Beelink L55 Review – An Intel Core i3-5005U Mini PC Tested with Windows 10 & Ubuntu 18.04

Beelink L55 Review

With the shortage of Gemini Lake processors, some manufacturers have taken to releasing new mini PCs using older CPUs. The Beelink L55 is one such example and uses Intel’s Broadwell I3-5005U CPU which is a dual-core 4-thread 2.00 GHz processor with Intel’s HD Graphics 5500.

The L55 is a ‘NUC’ style mini PC and physically consists of a 128 x 126 x 46 mm (5.04 x 4.96 x 1.81 inches) box case with a front panel that includes the power button, a headphone jack and a couple of 3.0 USB ports and then on the rear, two more 3.0 USBs, two 2.1 USBs, HDMI (1.4), DP and two gigabit Ethernet ports. Note that there is no SD or micro SD card slot. The full specifications include:

Beelink L55 Specifications
Click to Enlarge

The L55 comes with either a 256GB or 512GB (as tested in this review) M.2 SSD with pre-installed Windows 10 Home version 1809 together with a single slot of 8GB DDR3 RAM. The L55 also supports the addition of an optional 2.5” SSD:

Beelink L55 Motherboard
Click to Enlarge

In the box you get the SSD power and data connectors, a couple of HDMI cables, a mounting bracket for attaching the device to behind a monitor together with a power brick and cable (excluded from the picture below):

Beelink L55 Accessories
Click to Enlarge

Starting with a quick look at the hardware information shows it is aligned to the specification:

disk management
Click to Enlarge
L55 System Info
Click to Enlarge

HWinFO i3-5005U Mini-PC

With Windows fully updated to version 1809 build 17763.557 I ran my standard set of benchmarking tools to look at performance under Windows:

The results need interpreting carefully otherwise they could be misleading when compared to other Intel mini PCs. This is because mini PC benchmark results are heavily influenced by the quantity and type of memory and storage installed:

Windows 10 mini pc comparison: specs & benchmarks
Click to Enlarge

Broadly speaking however the L55 benchmarking performance appears on par with the latest Gemini Lake processors.

Having completed all the benchmarks Windows update then offered a feature update to version 1903:

Windows Features Update 1903So I both upgraded and re-ran all the benchmarks:

Comparing the results from the two Windows versions shows that they are both very similar and given the margin of error no substantial performance improvement or degradation was observed.

Next, I shrunk the Windows partition and created new a 100GB partition so I could install and dual boot Ubuntu using an Ubuntu 18.04.2 ISO:

Beelink L55 ubuntu partition

After installation a brief check showed everything working including Wifi, Bluetooth and audio:

Intel-Broadwell-ubuntu-sound
Click to Enlarge

I then ran some basic commands to look at the hardware in more detail:


Interestingly when I ran Octane the result was much lower compared to that of Windows at 14,613 vs. 18,680:

Beelink-L55-ubuntu-octane2
Click to Enlarge

Whilst the benchmarks indicate comparably good mini PC performance I next looked at real-world usage and the implications of using an older GPU by playing videos under Windows using both Edge and Chrome browsers and in Ubuntu using both Firefox and Chrome.

In Windows neither [email protected] nor [email protected] videos were available in Edge so the highest resolution were 1080p videos which played fine:

beelink-l55-windows-edge-browser-1080p-video
Click to Enlarge
beelink-l55-windows-edge-browser-1080p-at-60fps-video
Click to Enlarge

Whilst in Chrome [email protected] could load but not play:

l55-windows-chrome-browser-4k-video
Click to Enlarge

However, Chrome was able to play 1440 @ 30fps:

Beelink-L55-windows-chrome-browser1440-video
Click to Enlarge

Also available in Chrome were 4K @ 60fps and 1440p @ 60fps videos however they resulted in the frames being dropped and were unwatchable, and even at [email protected] the video was juddery with dropped frames:

Beelink-L55-windows-chrome-browser-1080p-at-60fps-video
Click to Enlarge

Playing videos in on Ubuntu was a similar story to other Intel processor-based mini PCs. With Firefox [email protected] was unwatchable:

ubuntu-firefox-browser-4k-video
Click to Enlarge

but was fine when played at 1440p:

ubuntu-firefox-browser-1440-video
Click to Enlarge

Firefox [email protected] and [email protected] videos resulted in the frames being dropped and were unwatchable,

ubuntu-firefox-browser-1440-at-60-video
Click to Enlarge

but fine when played at 1080p with all 45 frames being dropped right at the start of the video:

ubuntu-firefox-browser-1080p-at-60-video
Click to Enlarge

Chrome in Ubuntu was similar to Firefox. [email protected] was unwatchable but fine when played at 1440p:

ubuntu-chrome-browser-1440-video
Click to Enlarge

Again 4K @ 60fps and at 1440 were unwatchable. However, even at 1080p @ 60fps the video was juddery and dropped frames

Beelink-L55-ubuntu-chrome-browser-1080p-at-60-video
Click to Enlarge

The following table summarises the results:

Beelink L55 YouTube test results

Playing videos using Kodi on Windows with a VP9 codec encoded video software was used for decoding resulting in high CPU usage and juddery video:

Windows kodi vp9
Click to Enlarge

whereas an H.264 codec encoded video used hardware to decode and was fine:

windows-kodi-h264
Click to Enlarge

Videos encoded with H.265 or HEVC did not play with one simply crashing Kodi and the other juddery with skipped and dropped frames:

windows kodi hevc
Click to Enlarge

In Kodi on Ubuntu VP9 videos also used software to decode and resulted in some skipped frames:

ubuntu-kodi-vp9
Click to Enlarge

However, H.264 codec encoded videos again used hardware to decode and played fine:

ubuntu-kodi-h264
Click to Enlarge

Videos encoded with H.265 or HEVC both used software encoding and either kept stalling:

ubuntu-kodi-h265
Click to Enlarge

or were juddery with dropped and skipped frames:

ubuntu-kodi-hevc-juddery
Click to Enlarge

The lack of codec support and hardware decoding noticeably limits the use of the L55 as an HTPC.

The following table summarises the results:

Beelink L55 Kodi Video Test Results

As can be seen above when software decoding is used it creates high CPU usage resulting in higher internal temperatures. The L55 includes a fan which is slightly audible under these conditions as it cycles up and down:

Beelink L55 Fan
Click to Enlarge

I measured 31db right next to the device when the fan was at its lowest and up to 48db at its loudest when it made what can best be described as a whooshing sound.

Cooling, however, was effective and under Ubuntu, I ran both a VP9 and an H.264 video in Kodi for 20 minutes each and the internal temperature remained under control with an average of around 55°C for software decoding and 49°C for hardware decoding:

Beelink L55 fan effectiveness software encoding
Click to Enlarge
Beelink L55 fan effectiveness hardware encoding
Click to Enlarge

The external temperature of the device reached in places a maximum of 30°C however the ambient temperature was only 12°C being winter here.

I also briefly looked at gaming performance. Playing CS:GO on Windows with the default settings of:

 

Beelink L55 CSGO-windows-default-video-settings
Click to Enlarge

 

Beelink L55 CounterStrike windows-default-advanced-video-settings
Click to Enlarge

resulted in a frame rate average of a rather low 8fps:

Beelink L55 FPS windows csgo-default-settings
Click to Enlarge

Changing the video settings to low and dropping the resolution to 720p:

Beelink L55 CounterStrike low video settings Beelink L55 CounterStrike low advanced video settings

and the frame rate average improved to a playable 27fps:

Beelink L55 Gaming windows-csgo-low-settings

For CS:GO on Ubuntu and installed within a container, see the instructions.

CSGO ubuntu default-video-settings
Click to Enlarge
CSGO ubuntu default-advanced-video-settings
Click to Enlarge

the default frame rate was slightly higher at an estimated average of around 15fps:

Beelink L55 ubuntu csgo fps default-settings
Click to Enlarge

Changing the video settings to low and resolution to 720p the frame rate average improved to an estimated 25fps:

ubuntu-csgo-low-settings
Click to Enlarge

I also installed and ran the UNIGINE Heaven benchmark under both Windows:

windows-unigine-benchmarkand Ubuntu:

Beelink L55 ubuntu unigine

Interestingly but maybe not unsurprisingly the results differed with Windows scoring 141 at 5.6fps and Ubuntu scoring 116 at 4.6fps:

Unigine Windows 10 vs Ubuntu 18.04

Therefore the L55 offers limited gaming performance.

Given the optional storage expandability of the L55 I also briefly installed an internal SSD using the provided connectors:

Beelink L55 SSD Installation
Click to Enlarge

and confirmed it worked under both Windows:

windows ssd
Click to Enlarge

and Ubuntu:

ubuntu ssd
Click to Enlarge

In both cases, it was recognized as the first drive as ‘disk0’ in Windows and ‘sda’ in Ubuntu.

At some point in the removal of the SSD I must have knocked the memory board as after reassembling the device it refused to boot. Normally when the unit is switched on the blue light on the front comes on and there is a single beep before anything is displayed on the screen. The problem I had was that the blue light still came on however instead there were three beeps and nothing else happened. After raising the issue with Beelink they indicated that it was a memory issue and probably that the memory was not inserted completely. Although I then reseated the memory several times it did not fix the issue. Fortunately I had a spare 4GB memory module and on replacing the installed 8GB stick with this 4GB stick the device then rebooted. I then swapped back the original 8GB memory and the device has continued to boot fine.

As it has just been released I tried Debian 10 “buster” as a live image running from USB using both the Gnome and LXQT ISOs. Both had an issue in that there was no Wifi or Bluetooth OOTB as the Wifi card’s firmware needs to be updated.

Beelink L55 debian lxqt wifi issue
Click to Enlarge

Network connectivity throughput was measured on Ubuntu using ‘iperf’:

Beelink L55 network throughputUnder Ubuntu, I tried bonding both Ethernet ports together and enabled 802.3ad on the connected router however there was no further speed improvement.

Power consumption was measured as follows and is higher than comparable Gemini Lake devices:

  • Powered off – 1.5 Watts
  • Shutdown – 1.0 Watts (Windows)
  • BIOS*  – 12.5 Watts
  • Boot menu – 8.0 Watts
  • Idle – 8.2 Watts (Windows) and 7.9 Watts (Ubuntu)
  • CPU stressed – 14.1 Watts (Ubuntu)
  • Video playback** – 12.1 Watts (1080p in Windows) and 14.1 Watts (1080p in Ubuntu)

* BIOS (see below)
** The power figures fluctuate so the value is the average of the median high and median low power readings.

Finally, the BIOS is fully unlocked:

Beelink L55 BIOS main Beelink L55 BIOS advanced

I’d like to thank Beelink for providing the L55 for review. It currently retails at around $299 for the tested configuration and can be purchased on sites such as GearBest, GeekBuying or Aliexpress.

Support CNX Software - Donate via PayPal or become a Patron on Patreon

12
Leave a Reply

avatar
2 Comment threads
10 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
8 Comment authors
LinuxiumtkaiserFransMDiegoJean-Luc Aufranc (CNXSoft) Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
fse
Guest
fse

Thx for the review, I really like you mini-pc-results table every time!!!

FransM
Guest
FransM

i3-5005u processor. Launched in Q1 2015.
Intel NUC8i3BEH with i3-8109U can be purchased in NL for eur 265 (with 32G DDR4 but without disk) 500 G m.2 ssd starts at arund eur 50, so after adding import duties to the Beelink this is pricewise on par but the NUC is undoubtedly a better buy.

linuxium
Guest

If you prefer an actual NUC then Intel actually released a similar specification (but with only a single NIC) as the NUC513RHS (see https://ark.intel.com/content/www/us/en/ark/products/184691/intel-nuc-kit-nuc5i3ryhs.html) and it is still available on sites like Amazon although you have to add memory and storage.

rufik
Guest
rufik

Could you run tkaiser’s sbc-bench, please?

tkaiser
Guest
tkaiser

I don’t think that’s useful since ‘use case first’. Running sbc-bench is only useful if you have ‘server workloads’ in mind or are interested in thermal/throttling behavior. Mini PCs are normally used as… Mini PCs and as such sbc-bench numbers are of little to no meaning (Mini PC and ‘desktop’ use cases don’t depend on integer CPU performance and low memory latency but on many other different things like available HW accelerated video and 3D performance).

For a general performance assessment look at the Passmark scores in the benchmark table above. For the stuff sbc-bench focuses on this Mini PC might be 25% to 30% faster than a J4105 design like ODROID H2 for example.

Rufik
Guest
Rufik

I’d like to compare numbers to ODROID H2 (and others probably) to form an opinion of performance difference. Sure, I have “server workload” only in mind, I’m not interested in desktop usage at all.

Diego
Guest
Diego

Which seems to be a good decision anyways as the video engine serms to be slightly dated for desktop/htpc use and the two SSDs might make it interesting for such usecases.

Btw. as it’s winter at linuxiums place maybe testing a pi4 would have been the better choice, as it had lowered heating costs…

Jean-Luc Aufranc (CNXSoft)
Admin

I’ve actually run sbc bench in ODROID-H2 in Ubuntu 19.04, but I had to change one line to get it working without warning:

It turns out in Ubuntu 19.04 at least, /sys/devices/system/cpu/cpufreq/policy0/scaling_cur_freq is expected instead, so I changed the code in line 180 of the script to remove “cpufreq/” from the string as a quick fix, and started the benchmark again

linuxium
Guest

I also had to make the same change to line 180.

linuxium
Guest

sbc-bench v0.6.7

Installing needed tools. This may take some time… Done.
Checking cpufreq OPP… Done.
Executing tinymembench. This will take a long time… Done.
Executing OpenSSL benchmark. This will take 3 minutes… Done.
Executing 7-zip benchmark. This will take a long time… Done.
Executing cpuminer. This will take 5 minutes… Done.
Checking cpufreq OPP… Done.

Memory performance:
memcpy: 4979.6 MB/s (0.8%)
memset: 10336.4 MB/s (0.6%)

Cpuminer total scores (5 minutes execution): 22.74,22.73,22.71,22.70,22.69,22.68,22.67,22.66,22.65,22.64,22.63,22.62,22.61,22.60,22.56,22.53,22.52,22.51,22.50,22.49,22.48,22.44,22.34,21.97,19.94 kH/s

7-zip total scores (3 consecutive runs): 6030,5991,5888

OpenSSL results:
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-128-cbc 385793.77k 411749.87k 423892.65k 425905.15k 426713.09k 426497.37k
aes-128-cbc 386681.05k 412131.46k 423974.06k 425940.31k 426560.17k 426393.60k
aes-192-cbc 328462.58k 345861.91k 354516.31k 355989.16k 356300.12k 356144.47k
aes-192-cbc 328949.61k 346097.37k 354474.67k 355970.73k 356390.23k 356095.32k
aes-256-cbc 285603.17k 298168.68k 304614.91k 305725.10k 305957.55k 305714.52k
aes-256-cbc 284625.80k 298206.38k 304627.71k 305677.65k 305949.35k 305769.13k

Full results uploaded to http://ix.io/1OjK. Please check the log for anomalies (e.g. swapping
or throttling happenend) and otherwise share this URL.

tkaiser
Guest
tkaiser

> Full results uploaded to http://ix.io/1OjK. Please check the log for anomalies

Thank you. The obvious anomalies:

* wrong thermal sensor (the one chosen remains on 29.8°C all the time)
* the CPU cores are limited to 1.9GHz
* surprisingly low openssl and 7-zip scores. As you already noticed with your initial benchmarking here’s something wrong in Linux (maybe Spectre/Meltdown mitigations?)

Linuxium
Guest

Benchmark repeated with correct temperature sensing: http://ix.io/1OxL. I need to repeat Linux benchmarking on Gemini Lake SOCs post mitigations so I’ll include sbc-bench for comparison.