Quantcast
 


With Catalyst 14.1, AMD Unleashes Mantle

Author: SKYMTL
Date: January 29, 2014
Share |

AMDís Mantle has been discussed, dissected and theorized about since it was first announced last year. Now, with the release of the Catalyst 14.1 Beta, weíre about to see the first Mantle libraries packaged into a beta driver alongside plenty of other improvements. It wonít be available to the public just yet but expect it to be live in the next few days.

In and of itself the inclusion of Mantle compatibility into a driver canít be considered a drastic step forward since, up until today there simply weren't any applications which supported it. Without being able to experience AMDís marketing claims first-hand, the real benefits of this new low-level API remained nothing more than wistful thinking. Unfortunately, that portion of the equation ran into countless delays with Mantleís primary showcase, Battlefield 4, going through teething problems which ultimately destroyed a carefully crafted availability timeline.

This situation is about to change in a big way: Dice is rolling out their long awaited BF4 Mantle patch at 4AM today. At least thatís the plan right now but with the hiccups Origin is known forÖ.you never know.

As we have come to know Mantle a bit better over all these months, our understanding of its capabilities has evolved so now is as good a time as any to bring you up to speed.


A Quick Mantle Refresher


Before we go on, letís go through a quick refresher about what Mantle is and what it isnít. Unlike what some would have you believe, AMD is adamant that its primary design intent isnít to facilitate the porting of console games to the PC, nor is Mantle officially associated with consoles in any way. However, it can help game designers optimize their games in a more straightforward manner and that's a big deal as PC development time is at a premium.

Mantle acts as a low level facilitator between an application and the graphics subsystem, essentially replacing the primary API (be it DirectX or OpenGL) and display driver with Mantle equivalents, allowing developers clearer access to hardware functionality. This reduces software overhead and allows AMDís GCN architecture to reach its full potential, which brings us to our next point: Mantle is only compatible with GCN graphics cards and APUs.


Another interesting addition to Mantle is built-in compatibility with Crossfire. Instead of relying on AMD to release supporting drivers or application profiles, developers will now have full control over the multi GPU environment so resources can be shared more effectively. While this may not sound important, it could become a defining feature since Crossfire support is virtually guaranteed from day one of a gameís availability.

One of the most important elements associated with any revolutionary toolís launch is a broad support structure from developers. From our off-the-record conversations with several leading game development studios, many feel that Mantle has the capability to streamline the development process to such a significant degree that it could offer significant cost savings. If anything, that alone could spur development.


A New Understanding of Mantleís Offerings & Limitations


As AMD and developers have worked with Mantle, a number of realizations have come to light about its capabilities. First and foremost, todayís APIs boast a large amount of validation overhead which leads to CPU cycles being wasted and multi-core CPUs going underutilized. This potentially bottlenecks the communication process between the CPU and GPU which means in certain instances the graphics processor canít reach its peak performance.


Right about now youíre probably wondering why weíre talking about the processor rather than AMDís graphics cards. Isnít Mantle supposed to augment GPU performance in upcoming games? According to AMDís newly released documents, the answer to that is "yes"......but not in the way youíre expecting. By streamlining the way a game interacts with hardware Mantle allows CPU resources to be used more effectively which means, contrary to popular belief, the tertiary benefits don't directly target the primary graphics engine. Indeed, Mantle still uses the DirectX High Level Shading Language (or HLSL) so itís still tied at the hip to those inherent API-level limitations.

To paraphrase AMDís documents, Mantle uses several techniques to achieve more efficient CPU performance. They include:

- Low-overhead validation and processing of API commands
- Explicit command buffer control
- Close to linear performance scaling from recording command buffers onto multiple CPU cores
- Reduced runtime shader compilation overhead

In plain English, this means Mantle can offer extreme benefits in areas which feature CPU bottlenecks but will feature more limited performance increases in GPU-bound scenarios. This is an important distinction to make and AMD gave us a few examples of how systems will react. The results are eye-opening to say the least:


While these charts are based on numbers provided by AMD, they still represent two extremes. On one hand, Mantle can provide some excellent framerate increases when an entry level CPU is used alongside a higher end graphics card. However, once the graphics card becomes a bottleneck, Mantleís benefits become nearly non-existent.

The repercussions of this are multi faceted. Gamers with lower end processors paired up with better GPUs will be the primary beneficiaries, as will folks with high end systems playing at lower resolutions and detail settings (why they would do this in the first place is beyond us though).

Unfortunately, enthusiasts with overclocked processors and expensive graphics cards may be left out in the cold. Anyone who maxes out detail levels while maintaining playability (causing a GPU-limited scenario) wonít experience much in the way of improvements either.

So who or what is Mantle really targeting? At least initially gamers who invested a ton of money in their AMD GPUs and associated CPUs wonít feel any difference. We do however see this being a trump card for users of APUs and Intelís Pentium, i3 and i5 processors. It will be interesting to see where the tipping point between price and Mantle benefits lies, but weíll tackle that once more supporting games are available.

One of the largest limitations right now is GPU support on AMDís side. The 14.1 drivers wonít support Mantle on older GCN cards like the HD 7000 / HD 8000 series, the R9 280X and R9 270X. In addition, AMD has warned that multi GPU configurations may experience intermittent stuttering. This is a beta so donít expect miracles.


The Future of Mantle


While Mantleís purported benefits are limited to whatís arguably a large portion of todayís gaming market while somewhat shunning enthusiasts, there are many reasons to be excited about how it will develop in the future. Currently, thereís only single game which supports it through an add-in patch and therein lies the potential issue for anyone reading too much into these results: Battlefield 4 wasnít designed to natively support Mantle. Upcoming titles like Thief may have Mantle support built directly into their game engines which could facilitate other GPU-focused optimizations.

One example of AMDís much-ballyhooed Star Swarm demo (one of those ground-up Mantle developments) shows excellent improvements across the board due to its use of mass CPU-targeted AI algorithms. Is it realistic to expect Oxide Gamesí Nitrous Engineís optimizations will make their way into all games? Absolutely not but it shows some great potential and could at least point the way towards the future of RTS games.

Speaking of additional areas of improvements, Mantle hasnít been crippled from a purely graphics standpoint. Quite the opposite actually. AMD has built a number of rendering optimizations into their toolkit which developers will be able to take advantage of as time goes on.

- Reduction of command buffers submissions
- Explicit control of resource compression, expands and synchronizations
- Asynchronous DMA queue for data uploads independent from the graphics engine
- Asynchronous compute queue for overlapping of compute and graphics workloads
- Data formats optimizations via flexible buffer/image access
- Advanced Anti-Aliasing features for MSAA/EQAA optimizations

If you take anything away from this quick article, it should be this: Mantle is still in its infancy and the driver set is very much a beta. As more games are released, weíre sure to see a clearer picture emerge as to where this technology will take games and who it will ultimately benefit.

From Crossfire support to performance uplifts, as with all APIs, Mantle relies on the developers to take full advantage of what it brings to table. We take this to be a sign of real hope since close to the metal programming is what the creators of our favorite games have been asking for and if that helps along PC gaming, weíre all for it. Expect more Mantle-related articles in the coming weeks.
 
 

Latest Reviews in Video Cards
December 8, 2014
AMD and drivers have had a contentious relationship in the past but Catalyst Omega is supposed to change that in a big way. The new "Omega" drivers offer better stability, higher performance and more...
December 2, 2014
The ASUS GTX 980 STRIX is fast, quiet and only costs a few more dollars than NVIDIA's reference design. It may be the perfect high end graphics card....
November 23, 2014
EVGA's GTX 970 FTW is one of the fastest sub-$400 GPUs on the market and when paired up with the new ACX 2.0 heatsink, it also happens to be one of the quietest....