It depends on the model of xeon. Current production xeons are fine when matched up with i5 and i7 chips because they share almost all features and clock for clock are very close performance wise. Clocks from 5 years ago are not the same as clocks today, more commands are issued and processed in the same period of time now.
Old xeons, like the one the OP is using, rock a 45nm architecture, do not support turbo boost, do not support hyperthreading, do not support AVX commands, and were equivalent to the socket 775 core2 quads of the era.
Not to be rude, but how much of an improvement are i7's over the fastest core2 quad made (qx9770)? The answer is...A LOT. I had a qx9770 rig, and between chipset limitations, total supported bandwidth, ddr2, and 0 pcie lanes to the processor...probably more than twice as fast at this point. Those xeons have 820 million transistors on die vs 2.270 billion transistors on a 3930k at 6 cores 32nm or 1.4 billion transistors on a 3770k at 4 cores 22nm.
Not even considering all the bandwidth limitations native to those older systems, but modern xeons use QPI links to streamline the teaming of the processors which puts them in another category altogether.
It's apples and oranges. PassMark scores that processor at 3,742 passmarks. Assume that they have perfect doubling by adding a second (which they would not) and you get 7,484 passmarks. The i7-3930k gets over 12,000, and the 3770k gets 9,637. Take into account the extra bandwidth from pcie lanes being pumped directly into the CPU (great for HBA cards and graphics), ddr3, faster fsb clocks, updated mobo architecture etc and you can safely say a single i7 would be more than twice as fast as a pair of outdated xeons.
Take into account the money spent on old components to get them up and running, lack of modern driver support for the hardware and chipset under newer OS's...you get my point. While that benchmark isn't a real view of daily performance, it's a good starting point.
I'd love to see someone run a pcmark vantage and pcmark7 bench of a dual ls5420 system just to put some actual comparable numbers to it, since pcmark tests the system as a whole and not just the CPU.
As far as power efficiency, you're talking 50w x2 for 8 cores vs 77w on the 3770 for 8 cores (including HT) and 130w on the 3930k for 12 cores (including ht). If you underclock/under volt either of those you're still going to get better power efficiency just based (again) on the modern implementations of power saving features (speedstep, c-states, motherboard power, etc)
Edited by goros - 3/27/13 at 5:06am
