The Division Benchmark – GTX 1070 SLI vs. 1070, 1080 at 4K, 1440, & 1080p
The Division is intensive. We use its built-in benchmarking suite for analysis, but there are a few critical points to be aware of: As a game, The Division has pretty heavily varied FPS depending on effects in the environment. Walking through the smoldering ash clouds of an explosive will tank FPS hard for that brief instant in the cloud, but isn't representative of greater gameplay (and is only an issue with some settings enabled). The Division is also an MMO, which isn't something we can really test given the latency of the servers and the variability of players on FPS – we need a reproducible scenario.
And so we run the benchmark, but point-out these challenges. The benchmark tends to be representative of real-world, in-game performance (even with other players) as it is exceptionally heavy on the GPU despite a lack of other players.
The Division shows some reasonable gains for SLI. At 4K, we see a scaling of 29.27% against the single GTX 1070 FE, with clock-for-clock configurations. The SLI 1070s run 8.76% faster than a single GTX 1080 in average framerate. 1% and 0.1% lows struggle, meaning that frametimes and overall consistency of framerate (ms-to-ms) is worse than the single GTX 1080 or single GTX 1070.
1440p is a little more favorable for the dual-GPU setup. The GTX 1070s in SLI push 118.3FPS – they were likely choking on pixel count at 4K – and hold a substantial 32.76% lead over the GTX 1080. 1% and 0.1% lows might not be as high as they should be, but are still superior to the single GPU configurations. Against a single GTX 1070, the SLI configuration holds a 50.37% lead in average FPS.
1080p more-or-less sticks to this. The SLI 1070s distance themselves by 20.54% over the single GTX 1080, and 38.3% lead over the single GTX 1070. The average framerate at 1080p is 157.3FPS, with a >100FPS 1% low.
Mirror's Edge Catalyst Benchmark – GTX 1070 SLI vs. 1080, 1070 at Hyper
Mirror's Edge Catalyst came out this week. We ran benchmarks on day-one, our tests finding some odd behavior with low frametimes on AMD devices (as of 16.6.1 hotfix) and VRAM-abusive behavior by the Hyper settings.
ME Catalyst is one of the most GPU-intensive games we've tested in recent years, remaining true to its heritage, and even pushes the GTX 1080 into a struggle at 4K/High. When the game's fluid, it's beautiful – but the high-paced movement does make a higher framerate desirable for best gameplay (we found 50-60+ FPS to be the preferred window).
NOTE: Mirror's Edge is the one game where we ran an MSI GTX 1070 Gaming X as the single GTX 1070, rather than the Founders Edition. This is a change we will be making as we roll into the AIB partner cards, and ME Catalyst happened to be the game to initiate that transition. We try to move to AIB partner cards for AMD and nVidia once they ship, since we don't currently recommend the reference devices and because they're presently over-priced versus vendor pre-OC versions.
At 1080p/Hyper, the GTX 1070 SLI tops the bench with 105.67 FPS AVG and retains fairly strong 1% and 0.1% lows, despite being imperceptibly lower than the single GTX 1080. The SLI configuration scales 23.68% over a single GTX 1070 (83.3FPS), putting it within ~3FPS of the single GTX 1080 (Δ3.5%).
Moving to more intensive settings, 4K/High shows a gap of 4FPS between the 1080 and 1070 SLI, and a gap of 17FPS between the 1070 and 1070 SLI. As percentages, 4K/High distances the 1080 and 1070 SLI by Δ7.55%, but the 1% and 0.1% lows are much higher on the single GTX 1080 (Δ22.3% for 1% lows, Δ21.59% for 0.1% lows). This is a fairly constant trend for SLI/CrossFire. Against the 1070 non-SLI, we're seeing delta values of Δ36.56% for the average FPS, with the 1% and 0.1% lows marginally improving. Note again that the single GTX 1070 here is the Gaming X, so the gap is shrunken versus a reference 1070 (with lower clock-rate). These metrics provide hope considering what we just saw with DOOM.
At 1440p, we see a delta of 8.4% between the 1070 SLI and 1070 (remember – settings have changed from “high” at 4K to “ultra” at 1440p, which shifts load to be more intensive after the rasterization step). The 0.1% lows are a fair bit reduced for SLI versus the non-SLI setup, posting 48.67FPS 0.1% low vs. 63FPS 0.1% low.
At 1080p/ultra, the GTX 1080 pulls ahead by 4.47%, with the GTX 1070 SLI setup leading a single GTX 1070 (pre-OC) by 14.12%.
Black Ops 3 Benchmark – SLI GTX 1070 vs. 1070, 1080, more at 4K, 1440, & 1080
Black Ops 3 remains one of the best-optimized games we've worked with – perhaps a testament to Activision's experience in the space. The game has regularly posted strong performance for AMD devices as well, though the Fury X does still suffer at 4K.
At 4K, we're seeing averages nearing 100FPS on the dual GTX 1070s in SLI, with 1% and 0.1% lows performing near 60FPS – a strong output overall. The single GTX 1080 FE pushes 68FPS average, so that's a 35.15% difference versus the SLI config. Against a single GTX 1070, the dual setup runs ~58.15% faster in average framerates. The Fury X sustains very low 0.1% metrics because of the 4GB limitation, which is quickly saturated by BLOPS3 at 4K.
At 1440p, we're seeing 180.3FPS AVG on the SLI GTX 1070s – somewhat absurdly high – netting a ~28.52% difference over the GTX 1080 and 51.29% difference over the single GTX 1070. Good gains over all, with performance well above the magical 144Hz number for ultra-competitive gamers, and nearing 200Hz.
1080p is just comical, with the 2x GTX 1070s pumping 218FPS AVG against the 166FPS AVG of a single card. It's really not a use case that's realistic, though; at this point, most dual-1070 users will not be playing 1080p, anyway. Especially with 1440p pushing 180FPS.
GTA V Benchmark – SLI GTX 1070 vs. 1070, 1080, 980 Ti, Fury X at 4K & 1080
GTA V is becoming one of the longer-standing titles on our bench. It's been intensive, but reasonably well-optimized. We understand the graphics settings exceptionally well as a result of our graphics optimization guide, something which fuels our methodology.
GTA V at 4K also sees >50% scaling versus a single GTX 1070 (55.8%) in the AVG FPS department. 1% and 0.1% lows are also improved, and now land above 30FPS. Together, this makes 4K at Very High / Ultra completely playable on the SLI GTX 1070s, where a single GTX 1070 sees enough dips in performance (in the 1%/0.1% lows) to necessitate a very slight settings reduction. Versus a GTX 1080, the SLI GTX 1070 at 4K show a delta of 34.07%.
We hit a CPU bottleneck at 1080p, with the GTX 1080 and SLI 1070s – and effectively the 1x GTX 1070 – all landing at around the same ~120-125 limit.
Metro: Last Light Benchmark – SLI GTX 1070 vs. 980 Ti, Fury X at 4K, 1440, & 1080
Metro: Last Light will eventually be phased-out of our bench, but its reliability keeps it on for now.
At 4K, we're seeing 81.3FPS AVG in MLL, equating a 24.28% lead over the GTX 1080 and 42.69% lead over a single GTX 1070. 1440p has the GTX 1070s exceeding 100FPS, with nearly identical percent differences between the single 1080 and 1070. 1080p has us bumping against a performance wall at the top-end of the chart (1080 & 1070s in SLI).
Shadow of Mordor Benchmark – GTX 1070 SLI vs. 1080, Fury X at 4K, 1440, & 1080
The last chart for this FPS bench, we see Shadow of Mordor operating at 87.3FPS on the SLI setup, a good ~27FPS lead (at 4K) over the 1080, or 35.95%, and ~40FPS lead over a single 1070, or 72.8% scaling. That makes Shadow of Mordor one of the most aggressively scaled games, producing real, measurable, and tangible gains otherwise unachievable on other devices.