How AMD's Mantle Makes Games Better

Man, AMD and their money grabbing proprietary technology.

/sarcasm
 
Man, AMD and their open and free technology.

Fixed that for you.

AMD said if NVIDIA wants to implement Mantle is can.

NVIDIA don't even need to support full mantle just parts, games will use the parts that NVIDIA support and fall back to DX or OpenGL for parts not implemented.
 
That stuttering slow mess is supposed to show how great mantle is? That looks terrible.
 
Nobody doubts that Mantle would be awesome, but the problem is how Intel and Nvidia are going to respond to this. Even though Mantle is open, we all know Nvidia isn't going to support it, just to be dicks. Intel won't, cause they aren't about graphics.

The other problem is while Mantle is open, it's not showing it very well. This tech works exclusively on Southern Island based GPU's. Their own Northern Island's aren't even supported, so how exactly will this open standard work if AMD can't even get their other generation of GPUs to support it? Southern Islands isn't all that much different in performance, which leads me to doubt there's much difference in architecture. Even on the Linux side of things, a lot of the driver code used for 5000/6000 cards are used for the 7000/200 cards. The architecture isn't even vastly different and there's no mantle, but they expect Nvidia or Intel to adopt it? If someone like Silicon Graphics got Nvidia/AMD/Intel to come together and agree on a low level API standard, then I could see something serious. Something like an OpenGL extension that allows for platform specific optimizations.

Before people go around thinking that AMD is out to screw Nvidia, they're not. Mantle is purely to get customers to upgrade their video cards, rather then continue to use their HD 5000 or HD 6000 series cards. Their DX11 generation of cards have been going on for a long time, and saw very few upgrades. Instead of shifting their cards and prices, they renamed the 7000 series to 200 series. Mantle is their way to get people excited to go out and buy new graphic cards, and what better game then Battlefield 4.
 
That stuttering slow mess is supposed to show how great mantle is? That looks terrible.
That stuttering slow mess is a problem with your internet connection or video rendering.
Its very smooth here watching from UK.
I went to Youtube directly using the link provided.
 
Nobody doubts that Mantle would be awesome, but the problem is how Intel and Nvidia are going to respond to this. Even though Mantle is open, we all know Nvidia isn't going to support it, just to be dicks. Intel won't, cause they aren't about graphics.
Game engines are incorporating support for Mantle.
NVidia can create their own version and also incorporate it into game engines.

Before people go around thinking that AMD is out to screw Nvidia, they're not. Mantle is purely to get customers to upgrade their video cards, rather then continue to use their HD 5000 or HD 6000 series cards. Their DX11 generation of cards have been going on for a long time, and saw very few upgrades. Instead of shifting their cards and prices, they renamed the 7000 series to 200 series. Mantle is their way to get people excited to go out and buy new graphic cards, and what better game then Battlefield 4.
You could incorrectly say that about introducing any new tech that uses new hardware features.
 
time to switch to AMD?...maybe for my next VC upgrade depending on how Mantle performs vs how good Maxwell is
 
For me it's going to come down to future games that gain something from Mantle vs. existing games that already gain something from PhysX...and Nvidia's (IMO) better driver support.
 
Nobody doubts that Mantle would be awesome, but the problem is how Intel and Nvidia are going to respond to this. Even though Mantle is open, we all know Nvidia isn't going to support it, just to be dicks. Intel won't, cause they aren't about graphics.

The other problem is while Mantle is open, it's not showing it very well. This tech works exclusively on Southern Island based GPU's. Their own Northern Island's aren't even supported, so how exactly will this open standard work if AMD can't even get their other generation of GPUs to support it? Southern Islands isn't all that much different in performance, which leads me to doubt there's much difference in architecture. Even on the Linux side of things, a lot of the driver code used for 5000/6000 cards are used for the 7000/200 cards. The architecture isn't even vastly different and there's no mantle, but they expect Nvidia or Intel to adopt it? If someone like Silicon Graphics got Nvidia/AMD/Intel to come together and agree on a low level API standard, then I could see something serious. Something like an OpenGL extension that allows for platform specific optimizations.

Before people go around thinking that AMD is out to screw Nvidia, they're not. Mantle is purely to get customers to upgrade their video cards, rather then continue to use their HD 5000 or HD 6000 series cards. Their DX11 generation of cards have been going on for a long time, and saw very few upgrades. Instead of shifting their cards and prices, they renamed the 7000 series to 200 series. Mantle is their way to get people excited to go out and buy new graphic cards, and what better game then Battlefield 4.

Well, Valve is pushing OpenGL for steambox. So I'm guessing we will end up with Mantle for newer AMD, modified OpenGL for Nvidia (or maybe for anything), and DirectX as a fallback or for games that are close to Microsoft.
 
For me it's going to come down to future games that gain something from Mantle vs. existing games that already gain something from PhysX...and Nvidia's (IMO) better driver support.

Interesting potential for Mantle though is actually changing gameplay; like going from a standard to sport or performance bike. PhysX largely is purely cosmetic with no effects on cosmetics. Like lipstick on a pig.

I'm hoping developers adopt Mantle just because it will allow their games to do more. I have to say, I don't know if I like the way Oxide games is using the technology to be honest. It's interesting to conceptually imagine a space battle that has 6,000 space ships all battling each-other, each with their own laser fire, presumably differentiating stats, ai, etc. However, it's a bit hard to keep track (imo) of six thousand units and I'd imagine you'd leave the majority of your fleet on auto-pilot. Performing flanking maneuvers with 6k units would be hard. In which case, it becomes more like a crappy-cgi movie than 'a game'.

That being said... StarCraft 1 only had 200 unit 'supply' counters per person and 200 unit costs was very manageable to organize/move/play due to grouping and hotkeying. With air units, I would have loved to have been able to have 400-500 units at times. I think 6,000 units would be overkill though. At that point, I'd personally prefer less units and upgrading the graphical fidelity of individual units or ai capabilities.... than so many units.

That being said, it would be neat to have a GTAIV remake/mod for mantle where city streets are 'CROWDED' downtown with sidewalks and cross-ways having nearly row to row pedestrians when the lights change and each side of the road having a lot of pedestrians; like a real city. That would increase realism but not really negatively effect gameplay.... that is unless you liked driving on the sidewalks or diving straight through lights and nobody being on the crosswalk. There would once again need to be that fine balance stuck of better looking, better acting(ai) pedestrians vs sheer quantity.

More background people in L.A. Noire wouldn't hurt either, etc.
 
I think everyone should stop speculating and wait to pass judgement on Mantle once it's "out in the wild" with all of us.
 
I don't like how he skated around the comparison question ("turn things off and lower graphics"). How about you tell us hard numbers on an apple to apple comparison of this special demo. With mantle on and off. FPS vs FPS.

If it's 2FPS with it being off and 18FPS with it on, say that.
 
I don't like how he skated around the comparison question ("turn things off and lower graphics"). How about you tell us hard numbers on an apple to apple comparison of this special demo. With mantle on and off. FPS vs FPS.

If it's 2FPS with it being off and 18FPS with it on, say that.

No the fact is the GPU simply cannot produce the image without mantle.

Mantle helps remove a large bottleneck in gpus, and allows them to have much faster draw calls.
 
Interesting potential for Mantle though is actually changing gameplay; like going from a standard to sport or performance bike. PhysX largely is purely cosmetic with no effects on cosmetics. Like lipstick on a pig.

I'm hoping developers adopt Mantle just because it will allow their games to do more. I have to say, I don't know if I like the way Oxide games is using the technology to be honest. It's interesting to conceptually imagine a space battle that has 6,000 space ships all battling each-other, each with their own laser fire, presumably differentiating stats, ai, etc. However, it's a bit hard to keep track (imo) of six thousand units and I'd imagine you'd leave the majority of your fleet on auto-pilot. Performing flanking maneuvers with 6k units would be hard. In which case, it becomes more like a crappy-cgi movie than 'a game'.

That being said... StarCraft 1 only had 200 unit 'supply' counters per person and 200 unit costs was very manageable to organize/move/play due to grouping and hotkeying. With air units, I would have loved to have been able to have 400-500 units at times. I think 6,000 units would be overkill though. At that point, I'd personally prefer less units and upgrading the graphical fidelity of individual units or ai capabilities.... than so many units.

That being said, it would be neat to have a GTAIV remake/mod for mantle where city streets are 'CROWDED' downtown with sidewalks and cross-ways having nearly row to row pedestrians when the lights change and each side of the road having a lot of pedestrians; like a real city. That would increase realism but not really negatively effect gameplay.... that is unless you liked driving on the sidewalks or diving straight through lights and nobody being on the crosswalk. There would once again need to be that fine balance stuck of better looking, better acting(ai) pedestrians vs sheer quantity.

More background people in L.A. Noire wouldn't hurt either, etc.

HAHAHA.. I would just run over the peds in my way in GTAIV.
 
You could incorrectly say that about introducing any new tech that uses new hardware features.

Mantle isn't a hardware feature. It's purely a driver feature. That's the whole point, cutting out the overhead from OpenGL and Direct3D by making an API that's closer to the hardware.
 
I think everyone should stop speculating and wait to pass judgement on Mantle once it's "out in the wild" with all of us.

I think companies should stop making presentations for tech that isn't ready/finished yet. AMD got a bunch of buzz (and sales) off all this mantle talk, but we haven't seen real world results.
 
You're an idiot, or just blind.

Watch the video again, and look closely at the avg FPS display when zoomed in. You can clearly see massive FPS drops. You can also clearly see the FPS drops and stuttering if you watch the display behind the guys while they're talking(and moving quite fluidly, while demo is stuttering).

Using a demo that can't even maintain 30fps to try and show off an API is just terrible.

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.
 
Last edited:
You're an idiot, or just blind.

Watch the video again, and look closely at the avg FPS display when zoomed in. You can clearly see massive FPS drops. You can also clearly see the FPS drops and stuttering if you watch the display behind the guys while they're talking(and moving quite fluidly, while demo is stuttering).

Using a demo that can't even maintain 30fps to try and show off an API is just terrible.

I see very little stuttering. It has a max recent frametime ffs lol. The man also claims they havent done much optimization for it.
 
There is no stuttering at all when they're speaking, other than a certain beam effect which appears to be stuttery by its nature.
 
Fixed that for you.

AMD said if NVIDIA wants to implement Mantle is can.

I have yet to see evidence AMD definitively said Mantle will work on NV GPU's.

To date all info is Mantle requires GCN. So are you saying if NV decides to drop their current architecture and use AMD's GCN then it will work? That is being open?

Only evidence I have seen is a slide from APU 13, saying "multi-vendor support?", that was part of DICE's presentation on their impression of Mantle and wishlist for the future. Not an official AMD statement.

I am not flaming you, but you keep insisting AMD has allowed anyone and everyone to use Mantle. I don't buy it. Especially since AMD's own site claims Mantle is "Powered by AMD." Not one word of Mantle on other GPU's.

Why would AMD spend millions on a tech and just give it away? AMD are not and have not been nice guys just like NV are not nice guys. They are large corporations with investors and their own interests to protect. Plus neither side has ever played nice together. So why now?

I for one think Mantle can be amazing but I highly, highly doubt you will ever see AMD allow NV to use it. AMD has a better tech than NV and can keep it to themselves.
 
I for one think Mantle can be amazing but I highly, highly doubt you will ever see AMD allow NV to use it. AMD has a better tech than NV and can keep it to themselves.

agreed...here's this amazing technology we developed in-house...let's let our biggest competitor use it for free as well...if anything it will be borked...AMD saying it's open source or whatever is pure marketing speak meant for goodwill purposes only...look at us, we are saints and Nvidia is the devil for using proprietary tech like PhysX
 
http://i.imgur.com/KVESMwg.jpg?1

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.

I think I remember them saying submit is the time they're spending in Mantle -- so it aligns perfectly with what they're saying. Mantle is not the bottleneck.
 
You're an idiot, or just blind.

Watch the video again, and look closely at the avg FPS display when zoomed in. You can clearly see massive FPS drops. You can also clearly see the FPS drops and stuttering if you watch the display behind the guys while they're talking(and moving quite fluidly, while demo is stuttering).

Using a demo that can't even maintain 30fps to try and show off an API is just terrible.

I'll pass that one straight back to you.
I watched it again and no matter what the framerate was, it didnt look jerky or unsmooth.
It looked fine.
Sort your hardware out, it cant play video smoothly :p
 
at apu 13 , amd said it was still a work in progress so they weren't gonna make it an open standard.
could they? yes. Should day? Is a man not entitled to the sweat of his brow? -Andrew Ryan
Time will tell. Either way, i see amd taking 1-2 years exclusivity if they go open.
 
You're an idiot, or just blind.

Watch the video again, and look closely at the avg FPS display when zoomed in. You can clearly see massive FPS drops. You can also clearly see the FPS drops and stuttering if you watch the display behind the guys while they're talking(and moving quite fluidly, while demo is stuttering).

Using a demo that can't even maintain 30fps to try and show off an API is just terrible.

Should have watched the presentation...they said F2F time was the closest representaion of FPS

It was running at about 300 FPS if I remember what he said correctly. When it was on hawaii board.

This was on an UNDERCLOCKED fx8350 and over 30 FPS on an unknown board.
 
Same demo they showed off at APU13, would have been nice to see something new. They should probably be running a DX build next to the Mantle build for comparison purposes, would certainly help to make a proper judgement.

I'm still quite interested in the potential of the tech though, can you say non broken Total War?
 
You guys know youtube is max 30fps, right? No one can make a legitimate judgment about smoothness from a youtube video. Besides, it's a tech demo for an unreleased game. How can you knock it for low fps when it's being designed for high-end hardware in 2015/16?

And Mantle is certainly an open API, whether AMD likes it or not. There's no way they could get developers on board with developing for a new API if half the market is permanently locked out of it. That's why DICE's presentation a month ago stressed that other vendors (Nvidia) could add support for their hardware to Mantle.
 
Should have watched the presentation...they said F2F time was the closest representaion of FPS

It was running at about 300 FPS if I remember what he said correctly. When it was on hawaii board.

This was on an UNDERCLOCKED fx8350 and over 30 FPS on an unknown board.

The 300 number was an extrapolated estimate for performance that they would get if they had infinite GPU power. The real performance was lower (because they were GPU bound)
 
Should have watched the presentation...they said F2F time was the closest representaion of FPS

It was running at about 300 FPS if I remember what he said correctly. When it was on hawaii board.

This was on an UNDERCLOCKED fx8350 and over 30 FPS on an unknown board.

Which as I said, is a terrible way to demo it. If they want to show how great it is, they need to show a system with mantle, and a system without it. Not some supposedly underclocked hardware running a demo that apparently some people on this forum are incapable of seeing the framerate drop.you can clearly see the framerate tank at 1:57 while the dev is talking without and framerate issues visible due to the video playback. If it was a matter of my computer or internet connection(as if that would have mattered with a fully buffered video), then why doesn't the dev who is speaking look like crap if it was a matter of video playback framerate? It's hilarious that some people want to pretend that crappy stuttering video on youtube is somehow selective to screen area through an entire video.
 
GCN versions 1 and 2 are required to run Mantle.
http://en.wikipedia.org/wiki/Mantle_(API)

API's are purely software, and not hardware. AMD saying it requires GCN is strictly marketing. Ideally, to get the best performance out of Mantle you would need GCN type hardware, but that there lies the problem with Mantle. How does Nvidia expect to use this API and get acceptable performance out of it, when AMD's HD 5000 and HD 6000 series cards can't? It's marketing spin.

Technically Nvidia's hardware could benefit from Mantle if they write a well made driver, but AMD will always have the upper hand with Mantle, due to being closely made for GCN hardware. If that isn't enough, then someone could make a Mantle Wrapper to allow Nvidia cards to use it, even though a Wrapper wouldn't increase performance at all. Remember Glide and Glide Wrappers?
 
API's are purely software, and not hardware. AMD saying it requires GCN is strictly marketing. Ideally, to get the best performance out of Mantle you would need GCN type hardware, but that there lies the problem with Mantle. How does Nvidia expect to use this API and get acceptable performance out of it, when AMD's HD 5000 and HD 6000 series cards can't? It's marketing spin.

Technically Nvidia's hardware could benefit from Mantle if they write a well made driver, but AMD will always have the upper hand with Mantle, due to being closely made for GCN hardware. If that isn't enough, then someone could make a Mantle Wrapper to allow Nvidia cards to use it, even though a Wrapper wouldn't increase performance at all. Remember Glide and Glide Wrappers?

Mantle expects a certain hardware feature set, what is strange that they picked a recent architecture as a baseline?
 
API's are purely software, and not hardware. AMD saying it requires GCN is strictly marketing. Ideally, to get the best performance out of Mantle you would need GCN type hardware, but that there lies the problem with Mantle. How does Nvidia expect to use this API and get acceptable performance out of it, when AMD's HD 5000 and HD 6000 series cards can't? It's marketing spin.

Technically Nvidia's hardware could benefit from Mantle if they write a well made driver, but AMD will always have the upper hand with Mantle, due to being closely made for GCN hardware. If that isn't enough, then someone could make a Mantle Wrapper to allow Nvidia cards to use it, even though a Wrapper wouldn't increase performance at all. Remember Glide and Glide Wrappers?

So you are saying that my 6(?) year old 8800GTS can support DirectX 11 since the API is purely software?
 
API's are purely software, and not hardware.
GCN is a hardware architecture.
AMD saying it requires GCN is strictly marketing.
Can you back this up with any proof?

Ideally, to get the best performance out of Mantle you would need GCN type hardware, but that there lies the problem with Mantle. How does Nvidia expect to use this API and get acceptable performance out of it, when AMD's HD 5000 and HD 6000 series cards can't? It's marketing spin.
This isnt about NVidia, you are confused, Mantle is an AMD creation.
NVidia dont have to use it, there are other graphics APIs.
I imagine its a lot of hard work getting Mantle working with the hardware it was designed for.
Its no surprise that support for other slower architectures will have to wait.

Why point the finger at AMD when its them that is advancing performance.
Does Mantle somehow slow down NVidia performance?

Technically Nvidia's hardware could benefit from Mantle if they write a well made driver, but AMD will always have the upper hand with Mantle, due to being closely made for GCN hardware. If that isn't enough, then someone could make a Mantle Wrapper to allow Nvidia cards to use it, even though a Wrapper wouldn't increase performance at all. Remember Glide and Glide Wrappers?
So why dont NVidia make their own API that is closer to the bones that utilises parts of Mantle.
At least some of the work is already done for them.
They can design hardware to better utilise a Mantle type API then as well.
 
Fixed that for you.

AMD said if NVIDIA wants to implement Mantle is can.

NVIDIA don't even need to support full mantle just parts, games will use the parts that NVIDIA support and fall back to DX or OpenGL for parts not implemented.

AMD said that Mantle is tied the the GCN architecture. It is not open.

In fact it may be the most closed API out there. It won't even be used on the consoles.
 
Back
Top