NVIDIA GTX 1070 SLI FPS Benchmarks – Dx12, Vulkan, & OpenGL vs. DirectX 11
This page initiates our game benchmarking for the GTX 1070 SLI configuration. We do not include any Dx11-only games on this page, but we do provide comparative data between Dx11, Dx12, and Vulkan (where present) for tested games. Some of this data is presented as a percent change value or other unique values, like ms latency. These charts provide insight as to async compute performance and overall acceptance of new APIs by the hardware. Learn more about Pascal asynchronous compute in our GTX 1080 review.
OpenGL is also included through DOOM, which we originally benchmarked here.
OpenGL DOOM – GTX 1070 SLI vs. GTX 1080, 980, Fury X, 390X
We're still waiting on the Vulkan implementation within id Software's DOOM. For the time being, the game remains on OpenGL and makes promises of enabling Vulkan for all GPU vendors; id Software has gone to bat for both nVidia and AMD in the past few months for the GTX 1080 and RX480, and seems to be charged to deliver on the new API.
This should be obvious from one glance at the above, but DOOM does not have good SLI scaling at this time. SLI is not officially supported, and currently forces AFR2 to achieve function without any real gains.
At 4K, DOOM proves as intensive as ever for all devices under test (DUT). The GTX 1080 FE operates at 51.67FPS average, holding an imperceptible (but measurable) lead over the GTX 1070 SLI configuration (49FPS AVG). That lead boils-down to 5.3%, with the 1% and 0.1% lows being separated by a comparable margin. Over a single GTX 1070, the SLI configuration leads by 15.38% – that's not great scaling and would be pretty poor value for owners of 1070 SLI, but 4K is intensive. Let's move to 1440p and see if scaling improves.
No. Not really. At 1440p, now that the bars are a little more exaggerated, we can see how much of a hit the 1% and 0.1% low framerates take with SLI. Consistency isn't bad – there's no real, observable micro-stutter (as there is in some games) – but it's still worse than a single GTX 1070, with averages more-or-less identical (81 vs. 79FPS AVG, so 2.5% delta).
At 1080p, we see a delta of 3.4% between the averages, with the SLI cards again taking a hit (although mostly imperceptible) to 1% and 0.1% lows. The FPS overall remains above a point of jarring stutter, but there's also a measurable dip – and that means exceptionally poor value.
If DOOM is the only game you want to play, SLI just isn't worth it right now. There should be better support forthcoming, but we've heard that song and dance before – like with Just Cause 3 and Assassin's Creed Syndicate.
Comparative Dx12 vs. Dx11 FPS in Ashes of Singularity – GTX 1070 vs. 1080, 970 Fury X
SLI doesn't fare well in the Ashes of Singularity Dx12 benchmark.
Using the DirectX 12 API, the SLI GTX 1070s fall to the bottom of all charts, regardless of settings, showing dismal frametimes (42.54ms average at 1080/high) in Dx12. Against other nVidia devices, this is a major downturn and is a result of SLI. It would actually be better to disable SLI for Ashes of Singularity in Dx12, leaving the second card to rot. DirectX 11 performance is more reasonable; frametimes are closer to the single GTX 1070 (20.46ms for single 1070 vs. 23.02ms for SLI), and scaling is around 5% for 1080p/high. Not good – certainly not worth it – but better than the massive negative hit from Dx12.
As always, this is just one game – don't take too much away from this other than “SLI GTX 1070s are terrible in Ashes of Singularity.”
Comparative Vulkan vs. Dx11 in Talos Principle – GTX 1070 SLI vs. 1070, 1080, etc.
The Talos Principle shows massive gains in Dx11, but clearly unoptimized Vulkan/SLI performance.
At 4K in Dx11, we're seeing scaling of 48.65% in The Talos Principle (1070 to SLI 1070 at Dx11). Against a single 1080, the 1070s push a 28.63% higher average framerate. In Vulkan, we're seeing only a 9.13% gain over a single GTX 1070, with a 20.12% drag behind the single GTX 1080 – very poor value, considering the 1080 is cheaper than two 1070s. That said, this is with Vulkan in Talos Principle, which has already proven to be a maturing deployment.
SLI 1070s are pushing 185.37FPS AVG in Dx11 at 1440p, which nets a 23.9% lead over the single GTX 1080 and a 43.48% lead over a single GTX 1070. Not bad. With Vulkan, we see comparable results to before – negligible gains versus a single 1070 card, and dragging behind the single GTX 1080.