I've done a bit of research into socket 4 Pentium and PCI. For all the problems you may have with VLB, PCI in the early
days (1993) was a mess. It wasn't really until 1994 that most of it's problems were ironed out.
As far as I can tell, in the early days Intel was the only company providing PCI chipsets for the 486 and Pentium. The
Intel Saturn and Aries 486 chipsets were not particularly fast. If you wanted PCI and a Pentium you had one option, the Intel 82430LX (Mercury) chipset. I think the PCI based Intel "xPress Extended" was a later product.
The early Mercury chipset has a problem with the write-back cache of Pentium. This slows down PCI/ISA systems like the early Batman boards. It is particularly bad on the PCI/EISA systems. By Oct 04, no vendors were showing off PCI/EISA systems (InfoWorld Oct 4 1993 p8)
Intel seem to be fixing this by late October (Computerworld Oct 25, 1993 p45) Major manufactures have delayed shipping PCI systems due to this problem. Intel appears to have fixed this issue by Nov 29, when it release a new chipset revision (InfoWorld Nov 29, 1993 p1). Bus mastering, related to this issue is now fixed.
Other problems:
End of November, Intel redefines the Pentium spec to use 5.6v for the Pentium 66Mhz. A Non-trivial change in M/B design, further delaying Pentium adoption (InfoWorld Nov 29, 1993 p10)
Windows NT (3.1) Lacks PCI drivers as it was not available when MS shipped NT, "We didn't put it in because there wasn't any hardware to test with it" - Stork, "We just haven't been able to get manufactures to ship us PCI based systems so that we could test them." - Stork. MS has driver, but its not available yet. (InfoWorld Nov 22, 1993 p2).
There is also a problem with PCI-to-PCI bridges. From the later 450KX/GX datasheet: "Finally, if a P2P device in use is not fully compliant with the PCI 2.1 specification, the system is exposed to unresolvable conflicts between multiple bus masters issuing transactions attempting to cross between the hierarchical PCI busses. To eliminate the possibility of a resulting livelock failure, the system must operate with CPU-PCI write posting disabled. This will degrade the performance of outbound traffic such as graphics, but will not adversely affect the performance of bus mastering I/O devices." AFAIK this is describing shortcomings in PCI 2.0.
Conclusion:
PCI was very dependent on the 430LX chipset, and the early revisions have some serious issues partially caused by driver incompatibility, but also inherent in the chipset. The PCI/EISA variant particularly suffered. In addition, because of the easy adoption of VESA on the 486, manufactures were more inclined to build VESA addon cards. A PCI/ISA system built on the Intel Batman platform, side stepping the EISA issue, would have suffered from a lack of high end EISA add in cards. Everything fast would have to be on the PCI bus. However this has problems with bus mastering and the very useful PCI-to-PCI bridge, used on many later cards wont work. So it would be difficult to recommend PCI in 1993. It just wasn't an established technology. EISA would be the standard in the server world for at least a year.
An analysis of the the relevant datasheets.
If this seems confusing it's because it is. Intel keeps redefining the the spec. For example the 82378IB southbridge originally supports 4 PCI masters, then it's only two, but the 82378ZB supports 4, but then it's 6. See the notes section for a summary.
TL;DR
In short buy a VLB based 486 in 1993, and wait till '94 for a Pentium (as most did). Alternatively buy a Compaq or ALR in 1993.