Hi all, after some research it is becoming clear that the latest generation of Dell XPS BIOSes (9350/9360/9550/9560) are fitted with BootGuard, which means that even if we manage to successfully repack any modded ME or BIOS, the laptop won't boot.
Until we figure out a way how to bypass this protection we are limited to patching IFR variables via UEFI.
With this in mind, and following the requests by a number of users in this forum, I have attached the first set of variable changes (based on BIOS v2.1.0 ONLY) for cautious users to experiment with. Warning - these settings should not be taken lightly. The wrong settings can - and will - brick your board making recovery hard or impossible.
Here is a brief summary of what's working so far:
1) CPU and GPU undervolted by 50mV (works well on the i7-7500U). OSX does run at a lower temperature and with longer battery life (in my case around 15-20%). You *can* increase these settings but there is a significant risk that the board will not power up. Use with extreme caution.
2) Overclock enable. These settings allow you to use XTU freely within Windows, but most changes will not be preserved on boot into OSX. Use these to experiment with what settings work on your machine and then apply them via UEFI.
3) DVMT patch and MAX fix to remove framebuffer issues. Self-explanatory.
4) Ring Max OC. Normally set to Turbo-2, can be increased slightly beyond this, unknown if this has any real-world effect.
There are a couple of protection mechanisms to fall back on (but not guaranteed to work). If one setting renders the board unbootable (POST fail, and white/yellow flashes):
(i) disconnect the battery
(ii) disconnect the CMOS battery (on the right of the board)
(iii) press power for 30 sec to discharge EC/NVRAM
(iv) reconnect, press CTRL+ESC, plug in power
If the above fails there is also the (undocumented) possibility of putting the DELL BIOS on a USB stick, inserting it into the LEFT port and pressing CTRL+ESC before inserting power. No reports of this working successfully yet, but please post back if you have any results.
There are a lot more variables and configuration to experiment with, in particular how ConfigTDP works to overclock the multipliers for the CPU and GT, and forcing ASPM for the NVMe controller saving even more power.
Thus this project can now become one of contribution, where anyone who is willing to experiment (within defined bounds of risk) can post their successes and/or questions.
Again - this is not for most users. Those who are fully aware of the implications (and have successfully done this in the past) are welcome to try, experiment cautiously, and post back.