Power Testing – SLI GTX 1070s vs. GTX 1080, GTX 1070
We're not running thermals for this setup for a few reasons – mostly that the cards aren't the same design, and that we already ran extensive thermal testing on the FE card in our 1070 review. We'll be looking at the MSI card's thermals separately, when its standalone review is posted.
Power draw, however, was run through the ringer for this setup – and it's important, too. With SLI configurations, power draw increases greatly and can be the teetering point between one PSU and another, potentially offsetting price further when compared against single GPU configurations.
Here's a look:
The GTX 1080 FE runs (stock) at 270W for peak system load using GFX Test 2 in 3DMark FireStrike. A single GTX 1070 FE (stock) operates at 227.19W peak system load – a reduction of ~43W. Looking to the SLI configuration, we're now drawing more power than the single GTX 1080 – 347.72W peak system load. That's an increase of 41.93%, or ~120.53W. Whether or not the increase is important to your build is personal preference – but the power increase is often greater than the performance increase, depending on game, so it's worth keeping in mind.
Conclusion: Are SLI GTX 1070s Worth It or Single GTX 1080?
Answering this question is tough, but we face it with every SLI benchmark. In this instance, the question is whether or not there's better value in spending ~$600-$700 on a single GTX 1080 or $380-$450 on 2x GTX 1070s. Based on the AIB partner cards we know to be hitting the market, let's just make it easy and assume an average GTX 1080 price of $650, and an average joint price of 2x GTX 1070s at roughly $800-$850 – so that's a $150-$200 gap between the two main options. To be worth it, you'd want to feel like you're getting an additional $150 of performance out of two GTX 1070s.
In some games, such a gain is reasonably present. We saw upwards of a 30% lead over a single GTX 1080 in a few instances, which is no small jaunt considering the already impressive performance of the 1080. Still, though, SLI and CrossFire run the same risk they always have: Poor optimization earlier in launch cycles or a complete lack of any meaningful support in some games (like Just Cause 3, Assassin's Creed Syndicate, and Doom). Some of this will simply never be resolved – it's the way the game was made, and the devs aren't going to be able to do enough optimization to meaningfully impact performance. Some of the detriment is just launch cycles – nVidia tends to put out SLI optimizations over time, as do devs, and games can improve in overall performance throughput.
Inevitably, there will be times when multi-GPU configurations simply don't work well (or at all) and necessitate that the second card is disabled. That's never fun, but there are also gains – like in Mordor, with >70% scaling – that can counteract the occasional loss. Ultimately, it's up to you to look at data on a game-by-game basis and determine if SLI makes sense for you. If you're very heavily focusing on one game – maybe hoping to competitively play Infinite Warfare or BLOPS3 for a year or more – it could make sense to go SLI, with an occasional toggle for unsupported titles. If playing a wider swath of games, we'd recommend very seriously considering a single GTX 1080 instead – or waiting for the RX 480s to come out, and then seeing if AMD's claims of highly performant CrossFire framerates are valid across all titles.
Overall, it's still a mixed bag of results. Some games show less than 10% scaling – or even “negative” scaling, particularly with poor frametimes – and others show 40-50% (or, rarely, greater) scaling in SLI. Look through the data and make a decision.
That said, if you're the type who will be greatly upset every time SLI has to be disabled and you're left with a $400 brick in your PC (until it can be used again), maybe just avoid multi-GPU configurations altogether. A single card is certainly the “safe” and easy route, and the 1080 is a good single card as an alternative – and cheaper than two 1070s.
If you like our work, please consider supporting us on Patreon.
Editorial, Test Lead: Steve “Lelldorianx” Burke
Sr. Test Technician: Mike “Budekai?” Gaglione
Test Technician: Andie “Draguelian” Burke
Video: Andrew “ColossalCake” Coleman