The AMD Mantle Thread

[brandon888]

The first Mantle patch should be in late december. But they are awfully quiet about that and still no performance numbers. And the game to carry the first patch is having anything but success at the moment.

i bought amd card cause of mantle and bf4 ;/ if mantle will fail i will switch from my 7950 to 770 :/


performance will be about 20% + gain as i know

 

[NTMBK]

The first Mantle patch should be in late december. But they are awfully quiet about that and still no performance numbers. And the game to carry the first patch is having anything but success at the moment.

Yup, it's a shame that BF4 is crippled with bugs unrelated to Mantle; it's pretty unfortunate for Mantle's flagship game. Hopefully they can get it back into shape and turn it into an enjoyable experience.

Any news on what the next Mantle-enabled game after BF4 will be? Is Thief Mantle, or is that just TrueAudio? And will Mantle get ported to NFS:Rivals? (That last one I'd very much like, I'm really enjoying NFS:Most Wanted right now and will probably pick up Rivals in a sale.)

 

[Stewox]

The first Mantle patch should be in late december. But they are awfully quiet about that and still no performance numbers. And the game to carry the first patch is having anything but success at the moment.

xmas present

 

[brandon888]

Yup, it's a shame that BF4 is crippled with bugs unrelated to Mantle; it's pretty unfortunate for Mantle's flagship game. Hopefully they can get it back into shape and turn it into an enjoyable experience.

Any news on what the next Mantle-enabled game after BF4 will be? Is Thief Mantle, or is that just TrueAudio? And will Mantle get ported to NFS:Rivals? (That last one I'd very much like, I'm really enjoying NFS:Most Wanted right now and will probably pick up Rivals in a sale.)

all frostbite games will support mantle , even dragon age 3 ..... crossed fingers and hoped for the best. .... i gave it a chance to amd card cause of it ....

 

[ShintaiDK]

all frostbite games will support mantle , even dragon age 3 ..... crossed fingers and hoped for the best. .... i gave it a chance to amd card cause of it ....

You mean all Frostbite games got the option to support Mantle.

NFS:Rivals is already capped at 30FPS due to the slow consoles.

 

[ShintaiDK]

xmas present

So its coming today or monday?

 

[Silverforce11]

You mean all Frostbite games got the option to support Mantle.

NFS:Rivals is already capped at 30FPS due to the slow consoles.

Or it could, so crappy APUs will run it at 30 fps with maxed settings.

We just don't know yet.

 

[jj109]

If Mantle fails, at least AMD will be forced to enable DX11 command lists so it's not a complete loss.

 

[jj109]

Or it could, so crappy APUs will run it at 30 fps with maxed settings.

We just don't know yet.

Isn't NFS engine's simulation rate tied to render rate? Sounds like an awful design pulled from 15 years ago.

 

[NTMBK]

You mean all Frostbite games got the option to support Mantle.

NFS:Rivals is already capped at 30FPS due to the slow consoles.

Yeah, the 30FPS cap is a pain, but I'd be willing to accept it. I play plenty of console games at 30FPs, and it doesn't really bother me that much.

But reducing the CPU load would be a big bonus. I get some very nasty slowdowns in NFS:Most Wanted when I enter the city regions; I suspect that it is my crummy PCIe Gen1 motherboard, my terrible hard drive and my slightly elderly CPU all struggling to load the resources needed onto the GPU. (The slowdown persisted even with graphics settings turned right down, so I don't think it was GPU related.) If a more lightweight API reduces those sorts of slowdowns, I'll be happy.

 

[gorobei]

Yeah, the 30FPS cap is a pain, but I'd be willing to accept it. I play plenty of console games at 30FPs, and it doesn't really bother me that much.

But reducing the CPU load would be a big bonus. I get some very nasty slowdowns in NFS:Most Wanted when I enter the city regions; I suspect that it is my crummy PCIe Gen1 motherboard, my terrible hard drive and my slightly elderly CPU all struggling to load the resources needed onto the GPU. (The slowdown persisted even with graphics settings turned right down, so I don't think it was GPU related.) If a more lightweight API reduces those sorts of slowdowns, I'll be happy.

the oxide guys underclocked an 8core FX cpu into the sub 2ghz range(jaguar-esque) and they were still gpu bound at the 100k batch calls.

if NFS fps is capped at 30, mantle should eliminate any cpu bottlenecks.

 

[ShintaiDK]

the oxide guys underclocked an 8core FX cpu into the sub 2ghz range(jaguar-esque) and they were still gpu bound at the 100k batch calls.

if NFS fps is capped at 30, mantle should eliminate any cpu bottlenecks.

That doesnt say much in itself as prescripted benchmarks have shown.

A 2Ghz kaveri for example can get around 50-55FPS in Tomb Raider benchmark with a GTX680. But only around 15FPS ingame.

Same with Haswell. 800Mhz vs 4900Mhz.
74 vs 104FPS in benchmark.
22 vs 105FPS ingame.

Wast majority of the CPU limitations is outside the API.

 

[NTMBK]

That doesnt say much in itself as prescripted benchmarks have shown.

A 2Ghz kaveri for example can get around 50-55FPS in Tomb Raider benchmark with a GTX680. But only around 15FPS ingame.

Same with Haswell. 800Mhz vs 4900Mhz.
74 vs 104FPS in benchmark.
22 vs 105FPS ingame.

Wast majority of the CPU limitations is outside the API.

Wait, Kaveri Tomb Raider benchmarks? Did I miss something?

(Serious question, not sarcasm!)

 

[Skurge]

Yup, it's a shame that BF4 is crippled with bugs unrelated to Mantle; it's pretty unfortunate for Mantle's flagship game. Hopefully they can get it back into shape and turn it into an enjoyable experience.

Any news on what the next Mantle-enabled game after BF4 will be? Is Thief Mantle, or is that just TrueAudio? And will Mantle get ported to NFS:Rivals? (That last one I'd very much like, I'm really enjoying NFS:Most Wanted right now and will probably pick up Rivals in a sale.)

Next Frostbite game will be Plants vs Zombies Garden Warfare, Thief is also mantle. Rivals came out before Mantle was finished for Frostbite so I think games after PvZ will all have mantle as it will be part of the engine.

 

[gorobei]

That doesnt say much in itself as prescripted benchmarks have shown.

A 2Ghz kaveri for example can get around 50-55FPS in Tomb Raider benchmark with a GTX680. But only around 15FPS ingame.

source?

Same with Haswell. 800Mhz vs 4900Mhz.
74 vs 104FPS in benchmark.
22 vs 105FPS ingame.

the starswarm demo isnt a prescripted benchmark, its a real time demo with user camera control and full AI controlling all units, with individual navigation, individual targeting, individual effects with full render quality. its doing everything the game engine will do in a full game. its an in-house proof of concept test they cobbled together for other devs to test and play with.

Wast majority of the CPU limitations is outside the API.

source? or is this your opinion? im guessing you are making some sort of blanket statement about cpu bottlenecks. the oxide video opening and the APU13 dev QA session suggest the opposite.

there are games that are cpu bound and others that are gpu bound. this is entirely a function of the game engine, gameplay type, render pipeline, asset format when they are all using the same dx3d api.

NTMBK's example suggest his cpu and bus system are failing to send the game assets to the gpu. this is because the cpu is chugging hard on the dx overhead just to get the assets loaded into gpu memory and the draw calls to the gpu driver.
the oxide video shows they can get anywhere from 5 to 10 times the batch calls out with mantle over dx. this is on rudimentary alpha code. they arent even touching the direct to metal gcn code yet.

 

[krumme]

the oxide guys underclocked an 8core FX cpu into the sub 2ghz range(jaguar-esque) and they were still gpu bound at the 100k batch calls.

if NFS fps is capped at 30, mantle should eliminate any cpu bottlenecks.

The consoles playing same bf4 at 60fps on slow jaguar cores in a first person shooter - not even batch limited like a rts- shows how extreme the gains can be with mantle.
If anything is hard data thats it imho.

Go see the mantle execution model and presentation; There is a far better paralalization. That explains why the consoles can work fine even though they are not batch limited like a rts.

The crazy thing is as said before, we can even have situations in rts games where a slow quad core kaveri can play a game using mantle and an highend cpu on 780 can not do the same. Because of dx.

We are talking differences of a factor 30 or 3000%. People comming with 20% or 50% difference for lacking dx11 optimization should compare the 20-50% to 3000%.

I want the 3000%. Because it can give me completely new gaming experience. To be frank i dont care about ms or nv or amd financials or the feelings of some brand loyalists. I want some freaking new games that kick some a.. instead of the same stuff year after year. I want Dune e 3 in uber 3d with trueaudio created by a modern Mozart. Thousands of configurable units in a huge online war !!!!! 🙂

Unfortunately we will probably tave to wait for gen2 mantle engines to provide that. Lol.

 

[ShintaiDK]

source? or is this your opinion? im guessing you are making some sort of blanket statement about cpu bottlenecks. the oxide video opening and the APU13 dev QA session suggest the opposite.

I already gave you an example with Tomb Raider. 74FPS benchmark, 22FPS ingame for an 800Mhz Haswell. Both giving the same FPS without a CPU bottleneck with Haswell at 4900Mhz. 104FPS vs 105FPS.

Then I dont care whatever some promotion person said. This is a real life example.

 

[Bubbleawsome]

Yeah, the 30FPS cap is a pain, but I'd be willing to accept it. I play plenty of console games at 30FPs, and it doesn't really bother me that much.

But reducing the CPU load would be a big bonus. I get some very nasty slowdowns in NFS:Most Wanted when I enter the city regions; I suspect that it is my crummy PCIe Gen1 motherboard, my terrible hard drive and my slightly elderly CPU all struggling to load the resources needed onto the GPU. (The slowdown persisted even with graphics settings turned right down, so I don't think it was GPU related.) If a more lightweight API reduces those sorts of slowdowns, I'll be happy.

Off topic, but I had the same issue and all I had to do was max the game then turn off super sampling and HQ textures, then turn the rest down to medium and it was buttery smooth. Anything else and it lags. It's worth a shot.

 

[gorobei]

I already gave you an example with Tomb Raider. 74FPS benchmark, 22FPS ingame for an 800Mhz Haswell. Both giving the same FPS without a CPU bottleneck with Haswell at 4900Mhz. 104FPS vs 105FPS.

you gave some numbers, you didn't give where you got those numbers. when we say source, we want the original web address.

 

[ShintaiDK]

you gave some numbers, you didn't give where you got those numbers. when we say source, we want the original web address.

http://forums.anandtech.com/showpost.php?p=35371206&postcount=7

 

[Enigmoid]

The consoles playing same bf4 at 60fps on slow jaguar cores in a first person shooter - not even batch limited like a rts- shows how extreme the gains can be with mantle.
If anything is hard data thats it imho.

Go see the mantle execution model and presentation; There is a far better paralalization. That explains why the consoles can work fine even though they are not batch limited like a rts.

The crazy thing is as said before, we can even have situations in rts games where a slow quad core kaveri can play a game using mantle and an highend cpu on 780 can not do the same. Because of dx.

We are talking differences of a factor 30 or 3000%. People comming with 20% or 50% difference for lacking dx11 optimization should compare the 20-50% to 3000%.

I want the 3000%. Because it can give me completely new gaming experience. To be frank i dont care about ms or nv or amd financials or the feelings of some brand loyalists. I want some freaking new games that kick some a.. instead of the same stuff year after year. I want Dune e 3 in uber 3d with trueaudio created by a modern Mozart. Thousands of configurable units in a huge online war !!!!! 🙂

Unfortunately we will probably tave to wait for gen2 mantle engines to provide that. Lol.

https://www.youtube.com/watch?v=M4gaVvHXNC8

Neither can hold 60 fps min in the campaign. And BF4 campaign is hardly CPU limited.

On MP its more like 50 avg (jumps from 40-60)

https://www.youtube.com/watch?v=dkjS4wLEQX0

Which is around i3 level (ultra settings). When you take into account the 6 core and the fact that the console is not running ultra the framerates are comparable. I'm not seeing much if any console optimization here.

Similar framerate on a slightly stronger CPU at more demanding settings.

 

[Imouto]

The difference here is that the Oxide presentation isn't a scripted benchmark.

 

[blackened23]

The consoles playing same bf4 at 60fps on slow jaguar cores in a first person shooter - not even batch limited like a rts- shows how extreme the gains can be with mantle.

No. I thought the PS4 version of BF4 looked pretty good at first. Playing 20 feet away from a HDTV makes the perception good I guess? Which is how most console games are played. Anyway, upon closer inspection, the PS4 version (which is better than the XB1 version) runs at 900p and isn't at 60 fps in multiplayer. In fact, it dips to 40 fps with frequency, and it is not on ultra settings. The PS4 version is equal to low-medium settings on the PC side. Of course then you have the XB1 which is far worse than that, with nearly every XB1 title being 720p or 720p upscaled.

Despite this, they are a big jump from PS3 and Xbox 360. The versions of BF4 on those machines are 480p. XB1 is 720p. PS4 is 900p. None are running at 60 fps solid in multiplayer.

Granted, that *might* be pretty good given the anemic hardware in the PS4. But it is hardly amazing compared to how good the xbox 360 was at launch. The xbox 360 was straight up better in terms of GPU horsepower compared to PC at launch in 2005 and was actually a legitimately interesting console. (although the PC surpassed it VERY fast) The same cannot be said of PS4 or XB1.

It should also be noted your comparison to Mantle with the next generation consoles is weak at best. Not only are the next gen consoles not performing as well comparatively as their prior gen brethren did at launch (the 360 was actually REALLY good at launch), the API situation is completely different. They are not using Mantle.

 

[gorobei]

http://forums.anandtech.com/showpost.php?p=35371206&postcount=7

quoting bala on the difference between ingame and a prescripted bench doesnt qualify as a source for your assertion that

Wast majority of the CPU limitations is outside the API.

it is one example that you are taking to represent all games, which is an obvious fallacy.

and entirely irrelevant as i(and others)told you that starswarm is a realtime demo and not a scripted bench.

the starswarm demo isnt a prescripted benchmark, its a real time demo with user camera control and full AI controlling all units, with individual navigation, individual targeting, individual effects with full render quality. its doing everything the game engine will do in a full game. its an in-house proof of concept test they cobbled together for other devs to test and play with.

The difference here is that the Oxide presentation isn't a scripted benchmark.

also we asked for your kaveri source.

Wait, Kaveri Tomb Raider benchmarks? Did I miss something?

(Serious question, not sarcasm!)

That doesnt say much in itself as prescripted benchmarks have shown.

A 2Ghz kaveri for example can get around 50-55FPS in Tomb Raider benchmark with a GTX680. But only around 15FPS ingame.

source?

 

[ShintaiDK]

Wait, Kaveri Tomb Raider benchmarks? Did I miss something?

(Serious question, not sarcasm!)

Kabini obviously. With a GTX680 to remove GPU bottlenecks.

http://adrenaline.uol.com.br/biblioteca/analise/784/amd-a6-5200-kabini.html?pg=05

The API benefits is hugely overestimated.