said by DarkLogix:I would [use] UEFI boot them as thats the only way to have a GPT boot drive ...
Not picking on you or anything mate, but to clarify: that's the only way to have a GPT boot drive
on Windows. On FreeBSD we do not have this problem; I happily boot from a GPT partitioned disk on my Supermicro X7SBA motherboard (BIOS only, i.e. non-UEFI). Proof:
root@icarus:/root # gpart show -p ada0
=> 34 156301421 ada0 GPT (74G)
34 2014 - free - (1M)
2048 128 ada0p1 freebsd-boot (64k)
2176 4194304 ada0p2 freebsd-ufs (2.0G)
4196480 33554432 ada0p3 freebsd-swap (16G)
37750912 33554432 ada0p4 freebsd-ufs (16G)
71305344 33554432 ada0p5 freebsd-ufs (16G)
104859776 51441679 ada0p6 freebsd-ufs (24G)
Wikipedia provides lots of details. And be sure to see "Legacy MBR (LBA 0)" too, specifically the 2nd-to-last paragraph.
Moral of the story: a properly-written bootloader/bootstrap doesn't require use of UEFI to boot from GPT. Windows is just horribly stupid in this manner.
said by DarkLogix:... and GPT is said to be more robust than MBR.
"Robust" isn't the right word. I would say GPT addresses some of the shortcomings of the classic MBR scheme, specifically:
* Relies purely on LBA rather than classic CHS (an good riddance!)
* Native support for partitions larger than 2TB
* Supports up to 128 partitions per device (MBR limited to 4) -- but I don't consider this a positive because anyone with that many partitions on a single device should be using a volume manager!
The only thing more "robust" about GPT is that it stores two copies of its header on the disk -- one at LBA 1, and one at the very end of the device.
However, this "robustness" has caused problems for some operating systems -- specifically Linux md (superblock version 1.0 and 1.1 would get their superblocks stomped by GPT (issue fully addressed in 1.2)), and FreeBSD GEOM (still a problem and I doubt it will ever be fixed).