AMD Mantle vs. DirectX Benchmarks with Battlefield 4 and Star Swarm

As we mentioned a few days back, AMD’s been working on a new set of drivers for its GCN-based graphics cards and APUs, which enable Mantle and a number of other features as well.

Just to quickly recap, the Catalyst 14.1 betas (which were just released and are available here) are the first publicly available drivers from AMD that support Mantle, AMD’s “close to the metal” API that’ll let developers wring additional performance from systems equipped with GCN-based GPUs. The new drivers, however, will also add support for the HSA-related features introduced with the recently released Kaveri APU, and will reportedly fix the frame pacing issues associated with Radeon HD 7000 series CrossFire configurations as well, at resolutions above 1600p. AMD introduced frame pacing into their graphics drivers a few months ago, but resolutions above 2560x1600 were not supported, which meant Eyefinity or 4K monitor owners were out of luck. In addition, the Catalyst 14.1 beta drivers will reportedly fix frame pacing issues with dual-graphics (APU + Discrete) configurations too.

We had a chance to play around with the Catalyst 14.1 beta drivers on a couple of different system configurations to see how using Mantle affected performance. We tested two applications, the popular game Battlefield 4 and a new benchmark called Star Swarm, which leverages the Nitrous 3D engine. You can read more about Star Swarm and see a video of it in action here.

Battlefield 4 using Mantle

We tested the Catalyst 14.1’s with Mantle on an AMD A10-7850K Kaveri APU-based system, with and without discrete graphics, and with a Core i3-4330-based system equipped with a Radeon R9 290X discrete GPU. Both systems were outfitted with 8GB of DDR3-2133 system memory and were running Windows 8.1. We chose to test on these particular configurations because Mantle will show the greatest benefit in CPU bound workloads. There are some efficiencies on the GPU side as well, which will increase performance in GPU bound workloads too, but the greatest gains will be realized when Mantle can minimize the CPU overhead associated with DirectX, freeing up the CPU to process more data and keep the GPU fed. AMD has an excellent explanation as to how and why Mantle affects performance in this blog post.

As you can see in this first batch of numbers from Battlefield 4, Mantle has a significant impact on performance. On the GPU-bound A10-7850K, when tested with its on-processor Radeon R7 graphics, BF4’s performance increased roughly 10%. Pop in a powerful discrete GPU, however, and the gains are much larger. When the A10-7850K was paired to a Radeon R9 290X, utilizing Mantle resulted in a 28% increase in performance over DirectX. Performing the same test with that Radeon R9 290X installed in the Core i3-4330-based system showed Mantle outperforming DirectX by about 13%.


In the Star Swarm benchmark, the performance differences between DirectX and Mantle are much more dramatic. When testing Star Swarm with its built-in RTS demo on the Core i3 + R9 290X system, Mantle outperformed DirectX by margins of 181% (low quality), 168% (medium quality), and 106% (extreme quality).

The A10-7850K-powered system showed similar sized performance gains with Mantle. On the A10-7850K system, Mantle outpaced DirectX in Star Swarm’s RTS demo by margins of 78% (low quality), 92% (medium quality), and 70% (high quality). The Star Swarm demo is a little too much for the A10-7850K’s integrated Radeon R7-class GPU, but at the low quality setting, using Mantle at least makes the animation into playable framerate territory.

Mantle is still in its very early stages and Battlefield 4 and Star Swarm are the only applications currently available that leverage the technology. In time though, we should see more games that use Mantle and we suspect AMD will be able to further optimize its software to improve performance even more. Time will tell if Mantle makes a significant impact on PC gaming, but based on these limited tests, it’s clear that the low-overhead associated with the API can result in some monster performance gains in the right situations.

What’s also clear is that Microsoft hasn’t done enough to optimize graphics performance on the PC. DirectX has evolved greatly over the years and has enabled developers to create some fantastic looking games. But with overheard so high, there’s too much performance being sacrificed, likely for no good reason. Hopefully, at the very least, Mantle convinces Microsoft to do some tuning to DirectX to push the PC platform even further ahead of everything else out there.

Via:  HotHardware
HenryJones 10 months ago

AMD 4 lyfe

MarvinVelasco 10 months ago

sounds good!!

RonKurkowski 10 months ago

If mantle was used in more games I would be selling my 780 Ti's but for now they will remain.

TRoemhildt 10 months ago

Mantle 1.0 vs. DirectX 11... not bad at all. Even in games not initially built for Mantle, they're yielding a good $20-$30 worth of performance for FREE. Even more if you count the smaller amount of money needed to spend on the CPU. NVidia, if you adopt this I will buy one of your cards for the first time since 2007 as a reward. Please don't let me down, NVidia.

I wonder how Mantle 11 will fare against DirectX 22!

NickAaronson 10 months ago

Wow, above 30fps at ultra settings on an apu alone? That's damn impressive.

RCastaneda 10 months ago


basroil 10 months ago

In other words, Mantel really sucks when the developers actually programed the game properly in the first place. It was supposed to allow developers to use ten times the calls, but looks like good developers don't even need to use those calls. The only place where the API actually works is with physics and graphics on one GPU, and only because the developers didn't do it properly to begin with.

Mag 10 months ago

No, not really. On consoles the high number of batch count is very common, but during the directX optimalization phase, they have to radically reduce to fit in the "sweet spot" of the directx environment. With mantle you dont have to reduce that much, so free up some time and money. actually the devs say that is cheaper to implement mantle than optimize the code to directx (so this is not about the developer programming skills, its about the environment difference)

And another thing: not every dev can start developing for mantle. since the mantle api gets rid of a lot of failsafe mechanism, the responsibility of the devs are lot stronger. So the amd gives mantle only to the very good programmer teams

PafuncioSembranelli 10 months ago

How can it sucks, it´s for free and gives a really nice performance boost!

JaroslavDolej 10 months ago

I have actually kind of worse performance with mantle (Radeon 7750 + I3 3220), much higher CPU spikes making it very unpleasant to play. GPU performance is a bit smoother than with directx, but its ruined by bad CPU performance.

CDeeter 10 months ago

Quote from article "We tested the Catalyst 14.1’s with Mantle on an AMD A10-7850K Kaveri APU-based system"

RicoFrost 10 months ago

I;ll wait to see how it runs with games I play, I upgrade my video card every 1-2yrs anyway.

Looks like only BF4 is supported currently and its a 10% increase which is nice but not worth upgrade from a 780.

JakeMetica 10 months ago

Mantle will bring even more great things in the pc world.. this is the start.. GO GO GO AMD!!! KEEP IT UP!!!

Post a Comment
or Register to comment