Forum

With Xbian now being supported on 3 different hardware (Cubox-i, Hummingboard and of course RPi) it dawned on me that we should have a standard to work with, so that with future upgrades and possibly new hardware (Banana Pi?) we would have standard performance details for all possible configurations.

So I had a look around and came up with Sysbench.

Full details are available in the link including installing and running the CPU and I/O tests.

For the CPU use sysbench --test=cpu --cpu-max-prime=2000 run (As 20000 might take the RPi a while)
For the I/O use sysbench --test=fileio --file-total-size=1G prepare (As no one has 150G spare!!!)
For the execution use ... sysbench --test=fileio --file-total-size=1G --file-test-mode=rndrw --init-rng=on --max-time=300 --max-requests=0 run

To clear the file run sysbench --test=fileio --file-total-size=1G cleanup

CPU is total time
I/O is Kb/sec value

So if users could run the test on their machine, we will get a benchmark for different machines, which will assist in seeing if upgrades are faster, or if someone's machine is running slow (or faster for that matter). I don't have a basis yet for the norm, but I guess 10% either way is acceptable.

My results

Because the forum doesn't do formatting: Use
Model, Xbian Version, SD Card(speed), CPU result, I/O result

So:

Cubox-i4 Pro, RC3 (20141218), Kingston 4g(10), 23.9741s, 729.01Kb/sec

Already did this once for a Dutch forum. Once single threaded, once with all supported cores.

Radxa Rock Pro - 1 Thread

Radxa Rock Pro - 4 Threads

HummingBoard i2eX Dual Core - 1 Thread

HummingBoard i2eX Dual Core - 2 Threads

Hummingboard i1 Single Core - 1 thread

Hummingboard i1 Single Core - 2 threads

Raspberry Pi Model B OC (arm=950, core=450, sdram=500, overvolt=2)

Just because i can:
Raspberry Pi Model B OC (arm=950, core=450, sdram=500, overvolt=2) - 2 Threads

Raspberry Pi Model B No OC

Raspberry Pi Model B No OC - 2 Threads

BananaPi Pro - 1 Thread

BananaPi Pro - 2 Threads

Raspberry Pi 2 - 1 Thread

Raspberry Pi 2 - 4 Threads

Raspberry Pi 2 OC - 1 Thread

Raspberry Pi 2 OC - 4 Threads

Hummingboard 2 Edge - 1 Thread

Hummingboard 2 Edge - 4 Thread

ClearFog Pro - 1 Thread

ClearFog Pro - 2 Threads

Conclusion:
* XBian default OC
** Default Turbo OC

So now you can add to your list.

Cubox-i4 Pro 302.6553s

Threads 4!
Cubox-i4 Pro 155.7358s

Better you do it in the First Post.

No I wanted a standard benchmark that was applicable to Xbian on the 3 platforms. Unless it is now running on an Intel G620T!! Also I'm mainly interested on the Xbian version, to see if any upgrades affect the performance - for the better or the worse.

G620 was to put things into perspective.

I did the test again on a Hummingboard i1 Single Core and the results are not significantly different (+/- 3 points) between now (latest XBian fully updated) and may (unusable XBian). So i think all results will still stand.

It would seem that, not surprisingly that the hummingboard and the Cubox-i fair the same with duel threads but ramp it up to 4 and the CB wins.

So in perspective, it is NOT a fair comparison, as the thread count on multi cores machines makes a difference,

I wonder if a RPi is run with thread=8 (or 4) gives the same results as thread=1. If so, then I need to change the test to thread=max (if that option exists)

No, the more threads you use on a single core machine the worse it's going to perform. So indeed, especially the cores are the thing that count most.

Just added the Radxa Rock Pro

@IriDium,

prime number computation is the most CPU only test ever.
it is from 99.9% dependent on CPU frequency and CPU internal architecture (nr of math units, pipes in that case so CPU can do X multiplications in one clock tick even if being single core).

that's why 5000 primes it takes 90s on CortexA9(1ghz) core, 4s on i5(3.4ghz) and only 3s on i7(being only at 2ghz).
there is no involvement of memory, not even internal processor caches.

but try that :
sysbench --test=threads run

on 3.17 on RPI and 3.14 cubox, 3.10 cubox, 3.0.35 cubox. this test can make use of effective scheduler/locks handling.
and RPI on 3.17 gives ~~ as 3.14 cubox (with lower freq and inferior CPU design).

so context switching and wasted time on internal management in kernel
translated to user feel - responsiveness.

we used to use 2.4 kernels... in last century. don't remember it much but remember 3.0.35 on cubox. pain (but most Androids run it).

Raspberry Pi 2 Single thread

Raspberry Pi 2 multiple threads

Why is your faster?

No idea but I ran it in Raspbian will a full upgrade. I also deleted/removed a lot of the unwanted programs.

Try upgrading and run it again. I know there was talk of some optimization of the code, so it may have been reflected in the upgrade.

That is a bit of a difference between the two.

Just rerun the test and the results are similar to my first run.

otal time: 195.8444s

I noticed that mine was slower because i had it downclocked with XBian default OC