I have tried the new FreeNAS 9.3 on two older servers and neither boots off. GPT formatted USB sticks seem to be the culprit.
The symptoms: Either via GUI, or via USB boot or via CD boot (everything works), I do "install" FreeNAS ver 9.3 on the USB stick. While booting, the machine hangs at the non-blinking cursor and never boots. The USB stick has been created, is a perfect one, i.e. it is able to boot at newer machines. Various USB stick brands and sizes have been tried ( like idVendor=14cd, idProduct=121f , 8.04 GB/7.49 GiB), so this is certainly not the issue.
About the cursor - this is precisely the moment when the stage 1 bootloader should be launched. Instead of this, *nothing* happens. The machine will not freeze - CTRL-ALT-DEL helps, but there is no boot. The message "no suitable video mode" characteristic to GPT menu, will never appear. Using ENTER or other keys for "sleep" boot won't help.
The issue very evidently is in the principal inability of the older hardware to boot GPT sticks. At the same time, the hardware is absolutely capable to boot normal "ISO" sticks.
Machine no 1: HP/Compaq ML-350 G4p. BIOS version D19 02/15/2006. Very naturally - (U)EFI was not the mainstream artefact these years. BIOS has no option to tweak, at least HP SPP 2012.01.0 includes no upgrade for the particular BIOS. (Needless to say, but some purists think they are wiser asking irrelevant questions - 12GB of ECC mem, 5x2TB disks in pool, HP P600 card, FreeNAS 9.2.1.9 works brilliantly. The particular HW has been used 24/7 for FreeNAS over a year, so far no complaints).
To be even more precise - After F9/F10/F11 prompt, there is a little timeout, then two fast beeps as normal, but then (instead of the usual"Attempting to boot from CD") nothing happens, only a non-blinking cursor stays visible (and forever so).
Machine no 2: Fujitsu Siemens RX300 S3. The same behaviour - the server will start CD and is able to correctly create the GPT stick (it boots on newer HW). However, due to a known HW bug, it seems to be impossible to install from USB to USB on this hardware - only lower frontside USB connector is fully usable.
I am afraid I have to say out a very unpopular thought (I even registered as the forum user to do so) - it seems that the decision to use GPT layout on the USB stick was a bit of premature. It breaks booting on the otherwise normal legacy servers of a certain period. To use a dedicated SSD for booting is theoretically possible but painful.
My proposal is: somebody should amend the documentation saying it in clear and open - the premature decision of using GPT boot sticks has the unpleasant outcome that old servers of certain period (most likely 2004-2007) cannot boot these sticks and are thus out of the FreeNAS game.
Seems that the phenomena is not so unknown:
https://forums.freenas.org/index.php?threads/tried-9-3-back-to-9-2-1-9.25560/ . The citation: "
We suspect, but cannot prove, that a number of older motherboards (with older BIOSes that cannot deal with GPT labels) and/or older USB sticks that lock up under ZFS may be the culprit, but like I said, we've never seen it.". I very much hope someone is able to explain the phenomena of not booting GPT sticks more scientifically.
(Aside from the story, I am already using FreeNAS 9.3 on a newer machine. For my simple home installation, I haven't yet seen bugs.)