The validation and accuracy of internet protocol version 6 (IPv6) performance using jumbo frames is still not perfect, due to peer-to-peer connections testing within the same operating system and between operating systems. Therefore, inaccurate data test results. To mitigated, testing with a wider platform is recommended, a medium-scale network connection is proposed such as metropolitan area networks. In this works, a connection between computer devices connected by three proxy routers are made, with different IPv6 segments on each port. Then each computer device sends traffic data to each other using a traffic-generator application. The first test through three routers without tunnel connection is carried out as the first scenario to compare performance with tunnel-based testing. Three parameters have been used in this test, such as maximum transfer unit (MTU) 1500 bytes, MTU 400 bytes and MTU 9000 bytes. The results of the tests conducted show that the use of jumbo frames using a proxy is less effective, even though it produces a larger throughput when using the MTU 4000, but there is fragmentation in the packet passing through the proxy because the packet passing through the proxy is split into 1500 byte sizes.