Does nvidia have any memory bandwidth testing tools and monitoring tools, such as intel’s PCM and MCL tools
Does nvidia have any memory bandwidth testing tools and monitoring tools, such as intel's PCM and MCL tools
The short answer is yes, that you would probably want to use NVML or DCGM tools for that. But I am not the expert. I have reached out to the NVML expert to give you a more detailed answer here.
Ok thank you, please take a look at it in detail, I want to calculate the bandwidth when I use it
Hello，How to test the memory bandwidth on the soc? Is there any official tool from Nvidia, such as the memory bandwidth on orin?
I’m working on a project that involves monitoring and optimizing the CPU performance of my servers. I’m aware of tools like Intel’s PCM for monitoring memory bandwidth, but I’m wondering if there are similar tools or plugins available for monitoring CPU performance metrics in Grafana. I’d love to track metrics like CPU utilization, temperature, power consumption, and clock speeds to gain insights into my server’s performance and identify any potential bottlenecks.
@gupadhyaya do you have any insight here?
NVML and nvidia-smi are intended only for GPU monitoring; they don’t contain any APIs/hooks to do CPU performance monitoring. For CPU monitoring your best bet would be to use tools provided by the CPU manufacturer (e.g. Intel VTune profiling tools for Intel CPUs).
NVIDIA Nsight Systems does do CPU performance monitoring, for x86 and ARM server systems but we do not currently cover temperature or power consumption.