EIC7700X Results

[Fishwaldo]

I just got the EVB board from eswin for the new EIC7700x CPU. Results are here: https://0x0.st/Xyfg.bin

| ESWIN EIC7700 | 1400 MHz | 6.6 | Debian GNU/Linux trixie/sid riscv64 | 5390 | 1485 | 49340 | 4570 | 8700 | - | https://0x0.st/Xyfg.bin |

[ThomasKaiser]

Thank you! lm-sensors reports temperatures of stuff like CPU/GPU but no corresponding sysfs seem to be defined.

Are you able to provide output of cat /sys/class/hwmon/hwmon?/name in search of a hwmon node representing at least the CPU temperature?

[Fishwaldo]

Hi, it has a weird setup (or maybe its just the BSP kernel drivers?) - hwmon6 is the CPU temp, as well as voltages etc:

drwxr-xr-x 3 root root    0 Aug 25 23:37 .
drwxr-xr-x 3 root root    0 Aug 25 23:37 ..
lrwxrwxrwx 1 root root    0 Aug 24 01:54 device -> ../../../50b00000.pvt
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in0_input
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in0_label
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in1_input
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in1_label
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in2_input
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in2_label
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in3_input
-r--r--r-- 1 root root 4.0K Aug 25 23:37 in3_label
-r--r--r-- 1 root root 4.0K Aug 25 23:37 label
-r--r--r-- 1 root root 4.0K Aug 25 23:37 name
lrwxrwxrwx 1 root root    0 Aug 25 23:37 of_node -> ../../../../../../firmware/devicetree/base/soc/pvt@0x50b00000
drwxr-xr-x 2 root root    0 Aug 25 23:37 power
lrwxrwxrwx 1 root root    0 Jan  1  1970 subsystem -> ../../../../../../class/hwmon
-r--r--r-- 1 root root 4.0K Aug 25 23:37 temp1_input
-r--r--r-- 1 root root 4.0K Aug 25 23:37 temp1_label
-rw-r--r-- 1 root root 4.0K Aug 25 23:37 temp1_offset
-r--r--r-- 1 root root 4.0K Aug 25 23:37 temp1_type
-rw-r--r-- 1 root root 4.0K Aug 25 23:37 uevent
-rw-r--r-- 1 root root 4.0K Aug 25 23:37 update_interval
root@rockos-eswin:/sys/class/hwmon/hwmon6# cat temp1_input
30908
root@rockos-eswin:/sys/class/hwmon/hwmon6#

I'm not sure how accurate it is, other than to say, by touch, the Heatsink doesn't not even get warm, so i'm going to go with a "its pretty accurate"

Let me know if you are interested in anything else:

root@rockos-eswin:/sys/class/hwmon# cat /sys/class/hwmon/hwmon?/label
npu_vdd
cpu_vdd
soc_vdd
lpddr_vdd
dc_vdd
fan_control
pvt0
pvt1
root@rockos-eswin:/sys/class/hwmon#```

[Fishwaldo]

(pvt1 is the DDR temp/voltages)

[Fishwaldo]

I've also been told I can flip a dip switch to get it to clock at 1.8Ghz... its just that I am still pending hardware documents/schematics. When I get that, I'll re-run the test if it is true that we can change the max frequency.

[ThomasKaiser]

maybe its just the BSP kernel drivers?)

Sure, they should try to let these thermal readouts not only appear below /sys/class/hwmon but also as /sys/devices/virtual/thermal/thermal_zoneN, at least for the SoC/CPU temperature labeled as cpu-thermal or soc-thermal.

Looking through sbc-bench results it's quite a variety of node names:

tk@mac-tk results % grep "^Thermal source" *.txt | awk -F" " '{print $4}' | sort | uniq -c | sort -n -r
  38 (cpu-thermal)
  31 (cpu_thermal)
  21 (soc-thermal)
  14 (x86_pkg_temp)
   9 (soc_thermal)
   9 (cpu0-thermal)
   6 (cpu)
   4 (soc)
   4 (cpu_thermal_zone)
   4 (SOC2-therm)
   3 (cpu6-thermal)
   3 (cpu-thermal-zone)
   2 (soc_dts0)
   2 (scpi_sensors)
   2 (cpu7-thermal)
   2 (cpu0_thermal)
   1 (thsens_cpu)
   1 (soc-thmzone)
   1 (imx_thermal_zone)
   1 (iio_hwmon)
   1 (f10e4078.thermal)
   1 (cpu_little4)
   1 (cpu-1-2-usr)
   1 (ap-thermal-cpu3)
   1 (acpitz)
   1 (CPU-therm)
   1 (120e0000.tmon)

hwmon6 is the CPU temp

May I still ask for /sys/class/hwmon/hwmon6/name since then it's just adjusting a grep pattern for now to let the CPU temperature been included in your next sbc-bench run at 1.8 GHz (which I prefer to include into results lists since if the HW is capable of 1.8 GHz this should be reflected by results).

Once I got /sys/class/hwmon/hwmon6/name I will push a small commit fixing SoC detection and thermals, so after an sbc-bench -u you can retest at 1.8 GHz :)

[ThomasKaiser]

Closed by https://github.com/ThomasKaiser/sbc-bench/commit/76420fff0913e7626224879291295b1d63dc4fcd

[Fishwaldo]

Hi , here is the name of the sysfs node:

root@rockos-eswin:/sys/class/hwmon# more /sys/class/hwmon/hwmon6/name
pvt

very non-descriptive!

[ThomasKaiser]

very non-descriptive!

Yeah :)

But since the only other hwmon node labeled the same is in TH1520's BSP kernel where CPU temp is also exported as /sys/devices/virtual/thermal/thermal_zone0/ https://github.com/ThomasKaiser/sbc-bench/commit/236f4a05169958e58c8da948e14cd44dd8abc640 should now give thermal readouts on this platform w/o harming others.

But likely it will fail if /sys/class/hwmon/hwmon7/name will also read pvt.

As such after an sbc-bench -u you're ready for retesting at 1.8 GHz :)