I've posted my review of the latest Intel Compute Stick (STK2M3W64CC):
Intel Core m3 M3-6Y30 CPU with Pre-Installed Windows 10 here.
I look at the BIOS, performance under Windows, fan noise, temperature, and installing alternate OS including Ubuntu, Chromium, Android and OpenElec.
tl;dr I highly recommend it.
Wednesday 30 March 2016
Sunday 27 March 2016
Booting Windows from microSD card on the STK2M3W64CC
Whilst the latest Intel Core M Compute Stick comes pre-loaded with Windows 10 it is possible to dual-boot. This is typically achieved by reducing the partition space allocated to Windows and creating a new partition for Ubuntu or similar. The assumption has always been that the faster eMMC storage would be reserved for the OS and any additional data storage could be gained from using either a USB or a microSD card or both.
But what if the eMMC storage is totally dedicated to the primary OS and the secondary OS runs from the supplementary storage? Again this approach has been successfully with booting Ubuntu installed on a USB or a microSD card thereby leaving Windows untouched on the 'disk'.
However for some the requirement is for Ubuntu to be the primary OS and Windows the secondary one. There would obviously be no problem in installing and running Ubuntu but what doesn't seem to have discussed previously is whether it is possible to run Windows from the microSD card and what the performance is like.
Because the Windows 10 license key is now embedded in the BIOS and also because Microsoft allows the installation (or re-installation) of Windows 10 using a USB the creation of a bootable Windows microSD card is very simple. And if a fast enough microSD card is used then the performance is actually quite acceptable.
To demonstrate this I've used a Sandisk Extreme Pro microSD card. After installation to the microSD card Windows both automatically activates and boots by default:
First looking at the I/O performance:
the biggest issue is the 'write' speed as this is substantially slowed than for eMMC.
But this won't adversely affect everything. Taking a look at the graphics performance using the 3D Mark tests we see very similar results to those obtained when running the default Windows configuration:
and interestingly there is a slight improvement for PC Mark 8 Home (accelerated):
It only really becomes apparent in Passmark's benchmark where disk performance is a significant factor in the overall rating :
However JavaScript still performs well:
To choose whether to boot Ubuntu or Windows an NVRAM option needs to be added for Ubuntu and GRUB reconfigured. Here's a video demonstrating booting:
Overall the performance is perfectly acceptable for a secondary OS and certainly a workable configuration.
The mini PC comes of age
Nearly four years ago the world was introduced to the MK802, an Android 'stick' originally billed as:
and so the mini PC was born.
Although it used an Allwinner A10 processor capable of a maximum speed of 1.5 GHz it was clocked at 1 GHz. Immediately attention turned to running Linux on the device but because Android took over half of the 4 GB of available storage Linux had to be booted from a micro SD card that was supported by the included 'T-Flash card slot'.
Performance was dreadful as this early video of Ubuntu 10.04 shows even though it was optimistically talked up at the time:
Almost immediately an upgraded model, the MK802+, was released with 1 GB of memory. Then came the MK802 II which defined the form-factor we are familiar with today including a full-size HDMI connector located at one end of the device and a side located full-sized USB port.
Since then the mini PC has evolved using more powerful ARM processors to most recently with the introduction of Intel processors. From a Linux perspective the Intel processors were welcomed because they overcame the restrictive shortfall of lack of HD graphics due to closed source drivers.
Three years ago I started benchmarking the performance on mini PCs running Ubuntu and the performance improvement since then has been dramatic. It can best be seen by comparing the first MK802+ against the latest mini PC, Intel's Core M Compute Stick, the STK2M3W64CC:
First the system information:
Next a performance comparison using my standard set of benchmarking tests from the Phoronix Test Suite run on Ubuntu:
Which when viewed graphically highlight the magnitude of improvement:
As further comparison the following is a 're-enactment' of the above MK802+ video using the STK2M3W64CC:
It is with this latest evolution that the mini PC has come of age.
Friday 25 March 2016
Performance on Intel Compute Sticks running Ubuntu
Since the introduction of the Intel-based mini PCs there have been plenty of performance and benchmark comparisons under Windows. Because I've previously used a set of benchmarking tests from the Phoronix Test Suite to compare ARM-based mini PCs running Ubuntu I've decided to run them on the Intel Compute Sticks.
To create the baseline I performed a fresh install of Ubuntu 14.04 to eMMC on each device and upgraded each to the latest available packages:
The tests I've chosen aim to show CPU, RAM and I/O performance and include:
The results:
show just how more powerful the latest Core M device is.
The following is a brief video showing basic system information about Ubuntu 14.04 as installed on this latest Core M Intel Compute Stick (STK2M3W64CC) together with a demonstration of booting the latest "Ubuntu 16.04 (Xenial Xerus) Daily Build" ISO direct from eMMC as described in http://www.linuxium.com.au/how-tos/simplewaytobootanisoinwindowsorubuntuwithouttheneedforexternalmedia
To create the baseline I performed a fresh install of Ubuntu 14.04 to eMMC on each device and upgraded each to the latest available packages:
The tests I've chosen aim to show CPU, RAM and I/O performance and include:
- CacheBench – Memory and cache bandwidth performance benchmark.
- CLOMP – C version of the Livermore OpenMP benchmark developed to measure OpenMP overheads and other performance impacts due to threading.
- 7-Zip compression – Uses p7zip integrated benchmark feature.
- dcraw – This test measures the time it takes to convert several high-resolution RAW NEF image files to PPM image format using dcraw.
- LAME MP3 encoding – This test measures the time required to encode a WAV file to MP3 format.
- FFmpeg – Audio/video encoding performance benchmark.
- GMPbench – Test of the GMP 5.0.3 math library
- OpenSSL – Measures RSA 4096-bit performance of OpenSSL.
- PHPBench – Benchmark suite for PHP.
- PyBench – Python benchmark suite.
- SQLite – This test measures the time to perform a pre-defined number of insertions on an indexed database
- Stream – This benchmark tests the system memory (RAM) performance.
- TSCP – Performance benchmark built-in Tom Kerrigan’s Simple Chess Program.
- Unpacking the Linux kernel – This test measures the time it takes to extract the .tar.bz2 Linux kernel package.
- IOzone – This benchmark tests the hard disk drive / file-system performance.
The results:
show just how more powerful the latest Core M device is.
The following is a brief video showing basic system information about Ubuntu 14.04 as installed on this latest Core M Intel Compute Stick (STK2M3W64CC) together with a demonstration of booting the latest "Ubuntu 16.04 (Xenial Xerus) Daily Build" ISO direct from eMMC as described in http://www.linuxium.com.au/how-tos/simplewaytobootanisoinwindowsorubuntuwithouttheneedforexternalmedia
Wednesday 23 March 2016
A pseudo Intel Chromium Stick
In exploring what can be done with the latest Intel Compute Stick (the 'Cedar City' version with a Core M processor) I decided to install my latest build of Google's Chromium OS software directly to eMMC (the device's internal storage) in order to given an indication of actual performance and to show an example of using an alternative OS to Windows 10. As I've removed Windows 10 as part of this installation I've effectively created a pseudo Intel Chromium Stick.
The video starts after powering on the Intel Compute Stick and shows entering the boot menu. It then goes on to show Chromium OS boot and the login process with the auto-connection of wifi. Next are two usage examples. The first shows running Ubuntu using crouton and then how the Chromium browser performs running Octane 2. In the second example I briefly run an sample YouTube video. I've also included a couple of commands showing technical details about the system including OS version and disk utilization. There are two points in the video that demonstrate working audio. You can hear the shutter when the screenshot is taken and a couple of chimes from the software being demonstrated in the YouTube video. Okay it is more of a token gesture to audio than high fidelity but hopefully preferable to some annoying muzak.
The impressive Octane score of 22010 demonstrates how much more powerful the latest Intel Compute Stick is compared to earlier models and alternative mini PC sticks. Here is the actual screenshot as captured in the video:
Tuesday 15 March 2016
First look at the STK2M3W64CC
The latest mini PC 'stick' computer, Intel's Compute Stick Core-m3 STK2M3W64CC, should remove any doubts about the form-factor and its usability. The first commercially available mini PC 'stick' with 4GB ram, 64GB storage, dual-band wifi, bluetooth and USB3.0 has performance that completely exceeds any device so far.
Just take a look at the initial results on Windows:
And on Ubuntu:
But there's more. You can run Ubuntu from micro SD card:
Or from an ISO (see http://goo.gl/2PeFC4):
Or use Chromium OS from USB:
including crouton:
A full review can be found here: http://www.linuxium.com.au/reviews/stk2m3w64cc
Sunday 13 March 2016
Performance testing microSD cards
Here's an interesting article on 'The Best microSD Card' (see http://thewirecutter.com/reviews/best-microsd-card).
It's interesting because it includes:
- testing results from running CrystalDiskMark on leading quality branded cards
- good background information and additional useful material
- my favourite cards were excluded due to being too expensive and faster than most people need
In addressing the last point to afford the fastest I reduce the storage. So I currently use 16GB SanDisk Extreme PRO and 32GB Samsung PRO.
However with such comprehensive test data any future purchases will definitely be well-informed.
Sunday 6 March 2016
The cost of converging mini PC and tablets
If you thought that for Linux devices the next round of Cherry Trail mini PCs would be too expensive then take a look at the latest tablet from ASUS.
ASUS VivoBook E200HA-US01 Intel Atom Quad Core 2GB 32GB eMMC 11.6-inch Laptop
Processor: Intel® Cherry Trail Quad-Core Z8300 1.84 GHz Processor
Operating System: Windows 10 Home with Office 365 1 year included
Chipset: Integrated Intel® CPU
Memory: OnBoard Memory 2 GB
Display: 11.6" 16:9 HD (1366x768)
Graphic: Integrated Intel® HD Graphics
Storage: 32GB eMMC
Card Reader: card reader (Micro SD Micro SDXC Micro SDHC )
Camera: VGA Web Camera
Networking: Integrated 802.11 ac and Built-in Bluetooth™ V4.1
Ports:
1 x Microphone-in/Headphone-out jack
1 x USB 3.0 port(s)
1 x USB 2.0 port(s)
1 x micro HDMI
1 x AC adapter plug
1 x micro SD card
Audio: Built-in Speakers And Digital Array Microphone
Battery: 2 Cells 38 Whrs Polymer Battery
Power Adapter: Output : 19 V DC, 1.75 A, 33 W Input : 100 -240 V AC, 50/60 Hz universal
Dimensions: 11.26 x 7.6 x 0.69 inch (WxDxH)
Weight: 2.16 lbs (with 2 cell battery) (with Polymer Battery)
Price: $199.00
Ref: http://store.asus.com/us/item/201602AM020000390
Are we heading towards a bikini pricing model (see http://www.dailyfinance.com/2011/06/09/why-teeny-bikinis-have-such-big-price-tags) for mini PCs?
Mini PCs are costly to make when compared to other devices as whilst "The labor is the same... " "...and there's only slightly less material..." " ...it's the lack of ability to commoditize it or to use economies of scale" where "the cost savings that come from mass production".
Or, because of the short lifestyle of the product, "...retailers quickly begin reducing their prices to keep consumers buying. But to afford those reductions, manufacturers have to set their original prices relatively high. You have to mark it up to be able to mark it down".
Either way the price of mini PCs compared with other devices doesn't seem overly attractive.
Want another opinion? See http://shiritai-koto.net/?entry=74 (hint: use Google translate).
Zotac PC Stick
Interesting as it is a z5-Z8300 stick with an RJ45 port ...
- Intel Atom x5-Z8300
- Intel HD Graphics
- 32GB eMMC, 2GB DDR3L RAM
- 1x USB 2.0, 1x microSD
- 10/100 Mbps RJ-45 and 802.11ac Wi-Fi, BT 4.0
- HDMI output (male)
See http://liliputing.com/2016/01/zotac-pc-stick-is-a-cherry-trail-computer-on-a-stick.html
Intel have announced a 'pre-configured' barebones with OS mini PC which is priced in the same ballpark ... see http://www.intel.com/content/www/us/en/nuc/nuc-kit-nuc5pgyh.html ... so many swings and roundabouts or maybe snakes and ladders for the average consumer.
Mini PCs vs tablets - do you get value for money?
So far the new Cherry Trail mini PCs are all pre-release or crowd-funded so in comparing specification versus price I thought it would be interesting to look at the ASUS Transformer Book T100HA.
You get an x5 Z8500 CPU, 4GB RAM, 64 GB eMMC, 2.4/5.0 GHz Wifi, 4.0 BT, 1xType-C USB 3.1 (5Gbps), 1xMicro 2.0, 1xMicro HDMI, Windows 10 plus a screen, a keyboard, a battery, a couple of cameras etc. all for $299 and available now.
I've already mentioned the convergence of mini PCs with barebone devices and now it seems similar for notebooks/tablets. Effectively suggesting that mini PCs are becoming over-priced.
Refs: http://store.asus.com/us/item/201508AM070000885/A38616,http://promos.asus.com/US/PR_2015/T100HA/index.html, http://www.asus.com/us/2-in-1-PCs/ASUS_Transformer_Book_T100HA/specifications/
The cost of converging mini PC and barebone PCs
Will the next round of Cherry Trail mini PCs be just too expensive as Linux boxes?
Initially I compared the Tronsmart Ara X5 with the ECS Liva X2 using the Intel Compute Stick as a baseline. The performance and functionality improvements showed that for only $20 extra the Braswell SOC coupled with the extra USB 3.0 and Gigabit LAN was much better than the Cherry Trail device.
Which led me to then compare against the Intel NUC (see the table I've created showing key highlights). Whilst the NUC is barebones, I already have spare RAM and SSD and I want Linux and not Windows, so the price now becomes relevant. And the NUC is extendable. So for a Linux box, I think we are now at the convergence point between cheap mini PCs and barebone PCs. Where the tendency for mini PCs is to increase in price for improved functionality, barebone PCs are getting cheaper and better.
My conclusion is that for Linux boxes the next round of Cherry Trail mini PCs will be too expensive and that barebone PCs are now viable.
Reference Links
Intel Compute Stick benchmarks: http://www.anandtech.com/show/9167/intel-compute-stick-reviewTronsmart Ara X5 benchmarks: http://www.cnx-software.com/2015/08/24/review-of-tronsmart-ara-x5-windows-10-mini-pc-with-intel-atom-x5-processor/http://hometheatrelife.com/tronsmart-ara-x5-review-windows-10-cherry-trail/ECS Liva X2 benchmarks: http://www.anandtech.com/show/9554/ecs-liva-x2-review-a-compact-fanless-braswell-pcNUC NUC5CPYH benchmarks: http://nucblog.net/2015/07/braswell-nuc-nuc5cpyh-review-windows-performance/Linux benchmarks (Z3735F vs N3050): http://www.phoronix.com/scan.php?page=article&item=intel-celeron-n3050
Pricing Links
Intel Compute Stick: http://www.newegg.com/Product/Product.aspx?Item=9SIA24G36F9195&cm_re=intel_compute_stick-_-83-800-004-_-Product
Tronsmart Ara X5: http://www.geekbuying.com/item/Tronsmart-Ara-X5-Windows-10-TV-Box-Cherry-Trail-Z8300-Quad-Core-1-8G-Gen-8-Graphics-2G-32G-2-4Ghz-5Ghz-351938.html
ECS Liva X2: http://www.newegg.com/Product/Product.aspx?Item=N82E16856501013
NUC NUC5CPYH: http://www.newegg.com/Product/Product.aspx?Item=N82E16856102140&cm_re=NUC5CPYH-_-56-102-140-_-Product
Kingston 2GB RAM: http://www.newegg.com/Product/Product.aspx?Item=N82E16820239853&cm_re=SODIMM_2GB-_-20-239-853-_-Product
Mushkin 60GB SSD: http://www.newegg.com/Product/Product.aspx?Item=N82E16820226677
Oscar Wilde: “Imitation is the sincerest form of flattery that mediocrity can pay to greatness”
Somebody wrote to me today complaining that a Meegopad T02 with Ubuntu they bought from Geekbuying did not work well. The 'dmesg' shows that Meegopad had installed one of my earlier Linux builds on the devices and it had been sold as the 'Meegopad T02 Ubuntu Mini PC Compute Stick'. Selling devices based on my development work without telling me is pretty poor form in my opinion. And it is not very good for customers given it was an old version with known issues. Fortunately this customer found my work and having reinstalled Ubuntu is now very pleased.
Perhaps Meegopad could install the following image as the default wallpaper?
Perhaps Meegopad could install the following image as the default wallpaper?
OEM pricing for "Windows 8.1 with Bing"
Interesting article on OEM pricing for "Windows 8.1 with Bing" here ... http://www.zdnet.com/article/how-much-does-microsoft-make-from-pc-makers-with-windows-8-1 ... it is not free but zero dollars!
Comparing Android performance between Allwinner A80 Optimusboard, Samsung Exynos 5422 ODROID-XU3 Lite and Rockchip 3288 Firefly
Android performance testing with 'No Frills CPU Control' used to set governor and I/O scheduler and results shown for 'AnTuTu', '3D Mark' and 'AnTuTu X' except for the Odroid-XU3 Lite where 'AnTuTu X' is not available. Interesting the difference between 'AnTuTu' and 'AnTuTu X'.
Allwinner A80 Optimusboard
Governor: performance
I/O Scheduler: noop
AnTuTu v5.2.0: 57120
3D Mark: 14096
AnTuTu X v5.1: 33507
Samsung Exynos 5422 ODROID-XU3 Lite
Governor: performance
I/O Scheduler: noop
AnTuTu v5.2.0: 46136
3D Mark: 14973
AnTuTu X v5.1: N/A
Rockchip 3288 Firefly
Governor: performance
I/O Scheduler: noop
AnTuTu v5.2.0: 38046
3D Mark: 13308
AnTuTu X v5.1: 35225
Allwinner A80 Optimusboard
Governor: performance
I/O Scheduler: noop
AnTuTu v5.2.0: 57120
3D Mark: 14096
AnTuTu X v5.1: 33507
Samsung Exynos 5422 ODROID-XU3 Lite
Governor: performance
I/O Scheduler: noop
AnTuTu v5.2.0: 46136
3D Mark: 14973
AnTuTu X v5.1: N/A
Rockchip 3288 Firefly
Governor: performance
I/O Scheduler: noop
AnTuTu v5.2.0: 38046
3D Mark: 13308
AnTuTu X v5.1: 35225
Introducing the MK802V
The so called MK802V otherwise know as the Kemico MK802V or Jesurun T034 or Unuiga U33-4R (RK3288) or ...
Cosmetically the device is slightly larger than a typical RK3188 stick/dongle. I've included comparison pictures against an RKM MK802IV and a PQ Labs istick A300 (specifically the A350-SSD which is one of the first stick/dongles to include an RJ45 port).
Whatever it is called it is an RK3288 stick/dongle with a built-in RJ45 port. The board is stencilled with T034_V1_20140726 and QL2014.31. It comes installed with Android 4.4.2 (Kernel 3.10.0) which has build number rk3288-eng 4.4.2 KOT49H eng.ant.20140910.191101 test-keys.
Internally the device includes a heat-sink consisting of a thin sheet of metal (aluminium?) painted black on one side and covered in an extremely sticky adhesive on the side that covers the SOC. This certainly works for short periods of heavy usage like running Antutu. With Linux installed the device initially ran idle at around 42 degrees C but heated up to 104 degrees C when under load (running 'stress' with all 4 CPUs at 100% load) when it then crashed after about 10 minutes continuous load.
Comparing the Tronsmart Orion R28 and the Kemico MK802V with both having their case top removed (thereby exposing the respective heat-sinks to the open to help with ambient cooling), the MK802V registered around 5 degrees C higher than the R28. When running a comparison stress test, the R28 had a starting idle temperature of 33 degrees C and the MK802V had a starting idle temperature of 38 degrees C. On running the 'stress' test in parallel on each machine, after a couple of minutes the temperature had climbed to over 50 degrees C on the R28 and over 70 degrees C on the MK802V.
As expected, the temperature continued to climb reaching 65 degrees C on the R28 with the stress test completing successfully, whilst the MK802V climbed to 92 degrees C with the stress test just finished before the device crashed.
By removing the metal heat-sink from the MK802V and adding an external fan it was possible to repeat the stress test with the temperature keeping below the threshold limit for the duration of the test. The highest registered temperature with the MK802V was 83 degrees C (compared with [and running in parallel] the R28 registering only 65 degrees C).
Whilst the temperature findings may prove significant in running the MK802V as a media device under Android, as a basic Linux 'alternative' PC (with known h/w video acceleration limitations) so far it this has not impacted the performance.
An initial Antutu run gave a score of 34703. And whilst it came unrooted it had both Google Play Store and XBMC (13.0-Alpha12Git:20140227-b8512bf [Compiled: Aug 22 2014]) installed.
Cosmetically the device is slightly larger than a typical RK3188 stick/dongle. I've included comparison pictures against an RKM MK802IV and a PQ Labs istick A300 (specifically the A350-SSD which is one of the first stick/dongles to include an RJ45 port).
Whatever it is called it is an RK3288 stick/dongle with a built-in RJ45 port. The board is stencilled with T034_V1_20140726 and QL2014.31. It comes installed with Android 4.4.2 (Kernel 3.10.0) which has build number rk3288-eng 4.4.2 KOT49H eng.ant.20140910.191101 test-keys.
Internally the device includes a heat-sink consisting of a thin sheet of metal (aluminium?) painted black on one side and covered in an extremely sticky adhesive on the side that covers the SOC. This certainly works for short periods of heavy usage like running Antutu. With Linux installed the device initially ran idle at around 42 degrees C but heated up to 104 degrees C when under load (running 'stress' with all 4 CPUs at 100% load) when it then crashed after about 10 minutes continuous load.
Comparing the Tronsmart Orion R28 and the Kemico MK802V with both having their case top removed (thereby exposing the respective heat-sinks to the open to help with ambient cooling), the MK802V registered around 5 degrees C higher than the R28. When running a comparison stress test, the R28 had a starting idle temperature of 33 degrees C and the MK802V had a starting idle temperature of 38 degrees C. On running the 'stress' test in parallel on each machine, after a couple of minutes the temperature had climbed to over 50 degrees C on the R28 and over 70 degrees C on the MK802V.
As expected, the temperature continued to climb reaching 65 degrees C on the R28 with the stress test completing successfully, whilst the MK802V climbed to 92 degrees C with the stress test just finished before the device crashed.
By removing the metal heat-sink from the MK802V and adding an external fan it was possible to repeat the stress test with the temperature keeping below the threshold limit for the duration of the test. The highest registered temperature with the MK802V was 83 degrees C (compared with [and running in parallel] the R28 registering only 65 degrees C).
Whilst the temperature findings may prove significant in running the MK802V as a media device under Android, as a basic Linux 'alternative' PC (with known h/w video acceleration limitations) so far it this has not impacted the performance.
An initial Antutu run gave a score of 34703. And whilst it came unrooted it had both Google Play Store and XBMC (13.0-Alpha12Git:20140227-b8512bf [Compiled: Aug 22 2014]) installed.