Performance Payoff - The 10 GbE Changeover
As mentioned, Intel’s Data Direct I/O architecture provides a significant performance boost, but there are some other secondary benefits, as well. With fewer components in the packet processing chain (thanks to 20 MB of CPU cache in the E5 taking over the old role of system memory), total system power consumption drops by up to 65W. Additionally, roundtrip signal latency drops by roughly 7%. While these benefits are admittedly proprietary to Intel NIC and CPU combinations, they show how 10 GbE can be more effectively implemented as part of a tightly integrated platform.