Comparison

Performance: LeMaker Guitar VS Raspberry Pi 2

6 2795
Edited by MKT at Sun Nov 8, 2015 23:58

Thanks tkaiser, the previous test numbers is not reliable and detail, so I decide to delete it. And thanks tkaiser agian, he did a very detail test on Guitar to compared to Raspberry Pi, Odroid-C1, Banana Pi, Wander Board.

The link is below:
http://forum.armbian.com/index.p ... kers-guitar/?p=1998

Please read tkaiser test careafully, it is really very useful.


By Tony Zhang
2015.10.10





Rate

1

View Rating Log

tkaiser  
1) CPU Performance: The sysbench results seem questionable unless you provide also the cpufreq settings used (and using an ARMv6 distribution on Banana Pro with insane cpufreq settings -- still fantasy governor? -- seems like a bad joke).

2) GPU Performance: Where are informations regarding drivers used and hardware acceleration? It can make a huge difference which drivers/frameworks are used.

3) RAM Performance: Just did a quick check with a Banana Pro. I got 567.323 MiB/s using 'mbw -t1 256' so this is far more than the 320 you report and even more than the guitar's value.

4) SD Performance: Which class had the SD-card (since write performance on Banana Pro is below 16 MB/s the card must be the bottleneck)? Where are reference benchmark values written/read on a system known to not be the bottleneck? I would assume the guitar board you used had the optional 2 GB RAM? Please have a look how to use dd correctly (that means really testing I/O and not memory bandwidth like you did partially): https://romanrm.net/dd-benchmark

5) Ethernet Performance: Testing only one direction is weird (since it's a known fact that eg. on A20 based boards like the Banana Pro results vary heavily between RX/TX). Also using an UDP test with fixed bandwidth (-u -b 1000M) is really strange.

I made a mistake, the cpufreq is 1.3GHz, not 900MHz.  GPU hardware acceleration was turned on,

tkaiser  
Edited by tkaiser at Wed Aug 19, 2015 00:43

If you add this information only later in the thread it won't be noticed. Shortly many reviewers, blogs and publications will simply reuse your screenshots and spread the information that the Guitar is that fast at 900 MHz (and can be also clocked with 1.3 GHz so people will believe the Guitar is even 40 percent faster if clocked at the upper limit). Well done ;-)

And you really know what's meant with GPU acceleration... people aren't interested in gtkperf or how fast windows move within X11 but they ask about gaming and hardware accelerated video decoding.

What's possible regarding HW accelerated video encoding/decoding within Linux (on your specs page and in the first reviews published there are theoretical possiblities listed like "4K HD video decoding, 1080P@60fps HD video encoding, supports MPEG-4, H.263, H.264 and other mainstream video formats" -- since Actions Semi provides an SDK based on kernel 3.10 the question is: which of the above applies to linux and not only to closed source Android stuff?)

deenbee  
Good job Team

tkaiser  
As part of my review of LeMaker's guitar I repeated some of the tests made. Might be interesting that the results published here are partially wrong and questionable in many regards. I also added the ODROID-C1+ to the list of tested boards. For details have a look at http://forum.armbian.com/index.p ... kers-guitar/?p=1998

tkaiser replied at Fri Oct 9, 2015 03:50
As part of my review of LeMaker's guitar I repeated some of the tests made. Might be interesting tha ...

I have revised the previous thread, and suggest users to refer to your link.
You did really a great job. Thanks!

You have to log in before you can reply Login | Sign Up

Points Rules