Update: Our Black Ops III graphics optimization guide is now live for per-setting analysis.
We tested using our 2015 GPU test bench, detailed in the table below. Our thanks to supporting hardware vendors for supplying some of the test components.
The latest AMD Catalyst drivers (15.11 beta) were used for testing, including the Battlefront and Call of Duty patches. NVidia's 358.87 drivers were used for testing, including the Battlefront and Call of Duty patches. Game settings were configured to "Extra" (effectively “Ultra”), "High," and "Medium" presets at 1080p, 1440p, and 4K resolutions. Once we determined which settings provided a reasonable level of load for appropriate video cards, we forged forward testing those configurations on our suit of GPUs. "Medium" VRAM tests were conducted with Dynamic Shadows and Subsurface Scattering disabled. Anti-aliasing was configured to FXAA for these tests.
We differed our test for VRAM and graphics settings benchmarking, spawning 17 NPCs (allied) and running a fixed course alongside the AI, creating more of a worst-case scenario. Some tests were conducted in singleplayer, in which event they were executed on the first campaign level -- the most intensive, from our findings -- and done on a fixed path.
|GN Test Bench 2015||Name||Courtesy Of||Cost|
|CPU||Intel i7-4790K CPU||CyberPower
|Memory||32GB 2133MHz HyperX Savage RAM||Kingston Tech.||$300|
|Motherboard||Gigabyte Z97X Gaming G1||GamersNexus||$285|
|Power Supply||NZXT 1200W HALE90 V2||NZXT||$300|
|SSD||HyperX Predator PCI-e SSD||Kingston Tech.||TBD|
|Case||Top Deck Tech Station||GamersNexus||$250|
|CPU Cooler||Be Quiet! Dark Rock 3||Be Quiet!||~$60|
Average FPS, 1% low, and 0.1% low times are measured. We do not measure maximum or minimum FPS results as we consider these numbers to be pure outliers. Instead, we take an average of the lowest 1% of results (1% low) to show real-world, noticeable dips; we then take an average of the lowest 0.1% of results for severe spikes.
Resmon was used for monitoring basic system resources. GPU-Z and Afterburner validated GPU-specific activity.
Here are two screenshots of all settings present. The settings applied below do not necessarily reflect test settings (see text in methodology sections for that):
Call of Duty: Black Ops 3 VRAM Consumption – Multiplayer & Singleplayer
Our multiplayer testing featured 17 allied bots and a reproduced test run for every settings configuration. We used the Redwood map, having previously found that the map is among the most intensive multiplayer scenarios, and ran a full course through the environment.
As shown above, running 4K with “Extra” settings forces more than 8.6GB of memory consumption on our Titan X – which may actually present that our 4K/extra benchmarks on the 980 Ti could have been somewhat VRAM-limited. We will conduct additional testing out of pure interest to determine if there is more of a VRAM bottleneck than a GPU bottleneck.
Dropping to 1080/extra still sees 6.3GB saturation of our GDDR5 memory. Running 1080/high drops substantially – primarily because of texture resolution, something we'll show in the next post – down to 3.5GB. All settings hover above 2GB, aside from SP Low at 1080p.
A look at singleplayer's first campaign level:
At 1080/extra, it seemed as if VRAM consumption got “stuck” at 4200MB – we never saw usage tick above that in singleplayer on our Titan X. Moving to 4K/extra, though, was another story. We believe the disparity between singleplayer 4K/extra and the multiplayer equivalent can be attributed to overall higher-quality character models and cinematic graphics. A lot more time is spent looking at and seeing high-poly faces (without obfuscation by masks or other gear) in singleplayer, and some objects do appear to generally exhibit a higher quality than found in multiplayer prefabs. We had more simultaneous on-screen actors in multiplayer than singleplayer (which still had a high enemy count, but never as many 17 at once), so this likely contributed to additional VRAM consumption during multiplayer testing. The map can also impact component utilization.
Here's the important thing: You may have seen in our benchmarks that even cards with 4GB of VRAM (like the GTX 980 we used) are able to run 1080/extra above 100FPS. For example, the 1080/extra chart from our COD GPU benchmark:
Just because the memory is consumed when present does not necessitate that your GPU hosts 6-12GB VRAM. A huge amount of this memory is consumed by textures, so operating on a card with less VRAM will primarily impact the cleanliness of pop-in on non-player objects and the rapidity with which textures render higher resolutions. Here's a look at texture VRAM consumption, ahead of publication for our other piece:
In the above, you're looking strictly at delta VRAM consumption (not an absolute value) because we can't directly measure VRAM consumed only by textures. Measuring the delta allows us to see the change between texture quality (resolution) settings. No other settings were changed in the above texture benchmark, unlike the previous VRAM charts where all settings were down-classed for lighter-weight configurations.
Alongside its huge system memory consumption, Call of Duty: Black Ops III runs on the aggressive side regarding VRAM utilization. The game is fully playable on 2GB and 4GB cards – though 2GB should assign lower texture qualities – but will overall appear “cleaner” on the rendering with higher VRAM capacities at higher settings.
- Steve “Lelldorianx” Burke.