Hardware Showdown: GTX 1060 6GB vs. RX 580 8GB in 2019

In an era of Nvidia Turing and AMD Navi cards, old Pascal-based GTX 1060 and Polaris-based RX 580 still seem to provide enough graphics processing power in order to achieve enjoyable frame rates at 1080P and even 1440P resolutions. Although discontinued, refurbished GTX 1060s with warranty can be bought for 200€ (6GB model) on eBay in Europe, while brand new RX 580s go for as low as 170€ at retailers such as German Mindfactory.de or Czech-Slovak Alzashop.

Taking into account how cheap these cards are now, they are still worth considering to buy in 2019, especially if you plan to build a budget PC or if you plan to play games on 1080P resolution. Both cards have their advantages and disadvantages so main goal of this article will be to provide detailed analysis of graphical performance of these cards and determine which one is a better option for buying.

Nvidia GTX 1060 6GB and AMD RX 580 8GB specifications

Methodology

Since I personally don’t own Nvidia GTX 1060 6GB model, nor do I own enough hardware equipment or video games to properly test performance of GTX 1060 and RX 580 cards, for purposes of this study I have chosen a method of meta-analysis with which I would be able to conduct a proper quantitative research. What this means is that I basically chose to combine benchmark results from other web-sources and analyze them using basic inferential statistics. First step was, however, formulating null hypotheses which are the following:

1. There is statistically no significant difference in graphics performance between Nvidia GTX 1060 vs AMD RX 580 at 1920×1080 resolution.
2. There is statistically no significant difference in graphics performance between Nvidia GTX 1060 vs AMD RX 580 at 2160×1440 resolution.

After formulating research hypotheses, next step was defining inclusion and exclusion criteria by which benchmark web-sources will be selected and analyzed for purposes of this study. After careful thinking, these are the criteria that I adopted for conducting this study:

• Benchmark results must contain data that refer to Nvidia GTX 1060 6GB or AMD RX 580 8 GB graphics cards.
• Benchmark results must refer to video game titles released in period from 2015 to 2019.
• Data should describe performance at maximum settings and at 1080P or 1440P resolutions.
• Published benchmark results must not be older than 1st January 2019.
• Testing rigs used for benchmarking must be based on 4th generation of Intel Core i7 processors or newer, with overall PassMark Score of at least 11500 or higher, and Single Thread Rating of at least 2000 or higher.
• Testing rigs must contain at least 16GB DDR4 RAM and must run on Windows 10 OS.

Defining inclusion and exclusion criteria for benchmark testing rigs was in my personal opinion very important, since graphics card performance can differ very much depending on whether CPU bottlenecks the card or not. My criteria included even the old Intel Haswell-based CPUs, which might seem problematic, however, I only included CPUs with PassMark Score higher than 11500 which are on par with newer models such as i5-8400 and as such seem to be viable for this research.

I excluded benchmarks based on AMD processors, but not because they are not worthy, but because of the reason that many testers still only use Intel processors and because I wanted to achieve higher internally validity since all of my results were going to be based on one processor manufacturer. Finally, it is worth to say that I excluded all benchmarks older than 1st January 2019 since benchmarks prior to that date definitely used older GPU drivers and graphics card performance was probably worse in some video game titles compared to performance based on newer drivers.

After defining inclusion and exclusion criteria, I have determined additional research strategies that would be implemented in my research. Here they are:

• Sample of benchmarked video game titles should consist of titles that were published in period from early 2015 to late 2019. For each year there should be analyzed six titles that are different by genre and were published by different publishers.
• Data referring to titles published before 2017 should be shown separately from newer titles since performance of GTX 1060 vs RX 580 is expected to be higher in these titles because at the time these cards were considered to be upper tier cards.
• Average calculated results should be based on at least two different benchmark sources.

Next step was to locate and identify web-sources from which benchmark results would be extracted and analyzed. After careful exploration and examination, these are the sources that were selected for data extraction:

• https://www.eurogamer.net/
• https://www.techspot.com/
• https://www.pcgamesn.com/
• https://www.guru3d.com/
• https://www.pcgamer.com/
• https://www.anandtech.com/
• https://www.tomshardware.com/

From these eight sources, Guru3D and TechSpot had priority in data extraction since both of these websites had a large amount of benchmark data and as such proved to be very reliable. Data from other sources had priority if it was more recent or if there was no data to be found for a particular video game title at primary sources.

After extracting data from sources, I conducted a descriptive and inferential statistical analysis of obtained data. In the continuation you can see the results.

Performance at Full HD (1920×1080) Resolution

When talking about 1080P performance in pre-2017 video game titles, both GTX 1060 6GB and RX 580 8GB offer similar performance. Both cards are able to achieve more than 60 or even 70 FPS in almost all of the titles. Only exception seems to be Deus Ex: Mankind Divided in which RX 580 barely surpasses 60 FPS on average, while GTX 1060 struggles to obtain 55 FPS on average. Since I wasn’t able to obtain fresh data for at least six relevant titles from 2015, I have decided to include recent results from 2013’s Crysis 3 which still seems to be a relevant video game for benchmarking. Both cards easily achieve 70 FPS in that title, but GTX 1060 seems to have an upper hand in performance. Overall, Grand Theft Auto V seems to favor Nvidia’s card, while Battlefield 1 performs much better on AMD card. Largest performance difference between two cards is best seen in the case of 2016’s DOOM, in which RX 580 outbeats GTX 1060 by a large margin.

Contrary to pre-2017 titles, GTX 1060’s and RX 580’s performance in 2017’s and later video game titles is somewhat different. Both cards still prove to be solid 1080P performers at ultra settings, however, certain newer titles seem to put a lot of pressure on them. In particular, games such as Assassin’s Creed Odyssey, Metro Exodus and Red Dead Redemption 2 run very poorly on these cards.

One reason for that is that these games seem to be very badly optimized console to PC ports. The other reason is that these games probably utilize newer rendering engines that don’t run well on pre-2018 graphics cards. Clear example of badly optimized game is Borderlands 3, which, considering its visuals, offers very disappointing performance on both cards. However, disabling volumetric fog effect in this game seems to boost performance a lot, so results shown here shouldn’t be taken for granted.

Overall, both cards seem to show similar performance, but RX 580 is able to securely pass the 60 FPS mark in most titles, while GTX 1060 is not. Finally, it is worth to note that RX 580 performs better than GTX 1060 in newer titles, especially in games that utilize DirectX 12 and Vulkan APIs. Relative performance of RX 580 compared to GTX 1060 for all 28 analyzed games is summed up in chart below:

Analyzing all of the benchmark data at 1080P, RX 580 proves to be 2,5% faster than GTX 1060 in most titles. This is a little bit surprising since RX 580 was performing worse than GTX 1060 in its early 2017’s days. Over time, AMD clearly perfected its drivers in order to utilize better its own hardware and thus achieve better performance overall. RX 580 proves to be better option when playing Battlefield series games, while GTX 1060 performs better in titles such as Destiny 2 and Final Fantasy XV.

Compared to Nvidia’s card, RX 580 is somewhat better suited for running newer titles, especially ones that utilize newer APIs. This fact gives RX 580 an upper hand in 2019, especially if combined with FreeSync monitor, because it will easily achieve smooth gameplay in newer video game titles. When visualizing just how much RX 580’s performance is better (or not) than GTX 1060’s, it is worth to look at the following graph showing normal distribution curve of their benchmark results:

As you can see, both curves practically overlap each other, which means that both cards preform almost identically in 28 benchmarked titles. Average total frame rate for GTX 1060 6GB was 65,84 FPS, while for RX 580 was 67,62. RX 580 does have slightly higher 28 game average, however, that difference is barely noticeable and in the end could be attributable to measurement error. In order to really determine whether RX 580 is significantly better than GTX 1060 6GB, I have decided to conduct a student’s t-test on obtained data. Here is a formula I used and a table showing all relevant information for interpreting my t-test results:

Since calculated t-score didn’t surpass relevant t-value, which at 27 degrees of freedom and two-tail significance level of 5%, has amount of 2.052, then null hypothesis is not rejected. What this means is that at 5% risk of error, it is safe to conclude that there is no statistically significant difference in performance between these two cards.

Conclusion No. 1: GTX 1060 6GB and RX 580 perform about the same at 1080P.

Performance at Quad HD (2160×1440) Resolution

When taking a look at 1440P performance in pre-2017 titles, both GTX 1060 and RX 580 perform quite well. Both cards are able to run Fallout 4, Battlefield 1, Grand Theft Auto V and DOOM above 60 FPS. RX 580 tends to fare better in Tom Clancy’s The Division compared to GTX 1060, while Nvidia’s card has overall better frame rate in GTA V.

Performance of analyzed cards in 2017’s and newer titles at 1440P proves to be troublesome. Both cards have trouble achieving 60 FPS on average and tend to fluctuate in 35 to 45 FPS area. Extremely miserable performance is seen in titles such as Red Dead Redemption 2 and Metro Exodus. Unpleasant surprise is Total War: Warhammer II, which on both cards performs around 35 FPS on average. Both cards easily hit 60+ FPS in Forza Horizon 4, while RX 580 outperforms GTX 1060 in Battlefield V for around 10 FPS more on average.

When looking at relative 1440P performance chart, it can easily be seen that RX 580 beats GTX 1060 in some titles for around 20%, and in a lot of them for 10 to 14 percent. GTX 1060 6GB, however, shows more strength when dealing with games such as GTA V or Warhammer II. Some titles, such as Assassin’s Creed Odyssey or Tom Clancy’s Ghost Recon: Wildlands tend to perform the same on both cards at 1440P. On average, RX 580 proves to be 3% faster than GTX 1060 6GB at 1440P. Benchmark results distribution curve for both cards proves to be somewhat different at this resolution than was in the case with 1080P. Here is how it looks:

Although curve for GTX 1060 has higher peak than the curve for RX 580 (which is here insignificant since it is tied to the surface under the curve which must be equal to 1), it tends to fall down much quicker after its mean, than the curve for RX 580. This means that RX 580 seems to achieve higher frame rates than GTX 1060 at 1440P resolution. Average 28 games frame rate for GTX 1060 at this resolution was 45,87 FPS, while for RX 580 was 47,32 FPS. In order to determine whether is RX 580 significantly better than GTX 1060 6GB, I have conducted another t-test with the same formula mentioned above. Here is the table showing the data:

Since calculated t-score surpassed t-value, which for its circumstances amounts to 2.052, then we reject null hypothesis and accept the fact that there is a statistically significant difference between average results of two cards at 1440P with a level of error of 5%.

Conclusion No. 2: RX 580 performs better than GTX 1060 6GB at 1440P resolution.

Conclusion

So, overall benchmark results for GTX 1060 6GB and RX 580 in 28 games, and 1080P and 1440P resolutions can be summed up in this chart:

Statistically, both cards perform the same at 1080P resolution, but RX 580 is a somewhat better performer at 1440P resolution. Both cards are great mid tier mainstream products that offer great price to performance value. It is necessary to stress that GTX 1060 6GB is almost one year older than RX 580, so anyone who bought it back then had the ability to enjoy almost the same level of performance that RX 580 delivered 11 months later. GTX 1060 6GB is thus a great product that paid itself off really well. However, in 2019 it does show its age, as overall frame rates dip below 60 FPS on average at ultra settings. RX 580, on the other hand, still shows to be a solid performer that even in newer titles, based on DirectX 12 and Vulkan APIs, is able to achieve satisfactory performance.