Rendimiento en DEBIAN 12

Como sin duda sabéis siempre nos gusta «cuantificar» de alguna manera el funcionamiento de los equipos que revisamos. En el caso de los SBC el test típico que llevamos usando desde hace tiempo es el conocido SBC BENCH de Thomas Kaiser y así poder comparar el funcionamiento en un ambiente controlado.

pi@NanoPi-M6:~$ sudo /bin/bash ./sbc-bench.sh -c

[sudo] contraseña para pi:

Average load and/or CPU utilization too high (too much background activity). Waiting…

Too busy for benchmarking: 19:39:11 up 24 min, 3 users, load average: 1.88, 1.56, 1.05, cpu: 43%

Too busy for benchmarking: 19:39:16 up 24 min, 3 users, load average: 1.73, 1.54, 1.04, cpu: 0%

Too busy for benchmarking: 19:39:21 up 24 min, 3 users, load average: 1.59, 1.51, 1.04, cpu: 1%

Too busy for benchmarking: 19:39:26 up 24 min, 3 users, load average: 1.47, 1.49, 1.03, cpu: 0%

Too busy for benchmarking: 19:39:31 up 24 min, 3 users, load average: 1.35, 1.46, 1.02, cpu: 0%

Too busy for benchmarking: 19:39:36 up 24 min, 3 users, load average: 1.24, 1.44, 1.02, cpu: 0%

Too busy for benchmarking: 19:39:41 up 24 min, 3 users, load average: 1.14, 1.41, 1.01, cpu: 0%

Status of performance related governors found below /sys (w/o cpufreq):

dmc: dmc_ondemand / 534 MHz (rknpu_ondemand dmc_ondemand vop2_ondemand userspace powersave performance simple_ondemand / 534 1320 1968 2400)

fb000000.gpu: simple_ondemand / 300 MHz (rknpu_ondemand dmc_ondemand vop2_ondemand userspace powersave performance simple_ondemand / 300 400 500 600 700 800 900 1000)

fdab0000.npu: rknpu_ondemand / 300 MHz (rknpu_ondemand dmc_ondemand vop2_ondemand userspace powersave performance simple_ondemand / 300 400 500 600 700 800 900 1000)

fdd90000.vop: vop2_ondemand / 500 MHz (rknpu_ondemand dmc_ondemand vop2_ondemand userspace powersave performance simple_ondemand / 500 750 850)

sbc-bench v0.9.68

Installing needed tools: distro packages already installed. Done.

Checking cpufreq OPP. Done (results will be available in 21-32 minutes).

Executing tinymembench. Done.

Executing RAM latency tester. Done.

Executing OpenSSL benchmark. Done.

Executing 7-zip benchmark. Done.

Executing cpuminer. 5 more minutes to wait. Done.

Checking cpufreq OPP again. Done (23 minutes elapsed).

Results validation:

* Advertised vs. measured max CPU clockspeed: -1.4% before, -1.4% after -> https://tinyurl.com/32w9rr94

* Background activity (%system) OK

* Too much other background activity: 0% avg, 3% max -> https://tinyurl.com/mr2wy5uv

* No throttling

Memory performance (all 3 CPU clusters measured individually):

memcpy: 5858.4 MB/s (Cortex-A55)

memset: 21873.5 MB/s (Cortex-A55)

memcpy: 12563.8 MB/s (Cortex-A76)

memset: 27853.4 MB/s (Cortex-A76)

memcpy: 12518.8 MB/s (Cortex-A76)

memset: 27820.9 MB/s (Cortex-A76)

Cpuminer total scores (5 minutes execution): 21.46,21.45,21.44,21.43,21.42,21.41,21.34 kH/s

7-zip total scores (3 consecutive runs): 15079,15030,15049, single-threaded: 2584

OpenSSL results (all 3 CPU clusters measured individually):

type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes

aes-128-cbc 158769.46k 473743.47k 929576.79k 1229990.91k 1357179.56k 1366851.58k (Cortex-A55)

aes-128-cbc 596495.18k 1223657.00k 1624084.65k 1758762.33k 1812441.77k 1817903.10k (Cortex-A76)

aes-128-cbc 600417.93k 1231792.11k 1634646.36k 1770101.08k 1824189.10k 1829573.97k (Cortex-A76)

aes-192-cbc 152370.02k 423691.03k 758384.81k 947553.28k 1021777.24k 1027391.49k (Cortex-A55)

aes-192-cbc 568518.02k 1093940.54k 1384255.83k 1469783.04k 1511576.92k 1515132.25k (Cortex-A76)

aes-192-cbc 572282.04k 1101187.11k 1393372.67k 1479385.77k 1521582.08k 1525148.33k (Cortex-A76)

aes-256-cbc 148459.82k 389443.09k 656567.72k 792355.50k 842768.38k 847178.41k (Cortex-A55)

aes-256-cbc 551302.25k 980442.52k 1201969.92k 1269485.91k 1296673.45k 1299327.66k (Cortex-A76)

aes-256-cbc 554889.03k 986522.62k 1209252.27k 1276968.28k 1304840.87k 1307546.97k (Cortex-A76)

Full results uploaded to https://0x0.st/XU_7.bin

WARNING: The DMC governor settings of this device are not adjusted for performance

and as such DRAM bandwidth and latency might have been severly harmed. Your current

settings: dmc_ondemand (upthreshold: 40, downdifferential: 20)

For a discussion wrt settings see https://github.com/ThomasKaiser/Knowledge/issues/7

It is strongly advised to switch to maximum performance and run sbc-bench again to be

able to compare real hardware performance with default settings of this distro/kernel.

You might want to execute this now and call sbc-bench again directly afterwards:

echo performance | sudo tee /sys/devices/platform/dmc/devfreq/dmc/governor

Alternatively you can simply use review mode: sbc-bench -r

En este caso nos ha sorprendido favorablemente que no necesite bajar la velocidad del reloj para no recalentarse (No throttling). Sin embargo no nos ha gustado el último párrafo en el que parece decir que los propios creadores han ajustado / reducido el rendimiento del dispositivo lo que justifica nuestra impresión de que parece que funciona más lento que los anteriores SBC de este fabricante con el mismo procesador.

¿Será que han decidido proteger más de lo habitual el sistema al perder parte de la refrigeración al colocar la pantalla en la parte superior de la caja que además genera energía? Lo preguntaremos.