Our testing methods
Laptops present some difficulties for benchmarking, especially for those interested in the specific performance of one component in the system. When we test a graphics card, we generally try and control for as many variables as possible by placing it on a test bench. That allows us to do apples-to-apples comparisons between different cards.
Since that wasn't possible here—we couldn't exactly yank out the 1070 in the laptop and plug it into a handy motherboard—we submitted the 1070 to an completely unfair comparison. We ran our friendly neighborhood MSI GeForce GTX 1070 Gaming Z (the non-mobile kind) through the same benchmarks as Gigabyte's P57X using TR's standard graphics-card testing rig from recent reviews.
The machine that we used to test the MSI GTX 1070 has some major advantages over the P57X. Notably, the desktop GTX 1070 was paired up with a Core i7-6700K, an eight-threaded monster with a 4.2 GHz max turbo frequency and a 91W TDP. Our test rig also has 16GB of DDR4-3000 RAM compared to the P57X's DDR4-2133. Most notably, the CPU and graphics card for our desktop system have huge coolers that let them run without fear of thermal throttling.
Even though Gigabyte's cooling apparatus inside the P57X is an effective one from our past experience, there's no way it can compete with a desktop. The 45W Core i7-6700HQ inside the P57X is no slouch, but its tighter thermal envelope limits its peak performance. The same is true of the mobile GTX 1070—we'll be watching its clocks to see just how effective the P57X's heatpipes and fans are for cooling a desktop-class graphics card.
For some extra perspective about just where the mobile GeForce GTX 1070 slots into the graphics-card pantheon, we also retested our EVGA GeForce GTX 1060 6GB card using the same settings as we did for the mobile GTX 1070 on our desktop rig. Let's see if the P57X can run with the big dogs.
Maybe the best sequel since The Empire Strikes Back, Bethesda's 2016 reboot of the Doom franchise is a fast, frenetic fragfest. For these benchmarks, we ran the game at 1920x1080 on the game's Ultra preset, but with 8x TSSAA and 16x anisotropic filtering. Doom has a lot of graphics settings, so apologies in advance for the wall of screenshots.
Our "Inside the Second" benchmarking techniques gather frame-time data from our games of choice using a utility called PresentMon. Detailed analysis of the time it takes for the machines to generate each frame allows us to characterize the performance far more accurately than we could by simply presenting average frames per second. In a thermally-constrained chassis where temperatures, power limits, and clocks are constantly shifting and interacting, our methods could reveal any resulting roughness in-game.
That's what you call a fast start, ladies and gentlemen. All three units deliver fantastic results in the first game of our testing suite. The desktop GTX 1070 is the clear winner, with a commanding lead in average FPS and 99th percentile frame time. However, that doesn't diminish the results put in by the mobile GTX 1070 and the 1060. A 99th percentile frame time result of 13.5 ms is excellent, as that puts 99% of the rendered frames well north of 60 FPS.
So why does the desktop GTX 1070 produce about 32% more frames per second than the mobile GTX 1070? We surmise that we're seeing that result because of the resolution. At 1920x1080, the advantages of the desktop system are able to take over. Its CPU and RAM are just better at feeding the graphics card and keeping it busy.
Our "time spent beyond x" graphs provide a quick reference for how much time a graphics card spends working on tough frames past certain critical thresholds in our tests. The graphs that gamers should care about the most are the ones that show how many frames were above the 50-ms threshold and the 33.3 -ms threshold, as those reveal how often the game dropped beneath 20 FPS and 30 FPS, respectively. When frame rates drop that low, gamers are likely to notice, and perhaps lose their sense of immersion.
None of our three contenders yield any frames underneath these thresholds, however, only showing a difference when we look at frames that took longer than 16.7 ms to generate. Those frames are the ones that dip below 60 FPS. Even there, very little separates the three systems. The mobile GTX 1070 had a couple tiny spikes in the benchmarking runs, and that's what we're seeing here. Even so, the P57X spent less than a tenth of a second beyond this point—an excellent gaming experience.
The chart correlates well to our experience with Doom on the P57X. Overall, it provided a smooth, high-quality experience, but there were infrequent moments when the machine didn't quite keep up with our high expectations. Gamers who aren't actively watching for such moments might not even notice.