why spend more resources reinventing the wheel? This has potential to increase performance. But it also has potential to become a compatibility headache. The closer you get to the metal. The more complicated it gets. GCN is a hardware framework that will change with each iteration. A code path optimized for todays arch may not perform so well with the next iteration if it gets close enough.

DX, OpenGL sacrifice this performance for compatibility. Its code will work on any hardware.

That said Mantle is not being used in consoles. Consoles will have their own API that surely is closer to the metal. And that is fine because it is a closed system with defined hardware.

 

[144HzGamer]

When was this? I've heard AMD and DICE say that Mantle can be expanded to support other IHVs, but I don't recall them saying they were actually doing it yet..

#1: http://www.youtube.com/watch?v=tDPgJB2x7dQ&feature=*********&t=2m9s

#2:



One thing I would like to add is:

We have to remember that this is something that's being driven by developers, and they're not gonna be stupid about this, they would not be going all in with this if that didn't meant adding NVIDIA & Intel into the equation once everything is said and done, the less APIs the have to worry about, the better for them. The sad thing is that it will come down to Politics.

What issues are they having with Direct3D? BF4 performs brilliantly on any mid to high machine, achieving close to, or over 100 FPS, even at ultra settings..

The DX11.1 renderer in BF4 isn't holding back performance imo. How could it when we are achieving such high frame rates?

EDIT: Btw you're looking it all wrong, what does frame rate have to do with an API having limitations/is obsolete?

 

[3DVagabond]

Having played that game for 3 years, I can say that EVE's awful client performance has nothing to do with draw calls. Sometimes the CPU and GPU is going to be almost idle and the client is pushing only 13 fps with a 4770K. In huge blob fests the major limiting factor seemed to be the overview, which is the spreadsheet that lists everything on the screen. I could turn LoD to minimum and zoom all the way out, but still stuck at 10 fps with absolutely nothing rendering.

I was more impressed with how well Supreme Commander handled massive numbers of objects on screen for a DX9 game. That game came a few years too early though. Most people couldn't play the game as it was meant to be played without the simulation rate dropping to 1/5 speed.

You do realise that the problem with DX is you will be CPU limited because it is only able to use a small amount of the available CPU power.

 

[desprado]

#1: http://www.youtube.com/watch?v=tDPgJB2x7dQ&feature=*********&t=2m9s

#2:



One thing I would like to add is:

We have to remember that this is something that's being driven by developers, and they're not gonna be stupid about this, they would not be going all in with this if that didn't meant adding NVIDIA & Intel into the equation once everything is said and done, the less APIs the have to worry about, the better for them. The sad thing is that it will come down to Politics.





EDIT: Btw you're looking it all wrong, what does frame rate have to do with an API having limitations/is obsolete?

Which Dev are u talking about?Where are those Dev which asked for called low level API?These all Dev are AMD evolved partners and other than that there are no dev that announced that they will be start adopting mantle.

These all are PR and Marketing Slides which are made AMD and its partners.No actual result or demo but only PR talking which is AMD style.

 

[VulgarDisplay]

Which Dev are u talking about?Where are those Dev which asked for called low level API?These all Dev are AMD evolved partners and other than that there are no dev that announced that they will be start adopting mantle.

These all are PR and Marketing Slides which are made AMD and its partners.No actual result or demo but only PR talking which is AMD style.

You realize that repi appeared at an nvidia event like a week after the mantle announcement right? If he was as far up amd's....uh...payroll as you seem to think there is no way he would have done that.

 

[Stuka87]

The DX11.1 renderer in BF4 isn't holding back performance imo. How could it when we are achieving such high frame rates?

You can only get those very high frame rates when you have a heavily overclocked recent Intel CPU. Even 4.5GHz haswells run at 90-95% CPU. For the majority of gamers that do not have this setup, lower CPU usage is a big deal.

 

[VulgarDisplay]

The DX11.1 renderer in BF4 isn't holding back performance imo. How could it when we are achieving such high frame rates?

Why are there such horrible lod and draw distance issues in the game then? Dice have truly optimized the crap out of the game to get it to run and look as good as it does, but its anything but perfect.

What if mantle let's them completely draw everything in the world with still great performance? No more pop in, no more lod switching. That would be a killing blow to nvidia if its the case. I don't know if that is the type of thing mantle can make possible but if it is no one who is serious about gaming will use nvidia hardware anymore.

 

[desprado]

You realize that repi appeared at an nvidia event like a week after the mantle announcement right? If he was as far up amd's....uh...payroll as you seem to think there is no way he would have done that.

U realize AMD has a deal with DICE.So ofcourse they going many sweat thing they can

 

[jj109]

You do realise that the problem with DX is you will be CPU limited because it is only able to use a small amount of the available CPU power.

Obviously nothing else I posted on the last page has caught on.

 

[Carfax83]

#1: http://www.youtube.com/watch?v=tDPgJB2x7dQ&feature=*********&t=2m9s

#2:

Marketing slides? I'll believe it when I see it..

EDIT: Btw you're looking it all wrong, what does frame rate have to do with an API having limitations/is obsolete?

While there's no doubt that the DirectX API has limitations that should be addressed, that "marketing slide" makes it appear much worse than it really is.

Look at BF4. On PC, you get far superior frame rate, graphics and overall experience compared to the PS4 and Xbox One. The PS4 and Xbox one cannot even do 1080p with scaled back visual fidelity, and they still regularly fall far below 60 FPS in large multiplayer battles despite having higher draw call capacity and supposedly greater efficiency.

 

[moonbogg]

So is this now a 2014 thing?