Gears 5 Review: UE4 Performance, Guts And Glory Explored
Gears 5: Performance at 1080p and 1440p
Gears 5 has a built-in benchmark that provides a surprising amount of performance detail. The test itself replays a scene from Act II of the campaign (sans voice acting) set in a blizzard-stricken forest. The test uses the full game engine, and each run is slightly different. In one instance, a Juvie (a young member of The Swarm) exploded too close to an ice block where one player took cover, and as a result the ice block was destroyed. Despite all the benchmarking runs we performed, it never happened a second time.
The benchmark test is a snowy firefight. Note all the frame details in the upper-left.
First up we tested the High and Ultra settings at 1080p on our AMD Radeon RX 5700XT and an NVIDIA GeForce RTX 2060 Gaming Z built by MSI. While these cards aren't direct competitors, they're priced within 30 to 40 dollars of each other and represent what performance you can get from a graphics card that costs around $400. We didn't have an RTX 2060 Super handy for testing, however. If you recall from our GeForce RTX 2060 Super and 2070 Super review, the newer card is around 15 to 20% faster than the stock RTX 2060. That gap should be closed somewhat—but not entirely—by the hot clocks on our MSI model.
Out of the gate, we see that the Radeon RX 5700XT performed exceptionally well at these settings. On the Ultra settings, the AMD card performed around 20% faster than our RTX 2060—in fact, the 5700XT's Ultra performance was better than the GeForce on High. AMD's fastest mainstream card on running the High preset beat the frame rate of equivalent settings on the GeForce by a whopping 35%. This game was optimized for AMD graphics cards thanks to its console roots (GCN isn't RDNA, but they're still pretty closely related), and that seems to carry over to the PC version.
Frames per second only tells part of the story, however. Let's see how the two cards perform across the entirety of the run by checking frame times for long pauses or inconsistent performance.
The AMD card did pretty well across the whole run. A few frames early on spike as high as 30 milliseconds. That translates to an instantaneous frame rate over 30 frames per second, so we don't think anybody will notice. Once the card settles in, the Radeon RX 5700XT never gets above 16.7 milliseconds for the remainder of the run, and much of that was under 10 milliseconds per frame. That's outstanding performance at this resolution.
While the GeForce RTX 2060 didn't do as well as the Radeon RX 5700XT, NVIDIA's card still performed admirably with maximum settings at 1080p. Aside from a couple of outliers early on, the frame times stay below 16.7 milliseconds. This is still a very smooth experience, and anybody still gaming on a 60 Hz display would never be able to tell a difference.
That's an awful lot of data to take in, so let's see how we can summarize performance by grouping frame times together.
Folks with fast 1080p gaming monitors will definitely feel a smooth experience, since both cards managed to stay above 60 frames per second basically the entire run. More than half the time, the Radeon RX 5700XT turned in sub-10 millisecond frame times. If your display can't handle more than 60 Hz, these cards leave a lot of performance on the table.
It's time to turn the screws on these cards and nearly double the number of pixels they have to render in each frame.
Again, the Radeon RX 5700XT outperforms its price difference in Gears 5, but both cards manage to keep the average above 60 frames per second. At High settings, the Radeon beats the GeForce RTX 2060 by 26%. On the Ultra setting, the Radeon's advantage shrinks to 19%. Both of those figures are a far cry from a 5% price difference, making the RX 5700 XT look like a good value. Let's take a look at frame times.
Again, the Radeon RX 5700XT turns out excellent, consistent frame times. Subjectively, we felt the game runs great at these settings, and the data backs up our evaluation.
The GeForce RTX 2060 had few problems at 1440p, too. A couple of frames right together near the beginning hopped up over 40 milliseconds, but it goes by so quickly we don't think you'd feel but a tiny blip. However, the early part of the graph shows the RTX 2060 actually slightly above 16.7 milliseconds, going up as high as 20. For a decent chunk of the run, the GeForce couldn't manage 60 frames per second, and that's a difference we think you can feel. Based on the 99th percentile frame rate above, you can still get a smooth 60 frames per second basically all of the time by dropping the settings to High.
Just like before, we can summarize the data by grouping the frame times.
To be sure, there are a few spikes above 16.7 milliseconds on the Radeon RX 5700XT, but they almost never rise above 25 milliseconds. That's still incredibly solid. The GeForce isn't doing badly, but it just can't keep up, since 30% of all frames take longer than 16.7 milliseconds.
Let's see how these cards can handle 3840x2160 UHD next.
The benchmark test is a snowy firefight. Note all the frame details in the upper-left.
When the test run completes, a summary screen shows frame times for the entire run, along with some interesting stats. Unfortunately, the benchmark results are lost to the ether after you exit this screen, so we had to capture frame time data the old-fashioned way with CapFrameX. Still, we applaud The Coalition for including so much performance data after a test run, and would encourage other developers to continue this work in the future.
Tons of detail from the Gears 5 benchmark, including a combined GPU and CPU frametime graph
Tons of detail from the Gears 5 benchmark, including a combined GPU and CPU frametime graph
For our own performance tests, we endeavored as always to obtain clean runs. Each benchmark was run three times using the built-in graphics test, and the median number is reported here. Fortunately we didn't see a ton of fluctuation from run to run, so the simulation and physics didn't have a huge impact on performance. After each summary we also have a detailed frame time graph for the duration of the run represented in the performance summary.
Enough talk—it's time to see how our test subjects handle The Coalition's latest title.
Enough talk—it's time to see how our test subjects handle The Coalition's latest title.
 |
|
First up we tested the High and Ultra settings at 1080p on our AMD Radeon RX 5700XT and an NVIDIA GeForce RTX 2060 Gaming Z built by MSI. While these cards aren't direct competitors, they're priced within 30 to 40 dollars of each other and represent what performance you can get from a graphics card that costs around $400. We didn't have an RTX 2060 Super handy for testing, however. If you recall from our GeForce RTX 2060 Super and 2070 Super review, the newer card is around 15 to 20% faster than the stock RTX 2060. That gap should be closed somewhat—but not entirely—by the hot clocks on our MSI model.
Out of the gate, we see that the Radeon RX 5700XT performed exceptionally well at these settings. On the Ultra settings, the AMD card performed around 20% faster than our RTX 2060—in fact, the 5700XT's Ultra performance was better than the GeForce on High. AMD's fastest mainstream card on running the High preset beat the frame rate of equivalent settings on the GeForce by a whopping 35%. This game was optimized for AMD graphics cards thanks to its console roots (GCN isn't RDNA, but they're still pretty closely related), and that seems to carry over to the PC version.
Frames per second only tells part of the story, however. Let's see how the two cards perform across the entirety of the run by checking frame times for long pauses or inconsistent performance.
The AMD card did pretty well across the whole run. A few frames early on spike as high as 30 milliseconds. That translates to an instantaneous frame rate over 30 frames per second, so we don't think anybody will notice. Once the card settles in, the Radeon RX 5700XT never gets above 16.7 milliseconds for the remainder of the run, and much of that was under 10 milliseconds per frame. That's outstanding performance at this resolution.
While the GeForce RTX 2060 didn't do as well as the Radeon RX 5700XT, NVIDIA's card still performed admirably with maximum settings at 1080p. Aside from a couple of outliers early on, the frame times stay below 16.7 milliseconds. This is still a very smooth experience, and anybody still gaming on a 60 Hz display would never be able to tell a difference.
That's an awful lot of data to take in, so let's see how we can summarize performance by grouping frame times together.
| 1080p Performance Summary | GeForce RTX 2060 1080p | Radeon RX 5700XT 1080p |
| % below 10 (above 100 fps) | 11.02 | 60.59 |
| % between 10 and 16.7 ms (100 and 60 fps) | 88.33 | 39.22 |
| % between 16.7 and 25 (60 and 40 fps) | 0.61 | 0.16 |
| % between 25 and 33.3 (40 and 30 fps) | 0.02 | 0.03 |
| % above 33.3 (below 30 fps) | 0.02 | 0.00 |
Folks with fast 1080p gaming monitors will definitely feel a smooth experience, since both cards managed to stay above 60 frames per second basically the entire run. More than half the time, the Radeon RX 5700XT turned in sub-10 millisecond frame times. If your display can't handle more than 60 Hz, these cards leave a lot of performance on the table.
|
|
It's time to turn the screws on these cards and nearly double the number of pixels they have to render in each frame.
Again, the Radeon RX 5700XT outperforms its price difference in Gears 5, but both cards manage to keep the average above 60 frames per second. At High settings, the Radeon beats the GeForce RTX 2060 by 26%. On the Ultra setting, the Radeon's advantage shrinks to 19%. Both of those figures are a far cry from a 5% price difference, making the RX 5700 XT look like a good value. Let's take a look at frame times.
Again, the Radeon RX 5700XT turns out excellent, consistent frame times. Subjectively, we felt the game runs great at these settings, and the data backs up our evaluation.
The GeForce RTX 2060 had few problems at 1440p, too. A couple of frames right together near the beginning hopped up over 40 milliseconds, but it goes by so quickly we don't think you'd feel but a tiny blip. However, the early part of the graph shows the RTX 2060 actually slightly above 16.7 milliseconds, going up as high as 20. For a decent chunk of the run, the GeForce couldn't manage 60 frames per second, and that's a difference we think you can feel. Based on the 99th percentile frame rate above, you can still get a smooth 60 frames per second basically all of the time by dropping the settings to High.
Just like before, we can summarize the data by grouping the frame times.
| 1440p Performance Summary | GeForce RTX 2060 1440p | Radeon RX 5700XT 1440p |
| % below 10 (above 100 fps) | 0.14 | 0.27 |
| % between 10 and 16.7 ms (100 and 60 fps) | 69.24 | 94.86 |
| % between 16.7 and 25 (60 and 40 fps) | 30.40 | 4.77 |
| % between 25 and 33.3 (40 and 30 fps) | 0.19 | 0.09 |
| % above 33.3 (below 30 fps) | 0.03 | 0.00 |
To be sure, there are a few spikes above 16.7 milliseconds on the Radeon RX 5700XT, but they almost never rise above 25 milliseconds. That's still incredibly solid. The GeForce isn't doing badly, but it just can't keep up, since 30% of all frames take longer than 16.7 milliseconds.
Let's see how these cards can handle 3840x2160 UHD next.
