AMD’s performance-boosting ‘Mantle’ PC gaming tech launches with Battlefield 4 – PCWorld
On Thursday, AMD’s highly anticipated Mantle API technology is expected to go live, courtesy of a Mantle-enabling Battlefield 4 update and a new set of AMD Catalyst 14.1 beta drivers promised to go live sometime today.
Mantle, in case you missed it between all the buzz over the new R7 and R9 series Radeon graphics cards, is an application programming interface (API) that’s supposed to better leverage AMD hardware for increased performance—provided you’re running an AMD graphics card or APU.
There are two APIs traditionally used for games: DirectX and OpenGL. These high-level APIs are great for game developers because they run with few differences across a broad spectrum of hardware—Nvidia, AMD, Intel, whatever, DirectX and OpenGL can handle it.
But this creates a jack-of-all-trades-master-of-none scenario. When you’re optimizing a game for DirectX or OpenGL, you’re optimizing the game to run across a swath of hardware.
AMD’s Mantle is built on specificity. Think of it like a tailored suit versus something picked up off the rack. Mantle says “Hey, you’re using AMD hardware to power your graphics? I know exactly how to take advantage of that architecture. That’s all I’m good at.” As a result, you get some performance boosts just because the game is better optimized for the machine you’re using instead of, well, every machine.
AMD says Mantle users will see the biggest performance increase in CPU-limited set-ups—any machine where a low- to mid-tier CPU is bottlenecking graphics output. Those running discrete graphics cards will see a smaller change, though AMD still calls it “noticeable.”
How noticeable is noticeable?
It’s hard to know what kind of performance increase you’ll see on your personal rig because AMD and EA’s own numbers vary so widely. Performance seems to increase more in systems that are bottlenecked from CPU performance, while the gains are much more modest in systems where the graphics card is weaker than the processor.
On the high end, we have EA’s incredibly expensive dual-R9 290x 4GB rig, with an i7-3970x Extreme processor clocked at 3.5GHz. It’s a beast of a machine, and with two powerful graphics cards the processor struggles to keep up. With that setup, EA reports the total Mantle performance increase in Battlefield 4 be to a whopping gain of nearly 50 frames per second (from 78 up to 122).
Likewise, numbers provided by AMD show a 40 percent gain in a system pairing a ho-hum AMD A10-7700K APU with a Radeon R9 290X.
But let’s say you’re running an i7-4960x with just a single R7 260x card (a somewhat absurd but still illustrative example provided by AMD). Now we’ve entered the realm of GPU-bound systems—that lone R7 isn’t pumping out enough power to bottleneck the processor. And your reward for installing Mantle in this case? A mere 2.7% increase in performance.
In other words: Don’t necessarily expect miracles. Know what your hardware is, know where your system’s bottlenecks are, and don’t be surprised if Mantle fails to be a magical cure-all for your PC’s gaming performance.
Someone still loves you, DirectX
Mantle still faces an uphill battle for adoption, even if it does improve performance.
Developers aren’t going to suddenly stop making most every game in DirectX—at least, not yet. There are too many PC setups out there, and Nvidia and Intel are (much to AMD’s dismay) still the dominant graphics card companies as far as PC gaming is concerned.
While AMD says Mantle is an open platform—in other words, “Sure, Nvidia could take advantage of our new API”—don’t expect to see Nvidia crawling to Mantle’s door anytime soon. And that’s not just because of longtime rivalry between the two companies; the architectures behind AMD and Nvidia’s cards are so different that it’s doubtful Nvidia could even use the system without huge rewrites.
As a result, developers will have to opt in to Mantle, including it alongside something more popular like OpenGL or DirectX. Otherwise they’d lose a huge section of the market. Mantle will presumably always be an “also-ran” API instead of a primary focus, unless AMD’s successful push into the console market is a much larger factor than anticipated.