AMD Mantle Benchmark | DX11 vs Mantle API On Mid Range GPU


AMD’s shiny new API is out in the wild, already developers such as DICE are embracing Mantle and users who’ve opted for certain AMD GPU’s are taking advantage of the performance gains. But let’s say you’re torn between buying an AMD Radeon GPU or one of Nvidia’s Geforce lines – should Mantle be a deciding factor on your purchase, or is it merely a marketing tool and gimmick?

While Nvidia and AMD are promising full DX12 support, it’ll not be available to gamers until some point in the later half of next year (according to Microsoft), and while the API is promising, it’s a long wait. Mantle is available now. So after so much hype before and upon its release, the question is, does AMD’s Mantle live up to the expectation and what type of kit will be needed to take advantage of it?

What is the Mantle API?

Tempting as it is to jump straight into tests, it’s vital we’re all on the same page. We’ve discussed both DX12 and Mantle before, and indeed AMD have even given us an interview regarding Mantle, but we’ll go over the basics again here just for the sake of completeness. API is an acronym, standing for Application Programming Interface. Unlike a lot of technical names, this one actually tells you what’s in the can. The Application (in this case, games) require something to ‘talk’ (Interface) with the hardware on its behalf.

Why? Because if it didn’t, developers would have to code for the platform themselves, and would have to worry about interfacing directly with the GPU’s drivers. It’s a tall order, and would make PC development a nightmare. So now we have an API which sits between the game and the GPU’s driver. The game interfaces with the API (say DirectX), and DX then calls out to the GPU driver and says “Hey, I need you to draw this”. This is called an ‘abstraction’ layer. Effectively, it’s translating the desire of the game engine, but there’s a natural cost associated with all of this – it hides many of the GPu’s specific functions in the name of compatibility.

The two traditional API’s have been DirectX and OpenGL. Developers who’ve been making PC games have grumbled for some time that both of these API’s have certain issues – mostly their roots stem before multi-core CPu’s were the normal. So despite later versions of DX11 helping things somewhat, the problem with the rendering thread being only put across one CPU core hasn’t really gone away. While DX12 does fix this – it’s not here yet.


If you’ve been following coverage of either Microsoft’s Xbox One or Sony’s Playstation 4 (or better still, following our coverage – shameless plug) you’ll have phrases such as “Closer to the metal” and “low level API” burned into your brain. Mantle in this effect does just that – it’s a smaller, lighter API aimed at providing lower level access to specific GPU features, while improving multi core rendering, draw calls and finally is pretty easy to make games run on it.

Benchmarking Hardware and Software

For the graphics card, we’ll be using AMD’s R9 280. We’ve already conducted a full review of the card vs a GTX 760 with impressive results. For the rest of the test system, we’re using Windows 8.1, 16GB of DDR3 RAM, and two CPU configurations, both of which are set to stock. The first is an Intel I7-4770K, and the second is an Intel I5-4670K. The theory is we can get an idea how Mantle handles fewer CPU cores. For the software, we used the latest Catalyst drivers (14,7 BETA’s) and the latest patches (automatically downloaded) for both BF4 and Thief.

As one would expect, we’re targeting the resolution of 1080P – which let’s face it, is really the minimal resolution you’ll be wanting to game at on the PC.

Thief Benchmark Mantle Vs DX11

For our thief tests, all graphics options were cranked to their highest with the exception of SSAA (because it’s an extremely expensive form of Anti-Aliasing). The only settings changed between Mantle and DX11 was the renderer – in other words, simply flicking between Mantle and DX11.

PC Thief Benchmark on an AMD R9 280 with DX11 and Mantle

Starting things off with Microsoft’s tried and trusted DX11 and we were pretty happy with the results in Thief. With the R9 280 we hit 58.9 FPS, which is within the margin of error from our R9 280 full review (where we scored 58.8 FPS.) Not too bad, but let’s try messing about with the number of CPU cores. Dropping to a I5 4670K (stock) from our I7 we re-ran the exact same tests. This time we managed to achieve 53.8 FPS.

Now we re-run the entire series of tests again, this time using Mantle. Starting things out, we use the I7 and hit 70.1 FPS, (a slight dip of .3 FPS from our review, but within the same margin of error). Immediately we see a 20 percent increase from the DX11 results. Not bad considering all we had to do was tick a box. “What about with 4 cores?” you’re probably asking, and you’d be right to. 66.3 FPS, a drop of 4 frames per second. Not too bad,

For point of reference, using the GTX 760 with the I7 netted us 52.2 FPS. The results here are fairly clear – Mantle provides a tangible, real boost in performance. When overclocked, the R9 280 in the same Mantle vs DX11 tests reached 81.1 FPS for Mantle, and DirectX 11 62.8. Though the difference is tiny, it appears that Mantle scales a little better with the extra GPU speed in terms of raw FPS.

BattleField 4 Benchmark Mantle vs DX11

While BF Hardline is being developed, it’s still in Beta and isn’t fair to compare results on something still in development. To this end, we’ve stuck with BF4, which was one of the first titles announced to support AMD’s new API. DICE had put out several blog posts regarding its development and excitement over AMD’s tech, but gamer’s had to wait for a patch (dubbed by some as the Mantle patch) to play it. What are the results?

Firstly, we’re playing BattleField 4 with all of the graphic settings set to ultra, with the exception of MSAA, which is set to two times rather than four. If you’re overclocking the R9 280, x4 is fine, but we’re testing on stock for the moment and x2 dips just a little too much, especially with DX11.

Battlefield 4 maxed out on the R9 280 with Mantle enabled for benchmarking

Just like our Thief tests, Battlefield 4 is consistently faster using Mantle than DX11 manages. It’s important to understand that DX11 doesn’t do a bad job – frankly speaking the Radeon 280 using mantle does a better job with frame rate and visuals than either the PS4 or X1 can muster (the same could be said for Nvidia’s mid range cards too). Instead, think of Mantle as providing additional icing on the cake. Clearly in a game like BF4 frame rates can vary quickly – but if you view the results long enough you’ll see that the frame rate for Mantle is higher. At times up to 10 FPS.

Another important factor is that Mantle (at least from our initial testing – we’ll do more soon) seems to scale better with GPU overclocking. Considering the R9 280 overclocked well in our review, this is only a benefit.

Mantle Conclusions

The big question is – if you’re buying a GPU, especially in the mid range, should you make Mantle point to consider? Well, the most important factor would be – which games are coming out that are planning on using Mantle? If there’s a few (for example BattleField Hardline, Mass Effect, Star Citizen) then Mantle is certainly a nice addition.

I feel both AMD and Nvidia have rather nice exclusive tech, I’m a big fan of Hardware Physx… I like my extra smoke and debris damn it. But, arguably Mantle is much more useful – you’ll actually be able to see a nice performance boost in your games. In the mid range this is crucial, but even if you’re the proud owner of a R9 290X, would you say “actually, I’d rather not have extra performance?” of course, the answer is no. Particularly when dealing with higher resolutions.

Then again – DX12 is being worked on, and how successful Mantle will be after the release of Microsoft’s DirectX 12 is an unknown factor. It’s arguable that AMD will have had a nice lead time by then, and if Mantle / DX12 code is similar enough, who knows? We’ve also got Valve and their SteamOS, a variant on Linux. Though neither AMD or Valve are saying much, the prospect of Mantle on Linux is quite exciting. It would make Windows to Linux ports quite easy (or at least easier than they are now).

If you doubt Mantle’s performance, don’t – games do perform better and smoother. After all the hype and speculation it’s nice it’s not a huge let down. Both DX12 and Mantle will gain in importance in future years – as the demands of game engines gets on the increase.

Buying the R9 280

If you’re thinking of buy the Radeon R9 280 (or other card) please consider using this affiliate link to help us out! Buy Radeon R9 280 (US) | UK Link

Comments are closed.