BF4 AMD Mantle Video Card Performance Review Part 1

AMD's Mantle API has been with us for just over a month now, and we have strapped a variety of video cards to the test bench to see what real world differences are being delivered to gamers within Battlefield 4. We will compare D3D11, Mantle, on various GPUs, looking at highest playable settings, frame times, and discuss our experiences.

Introduction

The long awaited, and much delayed, AMD Mantle API has been available in Battlefield 4 for a month now. We have taken this time to thoroughly investigate AMD's Mantle API within Battlefield 4 on AMD's complete lineup of current video cards. We did encounter an issue though, explained at the bottom of this page you need to read before continuing.

We have tested performance differences between DirectX 11 (D3D11) and Mantle. We have tested performance differences between D3D11, Mantle, and NVIDIA's complete lineup of video cards running in D3D11. We have tested performance in D3D11 today, compared to Battlefield 4's initial release at the end of October 2013. We've evaluated highest playable gameplay experiences differences from the game's launch to now. We have a lot of testing and real-world gameplay that has gone into this evaluation to give you our experiences and opinions about AMD Mantle in Battlefield 4.

We set out in this evaluation trying to answer some important information as it pertains to Mantle and BF4. Primarily, we wanted to know in which price segment of AMD GPUs does Mantle benefit gamers the most? Which GPUs receive the best performance increase, and what type of gameplay experience can gamers achieve with Mantle? Does Mantle favor high-end or low-end cards?

Let's back track just a little, to November 17th, 2013 when we published our Battlefield 4 Video Card Performance and IQ Review. In that evaluation, the game was brand new, drivers were newly optimized for the game, and all we had available to us was the DirectX 11 (D3D11) API within the game. We tested all the latest video cards from AMD and NVIDIA at the time. Our overall conclusion was that, at the time, AMD GPU equivalents compared to the competition were faster and better performing in the game.

However, as time has progressed, we've found NVIDIA's newer GPUs to be catching up, or providing the same gameplay experience compared to AMD. Therefore, it will be interesting to see how AMD's GPUs and NVIDIA's GPUs lineup now in D3D11, and then how Mantle will change the landscape.

AMD Mantle

Mantle was originally announced and briefed with the release of the R9 and R7 series of graphics cards back in October of 2013. AMD and DICE have made many positive statements about what Mantle will bring to PC gaming and the advantages it has over Microsoft’s DirectX API.

Many of the improvements that have been publicized paint Mantle as not the graphics or GPU boosting API we all thought at first, but rather a way to reduce CPU overhead and limitations with DirectX. Thus, this reduction in CPU overhead means greater performance in CPU limited situations. However, we still want to know how that affects graphics cards, from low-end, to high-end in this game. That is the goal of this evaluation today.

We published a preview of Mantle performance shortly after launch to compare DirectX 11 performance to Mantle performance to see if it was headed in the right direction. We observed increases in performance across the board in the testing scenarios that we ran, and were happy with the direction that Mantle was heading.

However, at the same time, we were a bit disappointed in the improvements brought at the high-end. We have now spent the past several weeks working with Mantle, measuring its performance across a wider variety of hardware and gathering data that we feel is most relevant to our readers.

To learn how to use AMD Mantle and the built in frame time capture abilities in the game read this blog: Mantle renderer now available in Battlefield 4.

Mantle’s Teething Issues

To say our experiences with AMD Mantle has gone without issue would be unfortunately not true. We did have issues with AMD Mantle that directly related to certain specific AMD GPUs. At the onset of this evaluation we had planned to show you AMD's complete lineup of video cards from top to bottom, in one evaluation. However, we encountered issues with AMD's mainstream to low-end GPUs that caused us to have to switch tracks on presenting you this evaluation.

We ended up having to split up our evaluation into two parts. Today, we are bringing you Part 1 which shows the comparison of the high-end cards, which worked without issue. Once AMD and EA fix the issue with the mainstream to low-end cards, then we will bring you Part 2 of this evaluation which will focus on low-end GPU performance comparisons.

The issues we were finding are that the 270X and the 260X were performing significantly worse under Mantle at settings that averaged 60FPS on the DirectX 11 API (1920x1080 at Ultra quality and Low AA Post for the 270X). Specifically, after a couple minutes of game play at the expected level of performance in Mantle, the bottom would fall out and it would become an unplayable lagfest.

This issue persisted through a few different settings that we tested on each card and we then got in contact with AMD to investigate the issue further. AMD was eventually able to replicate our problem and determined that it was a game side issue that would require DICE to patch to fix the issue on those GPUs. At that point, we decided to exclude those cards from this evaluation and we will revisit them once the issues have been resolved.

We also encountered an odd issue with Eyefinity configurations on our personal gaming rigs. For both the Catalyst 14.1 and 14.2 Beta drivers, we were not able to get all three monitors (3x Dell 2412m) running at the Windows desktop, two connected via DVI and the third via DP. After switching things around a bit we found that using a passive DP to DVI cable instead of the DP cable to the third monitor, all three would light up and we were in business. This is most likely just another driver issue to add to the mix with Catalyst 14.1 and Catalyst 14.2.