For some other folks doing this on a 16 Gb mSata just as I did, there is a way to reclaim quite a few disk space and end up with approximately 14 Gb of space instead of the 7 Gb you would normally get when installing ESXi.
Once installed you'll have to move the scratch space from the dedicated 4 Gb partition to a different area.
To do so I followed the instruction found in this KB article: https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1033696
In particular I performed this section:
Configuring a persistent scratch location using the VMware Host Client for ESXi 6.5
Connect to ESXi host using Host Client.
Right-click a datastore and click Browse.
Create a uniquely-named directory for this ESXi host.
Close the Datastore Browser.
Click the Manage tab
Under System, click Advanced System Settings.
The setting ScratchConfig.CurrentScratchLocation shows the current location of the scratch partition.
In the field ScratchConfig.ConfiguredScratchLocation, enter a directory path that is unique for this host.
For example, /vmfs/volumes/DatastoreUUID/.locker-ESXHostname.
Reboot the host.
I relocated my scratch space to the packages partition. I would avoid doing to on the bootbanks.
Subsequently I was able to edit the partition in ESXi 6.5 UI and delete the 4 Gb Data partition as well as 2.5 Gb Diagnostics partitions and create a new datastore on the free space which now showed around 14 Gb.
Of course not something you would bother do on a large mSata disk, but on a 16 Gb one it does make sense.
I certainly cannot guarantee what will hapen in the case of a version update when you did such a change, so as always with those things decide for yourself and be aware this can have consequences !
To be honest I went away from ESXi on my APU2C4, the network throughput just didn't work for me. I wanted to have an easy way to switch back and forth between 2 pfSense versions. But in the end it becomes a bottleneck for my internet connection and even when I go through the painfull VMXNet configuration the throughput will max out at 260 Mb / s.
As my connection will give me 291 Meg this is a waste of bandwidth I'm not willing to accept so ended up installing pfSense natively and went to the expected network performance.
I simply ran the script with a remove option: .\ESXi-Customizer-PS-v2.5.ps1 -remove vmkusb
After a while you'll find an ISO file in your running folder that you'll be able to boot from and install ESX 6.5. Don't forget to add the kernel option to the boot.cfg on the stick otherwise you won't see the output of your console.
The installation seems to hang at 27%, but be patient as is it actually copying the files over to your storage.
After that I recommend SSHing into vmware to edit boot.cfg rather than booting tinycore, as this screwed by installation.
If you need help or even the produced ISO file just let me know
They made some changes to vmkusb in the latest update, now it's working like it should without custom build (I upgraded from 6.0 to 6.5 update 1) !
https://esxi-patches.v-front.de/ESXi-6.5.0.html (I used ssh to upgrade)
Summaries and Symptoms
This patch updates the vmkusb VIB to resolve the following issues:
The ESXi host might fail with a purple screen when booting the ESXi host on the following Oracle servers: X6-2, X5-2, and X4-2, and the backtrace shows that is caused by pcidrv_alloc_resource failure. The issue is caused because the system cannot recognize or reserve resources for the USB device.
When a Keyboard is configured with a different layout other than the U.S. default, and later it is unplugged and plugged again to the ESXi host, the newly-connected keyboard is assigned with U.S. default layout instead of the user-selected layout.
On some servers, a USB network device is integrated in IMM or iLO to manange the server. When you reboot IMM by using the vSphere Web Client or an IMM or iLO command, the transaction on the USB network device is lost.
The XHCI related platform error messages, such as xHCI Host Controller USB 2.0 Control Transfer may cause IN Data to be dropped and xHCI controller Parity Error response bit set to avoid parity error on poison packet, are reported in ESXi VMkernel logs.
An ESXi host might become unresponsive with no heartbeat NMI state on AMD machines with OHCI USB Host Controllers.