Skull Canyon NUC Review 2/3: Benchmarks (NUC6i7KYK)

In the second part of our Skylake i7 NUC review we run some popular benchmarks on it to get an idea how does it perform compared to the other Skylake NUCs and the previous generation i7 NUC. If you came here via a search engine, I’d suggest you take a look at the first part of the review first. We’re going to run our standard setup which is 3DMark, Cinebench and PCMark plus some extras this time.


Benchmark Results

Intel Iris Pro graphics driver version, Windows 10, BIOS version 34 was used for the following benchmark results. The unit was equipped with 2x4GB HyperX DDR4-2400 CL14 memory modules and Samsung SM-951 NVME SSD drive.


3DMark is a popular benchmarking suite that benchmarks video and gaming performance of the computer. See below for the results and a comparison to the previous i7 NUC and the other Skylake NUCs:

Skull Canyon NUC Benchmarks - 3DMark

As you can see, the i7 NUC gets 45-65% better results when compared to the Skylake i5 NUC and is pretty much twice as fast as the previous generation i7 NUC!

Cinebench R15

Cinebench runs 3 separate benchmarks and gives us figures that are comparable. First a simple 3D car chase that measures mainly the GPU (OpenGL) performance. The result is in frames per second. Second there’s a rendering of 3D model with all cores. This stresses purely the CPU. Finally there’s another rendering of the same model, this time using just a single CPU core.

Skull Canyon NUC Benchmarks - Cinebench R15

The single-core performance is slightly improved when looking at the Broadwell i7, but this time there’s 2 more cores. In the multi-core CPU test the Skull Canyon NUC shows its power. 100% better result when compared to the Broadwell i7 NUC. Of course the Broadwell was a 28-watt CPU and this is a 45-watt CPU. In the OpenGL test the Skull Canyon is 33% faster than the Skylake i5 NUC and more than 100% faster than the previous generation i7 NUC, NUC5i7RYH.

PCMark 8

PCMark 8 is a benchmarking utility to test the performance of the computer in typical office and creativity-related tasks. Below you can see the results of my benchmarking. As you can see, when we leave the GPU-intensive gaming zone, the differences are more subtle. The raw CPU power of the Skull Canyon NUC guarantees the pole position in each of the tests, but the differences are not that big as they were on the previous benchmarks.

Skull Canyon NUC Benchmarks - PCMark 8


I’ve never run Geekbench before, but a reader requested Geekbench results earlier so here they are. Below you can find the 64-bit Geekbench score for NUC6i7KYK. 14028 for multi-core and 3744 for single-core tests. If you’re interested, you can read the full detailed results here.

NUC6i7KYK Benchmark Result, Geekbench 64-bit

ATTO Disk Benchmark

Previously, the disk performance has been locked at 1600 MB/s level. This was the case with the Broadwell i7 NUC and this was the case with the Skylake i3 and i5 NUCs. However, a BIOS update supposedly fixes this for Skylake NUCs in the nearby future (there’s a beta BIOS already). For some reason, I get “only” 680 megs of write speed, but I have not yet explored the reason. It’s the first time I use this drive, so I don’t know if it’s due to the drive or the NUC. Read speeds are in the healthy 2 gigs per second range.

NUC6i7KYK, ATTO results

Fan Noise

After seeing the strange fan behaviour of the Broadwell i7 NUC and considering the fact that the Skull Canyon NUC has a 45-watt CPU in a 0.68 liter case, it seems that fan noise is the question on everyone’s mind. Evaluating fan noise is difficult, as noise is such a subjective topic. I’ll try though. When the CPU is close to idling (ie. when you’re browsing internet, writing documents and doing typical productivity related tasks) the unit is pretty quiet. You can hear the fan, but it’s not annoying. Quite close to the other Skylake NUCs I would say. When you put more stress on the CPU, the fan speed gradually ramps up and it starts to get more noisy. It does not get NUC5i7RYH noisy, but it gets around to the same level as some laptops when under full stress. The CPU gets close to 100 degrees celsius – that’s pretty hot!

If you’ll be gaming with your Skull Canyon NUC or doing constantly some other CPU-intensive tasks, you’ll be hearing the fan. It’s a small unit and it’s a CPU that puts out 40+ watts. There’s no escaping that. It’ll be interesting to see if third parties do come up with fanless or more quiet cases for this NUC.

DDR4-2133 or Something Faster?

The NUC6i7KYK supports officially DDR4 SODIMMs in 2133 MHz variety. However, even Intel said that the unit supports 2133+ MHz DDR4 RAM. I tested the following 3 RAM modules in a dual-channel configuration and ran some benchmarks to see if it really makes a difference which RAM you choose.

I made sure that the memory was detected according to the speed and that I saw in Windows also that the memory speed was as expected. Based on the results, I’d say it really does not matter that much which RAM you choose. I’d say go for the DDR4-2400 CL14 RAM as that seems to be as cheap as the 2133 MHz RAM (and you can always run that in 2133 MHz mode if you like).

Skull Canyon NUC Memory Comparison

Keep on Reading

Other parts of the review:

20 Responses

  1. pat says:

    Amazing review, thanks

  2. johnmflores says:

    Good stuff. Thanks for doing this!

  3. Did you confuse 3DMark 11 with the newer 3dmark? There is clearly no 3dmark11 in your test, that’s a different 3dmark.

    • Olli says:

      Thanks for pointing it out! Don’t know what I was eating for breakfast – It’s been a few years since 3DMark 11 was the thing…

  4. I am confused by your 3dmark scores. Are these the results for the ‘Standard Score’ or the ‘Graphics Score’.

    Also, I am confused by the result of Cloud Gate, your cloudgate result for the iris 580 pro, is lower than the results of the last generation i7, with iris 6200 pro.. and we were expecting a 1.5x increase in cloud gate performance.

    Also it seems odd, you compare it to the skylake i5, with iris 540… would make more sense to compare it to an i5 with the same 580, or atleast the 550….
    And it seems even more odd that when you compare it to the last gen i7…. the braodwell i7.. you compare it to the Iris 6100 pro. Very poor comparisons.

    • Olli says:

      Hello there and thanks for the comment,

      The figures there are the total 3DMark scores. Ie. the “big number” that 3DMark reports after a benchmark run. Here’s an example of one early Sky Diver run that I did where the score is 8150:

      I’m only comparing figures that I have benchmarked myself before. I run this blog as a hobby and getting (buying/borrowing) each NUC for a test is already a struggle. Basically, I’ve compared the performance of the Skylake i7 NUC to the other current NUCs and the previous generation i7 NUC. In my opinion, you buy the NUC for the small form factor. Thus I did not find it meaningful to compare the results to the desktop processors or even to processors available only in laptops. However, the beauty of standardized benchmarks is that the results should comparable by the readers if they can source a result somewhere else.

  5. Also, for your Cinebench table, one column is R15 OpenGL, and there are 2 numbers separated by commas, what are those 2 numbers?

  6. Steven Cardella says:

    I haven’t found this anywhere: When displaying to a 4k display, how well does it deal with streaming video? Any stutters on Youtube or Netflix that are cpu/gpu sourced?

  7. analyticpartners says:

    Hi there – how should I interpret your memory comparison chart? Smaller number better or bigger number better? So for example, is memory read faster or slower on the 2133 compared to 2400 ?

    • nucblognet says:

      In theory you want as high bandwidth (the first number, ie. 2133 vs 2400) as possible, but as low CAS latency (CL) as possible. However, I think that results point out that it doesn’t make almost any difference which memory you get. I’d go with 2400 CL14.

      • analyticpartners says:

        Thanks, and I agree but my question still stands – are smaller numbers “better” (even if minimally), than the larger numbers?

        • nucblognet says:

          That would depend on the test. Basically bigger numbers are better for all of the tests here except the memory latency where you want as low number as possible…

          • analyticpartners says:

            Interesting, I would have thought the opposite. Thanks for clarifying!

  8. Mario T. says:

    Thanks for the Geekbench results. As a long-time Mac user this puts things into perspective for me :)

  9. yt75 says:

    I’m thinking of buying a kabylake NUC, hesitating between i5 and i7, would be to do some 3d design (Rhino 3d) and a bit of video editing (no games at all).
    The i5 would be fine right ? (not so much about the price, but I’m afraid of fan noise with the i7)

  10. Japex says:

    Hi… How is the Skull Canyon going? Any plans to test it with Razer Core or Asus ROG XG2? And also, do you know if there is any plans from intel to update Skull Canyon anytime soon? I’m thinking about getting one for Adobe PS/LR in exchange of my currently MSFT SP3 I7.


  1. May 20, 2016

    […] info dari, Processor Intel Core-i7 generasi Skylake adalah processor utama yang digunakan pada Intel NUC […]

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.