Nvidia’s GTX 1080 and 1070 may win the overall performance competition, but they don’t represent the bulk of the GPU market. According to both AMD and Nvidia, they sell far more GPUs in the midrange than the luxury market. Today we’ll be reviewing Gigabyte’s GTX 1060 G1 Gaming 6GB to see how it compares with a range of current and previous-generation hardware from both AMD and Nvidia.
The GTX 1060 is a testament to the idea that less is more. At first glance, this card doesn’t look competitive compared with what both AMD and Nvidia have previously shipped. The 1060 has 1,280 cores, 80 texture mapping units, and 48 ROPS (1280:80:48). Compare that to the RX 480, with its 2304:144:32 configuration, and it’s obvious we’ve got two fundamentally different 14nm architectures in play. Even allowing for differences in clock speed, AMD built a card that emphasizes texture throughput and memory bandwidth, while Nvidia’s GPU focuses on pixel fillrate.
The Gigabyte G1 Gaming we have on the bench today has a maximum boost clock of 1847MHz in OC Mode and a top clock of 1809MHz in Game Mode (keep in mind that Nvidia’s Boost technology can exceed these clock speeds under the right circumstances), while the RX 480 tops out at 1266MHz. The 6GB of memory is a step up from previous 28nm GPUs at this price point, but the memory bus is narrow, at just 192 bits. Total bandwidth for the GTX 1060 is 192GB/s compared with 256GB/s of bandwidth for RX 480 and 224GB/s for GTX 980.
If you’ve read our coverage of the Gigabyte G1 Gaming GTX 1070, you’ll recognize the GTX 1060’s cooler and fan layouts. Both of the G1 Gaming GPUs pull air from inside the case rather than through a blower, and cuts the number of fans from three on the GTX 1070 to two here. Otherwise, the two coolers are nearly identical. Both use heatpiped heatsinks, fans that spin in alternating directions, LEDs that light up when the fans are stopped at idle, and an identical fan blade design. Power is provided by a single eight-pin PCI Express connector for both the GTX 1060 and 1070. This is unquestionably overkill for the GTX 1060, which has a rated TDP of just 120W, but it could give the GPU a bit more oomph when overclocking.
The Gigabyte G1 Gaming GTX 1060 and G1 Gaming 1070 both use the same Xtreme Software Tuner, so I’ll refer you back to the GTX 1070 review if you have specific questions. Overall I quite liked the Xtreme Software Tuner — I find the UI easier and more intuitive to use than EVGA’s Precision X, though this is a matter of personal taste. You can customize the GTX 1060’s visual effects and patterns the same way you can with the GTX 1070.
The other thing the 1070 and 1060 share is support for Nvidia’s Turbo Boost 3.0, and that’s worth touching on again. While Gigabyte specifies a maximum boost clock of 1809/1847MHz depending on your chosen mode, that’s not the GPU’s actual max clock.
Nvidia’s Boost 3.0 allows the GPU to burst higher than its actual listed boost clock, sometimes by extensive margins. Our tests show that the GTX 1060 G1 Gaming can actually hit 1949MHz and hold this clock under load, though it’ll drop back to 1923MHz in some titles. This is a further 5.5% overclock above listed clock.
Test setup, comparison points
All GPUs were tested with a Haswell 5960X eight-core CPU and an Asus X99-Deluxe motherboard. Nvidia’s 368.81 and AMD’s 16.7.3 driver were used for the appropriate cards. All of our power consumption figures were run using an Antec 750W 80 Plus Gold PSU, not the larger-but-less efficient Thermaltake 1200 80 Plus we used for a number of older reviews.
Unlike the GTX 1070, the GTX 1060 faces multiple potential competitors from both Nvidia and AMD. We’ve compared the GTX 1060 against AMD’s RX 480, as well as the older AMD Fury (now down to as little as $ 310 due to price cuts), Nvidia’s own GTX 980, and the Hawaii-derived R9 390.
Hovering over any comparison of modern GPUs is the specter of street pricing, and this is where things get ugly. In theory, the RX 480 is a $ 240 GPU competing against a GTX 1060 6GB expected to start at $ 250. In reality, the RX 480 has been bouncing around between a low of $ 259 and spikes as high as $ 300. As we discussed earlier this month, AMD’s current GPU stack has skewed 10-15% above list price for nearly ever card and RX 480’s selling at the official SEP (Suggested E-tail Price) of $ 239 have been very hard to find.
As of this writing, the RX 480 tends to sell for roughly $ 259, the R9 390 starts at $ 289, and the Sapphire Nitro Fury is $ 309. The Gigabyte GTX 1060 G1 Gaming is a $ 289 card and can be purchased at that price. That’s a substantial discount over the GTX 980, which typically sold for $ 450+ when Maxwell cards were Nvidia’s top-end family. GTX 1070 cards are slipping towards their $ 379 MSRP but haven’t hit it yet; the least expensive cards available today are $ 399.
Our price analysis will focus on where these cards are priced today, but will acknowledge MSRP prices as well. We’re focusing on 1080p with this coverage due to the popularity of that resolution and the fact that neither the RX 480 nor GTX 1060 are really 4K-capable GPUs. We’ve included the GTX 1070 as an example of a “halo” card, to illustrate how much additional performance an extra $ 100-$ 120 can buy you.
4K is in a bit of an odd spot right now, particularly at this price point. AMD explicitly marketed the RX 480 as a GPU for 1080p to 1440p and our spot checks confirmed this. Historically, GCN GPUs have scaled better at 4K than their Maxwell counterparts. The RX 480, in contrast, does not, probably because of its 32 ROP limit and comparatively low pixel fill rate (the RX 480 is a 2304:144:32 design, while the R9 390 is a 2560:160:64 architecture). Whether any of these cards are good enough for 4K depends a great deal on how much detail you are willing to sacrifice for higher resolutions. The minimum GPU we’d realistically recommend for 4K would be the GTX 1070, but that doesn’t mean we can’t get some useful mileage out of comparing the GTX 980 against the GTX 1060 to see if any bottlenecks in Pascal become apparent. If you buy the Gigabyte GTX 1060 G1 Gaming, could the card’s modest specs become an issue a few years from now?
In a word: No. While we wouldn’t recommend any of these GPUs for high-end 4K, you aren’t giving up performance by going with the GTX 1060 as compared with the GTX 980. The Strix Fury is faster than either card in every test, but hold that thought — we’re going to swing back around to talk about the Fury.
We measure load power consumption via Metro Last Light Redux at 1920×1080 with Very High detail and SSAA enabled. We loop the benchmark three times and record our power consumption values during the third run to give the card time to heat up.
The R9 390 draws more power than even the Asus Strix Fury, even though that card packs considerably more punch. HBM saves AMD a considerable amount of power. Let’s see what happens when we factor power efficiency into the equation and calculate how many watts, on average, each GPU requires per average frame of animation.
I prefer watts-per-frame to frames-per-watt because it lets us talk about numbers that are larger than 1. In this chart, smaller numbers mean the card is more power efficient. The Radeon Nano remains the most power-efficient GCN GPU AMD has ever built and is within 90% of the GTX 1060’s absolute power efficiency. The GTX 1070 wins the overall power-efficiency prize.
Price / performance ratios
The GTX 1060 has debuted in a crowded field of still-potent older GPUs and fresh competition from AMD’s 14nm midrange. Both AMD and Nvidia have struggled to deliver their GPUs in quantity and at their official price points. Unfortunately for AMD, Nvidia has done a better job of ensuring its midrange GPUs are available and hitting their price targets. We doubt it’s an accident that the GTX 1060 has come down to MSRP so quickly, given that it’s also the only category where AMD is competing on 14nm at the moment, but we’ve got to make our recommendations based on what’s happening in the market today. Right now, the cheapest RX 480 you can buy is $ 259, not the $ 239 AMD targeted.
The Gigabyte GTX 1060 G1 Gaming is generally faster than the RX 480 across our entire suite of DX11 titles. The two cards are closest in Company of Heroes 2, where the GTX 1060 leads by 3%, and farthest apart in Total War: Rome 2, where the GTX 1060 leads by 26%. The R9 390 offers stiffer competition, but the GTX 1060 still leads the 28nm GCN GPU in all of our DX11 benchmarks, all while drawing much less power.
The most interesting match-up at the $ 300 price point is between the Gigabyte GTX 1060 and the Asus Fury Strix. At $ 310, the Fury is just 7% more expensive than the GTX 1060, and it offers higher performance in multiple games in our test suite. While it doesn’t win every DX11 benchmark, it ties in COH2, Dragon Age: Inquisition, and Total War: Rome 2. Shadow of Mordor, BioShock Infinite, and Metro Last Light Redux are all wins for Team Red.
The RAM limitation on Fury makes this a tougher call than it might be otherwise. All of AMD’s Fury and Nano GPUs are limited to 4GB of RAM. That’s fine for 1080p or 1440p today, but gaming memory requirements do tend to tick upwards as time passes. Both AMD and Nvidia have started equipping GPUs in the $ 200 to $ 300 price segment with more than 4GB of RAM, which implies they think larger frame buffers will matter to this price segment in the future.
If you’re the kind of buyer who buys a new GPU every 18 to 24 months, aren’t particularly concerned with power consumption, and plan to spend those months gaming at either 1080p or 1440p, there’s a good case to be made for the Asus Strix Fury. Its overall performance is good and the cost premium over the Gigabyte GTX 1060 G1 Gaming is small.
The GTX 1060 isn’t without advantages of its own, however. Its larger VRAM buffer means there’s less chance a cutting-edge title will run into RAM trouble a year or two from now, its power consumption is lower, and it’s a bit cheaper, at least in the US. If you plan to keep your GPU for three to four years and don’t want to run into memory issues, the GTX 1060 may be the better long-term bet.
As for DirectX 12 and Vulkan, as much as I’m tired of typing this, it’s still hard to get a fix on the long-term performance implications of these APIs. Tomb Raider’s DX12 version is still quite Nvidia-friendly on Pascal cards. Other titles, like Ashes of the Singularity and Doom, show very strong results for AMD. I’ve a long-standing policy of recommending GPUs based on where the market is now, not what benchmark scores might show six months from now, and while DX12 and Vulkan are coming along nicely, DX11 continues to be the dominant PC API.
The Gigabyte GTX 1060 G1 Gaming packs all of the features we liked about its GTX 1070 big brother in a lower price point. The card’s modest stats hide a GPU with a hefty punch. It’s performance is equal-to or better-than the GTX 980 in every test we ran, even at 4K, where we’d expect Maxwell to stretch its legs a bit more. The GTX 1060 is proof that specs don’t always tell the whole story and it punches harder than you’d expect as a result.
It’s also a GPU that’s difficult for AMD to match. Fury’s $ 310 price point competes well against the Gigabyte GTX 1060 G1 Gaming on price and performance, but is limited to a 4GB frame buffer and draws considerably more power. It’s also a physically larger card, which can limit its usefulness in some cases (pun intended). If the RX 480 was reliably selling for $ 239, the $ 50 price delta between Polaris and this particular iteration of Pascal would give AMD some breathing room. As things stand, Nvidia and Gigabyte have staked out a very strong position for themselves in the midrange market. We’re going to keep our eye on GPU prices as they continue to evolve, but for now the Gigabyte GTX 1060 G1 Gaming is in a class of its own.