Coffee Lake i5 NUC Review (NUC8i5BEK / NUC8i5BEH)

Performance and Benchmarks

Henri from Terastore, the company that borrowed me the NUC for this review, had preinstalled Windows 10 on the NUC before borrowing it for me, so I didn’t have to do that this time, however, as seen on the i3 review there are no issues during the installation of Windows 10 on the NUC. The graphics driver version and BIOS version 0048 were used for all the following benchmarks. The system was equipped with a 2×4 GB DDR4-2400 RAM modules ($79.99 on Amazon) and a Samsung 970 EVO 250 GB M.2 NVMe SSD.



3DMark is a popular benchmarking suite that benchmarks video and gaming performance of the computer.

In the Cloud Gate test we see significant increase over the i3 model that itself showed pretty respectable performance in these tests when comparing to the previous generation of NUCs. In the Sky Diver and Fire Strike tests there are only much smaller differences between the i3 and the i5 NUCs (around 10 percent). Both NUCs have the same Iris Plus 655 GPU, so it’s understandable that the results are similar.

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.

The OpenGL test that stresses mainly the GPU is a similar story again: the results are a bit better on the i5 vs. the i3, but only by approximately 10%.

The raw CPU power advantage of the i5 model becomes apparent in the CPU tests. The single core result is almost identical, but the multi-core test results is 100% better. This makes me think that the i3 and the i5 are really very similar CPUs the only difference being the number of cores – the i5 has double the number of them.

Geekbench 3

The Geekbench 3 result supports that logic. The single core result of 4154 is only a little bit better (probably due to 0.2 GHz higher max frequency), but the multi-core test (result 15509) is where the difference is again about 100% better.

For those interested, the Geekbench 4 results for the NUC8i5BEK were 4885/16912.

Passmark CPU test

The Passmark test is often used to compare the CPU performance. The score for PassMark CPU performance test was 11289. Here’s a link to the full results.


Using the Intel Extreme Tuning Utility (XTU) you can alter (among other things) the power limits of the CPU package. All the benchmarks above have been taken using the default settings. The default settings will limit the power to a certain rate. At that rate the CPU heats up to 80 degrees Celsius, so there’s no thermal throttling. At that setting the NUC eats approximately 50 watts of power. By increasing the power limit by 20 watts I’m able to make the NUC consume 70 watts of power and retain a higher clock rate. However, the CPU will heat up to 100 degrees Celsius before the thermal throttling kicks in. The possibility to use XTU to tune the CPU performance is probably interesting in the cases where you’re able to provide better cooling than the stock fan assembly does. In the standard case I would not like to yank it up from its default setting as I much prefer a cool PC over a bit faster but noisy one.

Thoughts on Performance

The four cores in this NUC perform each about on the same level as they do on the Coffee Lake i3 NUCs. There’s just two more of them. So in theory you should be able to see twice as much performance when using applications that can spread their load nicely over multiple cores. Of course this is not the case for many applications and then again, if you want to do some gaming the GPU is actually the bottleneck. The GPU is exactly the same as in the Coffee Lake i3 NUCs and in the Coffee Lake i7 NUC (ok, clocked a bit higher on i7). In some GPU intensive tasks the advantage was only about 10% for the i5 vs. the i3.

It’s also worth pointing out that this i5 NUC is about on par with the Skull Canyon NUC (NUC6i7KYK) that was Intel’s first gaming NUC.

If you’re interested in how this NUC does in a couple of games, I wrote a separate article on that topic.

38 Responses

  1. omnium says:

    Again: Thank You very much for this great review!

  2. Zokkel says:

    Thanks again for your review. Very curious about the fifa18 benchmarks. Already have my nuc8i3 and 970 Evo, but still waiting for my ram…

  3. Jason says:

    Another solid review and much appreciated! Will you be posting a review of the i7 version? If so, when do you expect that to be done?

  4. Jacob says:

    Much appreciated review! I find the performance with relation to power consumption very interesting. Especially when comparing to the i3 version, where the i5 perform more than twice the perfromance. I suppose the cinebench score of 779p is with power consumption equal the prime95 of 74W, i.e. with the higher power limit? What is the steady-state performance of the i5 version when consuming around 50W, which is also equal to the i3 version during the same load?

  5. TOM says:

    So if you’re looking to use the NUC primarily as an emulation machine, you can just go with the i3-version? Seeing how must emulators are not optimized for or don’t support a multi-core architecture.

    Btw, have you heard anything about a possible passive, fanless case for these new NUCs?

  6. Juan K says:

    As usual, a great review. I bought my Skylake i5 NUC 2.5 years ago after reading your review of that model, and I still read your blog entries, just in case I decide to upgrade my little box.

    I have a request, and I hope is not a weird one: as the CPU has 4 physical and 8 logical cores, could you disable hyperthreading, perform some tests and compare them to the “normal” (with HT enabled)? I’m mainly curious about temperaturas and power consumption.

    Many thanks in advance.

    • Olli says:

      Here’s NUC8i5BEK Geekbench 3 results with HT enabled/disabled:
      Cinebench multi-core result was 575 with HT disabled. Graphics related tests (Cinebench OpenGL, 3DMark etc) show only a minor drop as the GPU is the likely bottleneck there. When running applications that handle multiple cores, the power consumption and fan speed seem identical. For example when running Prime95 there’s no difference in power consumption nor fan noise. I suppose that is because the CPU core frequency gets adjusted as high as possible before a certain power consumption (read: heat generation) level is reached. If you disable HT, there might be less load on the core, but that pushes the CPU core frequency a bit higher so the power consumption becomes identical.

  7. TOMillr says:

    How about the included HDMI-CEC capabilities? Can the NUC automatically turn on a connected TV when cold booting? Or is this feature only available in sleep states?

    • Zokkel says:

      It turns my TV on (when cold booting), but turning off is another thing. Sometimes it works, sometimes it doesnt…
      I have a Samsung KS7000 TV

      • TOMillr says:

        Thanks for the feedback. Does it also switch the HDMI input to the one the NUC is connected to?

        • Chorizo says:

          Yes, you can choose the HDMI input in the BIOS of the NUC. So if you connect it to your TVs HDMI2 you should set HDMI2 in BIOS. Or you might be left wondering why the TV always goes to HDMI1 when you start the PC…

  8. Poli says:

    Great blog. Keep it coming! Looking to buy a NUC so this blog really aids my choices!

  9. Padaung says:

    Great review as always, thank you.

    If you find the time and means, I would be very interested in a performance review update for this machine (or the i3) with an external graphics card plugged in via the thunderbolt port (if these machine are compatible with such a setup of course…)

  10. SZQ says:

    I ended up buying the i5BEK from Amazon US as it received a little price drop recently to $369. I was planning to wait for the Black Friday sales but the temptation was too great 🙂

    Ordered yesterday and should receive tomorrow (US -> Japan, impressed with Amazon logistics). However I will buy the SSD and RAM locally as those parts may have a higher chance of issues. I’ll probably go with the popular Samsung 970 Evo 250GB NVMe. Not sure about the RAM but will get 4GBx2 for dual-channel.

    Thank you for the reviews, they really helped me a lot!

  11. din says:

    the new NUC mounts a microSDXC reader with UHS-I support: what if I insert a microsd with UHS-II support?

  12. Frisbee says:

    Wonder how people feel about the Lenovo ThinkCenter M720SFF. With a coupon, i5-8400, 8GB RAM, 512GB NVME SSD, & Win10 for ~ $600USD. I get that it’s a bigger than the NUC (but a has a lot more ports), but for similar configuration (if you need WinOS), it’s about $100 cheaper.

  13. cihi84 says:

    Do you know if there is any noise and cooling differences between the BEH and BEK versions? (with and without a 2.5″ drive)? the NUC8i5BEK (no 2.5″ drive) should, in theory have better cooling as the fan is not blowing directly into the 2.5″ metal cage.

    I have some older 2.5″ SSDs and I am wondering if it is even worth considering buying the NUC8i5BEH or if it is better to get the smaller NUC8i5BEK and just sell my SSDs on ebay.

    Thanks in advance everyone !

    • omnium says:

      What do you think, why is there no NUC8i7BEK? In my theory, the cooling is better on the H models.

    • Olli says:

      My gut feeling is that the BEH models have better cooling. I don’t understand your comment regarding the fan blowing into the 2.5″ cage. The fan is on the other side of the mainboard vs. the drive cage. I don’t think it makes a difference here…

  14. Robert says:

    Anyone with this or the i7 can tell me how many fps they get in world of warcraft Thank you

  15. Dilbert says:

    Could you by any chance be convinced to make a video where you can hear the fan noise at different load levels? 🙂

  16. dd4u says:

    Hi. I asked a question regarding USB C and Power Delivery (PD) in another article before. Additionally I read in an article of the current issue if the german magazine c’t (computer technology) that the USB C is capable to deliver up to 15 Watt. Now: As I understand USB C can deliver power in both directions and up to 100 Watt. Can the NUC be powered by a monitor? Or vice versa: According to the c’t article the NUC should be capable to power a monitor with up to 15 Watt. Maybe someone hat the possibility to test that?

    • hulabalooza says:

      The NUC cannot be powered via its USB Type-C connector. I just tested using my laptop’s USB Type-C charger, but the NUC doesn’t show any signs of life.

  17. Le Matmasta says:

    Guys, who can tell me about NUC8i7BEH –
    Can i raise CPU frequency manually in bios?
    I want to disable function “turbo boost” and raise manually base frequncy from 2.7mhz to 3.2mhz – it is possible?
    It would be great – if anyone make a screenshot of the tab “Performance” in bios!

  18. Alex segers says:

    Hello, I am looking to buy a small pc for Office mail internet and banking. Is a NUC like this fast enough to do al these tasks without having to wait a minute or so to open up like a big file on the internet?
    I want it to be fast like a normal desktop would be or like a fast laptop.
    As I do not understand those tests and benchmarks just asking here.
    Thank you for the answer.
    Great blog btw

    • nucblognet says:

      Hi, it certainly is. It would be about as fast as a high-end laptop and certainly more than fast enough for mail, internet, office applications etc.

Leave a Reply

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