There are also Xeons with no i7 equivalent, such as dual and multi processor Xeons. You cannot replace these with i7 chips.
On new architectures, Xeon seems to come out much later than i7.
The latter is deceptive and the former applies to the i7 as well. The "Core i7" label is applied to
two different designs that fit into different sockets. One is the "mainstream" design for that microarchitecture iteration and the other is applied to the "server/performance" design for that microarchecture. The i7 3960+ series is different from the i7 3600 series.
The i7 3960+ (sometimes 3800 and formerly 9xx ) series arrives around the same time as the Xeon E5 class does. ( There was a 3 month gap between first 3960+ and announce for E5 this last iteration but isn't an established pattern yet. ). A couple of months relatively isn't in the "much later" category. [ Technically the E5's shipped before the i7 did, but only to supercomputer system vendors. So no, there isn't a large gap for the moderately patient. ] The core i7 is all very muddled at this point.
The differences in i7 line up show up in socket type and/or whether iGPU is incorporated.
The same label applied to two design is what continues to fuel the myth that Mac Pro would be "oh so cheaper if just used the cheaper i7 and dumped Xeon". That is even less true now that the Xeon E3 line-up is in the mix.
I have heard that Xeons run cooler. If this is true, it is probably due to the Demand Based Switching.
Not really. There are Xeon variants that top out at higher core counts than the "Core i7 Extreme" variants. More cores means lower top end clock speeds which leads to lower individual package. However, in two package set-ups it is a higher system TDP. For example 2 * 95W ==> 190W versus the 135W of something with lower top end core count but higher clock speed.
Similarly there are some substantially underclocked versions with product numbers like xx02 xx05 xx10 that are intended to more highly temperature constricted telecom cages. For those, x86 corer performance is tossed to hit the lower TDP, but leverages the higher aggregate I/O bandwidth available. Sort of the same as the ULV versions in the mainstream line up ( same stuff just clocked much slower. )
Lack of ECC support does not make i7 chips "faster". Xeons don't have to use ECC either, they just have the option of doing so.
This is far more often a indirect argument against restricted overclocking than it a significant performance gap. But yes if primarily interesting in "drag racing" the chipset and CPU package the Core i7 option is better.
It is extremely unlikely Apple is going to ship a drag racing oriented system.
On some tick iternations it is often moot since the Xeon server focused designs move to a faster memory speed than the "mainstream" design will go with. Apple also uses Unbuffered ECC (and keeps the DIMM slot count down ) which also reduces the speed gap.
