If you're looking for an in-depth take on how the GTX 1080 performs in our advanced "Inside the Second" frame-time benchmarks, you should go read our GTX 1080 Founders Edition review. We won't be repeating that fine-grained testing here. Instead, we'll be relying on the scripted benchmarks from Middle-Earth: Shadows of Mordor and Rise of the Tomb Raider to gauge average frame rates at a variety of resolutions. These simple FPS-based tests should be good enough to separate the various GTX 1080s we have on hand.
Middle-earth: Shadow of Mordor
To put Middle-earth: Shadow of Mordor through its paces, we used the game's internal benchmark and the Ultra quality preset. Click the buttons below the FPS graphs to see how the three GTX 1080 cards perform at three common resolutions.
Interesting. Despite their similar core clock speeds, the Gigabyte GTX 1080 Xtreme Gaming ekes out a tiny but consistent lead over the Strix. Perhaps the Xtreme Gaming's higher-clocked memory is giving it the edge over the Asus card here. Both custom GTX 1080s open a wide performance advantage over the GTX 1080 Founders Edition card.
Rise of the Tomb Raider
Rise of the Tomb Raider is a recent, demanding title from our current graphics testing sute. We ran this game using the following settings:
Man. Far be it from us to call a winner here. Both custom cards are within about a frame per second of one another at all of the resolutions we tested. Both the Asus and the Gigabyte cards offer a nice performance boost over the Founders Edition card, to be sure, but whatever characteristic gave the Gigabyte card the edge in Shadow of Mordor isn't being exercised here.
Our noise level measurements don't make this race any less close. Going by pure dBA numbers alone, the Strix card is ever-so-slightly quieter than the already excellent cooler on the GTX 1080 Xtreme Gaming. To be perfectly honest, I doubt most people's ears are sensitive enough to hear the difference in absolute noise levels between these cards while they're operating. The Strix and the Xtreme Gaming cards both offer improvements over the GTX 1080 Founders Edition in this regard, as well.
dBA numbers never tell the whole story about noise, though. Despite its low dBA figure, the Strix's trio of fans produces a higher-pitched whoosh than the GTX 1080 Xtreme Gaming's 100-mm spinners do. Unfortunately, this sound isn't entirely smooth. It has a slight buffeting quality or roughness about it that puts a bit of scruffiness on an otherwise broad-spectrum noise character.
At their worst, the Gigabyte fans produce the slightest hint of a baritone note, but they usually don't sound like much of anything in operation. That means the Xtreme Gaming card's sometimes-prominent coil whine is often the only sound that'll clue you in to the fact that it's operating at speed. On the other hand, the Strix card produces barely any coil whine, so which card is "better" on a subjective basis will depend on whether you're bothered more by the sound of air moving or by the occasional sound of electronics switching at high speed.
Now that I've heard it with my own ears, I think we may have been a bit harsh on the GTX 1080 Founders Edition cooler in our original review of that card. The fan on the Founders Edition is one of the best-sounding blower coolers I've ever heard, but it still has a certain whiny, hissy quality about it that doesn't fade into the background as easily as the sounds of the Asus or Gigabyte cards do. Once again, the card isn't exactly loud, but it'll always make itself known in operation.
At idle, our test system is impressively frugal on power with all of the GTX 1080s we have on hand. The overclocked custom cards unsurprisingly consume slightly more power than the GTX 1080 Founders Edition, but that's to be expected.
Under load, the GTX 1080 FE consumes significantly less power than both the Asus and Gigabyte cards. The factory clock speed boost on both of the custom cards seems to come at a significant cost for power consumption. The Strix holds about a 10W advantage over the Xtreme Gaming card while running Unigine Heaven, but the two are so closely matched in every other regard that we'd call this test a wash.
GPU temperatures and observed clock speeds
In our experience with Pascal chips, the GPU Boost 3.0 algorithm tends to boost clocks to a peak speed before settling into an equilibrium once the card has had a chance to heat up. To take that behavior into account, we ran the Unigine Heaven benchmark for 10 minutes before observing clock speeds and taking temperature measurements.
Unsurprisingly, both custom coolers handily outperform the GTX 1080 Founders Edition's reference heatsink. Asus' custom card delivers a 14° C drop over the reference design, while the Gigabyte card's massive heatsink lets it run another 3° C cooler yet. If you're concerned about thermal headroom for overclocking, the performance potential of either of these coolers should offer plenty of wiggle room for a chip that might be thermally limited.
We also observed the sustained boost clocks that each card reached under our Unigine Heaven load. After 10 minutes, the Strix settled into a boost speed of 1972MHz, while the Xtreme Gaming card seemed content to run at 1987MHz. Both of those speeds are far in excess of the 1898-MHz boost speed that Asus and Gigabyte list in the specs for these cards, so that figure is at best a very loose guideline of what to expect.
The Founders Edition card, on the other hand, has to modulate its clocks in apparent service to its thermal limit. Recall that Nvidia's boost speeds are a range, not a ceiling. Despite Nvidia's specified 1733-MHz boost speed for the GTX 1080, the card doesn't just boost up to that speed and stop. Instead, we observed that the Founders Edition's clocks tended to oscillate around that speed in order to keep temperatures in check.
After being loaded with Heaven for a while and reaching its 82° C maximum, the Founders Edition card occasionally boosted as high as 1772MHz, but it also dipped as low as 1696MHz. The handy GPU-Z utility can expose why a card is hitting a performance ceiling, and unsurprisingly, it showed that the Founders Edition card was hitting a thermal limit instead of a power cap.
In contrast, neither custom card needed to vary its boost speed while under load. After building up some heat, both the Strix and the Xtreme Gaming cards maintained their boost speeds without a hitch. While we don't think most people will notice, that performance consistency could translate into smoother gameplay over time, and it's another reason to consider custom-cooled GTX 1080s instead of the Founders Edition card.