Game Benchmarks stub

Watch Dogs 2 CPU Benchmark - Threads Matter

Posted on February 20, 2017

With Ryzen around the corner, we wanted to publish a full CPU benchmark of Watch Dogs 2 in our test course, as we’ve recently found the game to be heavily thread-intensive and responsive to CPU changes. The game even posts sizable gains for some overclocks, like on the i5-2500K, and establishes a real-world platform of when CPU choice matters. It’s easy to bottleneck GPUs with Watch Dogs 2, which is something of a unique characteristic for modern games.

Watch Dogs 2 is a familiar title by now at the GN test bench, and while we’ve published a GPU benchmark and a more recent CPU optimization guide, we never published a comprehensive CPU benchmark. We’ve gathered together all our results here, from the 2500K revisit all the way to Kaby Lake reviews (see: 7600K review & 7350K review), and analyzed what exactly makes a CPU work well with Watch Dogs 2 and why.

In this Watch Dogs 2 CPU benchmark, we’ll recap some graphics optimization tips for CPUs and test whether an i7 is worth it, alongside tests of the 7600K, 7700K, 6600K, 7350K, FX-8370, and more.

 

Test Methodology

We’ve explained our methodology in previous Watch Dogs 2 coverage, but just to recap: we walk down a specific hill around 3-4PM in-game time, clear skies, and log framerate using FRAPS, then average these multiple runs (per test) together. The runs typically have near-identical average framerates and mild variation among 1% and 0.1% lows. We use the High preset as a baseline and for this comparative benchmark, as we’re not trying to overload the GPU, but still want to represent a real scenario. 1080p is used for resolution.

In terms of user-controlled settings, the most intensive options are “extra detail” and “geometry,” set to 0 and high respectively. More complex geometry means more draw calls added to the graphics pipe, requiring the CPU to work harder at distributing calls to the GPU. For more detail and performance comparisons, check our CPU optimization guide.

As much hardware as possible is kept consistent between runs, but obviously several motherboards and sets of RAM were necessary for compatibility with all the different CPUs tested. An EVGA GTX 1080 FTW ($590) was used in all tests in order to artificially create a CPU bottleneck (two 1080s in SLI gave worse performance scaling at the time testing began, though driver updates have since shipped). We’ve tested an AMD CPU this time around in the interest of completeness, but until the release of Ryzen, there are some serious limitations on AMD WD2 performance. Much of this is to do with core architecture.

Note also that we intend to benchmark the FX-6300 and other Athlon series chips (again) in future coverage, as we move ever closer to filling-in our CPU benchmark for the full Kaby Lake & inevitable Ryzen SKUs.

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.

Further note that some CPUs were overclocked during this testing, purely an artifact of when we reviewed or revisited those CPUs. Not every CPU on this bench has been reviewed or revisited recently, which means not every CPU has overclocked numbers. We add overclocking benchmarks incrementally per review or per revisit, as it is plainly unreasonable to overclock every single CPU for this benchmark (that doubles the workload, and thus becomes unprofitable/unsustainable as a business).

Hardware Tested

Intel CPUs:

  • Intel Core i7-7700K OC 5.1GHz HT1
  • Intel Core i7-7700K OC 4.8GHz HT1
  • Intel Core i7-7700K Stock HT1
  • Intel Core i7-6700K Stock HT1
  • Intel Core i7-4790K Stock HT1
  • Intel Core i7-7700K Stock HT0
  • Intel Core i5-7600K Stock
  • Intel Core i5-6600K Stock
  • Intel Core i7-2600K Stock HT1
  • Intel Core i5-4690K Stock
  • Intel Core i5-2500K OC 4.5GHz
  • Intel Core i3-7350K Stock
  • Intel Core i5-3570K Stock
  • Intel Core i5-2500K Stock
  • Intel i3-6300 Stock

AMD CPUs:

  • AMD FX-8370 Stock
  • More to come very soon!

Motherboards & Memory:

  • Intel Kaby Lake series: MSI Z270 Gaming Pro Carbon w/ Corsair Vengeance 32GB 3200MHz
  • Intel Skylake Series: MSI Z170 Gaming M7 w/ Corsair Vengeance 32GB 3200MHz
  • Intel Haswell/Devil’s Canyon Series: Gigabyte Z97X G1 WIFI-BK with HyperX Savage 32GB 2400MHz
  • Intel Ivy Bridge Series: MSI Z77 GD65 with HyperX Savage 32GB 2400MHz
  • Intel Sandy Bridge Series: MSI Z77 GD65 with HyperX Savage 32GB 2400MHz
  • AMD FX Series: Gigabyte G1 Gaming with 32GB Corsair Vengeance 1866MHz (unstable beyond this speed)

GPU & Drivers:

1 x EVGA GTX 1080 FTW with 376.33 drivers

OS:

Win 10 x64

Continue to Page 2 for the results.


Watch Dogs 2 CPU Benchmark Results – 1080p/High

We published a lot of Watch Dogs 2 results in our i5-2500K revisit, and those numbers remain the same; however, we’ve added several more CPUs to our lineup. As we noted back then, high-end i7s reach a cap of 110-114 FPS thanks to system limitations, but other CPUs are allowed to really push their limit and provide good comparative numbers.

All CPUs tested managed at least a near-60 FPS average, but a few dipped well below that in the 1% and 0.1% lows. Interestingly, every i7 (including the venerable 2600K) managed above 54 FPS or greater at their 1% lows, and the newer 6700K and 7700K CPUs reached an obvious limit around 113 FPS. Meanwhile, none of the lower threadcount i3s and i5s managed to surpass 85 FPS average, although that still meets the acceptable performance threshold for most users.

watch-dogs-2-cpu-benchmark 1

Our in-depth testing of the i5-6600K revealed that it was capable of handling the Very High preset while keeping 1% lows above 60 FPS, but not Ultra. The i7-6700K outperformed its i5 counterpart by just over 40% at the High preset, making it (and i7s in general) the clear choice for higher graphics in WD2. To support this theory, the i7-7700K dropped from 112.7 FPS average to 87.3 with hyperthreading disabled, a 22% decrease even despite the i7’s maximum FPS being limited by other factors. This kind of performance gap is unusual but encouraging, as it means WD2 is taking serious advantage of multiple threads, a valuable feature for both benchmarking purposes and actual gameplay.

This trend extended downwards to the i3s tested, which performed abysmally. The brand new i3-7350K ($180), when overclocked to 5.0GHz (as it should be), scraped by with 68 FPS average, barely outperforming the six-year-old overclocked i5-2500K. The i3-6300 ($150) at its stock frequency had a lower average FPS than every other CPU we tested, including the stock 2500K.

The obvious exception to the “more threads = better” rule is AMD’s FX-8370, which scored between the i5-2500K and i5-3570K (both stock) at 62 FPS average with 8 physical cores. This largely boils down to something that is still surprisingly unknown: An AMD “core” does not equal an Intel “core.” The two companies have vastly different definitions of what constitutes a core under each architecture, and further have vast differences in CPU designs that would impact gaming performance (different prefetch/branch prediction routines, cache configuration, so forth). With the older FX series of CPUs, we’re looking at 1FPU per 2INT units, so that FPU is getting split in ways that may not be agreeable to a floating point-intensive task.

Improvements based on frequency were somewhat less consistent. Overclocking the i7-7700K ($350) was almost completely ineffective, as it was already at the limit of what the system could handle—we managed only a 1 FPS increase in average framerate by pushing it to 5.1GHz. More dramatic results could be seen at the bottom of the spectrum, where overclocking the i5-2500K (from its boost clock of 3.7GHz to 4.5GHz) lifted it past the 60 FPS barrier from 58.5 to a more tolerable 67 average (with improvement in 1% and .1% lows as well). Obviously, frequency affects performance, but there are other factors at work—for instance, the 7350K barely improved at all when overclocked.

Generationally, FPS increases are relatively small. There was about a 7.1% performance gain from the 3.9GHz boost i5-6600K to the 4.2GHz boost i5-7600K ($240), just slightly under the percent difference in frequency. The 6600K in turn scored a 7.4% increase in FPS over the i5-4690K, but their boost frequency is the same: the main difference is the 6600K’s 14nm process, compared to 22nm in the older 4690K.

Which CPU to Buy for Watch Dogs 2?

Strictly for Watch_Dogs 2 performance, it seems that moving from an i3 to an i5, or from an i5 to even an older-model i7 is a more cost-effective way to improve performance than simply upgrading to the newest CPU. As of this writing, no i3, the overclocked K-SKU being the exception, could provide solid framerates even just at the High graphics preset. Watch Dogs 2 is definitely playable on the i5 lineup and even the FX-8370, as we’ve previously shown the game is a tolerable performer even at a sub-60FPS framerate (we’ve found 45-50FPS to be agreeable with this title, subjectively). The i3-6300 struggles to keep up a bit, but isn’t so bad.

Still, if you were to buy a CPU just to play this game, the i5 & i7 units do show substantial improvement over the i3 units and FX units. This is one of those rare titles where it wouldn’t be too hard to bottleneck even a GTX 1070 with an i5 CPU.

Editorial, Testing: Patrick Lathan
Host, Test Lead: Steve Burke
Video Production: Andrew Coleman