How AMD's Mantle Makes Games Better

I dont care what hardware it works on, or who is going to support it. All i care about now is whether or not that "demo" is an actual game in development that i will get to one day play! it looks amazingly fun.
 
Must have missed that. Have a link?

Sorry, cant find a direct one, I'll try later.
In the meantime, if you google the following text:
Mantle is not tied to AMD's GCN

There are many links to what was discussed at the AMD Developer Summit by Johan Andersson.
 
The did say they were running on a downclocked 8350 as well as the game being played by two AI players.
One of the major benefits of going Mantle is the lower CPU usage the GPU would have on it. Typical OpenGL/DX API's require a process that looks like this.

code->driver->hardware

The problem is the driver has to convert OpenGL/DX calls over to the native hardware. That conversion can be a real problem as the GPU can wait for the CPU and the CPU is loaded up with unnecessary work. That's why if the CPU isn't up to the task, the GPU can be limited in performance. With the way modern GPU's work, the demand from the CPU is going to be high.

Another example of this is Glide wrappers or how Wine gets Direct3D working in Linux. A wrapper adds another layer that requires extra CPU usage, and more delay for the GPU. That's why Windows games run so slowly on Linux, and why Wine developers are moving their D3D code to be multi-threaded. Nvidia and AMD both do heavy optimizations for this reason.

code->wrapper->driver->hardware

Mantle gives more control over to developers so the driver doesn't do as much work. Developers don't want the driver to do the work, and want that control. But the performance boost will probably effect faster graphic cards with slower CPUs. So the game would have to be very graphics intensive, and the CPU would have to be very slow. The 8350 is not a slow chip, so to showcase the benefits of Mantle they reduced the clock speed.

People with Radeon HD 7770's with a Intel i5 or AMD quad core chips probably won't benefit much if like Blizzard put Mantle in World of Warcraft. If you have a 7950 with a i3 or dual core AMD chip, you would benefit a lot in a game like Battlefield 4.
 
Did he say "AI"? thats what i want to hear. Sure eye candy is nice but when you start talking about 7,000 individual bots with "their own physics and AI" then i get excited.
Between Mantle and the Occulus Rift, PC gamers are getting some much needed love!
 
Sorry, cant find a direct one, I'll try later.
In the meantime, if you google the following text:
Mantle is not tied to AMD's GCN

There are many links to what was discussed at the AMD Developer Summit by Johan Andersson.

You mean Johan's wish list for what he would like to see from Mantle? Notice the "?" on the slide for multi-vendor? If it was a fact why the use of a "?" and not something more appropriate like an exclamation point?

A devs wish list does not indeed make it a fact AMD will allow other GPU devs access to Mantle. Could Mantle work on other GPU's, I don't see why not, but its AMD property designed for AMD cards.

Andersson stated from what he saw in the code he doesn't see why other GPU's couldn't run full or parts of Mantle. As of now and for the foreseeable future it is completely locked to GCN. AMD has yet to officially announce they are allowing other GPU devs to have Mantle. It has all been conjecture and maybes.

If you have an official statement from AMD saying NV will get access by all means let me know.

Even if AMD does allow Mantle access to NV or Intel it won't be on an open source basis. Maybe open API while still allowing AMD to have full control and ability to lock out features of the API to other GPU's. That way AMD gets to control how powerful Mantle will be on other hardware.

Just imagine if AMD allowed NV full access to Mantle and NV cards start destroying AMD cards because of Mantle. It would defeat the purpose of AMD spending millions to create and launch Mantle.

No one with any business sense can see a company shooting themselves in the foot on purpose just to be nice. AMD is not a "good guy" just like NV is not. They are in business to destroy each other. Mantle is one tool AMD has that gives them an edge. Much like Gameworks for NV. It sucks for us the consumer but business is business. Either company uses the tools they have to make the other look bad in order to sell more cards.

AMD needs Mantle to help AMD's bottom line and stock price.

Don't expect NV to get access to Mantle anytime soon if ever. It will remain proprietary to AMD. It gives AMD an edge they should gladly exploit to sell more GPU's.

If Mantle does what it is rumored to do it will be amazing and I think it would be great to have a single API for the PC but I am also a realist. Business is business.

But again if AMD has made an official statement saying it is open source or will be given to NV please link it.
 
Last edited:
You mean Johan's wish list for what he would like to see from Mantle? Notice the "?" on the slide for multi-vendor? If it was a fact why the use of a "?" and not something more appropriate like an exclamation point?
...

No, there was a slide produced during a presentation.
I finally found a copy of it, but I'm burned out researching this today to find the original.
Untitled11.jpg


The text is this:
Multi Vendor?

Mantle is designed to be a thin hardware abstraction
- not tied to AMDs GCN
- forward compatible
- extensions for architecture- and platform-specific functionality

Mantle would be a much more efficient graphics API for other vendors as well
- most Mantle functionality can be supported on todays modern GPUs


edit, I meant to say no thats not how it is written.
Note that he states what Mantle is.
The question being asked on the slide is whether it will be used by multiple vendors, not whether it is capable.
 
Last edited:
For developers, that's the whole point of Mantle. You don't want to alienate your customers because you can't get your game to run acceptably on their hardware. Mantle caters specifically to relatively new hardware, like the 7000 and 200 series cards. Even though the architecture of the 5000/6000 are very similar to the GCN hardware, it isn't supported. Not because it isn't best suited, but marketing.
No it isnt.
The point of Mantle is give the developers the tools they need to program graphics hardware more directly.
Developers asked for this.


Uh, that's the reason why Mantle exists.
Erm, thats why I said its why Mantle exists.

The drivers handles everything going on, which is why Mantle gives all that control over to developers.
Is this refuting anything I said?
It backs up my first point above.

OpenGL was actually made by SGI for their graphic cards, and then they later opened up the API. Glide, which is exactly just like Mantle, was created by 3DFX and became open source.
What does any of this prove?
If OpenGL could do what Mantle is doing then why havent they?

I don't like NVAPI or Mantle. Though to be fair, NVAPI is 5 years old and does support older hardware. You could use OpenGL and throw in some custom extensions. Both Nvidia and ATI have been doing that for years, before DirectX because the thing of the future.

For example, Nvidia has created a number of custom OpenGL extensions that ATI has to use just to be compatible, and vice versa. But Nvidia has created so many more and has gotten developers to use them more often, that it's no wonder why Nvidia is usually higher in performance then ATI was.

It's not like Mantle where you created an entire API, but it does give a decent performance boost without having developers to create whole new code.
I'm not a fan of an increase in the number of APIs if they are not universal.
But I am a fan of increasing functionality from hardware and advancing technology.
Mantle is needed to kick start much needed improvements, I dont see anyone else making such a change.
 
GCN versions 1 and 2 are required to run Mantle.
http://en.wikipedia.org/wiki/Mantle_(API)

This why Nvidia will not support it.

They will have to create a arch that will play nice with it. They're in a better position to work on their own version and roll it out faster.
Remember, single card multi-monitor support wasn't available until Kepler, Fermi couldn't support it. Also, they called it Surround instead of Eyefinity.
I doubt Nv will call their version Mantle, they'll make up a different name.
Mantle isn't a driver, the arch is what makes Mantle possible. Try to run Mantle on a 5870 and see where that get you.

:confused:

Worthless Source
>Wrong
>> More Wrong


I think many people are extremely inexperienced and looking at the code like it's a fixed product like a lemonade stand, or a refrigerator or something. It doesn't matter if GCN is used now, anything somewhat in the modern age can be suppored if they want to do it, but there's just a minimal support limitation, that's why no compatability with older hardware.

People just expect this thing called compatability like something usual, you want compatability, you'll always suffer in performance, that's mathematical certainty, that's how the physic logics work, that's how this universe works, and stop whining about it already. You want performance, you'll get rid of compatability, that's it. It's all a trade off, mantle was not made for robust DirectX protectors, Mantle was not made for OGL Unity fanboys and other mobile trendy arrogant twitter-twatter just-out-of-college stupid developers who think they're smarter than carmack and argue with him on twitter. Mantle was made for the top game developers to enable them what they want, they want to write their own code, not some other 3rdparty company that can only do a half-baked job, but it's not really their fault, because the driver guys can't do better, and developers can't do it either. The source of the problem is API, the DirectX and OGL are high-level, high-compatability APIs, no matter how much their fanboys rationalize and explain and try to downplay it does't change anything, it doesn't work for performance, their claims are BS, period.

It's not just the API here that's getting replaced, it's the CONTROL the developers will get, and with the control they can do much more deeper optimizations that WERE NOT POSSIBLE with driver developers using hacks, driver developers don't have access to game source code, they communicate what to do with the game developer off-site, it's miles apart, they aren't in the same building, it's 2 companies building one game basically, it's impractical, the constant communication back-and-forth clustermuck is very costly on time and money, it's a super inefficient way to optimize games, that's why there are issues on launch, that's why there are beta drivers on launch, that's why the launch always doesn't work as it should, all this will be GONE with Mantle, while using Mantle mode ofcourse, if the same game has DX same things will persist.

For someone to say "Architecture makes mantle possible" is probably the biggest incidactor we're dealing with someone like 13 years old or something, because if you are like 25 maybe, why would you even try to discuss if you're not sure. Because what I'm saying here is all what I'm 100% sure I'm saying holds. This is not my opinion here. And by the way, mantle contains a driver as well, the mantle driver is a fraction of the size and functionality if we compare it to the DX driver.

With the CONTROL the developers get, and the fact that they can fix stuff them selfs, this means, you don't need to wait for the GPU Vendor to release driver updates to improve 5% performance here and there and a little bug fix here and there in various games, no, each developer's game using mantle can FOCUS their effort on the game so much that performance updates could fly out every week, texture corruption, blinkering shadows, that can be fixed, the stupid stacraft 2 cursor corruption can be fixed, the developer doesn't have to keep pressuring the GPU Vendor to please get a fix into the next monthly driver release, they can do it them selfs in matter of days, or even hours, stuff like hotfix could become more popular again because with mantle changing something there won't bug everything else like with crappy DX when it's all unpredicatable with all the driver stuff, there won't be stuff in the driver to interfeer with the game so much, and not so much DX fault directly (but passively) because it's the driver that is a big big apparent causer of all the problems, the drivers are extremely mediocre hacks, whenever you see "optimizations" it's all A HACK, it's not real code, it's not real, it's a hack, it doesn't suppose to be that way, it just a stupid way to do things, limiting yourself with an high-level API and then correcting it with a bloated driver, the driver does is just pick and chooses when to do what to the GPU and it's a joke, and driver developers at GPU Vendors aren't game developers so that's also a cherry on the failcake.


All you consumers expect compatability, because you've been acclimated to it, you've been spoiled, I happen to be basically out of gaming so I don't play much at all, but it doesn't work this way if you want full performance, tradeoffing compatability and other things for performance, now to say that this is a downside of mantle, is extremely subjective again, the media will probably view it as a downside as well, let me explain, this is only a downside from a "mainstream consumer view" , the mainstream consumer wasn't even the target of Mantle, it's not designed for the mainstream casual consumer market as much as they want it to spread, it's not a downside for someone who wants absolute full performance, for me, and for people who waited for something like Mantle for years, we ofcourse don't see compatability as a downside by default, i don't even think about, that's just something that was expected all along, we see it as an ACCEPTABLE SACRIFICE, which means i don't give a rats butt about it, so it's not a downside, I don't NEED it, i don't miss it, I don't want it.
 
Last edited:
No, there was a slide produced during a presentation.
I finally found a copy of it, but I'm burned out researching this today to find the original.
Untitled11.jpg


The text is this:
Multi Vendor?

Mantle is designed to be a thin hardware abstraction
- not tied to AMDs GCN
- forward compatible
- extensions for architecture- and platform-specific functionality

Mantle would be a much more efficient graphics API for other vendors as well
- most Mantle functionality can be supported on todays modern GPUs


edit, I meant to say no thats not how it is written.
Note that he states what Mantle is.
The question being asked on the slide is whether it will be used by multiple vendors, not whether it is capable.

Please fully read posts before you comment back. In fact I acknowledged Andersson statement that Mantle may work on non-GCN. As of now it is tied to GCN because of AMD.

Andersson stated from what he saw in the code he doesn't see why other GPU's couldn't run full or parts of Mantle. As of now and for the foreseeable future it is completely locked to GCN. AMD has yet to officially announce they are allowing other GPU devs to have Mantle. It has all been conjecture and maybes.

I asked you to provide proof AMD intends to give other GPU devs full access to Mantle.
 
It still isn't really here yet, until then it is all just hyperbole.

You can say that for anything. Minimally we should be excited across the board for this simply due to their being tangible performance benefits for gamers.
 
It still isn't really here yet, until then it is all just hyperbole.

I'd say it's here because Dice and other devs are using it And have announced titles after BF4 going forward. I mean seriously, how dumb of a statement could you make saying it isn't here?
 
I am truly excited for this. It's been rumored that Mantle will allowed for unparalleled ease in porting games between consoles and PCs (on top of the now mutual use of x86 architectures between PCs and the next-gen consoles). Can't wait.
 
I am truly excited for this. It's been rumored that Mantle will allowed for unparalleled ease in porting games between consoles and PCs (on top of the now mutual use of x86 architectures between PCs and the next-gen consoles). Can't wait.

that porting bonus is only as long as PC GPUs use GCN, when AMD moves off GCN that's obsolete
 
Please fully read posts before you comment back. In fact I acknowledged Andersson statement that Mantle may work on non-GCN. As of now it is tied to GCN because of AMD.
Yeah I didnt realise we were talking about the same thing, but it wasnt a wish list.
It was a statement of facts and what may be possible,

I asked you to provide proof AMD intends to give other GPU devs full access to Mantle.
I havent said that AMD intends to give other vendors full access to Mantle.
 
I'd say it's here because Dice and other devs are using it And have announced titles after BF4 going forward. I mean seriously, how dumb of a statement could you make saying it isn't here?

Certainly less dumb than a people who believe in they've not personally used or something that hasn't actually been tested by independent reviewers etc.

Like all technology which arrives via demos, specs, abilities, etc by any company it is all hogwash and hyperbole until properly tested.
 
that porting bonus is only as long as PC GPUs use GCN, when AMD moves off GCN that's obsolete

That is why Raja Kodori called it the most scalable architecture on the GPU14 event :) . Because it does not scale and they will move of to something else really soon and all the "work" in Mantle will be the biggest waste of time.

Just to feed the Nvidia trolls AMD is evil like that ......
 
KVESMwg.jpg


That was the lowest FPS I saw it hit during the video and to be fair he stated that it was running at about 30fps. Watching it I saw it spike from 60 at its highest to 34 at its lowest which I don't feel is something that is that unusual as the high fps spots were pushed in views only having to render one ship and a couple of beams whereas the 34fps areas were rendering the entire battle.

Frankly its a new thing and it will take time to be optimized and implemented correctly, I'm not holding my breath and I don't expect massive gains from it. Right now I get the FPS that I want in the games I play, if I don't get a boost no biggs but if I do it will be a nice bonus.

EDIT:

The "Average Frames" is measured in MS.



Did you bother to read how many units that was in that one scene (6200)? Or how about the fact that over 20,000 batch calls are being dispatched. I guess you think in your delusional world that an Nvidia card using DirectX would somehow match this. LOL. In DirectX you cannot even do 20,000 batch calls and if 6200 units had to be rendered with DirectX, then the game would be lucky to be running at one FPS.

You are just crying butt hurt tears because of regret of buying Nvidia. LOL. Take it like man. AMD stock value is also rising. Nvidia is going to lose stock price value. Mark my words. Nvidia = KO'ed

1337364908044.jpg
 
You are just crying butt hurt tears because of regret of buying Nvidia. LOL. Take it like man. AMD stock value is also rising. Nvidia is going to lose stock price value. Mark my words. Nvidia = KO'ed

"Frankly its a new thing and it will take time to be optimized and implemented correctly, I'm not holding my breath and I don't expect massive gains from it. Right now I get the FPS that I want in the games I play, if I don't get a boost no biggs but if I do it will be a nice bonus."

If you actually read my post and what I was replying to you might understand it rather than being a blathering fanboi.

I'm running a 7950 and I'm quite happy with the performance vs price that I paid for it before the buying frenzy. If I don't see a increase in performance with Mantle it won't be a big deal for me as (just like I said) I'm happy with what the 7950 does as is. If I see a boost it will be a nice added bonus, I didn't buy the card specifically for Mantle or for the bundle attached. Both of those things fell far upon the wayside when it came to picking out a new card.

You need to go back and re-read both of those posts and not just make a kneejerk fanboy post as it brought absolutely nothing to the table except making you look like a giant tool.
 
all your arguing is fine and dandy, but am i seriously the only one that cares about what game that may be on display?
 
well, to be fair, the inview with the guy from oxide said he ported the nitrous engine in 2 months by himself.

Thats seems pretty fast to me.
 
all your arguing is fine and dandy, but am i seriously the only one that cares about what game that may be on display?

if you really care about that you would have listened to the presentation. IT is no game, it's just a scenario to test out their engine(though it could at one point develop into a game, it looks like homeworld atm)
 
actually, i think its going to be starcontrol III or a new Sins of a solar empire as all of the ship artwork came from stardock.
 
if you really care about that you would have listened to the presentation. IT is no game, it's just a scenario to test out their engine(though it could at one point develop into a game, it looks like homeworld atm)

being as it looks like something, though they never mentioned what it may have been based on, the last 5 words from your response were appreciated. First couple lines not so much.

actually, i think its going to be starcontrol III or a new Sins of a solar empire as all of the ship artwork came from stardock.

Thank you!
 
i would love for it to be a new sins, the outdated single threaded engine drives me batty when i cant get more than 20fps zoomed in on my 290 pro.
 
[

You are just crying butt hurt tears because of regret of buying Nvidia. LOL. Take it like man. AMD stock value is also rising. Nvidia is going to lose stock price value. Mark my words. Nvidia = KO'ed

AMD spent $8,000,000 to possibly get better performance in one game long after it was released.

Yeah I can see how NVIDIA should be worried. :rolleyes:

Also that tech demo on a laptop that they have stated they have no current plans to bring to market sounds promising.

I'm sure a lot of AMD "supporters" are rewarded for their work on forums, but they really need to brush up on what they are actually talking about.
 
oh geez, here comes the master nvidiot PRIME1.

If you actually keep up with thiis stuff, you would know that they arent bringing it to market, because there isnt anything to bring. They are using it as proof on concept to push for the new dp 1.3/eDP/VBI adoption by the OEMs so no new hardware is needed for dynamic refresh.

8M dollars to drive APU sales as a viable mid-end game platform? id say that money well spent.
 
Last edited:
oh geez, here comes the master nvidiot PRIME1.

If you actually keep up with thiis stuff, you would know that they arent bringing it to market, because there isnt anything to bring. They are using it as proof on concept to push for the new dp 1.3/eDP/VBI adoption by the OEMs so no new hardware is needed for dynamic refresh.

8M dollars to drive APU sales as a viable mid-end game platform? id say that money well spent.
no new hardware, except yes entirely new hardware which is why they need it in future standards including VESA because they can't do it with current hardware. freesync ofc they didn't solve the software side that actually makes it useful for games it's all just pr fluff they threw together in a couple months.

mantle api is worthless in the same, they'll have to pay developers, which unlike nvidia AMD doesn't do with as much resolve, to use mantle as it's an extra cost as no develop is going to spend money to port a game to mantle for no god dam reason that's extra work that they could use to improve their directX side of the game for all cards not just a section of AMD cards.
 
no new hardware, except yes entirely new hardware which is why they need it in future standards including VESA because they can't do it with current hardware. freesync ofc they didn't solve the software side that actually makes it useful for games it's all just pr fluff they threw together in a couple months.

mantle api is worthless in the same, they'll have to pay developers, which unlike nvidia AMD doesn't do with as much resolve, to use mantle as it's an extra cost as no develop is going to spend money to port a game to mantle for no god dam reason that's extra work that they could use to
improve their directX side of the game for all cards not just a section of AMD cards.

Its already in the spec, they are pushing adoption, not new hardware. They spent exactly zero resources on that demo. It already exists in laptops, as they use eDP, and even some desktop monitors if they stayed with the VESA spec and included VBI. Im willing to bet most monitors with a vga input still comply.

The developers ASKED for mantle. They approached nvidia and Intel as well, but they blew them off.

Any money spent to develop mantle will be recouped by apu sales. many times over. Once the api is set, there wil be no more costs for AMD other than occasional driver updates.

Any extra development cost by the developers will be happily paid because they got what they wanted. They get a massively better port than with DX, that's extra sales for them. Think about all the lousy ports out there, and how much better they would be if they could actually take advantage of all that power. The extra porting costs from console to mantle is minimal compared to Dx. The API was intentionally designed to mimic console APIs.

The whole point of the partnership with DICE was to make sure porting was easy as possible.

If you cant see the financial model here is potentially worth a fortune, assuming mantle does what its supposed to, then you need more business experience.
 
Last edited:
AMD spent $8,000,000 to possibly get better performance in one game long after it was released.

2months is long after release?
$8 million was marketing/game bundling, etc. IE not AMD paying for DICE to use Mantle.
I also doubt it was a cash deal, if that was indeed the deal.
 
If variable v-blank makes it into the DP 1.3 standard (not optional), then new hardware will be needed but it'll become common place and built into the price of a monitor, just like scalers, OSD menus, and controllers to handle multiple inputs. If that happens, the good news is that any video card manufacturer would be able to use variable v-blank for their own implementation of FreeSync. I fail to see how pushing for that by showing a proof-of-concept would be anything but a good thing.
 
i would love for it to be a new sins, the outdated single threaded engine drives me batty when i cant get more than 20fps zoomed in on my 290 pro.

oh, that sucks. I had heard about that game before and was about to check it out and see if its any fun, but if its going to run like crap than i might have to pass.
 
Back
Top