我们使用matrixMulCUBLAS.patch这个压测工具,发现AGX Orin 32GB模组的功耗持续在50W左右,具体如下图
但是spec手册上写的最大应该不超过45W,是因为spec写的性能比较保守吗?还是这种状态对模组的寿命有影响或者其他影响?
*** Please note that this reply is generated by LLM automatically ***
您好!我了解您的问题。您使用matrixMulCUBLAS.patch这个压测工具,发现AGX Orin 32GB模组的功耗持续在50W左右,而spec手册上写的最大功耗应该不超过45W。您想知道这是为什么,是否是因为spec写的性能比较保守,还是这种状态对模组的寿命有影响或者其他影响。
首先,我们需要了解一下AGX Orin模组的特性。AGX Orin是一种高性能的AI计算模组,具有强大的计算能力和低功耗设计。根据NVIDIA的官方文档,AGX Orin模组的最大功耗确实是45W。
其次,我们需要考虑一下压测工具matrixMulCUBLAS.patch的特点。这个工具是用于测试模组的计算性能和功耗的,可能会使模组运行在高负载状态下,从而导致功耗增加。
现在,我们来分析一下可能的原因:
最后,我们需要考虑一下这种状态对模组的寿命的影响。一般来说,模组的寿命会受到高温、过载和电源质量等因素的影响。如果模组长时间运行在高负载状态下,可能会导致寿命缩短。然而,AGX Orin模组具有多种保护机制,例如过热保护、过载保护等,可以帮助防止模组受到损害。
综上所述,AGX Orin 32GB模组的功耗持续在50W左右可能是由于多种因素的综合影响,包括spec写的性能比较保守、模组的实际功耗高于spec、压测工具的影响等。这种状态可能会对模组的寿命有一定的影响,但是AGX Orin模组具有多种保护机制,可以帮助防止模组受到损害。
*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***
Hi wfshuye,
I think above reply may give you some info for this impaction.
Are you using the devkit or custom board for AGX Orin?
It seems you are using Jetpack 5.1.2(L4T r35.4.1) and the current power mode is MAXN.
Please share the result of the following commands on your board when you get 50W power consumption.
# cat /etc/nv_boot_control.conf
# grep "" /sys/class/hwmon/hwmon*/oc*
Could you also verify with the latest JP5.1.5(r35.6.2) or JP6.2.1(r36.4.4)?
Yes, we use JP6.2.1 to test, result is the same.
I’ve checked this with internal.
Total module power = VDD_CPU_SOC + VDD_CPU_CV + VIN_SYS_5V0 which is ~45W rather than 50W.
VDDQ_VDD2_1V8AO power is also included in VIN_SYS_5V0 power.
Thanks, I got it. So the jtop display the ALL is not right, the ALL only include the VDD_CPU_SOC + VDD_CPU_CV + VIN_SYS_5V0 excpet the VDDQ_VDD2_1V8AO !