Another day, another Kaveri article! If you haven’t yet, be sure to first read our full Kaveri review, then perhaps move on to see how it performs when overclocked to 4.7 GHz with 1020 MHz GPU. Also if you’re curious, you may want to check out how gaming performance scales with memory speed, up to DDR3-2400.
Today is extra exciting, because we finally get to see some real Mantle performance figures. Battlefield 4 has been updated to support the Mantle API, but we’re still waiting for a public beta release of the Catalyst 14.1 driver that will enable its use. However we do have a private beta driver from AMD that will allow us to take a quick early look at Mantle performance on the Kaveri APU in Battlefield 4.
This driver feels very early still, which is probably why it’s not publicly available yet. However with some effort, we were able to update everything, and get it working to show you some benchmarks. We’ll be using our standard BF4 benchmark run, comparing it to DX11 results on a stock Kaveri 7850K APU with DDR3-2400.
|APU Specs||Radeon 7850K CPU
3.7 GHz Base
4.0 GHz Turbo
Radeon R7 Series IGP
AMD A88X Chipset
|Memory||8GB (2x4GB) Radeon Memory GAMER Series DDR3 2400 11-12-12-31|
|Hard Drive||OCZ Vector 256GB
|Motherboard Drivers||Catalyst 14.1 b1.5 Chipset
Catalyst 13.12 SB
|Video Drivers||Catalyst 14.1 b1.5|
|Operating System||Windows 8 Professional x64 RTM
All updates as of Dec 1, 2013
Benchmarking Battlefield 4 with Mantle is a little bit different, since FRAPS will not work without DirectX calls. Thankfully, the developers have put in a feature that records the frame time of every single frame delivered, similarly to how FRAPS works. Unlike FRAPS though, it does not support an auto-timed benchmark, or hotkeys. Instead, the command has to be entered into the console, timed manually, then disabled in the console again. It’s not ideal, but we were able to get consistent results using our standard BF4 benchmark run. I think I could do this in my sleep now…[youtube=http://www.youtube.com/watch?v=4BAB7Yg1QLw&rel=0]
All benchmarks are run 5 times, to find a standard performance baseline (a mean of frames delivered within the 60 second run). Once that is found, the run that delivers the amount of frames closest to that mean is used to represent performance.
Battlefield 4 on Kaveri with Mantle – 720p Low
First we’ll do our Low Quality run. This is for people who prefer frame rate over visual effects. For me, this is the best way to play Battlefield 4 on Kaveri – I find it very difficult to aim at 30 FPS with no assists. Especially the way BF4 plays, with the gun moving around on its own when looking down the scope, and bouncing all over the screen when you are being fired at. This is hard to overcome with a low frame rate.
Frame time charts can be clicked on for more detail.
Well there you go, an instant performance gain of about 10%. Also, the 99th percentile frame time (the top 1% of slowest frames, which is what makes games feel ‘stuttery’ on occasion, and can throw off Vsync) was improved to just over 17 ms.
Battlefield 4 on Kaveri with Mantle – 1080p Medium
We’re testing this setting for people who want to crank up the resolution and settings, and don’t mind playing at a lower framerates to do so. This might be suitable for casual living-room gaming with a controller, with aim assist turned on. This will give you an experience that looks better than current gen consoles like the PS4, but plays more like a previous gen console like the PS3.
This is where the hardware simply hits a wall, it seems. While the overall FPS is improved by about 6.3%, the actual 99th percentile frame time went. To see why, simply look at the actual frame times – Mantle seems to introduce these 50-100 ms frame spikes that occur at regular intervals. As for whether these spikes are noticeable in game play, I can’t really say. To me, 28 FPS feels bad whether there are frame spikes or not. These spikes were consistent, and occurred in every test we ran at this resolution.
We have seen some pretty lofty predictions on what kind of performance increase Battlefield 4 would receive from Mantle. On the discrete GPU side of things, if you are using an R9 290X with a slower CPU, you will probably see the best gains possible. For Kaveri, it’s nice to get a 10% performance boost at 720p. This is something that we sometimes see with Crossfire setups when a new driver is released, but for a 512 shader integrated GPU to get this kind of increase is pretty impressive.
Overclocking should help, as it always does, and we’ll have to take another look down the road when a more complete driver is released. For now, we have an idea of what Mantle will do for Kaveri in Battlefield 4 at least.
It was troubling to see these weird spikes that happened at regular intervals when Mantle was used, but we can probably chalk that up to the drivers being such an early beta.
Overall, it’s easy to recommend using Mantle and the new Catalyst driver when it comes out, on your Kaveri APU. If you play at 720p with reasonable settings, you can expect a nice boost in performance for free. If you upgrade to a discrete GPU down the road, things will get even better.