benchmark

[huichenggong]

I recently did some test and compared the Pi64 with raspbian. I'd like to share it hear. https://blog.csdn.net/weixin_38412284/article/details/79900850 I wrote it in Chinese. Here is a brief introduction to the result.

In general, after the 64-bit system was used, the integer operation (sysbench) fully utilized the performance of the processor and was greatly improved. Floating-point performance (linpack) remained almost unchanged with a slight decrease. Memory performance (memtester) has decreased due to 64-bit reasons. In the application testing section, the GMX was tested. The overall strength of the 64-bit system was still good, with a certain degree of improvement. Hope this will give you a brief insight into the performence of Pi64.

[ThomasKaiser]

Hope this will give you a brief insight into the performence of Pi64.

Unfortunately not. Every 'hardware performance' comparison including sysbench is just collecting numbers without meaning. Same with Linpack when done wrong (as here).

[huichenggong]

sorry for the confusion numbers, I just want to compaire it with Raspbian.Can you please tell me what did I do wrong in the Linpack. @ThomasKaiser

[ThomasKaiser]

Linpack is not a 'fire and forget' benchmark and you should never use distro packages (as it's done everywhere unfortunately): https://www.raspberrypi.org/forums/viewtopic.php?t=208057&start=25#p1288630

There are a few threads over in the forum discussing performance impact of 32-bit vs. 64-bit code. Sometimes 64-bit code runs slightly faster, sometimes slightly slower. Since Raspberries have only a laughable amount of DRAM memory constraints should also be taken into account, see e.g. https://github.com/nodesource/distributions/issues/375#issuecomment-290440706

[huichenggong]

thank you very much for pointing out my mistakes. I haven't done enough work on it.