Today my 3014 got the other hardware upgrade of the current portfolio: Upping the sample rate from 4 to 5 GSa/s. Kudos to user "memset".
Originally I wanted to do that together with the frontend upgrade, but ordered only one inductor, didn't knew there's another on the backside.
For the record, I used Coilcraft 0603CS-10NXGLU (10 nH), and of course removed the strapping resistor H2. I was a bit tempted to try 12 nH and hope this would do both frequencies, so I could change the sample rate back and forth, but from the ADF4360 datasheet this looked too marginal to be safe. The lock range of the oscillator is rather small.
After that mod, the trigger position was way off, the scope needs a new user calibration, which fixed it.
BTW, the calibration can be backed up with telnet and USB, it resides in the hidden folder "\secure\cal\", the two *cal.dat files. There's also a logfile from the calibration run, it seems Keysight does the same at their manufacturing site.
The scope seems to do well with the unusual rate from its biggest cousin, measured timings and display are correct. As expected, the effect of 25% finer sampling is not dramatic. The hardware rendering does a good job on sinc interpolation already. It's slightly visible on direct comparison. Again pictures of a "Jim Williams" pulse, 500 MHz bandwidth, before and after.
Slightly more amplitude of the pulse gets captured, the dip behind the pulse is a bit deeper. (We're talking about 4 vs. 5 samples per nanosecond division grid here, everything else is the sinc interpolation...)
Before, 4GSa/s:
After, 5 GSa/s: