The AMD Mantle Thread

[Adampa1006]

So you are saying there are not any enthusiast who run high end Intel CPU's with crossfire 290X's? Or are you just another person that completely ignored that 58% jump in performance for that use case?

LMAO this dude is blind or on some nice drugs, doesnt notice a perfectly acceptable/average enthusiast setup (High end intel CPU, crossfire 290x) and just ignores 58% improvement. Some people just cant be helped.

Warning issued for personal attack and inflammatory language.
-- stahlhart

 

[MutantGith]

That makes even less sense. If anything in your system is taking 14.02 ms to deliver a frame you can't have 116 FPS unless they're doing some space-time folding.

I have a feeling that that disparity might be due to looking at the average FPS vs the average frame time. If there are a lot of pauses or hiccups due to loading (beginning of level or new effects loading) that could spike the average frame times, without severely tanking the 'reported' frame rate, depending on how the FPS figure is being reported.

One of the big problems with spot data being reported. Hopefully, sometime veeeeeery soon, real full blown analysis will be done.

On a side note, I grabbed the StarSwarm Demo, AMD card, no Mantle driver. Extreme settings on a benchmark..... 6.5 FPS. Dropping the settings to the low preset, .....wait for it.... 11 FPS.

In percentage terms, a massive increase. Neither are playable (as much as that applies to a benchmark). Interestingly, other RTS titles with a respectible, detailed number of units run nowhere near as poorly, and those are actual games that can be interacted with.

So, ... congratulations on recreating FurMark? I'm curious as to how the results come in when Mantle drivers are out. I'm also not sure how much designing a game that completely hammers a rendering weakness in a different API as though it were its main purpose in life is a great example.

I'm very hopeful about the potential inherent in an RTS/RPG engine designed from scratch, but I'm not 100% on board with a system that is going to really only be viable if you own a specific combination of hardware and software.

 

[Midwayman]

Who in the world would pair those? So basically Mantle wasn't made for enthusiasts, yet it was beat to death on this enthusiast board.

If you're doing extremely high frame rates for VR or a strobing low persistence display you'll probably see some gains.

 

[Imouto]

So the idea is to spend $1000+ on GPU hardware to use a $100 CPU? I just dont see it.

The idea is removing CPU bottlenecks. If you haven't noticed with Mantle you can cut CPU frame times in about half for a 3960X.

But some people being so obtuse in this thread is not surprising at all.

 

[chimaxi83]

Looks like Mantle may be able to save people CPU/MB upgrades and allow a sizable performance increase with just a video card. Not bad :thumbsup:

 

[AtenRa]

So the idea is to spend $1000+ on GPU hardware to use a $100 CPU? I just dont see it.

According to your sing, you are already using a $200 CPU with $700-800 GPUs. Although nobody will pair a $1000 GPU with $100 CPU, with Mantle you will be able to get 90-95% of your $300-500 GPU performance using a $100 CPU. Something that is not possible today in modern high end games like BF4 MP.

 

[Imouto]

I have a feeling that that disparity might be due to looking at the average FPS vs the average frame time. If there are a lot of pauses or hiccups due to loading (beginning of level or new effects loading) that could spike the average frame times, without severely tanking the 'reported' frame rate, depending on how the FPS figure is being reported.

You have all the info on screen. Min, avg, max, current. It fits perfectly.

EDIT:

I corrected my previous post. Repi answered my tweet saying that those are individual GPU times, effectively divided by the amount of GPUs.

 

[krumme]

What kind of performance are you seeing?

Well it varies wildly. On my i5 ib 4c 4t at 4.1Ghz and a 7970 ghz i get from aprox 4 to 70 fps on extreme setting on most of the modes. Surely its drawcall limited, and needs a low level api like Mantle. You have to lower detail to medium og lowest level to get consistent good fps with the used number of units.
But never the less on dx its still very impressive what they have done. Very special and different for a rts thats for sure.

 

[Carfax83]

OC to 3.5. that's what they ran their 3960x at.

OK, I tested at stock clocks, which is 3.2ghz but the CPU was running at 3.5ghz in that scene due to turbo boost..

Honestly, I guess I'm surprised, but there was very little performance difference between 4.5ghz and 3.5ghz, which leads me to conclude that I'm not particularly CPU limited. The frame rate was steadier at 4.5ghz, and slightly higher, but it definitely wasn't a drastic difference. I stayed above 100 FPS most of the time.

I suppose the Frostbite 3 engine being well multithreaded does perform admirably on Intel hex core processors.. I monitored CPU usage during that stage, and the game was using at least 4 cores, plus hyperthreading judging by the task manager..

 

[ASM-coder]

Who in the world would pair those? So basically Mantle wasn't made for enthusiasts, yet it was beat to death on this enthusiast board.

So even though I build PCs, if I try to get the most performance out of the lowest cost, I am not an enthusiast?

 

[DamnedLife]

OK, I tested at stock clocks, which is 3.2ghz but the CPU was running at 3.5ghz in that scene due to turbo boost..

Honestly, I guess I'm surprised, but there was very little performance difference between 4.5ghz and 3.5ghz, which leads me to conclude that I'm not particularly CPU limited. The frame rate was steadier at 4.5ghz, and slightly higher, but it definitely wasn't a drastic difference. I stayed above 100 FPS most of the time.

I suppose the Frostbite 3 engine being well multithreaded does perform admirably on Intel hex core processors.. I monitored CPU usage during that stage, and the game was using at least 4 cores, plus hyperthreading judging by the task manager..

Wtih Dx constraints higher OC yields less percentile increase in performance, that's why you have to insan OC for 4 way SLI or Crossfire.
Mantle changes those situations for multi gpus, there is no more constraint and you don't even have to insane OC, that hex core intel i7 was running at 3.5GHz only it's turbo boost nothing more and yields %55 increase.

 

[mindbomb]

the least impressive results come from $1000 cpu and $100 gpu, so you can more accurately conclude that it is not for low end graphics cards. The conclusion that it doesn't help high end CPU's is not correct imo.

 

[stahlhart]

Several of you in here are attempting to derail this discussion. Stop it now, or I'm locking this thread permanently.
-- stahlhart

 

[bystander36]

Mantle clearly helps CPU limited situations. Whether your are CPU limited depends on a LOT of different things:
Crossfire makes it more likely for you to get CPU limited.
Your resolution plays a big part on whether you are CPU limited.
The settings you use will play a part on whether you are CPU limited.
And of course your CPU plays a big part.

 

[el etro]

True, but not in BF4. The Frostbite engine uses it's own sound drivers.

Thanks!

So the idea is to spend $1000+ on GPU hardware to use a $100 CPU? I just dont see it.

The final idea is to bring the cost of PC gaming more attractive vs the $400 consoles.
The tests only shows $100 cpus paired with $550GPUs, but what about $200CPUs driving GPUs two times more expensive?

 

[bystander36]

The final idea is to bring the cost of PC gaming more attractive vs the $400 consoles.
The tests only shows $100 cpus paired with $550GPUs, but what about $200CPUs driving GPUs two times more expensive?

The only problem with that is if you play non-Mantle games, then you'd be hurting on CPU performance, which is most games.

 

[Venomous]

If you're going to release the bf4 patch, shouldn't the driver have been available already? I cant even run the oxide demo in mantle because no driver!

 

[ASM-coder]

If you're going to release the bf4 patch, shouldn't the driver have been available already? I cant even run the oxide demo in mantle because no driver!

I think we can all agree on that. Anxiously awaiting, I want to hear/see some impartial benchmarks.

 

[moonbogg]

58% improvement with crossfire???? WOW!!!!! Why arent more people excited about this? I really cant wait to get a second 290X and a 120HZ monitor now

Everyone in this thread who loves gaming enough to take note of the above, reply right now with a big, fat "YAY!" People should be yaying this, but they aren't.

 

[smackababy]

Everyone in this thread who loves gaming enough to take note of the above, reply right now with a big, fat "YAY!" People should be yaying this, but they aren't.

When I can get another 290 for anywhere close to MSRP, I might care. =(

 

[JimmyH]

Battlefield 4 mantle patch- check. Star Swarm Benchmark- check. grabs popcorn and anxiously awaits 14.1

Haven't been this excited for a beta since Quake 3 back when Q3test hit and like Voodoo Extreme & Bluenews were blowing up w/ expectation. My clan plays alot of BF4 and most rox I5 and all radeon and alot of crossfire because cryptomining and all I5s cpu limited. I7s seem pull away. Most in my clan run 1080p low setting on 64 man servers w/ 120 or 144hz monitors so this outta really help out. My 3570k sits at 97% to 100% all the time and 7950 not pegged out at my prefered settings.

 

[Shmee]

If you're going to release the bf4 patch, shouldn't the driver have been available already? I cant even run the oxide demo in mantle because no driver!

Agreed LOL. Though I am liking the other changes, such as the DMR buff.

 

[DamnedLife]

Battlefield 4 mantle patch- check. Star Swarm Benchmark- check. grabs popcorn and anxiously awaits 14.1

Haven't been this excited for a beta since Quake 3 back when Q3test hit and like Voodoo Extreme & Bluenews were blowing up w/ expectation. My clan plays alot of BF4 and most rox I5 and all radeon and alot of crossfire because cryptomining and all I5s cpu limited. I7s seem pull away. Most in my clan run 1080p low setting on 64 man servers w/ 120 or 144hz monitors so this outta really help out. My 3570k sits at 97% to 100% all the time and 7950 not pegged out at my prefered settings.

Mantle will really help imo

 

[Ertaz]

Do I build a dual 290X rig just to see what this mantle stuff is all about? Hrm.. Guess so.

 

[Shmee]

Do I build a dual 290X rig just to see what this mantle stuff is all about? Hrm.. Guess so.

Lol ya I myself am hoping to get another 290 when they are cheaper.