We have tested benchmark application i7 & xeon e5-2620 server. Here are the results
i7 results -
[ INFO ] Latency: 12.6197 ms
[ INFO ] Throughput: 79.2411 FPS
Xeon e5-2620 results -
[ INFO ] Latency: 18.3562 ms
[ INFO ] Throughput: 54.4776 FPS
Not sure why performance of xeon server is lesser than i7. Do we need to fine tune any parameters for xeon server?
Appreciate any help to improve the performance on xeon server.
Xeon server details are attached.
Sudir it's very difficult to diagnose exactly why the benchmark app runs faster on an I5 versus I7. There are many variables involved such as memory configuration, OS version, kernel version, UEFI/BIOS settings, etc...It is impossible to make a general statement without knowing more.
However, one way to figure it out is to use performance tools (Intel has many to offer but there are also open source tools available) to see where the bottlenecks are in each configuration, then compare and contrast your findings.
Although I am not an Intel expert, it was a topic of great interest, so I examined the processor differences a little bit.
Processor Base Frequency：i7 (3.40 GHz) > Xeon (2.00 GHz)
Bus Speed：i7 (8 GT/s DMI3) > Xeon (7.2 GT/s QPI)
Number of processors：i7 (x1) < Xeon (x2)
# of Threads：i7 (x8) < Xeon (x12)
I am also very interesting about what is affecting performance.
I’m also very interested to know which is the best HW for a large scale deployment using OpenVino.
- PCI-e with many MyriadX?
- Future discrete intel GPU?