我们在同一块主板上适配了 r38.2 和 r39.2 :
r38.2 GPU 功耗可以到 60W
r39.2 GPU 功耗只能到 40W
压测指令都是:
nvpmodel -m 0
jetson_clock
stress-ng --cpu 14 --cpu-method fft &
./gpu_burn -tc 1000000
请问这个是什么原因?有排查思路吗?
另外问一下 r39.2 多出来的 Sensor 1 是什么?
Sensor 1 and Sensor 2 are not descriptive thermal-zone names. They are jtop/libjetsonpower generic labels for hwmon temperature inputs that do not expose a useful temp*_label.
Composite -> NVMe composite temperature
Sensor 1 -> NVMe internal sensor 1
Sensor 2 -> NVMe internal sensor 2
cpu -> /sys/class/thermal/thermal_zone*/temp where type=cpu-thermal
gpu -> /sys/class/thermal/thermal_zone*/temp where type=gpu-thermal
soc012 -> /sys/class/thermal/thermal_zone*/temp where type=soc012-thermal
soc345 -> /sys/class/thermal/thermal_zone*/temp where type=soc345-thermal
tj -> /sys/class/thermal/thermal_zone*/temp where type=tj-thermal
for h in /sys/class/hwmon/hwmon*; do
[ "$(cat "$h/name" 2>/dev/null)" = "nvme" ] || continue
for f in "$h"/temp*_input; do
[ -e "$f" ] || continue
b="${f%_input}"
label="$(cat "${b}_label" 2>/dev/null || echo "$(basename "$f")")"
printf "%-12s %s " "$label" "$f"
awk '{printf "%.2f C\n", $1/1000}' "$f"
done
done
Composite /sys/class/hwmon/hwmon0/temp1_input 43.85 C
Sensor 1 /sys/class/hwmon/hwmon0/temp2_input 43.85 C
Sensor 2 /sys/class/hwmon/hwmon0/temp3_input 46.85 C
请问这个问题有排查思路吗?是否是 BPMP 固件更新导致的?
Hi,
Could you share the gpu_burn test so we can check it internally?
Thanks
Hi,
Thanks for the update.
We will test this internally and provide more information to you later.
Thanks.
Upgrade your jtop
curl -LsSf https://raw.githubusercontent.com/rbonghi/jetson_stats/master/scripts/upgrade-jtop.sh | bash
Hi,
Is this issue related to jtop?
Thanks.