For a while now, NVIDIA and AMD have been churning out their respective refreshed 14nm architecture graphics cards. As established by their rather long histories, the two have been going in completely different routes in terms of releases and priorities. NVIDIA continued with it’s standard top-down refresh cycle, releasing its high end cards first, while AMD kicks off their releases with their mid-range competitor cards.
Historically, AMD has always been the go-to for performance to dollar value, providing excellent performance, at the cost of a little extra power consumption and heat, and slightly less optimized drivers. However, the RX 480 proves to be potent competition for NVIDIA’s previous-generation Maxwell architecture products, coming with competitive power consumption, more VRAM, and better overall performance.
However, AMD’s foothold has been shaken somewhat by the release of NVIDIA’s GTX 1060. NVIDIA’s card comes with two very bold claims: that it matches the performance of the GTX 980, and that it beats AMD’s RX480 by a hefty 15%.
Both cards deliver almost the same performance, although NVIDIA’s offering sees a slight performance advantage in a number of benchmark tests. The main difference between the two, however, is that the GTX 1060 is priced a bit higher than the RX 480, which is making it difficult for buyers to decide which one to get. The GTX 1060 is currently being offered at $249, while the RX 480 is relatively cheaper, coming out to be around $200.
However, if their first claim turns out to be true, its $249 price tag makes it a compelling upgrade, bringing with it an incredible generational performance leap. And to be perfectly honest, it’s a trend they’ve confidently established with the GTX 1080 and GTX 1070 cards.
However, these prices are almost nowhere to be found in the real world You can get an RX 480 4 GB card for $230 once the aftermarket builds really begin to roll out, and the 8 GB variant, for $250 or so.
The Red Team
More importantly, for AMD and for the future, at any rate, does it beat the Radeon RX 480 by 15 percent?
This is important, because AMD has effectively conceded the high-end graphics card war to NVIDIA until the release of their Vega architecture, and the inevitable release of the Radeon RX 490 — which is due to be released much later this year. This means it’s crucial to AMD that they claw back their desktop GPU market share at the midrange and lower end of the spectrum. If NVIDIA boxes them in at the sub-$300 price range, that could spell trouble for the already struggling Red Team.
The answer, looking at various benchmarks, is that the GTX 1060 scores an average of 13.5% higher performance score over the RX 480. The card also has a more rounded overclocking suite, allowing for more headroom for those who want to squeeze a little more performance out of the card. However, the RX 480 is most definitely still a worthwhile card.
DirectX 12 Benchmarks
Games utilizing DirectX 12 aren’t as widespread yet, but within 2 or 3 years, it will be the standard.
We start by looking at FutureMark’s new Time Spy synthetic benchmark, which is essentially the DirectX 12 equivalent of their ever-popular FireStrike benchmarking software. You will see that despite the rather skewed appearance of my chart, both the overall scores as well as the graphics scores of the GTX 1060 and the RX 480 are quite close.
As a matter of fact, the difference is well within the margin of error and variance that one can expect to observe between runs. Just like FireStrike, Time Spy renders a pretty accurate approximation of a system’s overall gaming performance.
Technically NVIDIA wins this one, but it’s pretty much neck and neck in terms of 3DMark scoring.
On to performance in DirectX 12 games:
The RX 480 pulls ahead of the GTX 1060 significantly in frames per second in Hitman.
Rise of the Tomb Raider comes along to upset what would be a pretty convincing DirectX 12 win for AMD. Here, the GTX 1060 pulls way ahead, registering a 22% performance increase over the RX 480.*
* although this may be because of the use of proprietary NVIDIA technologies, which make AMD cards struggle. Watch this video for more information.
Before we take a look at DirectX 11 game performance, let’s circle back and get some synthetic results from FireStrike and Heaven 4.0.
FireStrike is a synthetic benchmark that tests a system’s overall graphics capabilities, but it’s not a replacement for real-world gaming results. It does tend to provide a fairly accurate representation of the competitive advantage one graphics card can have over another, which is probably why it is so popular. You can’t hand down a verdict based solely on this test, but it’s definitely a solid indicator of expected performance.
Unigine’s Heaven benchmarking tool utilizes something called the “Unigine Engine”, and allows tests to be run in DirectX 10, DirectX 11, or OpenGL modes. Here we see the Heaven benchmark in DirectX 11 mode, utilizing a lot of tessellation and screen-space ambient occlusion (SSAO). Just like FireStrike, and prety much all synthetic benchmarking tools, it’s not always indicative of real-world gaming scenarios, but it does provide that performance snapshot we need to start identifying trends.
Will the GTX 1060 create a defining lead in DirectX 11?
It definitely does, and clocks approximately 23% better performance under Heaven 4.0 Extreme settings at 1080p, and about 15% more frames at 1440p. This can be largely attributed to the fact that NVIDIA has always had a really strong tessellation engine.
However, as far as synthetic tests go, FireStrike is generally more accurate as a measure of typical performance.
On to real-world gaming benchmarks:
DirectX 11 Benchmarks
The cards are neck and neck in Middle Earth: Shadow of Mordor, with a variance of one or two fps observed between them. Given the price difference between the two cards, though, it’s fair to say that the Radeon RX 480 has the edge in Shadow of Mordor
However, Grand Theft Auto V gives the advantage to the GTX 1060 at both tested resolutions.
While we can’t see into the future, GTX 1060 exhibits a little stronger performance in the SteamVR Performance Test than the RX 480.
Despite these apparent disadvantages, the RX 480 continues to be a great contender, simply due to the 25% savings buyers will be getting with the purchase. And after all, the card still performs very well, and beats the GTX 1060 in 9 out of 23 benchmark tests conducted.
If we segregate the list into DirectX 11 versus DirectX 12, AMD wins 21.7% of the DirectX 11 comparisons and 67% of the DirectX 12 / Vulkan comparisons. That’s a non-trivial difference, and it could speak to the long-term strengths of the GCN architecture over NVIDIA’s Pascal.
Deciding which of the two cards is better now simply boils down to preference. The GTX 1060 is most likely good for more seasoned builders and gamers who are planning to overclock their cards down the road, while the RX 480 is for those who want to get continuous performance without having to fuss over upgrading their gaming rigs anytime soon.
Either should be just about as capable. Get whichever one you can find first and comes with a cooler you like.
The RX 480
Potentially will perform better in upcoming DirectX 12 / Vulkan games (for example, the new Doom), and will have the ability to benefit from a Freesync monitor rather than paying a large premium for a Gsync equipped one in the future.
If you were to get the 4GB version of the card, the RX 480 hands down wins the price/performance competition, but nobody seems to be selling those.
The GTX 1060
Performs better in today’s DirectX 11 games, but does not benefit as much from DirectX 12 / Vulkan games as the AMD card.
The card features a slightly lower power consumption, and thus lower operating temperatures, and also boasts more overclocking overhead.
Aftermarket 480’s close the gap in DirectX 11 performance further, but the GTX 1060 still pulls ahead slightly.
The RX 470 has a similar appeal as the RX 480. Once prices stabilize, it’s going to be extremely hard to beat the value of the 4GB 470.
You can’t go wrong with either choice, really.