Creative Assembly’s Total War: Warhammer ($60 on Steam) seems like nothing short of a love letter to geeks. This game blends the thrilling tactical action of the Total War series with the beloved Warhammer universe, and as if that wasn’t nerdy enough, it will even be an early DirectX 12 title, with a DX12 patch planned to land sometime shortly after launch.
We’re still in the process of reviewing Total War: Warhammer, which launches Tuesday, May 24. Spoiler alert: The switch from historical factions to fantastical ones breathes new life into the Total War series, both in atmosphere and mechanics. It’s good! (Update: Here's our Total War: Warhammer review.)
But today, we want to give you a taste of what to expect when that DirectX 12 patch lands. While Creative Assembly’s technical optimization blog has only said “We’re really happy with the DX12 performance we’re seeing so far, so watch this space!” AMD shot us a build of an automated Total War: Warhammer DirectX 12 benchmark to show what the game’s capable of. Second spoiler alert: Radeon graphics cards come out on top against their GeForce opposites, as expected in an AMD Gaming Evolved title, but we’ll get into that more later.
Total War: Warhammer DirectX 12 benchmarks
The automated benchmark lets you choose from a wealth of resolution and graphical options. We ran all tests on the Ultra preset, on PCWorld’s test system with a fully patched version of Windows 10 installed.
The benchmark provided, dubbed “Empire vs. Greenskins,” runs through a brief scene showing Total War: Warhammer’s Empire and Greenskin factions locked in battle. When it’s done, it spits out a results screen showing the average frame rate as well as a graph of frame rate fluctuations. Since the AMD-provided benchmark only runs in DirectX 12, we couldn’t verify frame rates or times with a tool like FRAPS, which doesn’t work with DirectX 12 yet.
That’s not the only caveat to be mindful of. Because we can’t run the benchmark in DX11 mode, we have no idea how the frame rate results compare to standard DX11 performance for the game. Also worth mentioning: Nvidia’s drivers typically outperform AMD’s in DX11, so performance in that mode may well be tilted in Nvidia’s favor. We don’t know! Also, Nvidia hasn’t released drivers optimized for the game, and DirectX 12 is only available in Windows 10.
Finally, performance results may very well change once the final DirectX 12 patch for Total War: Warhammer is released. This is based on unfinished code.
So you’ll want to take these results with a big pinch of salt. All that said, we thought you’d be interested in seeing the early initial results.
We tested the three most popular gaming resolutions—1920x1080, 2560x1440, and 3840x2160—with AMD and Nvidia’s rivals in the $200 to $500 price range to make this an apples-to-apples comparison.
First up, we tested 1080p resolution with the Asus Strix GTX 960 DirectCU II ($205 at Newegg) and VisionTek Radeon R9 380 ($200 at Amazon). Our 1440p testing was a bit more crowded, as it felt worthwhile to benchmark all of the remaining cards at this resolution. We tested the EVGA GTX 970 FTW ($325 on Amazon), Sapphire Nitro 390 ($325 on Newegg), Nvidia’s reference GTX 980, and the MSI R9 390X Gaming 8GB ($420 on Newegg).
Finally, we tested the GTX 980 and MSI R9 390X at 4K, and we tossed in Nvidia’s new $600 GTX 1080 for good measure, because we had one handy. Even that powerhouse can’t come close to cracking the hallowed 60fps barrier in Total Warhammer at 4K/Ultra.
Now that the raw benchmarks are out of the way, with the Radeon cards beating their rivals across the board and the cutting-edge GTX 1080 destroying everything else (as expected), there’s one other tidbit to mention. The frame rate fluctuation graphs that appear at the end of the benchmark show AMD’s Radeon cards running more smoothly over the course of the scene…
…than their GeForce counterparts. Update: Pay close attention to the truncation of the Y-axis in those graphs, though; the scale isn’t the same from picture to picture. (I don’t see a way to disable the truncation in the benchmark options.) Though AMD cards still show less overall variation than GeForce GPUs, the difference isn’t as drastic as the images seem to show at first blush.
A dedicated Total War: Warhammer driver from Nvidia could likely alleviate the problem somewhat, but it’s unlikely to eliminate the gap completely. As a strategy title featuring dozens if not hundreds of on-screen units to keep track of, Total War: Warhammer—much like Stardock’s superb Ashes of the Singularity—relies heavily on CPU utilization to process everything. Moving to DirectX 12 lets the game handle those commands much more efficiently, and Total War: Warhammer’s DirectX 12 implementation leans on the dedicated asynchronous compute engine (ACE) hardware baked inside AMD’s Graphics Core Next architecture.
Nvidia’s GTX 700- and 900-series graphics cards don’t have that dedicated hardware, relying instead on a software-based task switching technique called pre-emption that isn’t as efficient as the Radeon ACEs. As a result, while AMD’s cards see massive performance gains switching from DX11 to DX12 in Ashes of the Singularity, Nvidia’s GTX 900-series cards see none whatsoever. In fact, the Ashes developer disables asynchronous computing when Nvidia cards are detected because it could actually reduce frame rates. (Sadly, we can’t compare DX11 vs. DX12 in this DX12-only Total War benchmark.)
Nvidia isn’t taking the threat sitting down. The new Pascal GPU in the GTX 1080 and GTX 1070 packs a bevy of new hardware and software features dedicated to improving async compute performance. In our tests, its AoTS DirectX 12 performance skyrocketed despite the async hooks’ ostensibly being disabled.
Just how well Nvidia’s new GPU architecture responds to async compute tasks remains to be seen—as does the DX11 performance of Radeon GPUs in Total War: Warhammer. Like we said, this was more an exercise in fun than anything else. Happy Greenskin hunting!