Heads up to anyone with these bodies, Nikon finally released the firmware update for these bodies today to allow the use of CFExpress cards in addition to XQD.

I'm happy about it.  I (extremely likely) won't get a performance boost on my D850, but extra card compatibility is nice.

Yeah, pretty much my thinking - but I have a ProGrade Cobalt card en route and plan to do some testing.

CFExpress is the same form factor as XQd, right?

Correct, just capable of faster read/write speeds and has more suppliers than XQD. Hopefully the more suppliers bit will help bring the price down after a while.

Damn, you went all in with that Cobalt lol. I can't bring myself to spend that kind of cash on a memory card just yet that I don't think will make that big of a difference (if any) in either of my compatible cameras.

CFexpress Type B is the same form factor. Be aware that there are Type A cards.

Yeah, I considered it an investment for my next camera, too.  

Results from my non-scientific testing of my D850 w/CFexpress...

Setup:
Camera: D850 w/firmware version C 1.20, LD 2.017, LF 1.00
XQD card: Sony G-series 240GB
XQD reader: Sony QDA-SB1
CFx card: ProGrade Digital Cobalt 325GB
CFx Reader: ProGrade Digital PG05.5 CFexpress B/SD (dual slot reader)
PC storage: Internal NVMe SSD

BLUF:

1. The CFx card functions correctly in the D850 after this firmware upgrade.
2. There is no noticeable difference in camera write speed between the the two cards (XQD, CFx).
3. There is a very significant difference in read speeds on the PC with the CFx nearly ~1.8x faster than the XQD.

Personal Observations:

1. I expected and was not surprised to see no noticeable difference in write speeds in the D850 since the throughput speed (bus speed) of its card port is much more dependent on hardware limitations that are unaffected by firmware updates.
2. Use a USB 3.1 Gen 2 port on the PC/Mac for maximum throughput.
3. Most casual users will likely never see a performance difference in read speeds between the ProGrade Cobalt and Gold series cards. I'm satisfied with my purchase since read speed was my main motivation and I expect CFx to be a storage mainstay. As always, YMMV.

Methodology for Write Test in Camera:

1. Set camera to record 14-bit compressed raw (NEF) images only. (No "+JPEG" storage.)
2. Remove the SD card to ensure no attempt to write to that slot.
3. Fill the image frame with a brightly lit white card. The goal is to avoid any detail that would cause additional processing/compression delays. (I chose a white image over a dark one to make it easier to count images on the camera [without having to move the card to the reader]).
4. Set the shutter mode to "continuous high".
5. Insert XQD card.
6. Format the card.
7. Press and hold the shutter until the first "stutter" indicates a buffer overrun condition.
8. Count the number of images captured.
9. Repeat 6-8 three times.
   
10. Average the number of images captured.
11. Insert CFx card.
12. Repeat 6-8 three times.
13. Average the number of images captured.
14. Set the camera to record 12-bit compressed raw images.
15. Repeat 5-13.

Key finding for write test:
Very similar numbers of images captured between runs for CFx and XQD.


Methodology for Read Test on PC:

1. Use the last series of images from the in-camera capture test and copy them to both cards (108 files, ~4.15GB total).
2. Use Windows Explorer to select all images from the XQD card.
3. Copy all files (Ctrl-A, Ctrl-C)
4. Navigate to a discrete folder on the internal SSD.
5. Launch a timer app on a smartphone.
6. As near as possible, simultaneously execute a paste operation (Ctrl-V) and tap the "start" on the timer.
7. Press "stop" on the timer as close as possible to the completion of the paste operation (as soon as the progress bar shows "done").
8. Delete all files from the target folder.
9. Note duration of the operation.
10. Repeat 2-9 three times.
11. Note the average time (~14.9sec).
12. Use Windows Explorer to select all images from the CFx card.
13. Execute steps 3-9.
14. Repeat 11-12 three times.
15. Note the average time (~8.1sec).

Key Finding for Read Test:
The CFx card is much faster than XQD for read operations (~1.8x) when using a USB 3.1 Gen 2 port.


Additional notes:
To demonstrate performance, here are averages for 12-bit files captured in camera via the CFx card. Again, the numbers captured were nearly identical with both XQD and CFx. For action sequences, I use 12-bit lossless compressed to balance performance with image detail and, with 7fps, I can reliably capture ~6 seconds of action if I choose to use only the XQD/CFx card. (Write speed for the Sony G-series SD card is about 75% that of the XQD/CFx in the D850.) The camera seems to require <3 seconds to empty the buffer to the card (though that was not a specific measurement I attempted).

1. 12-bit, compressed: 108 images (~38.0MB/file)
2. 12-bit, lossless compressed: 69 images (~46.4MB/file)
   
3. 12-bit, no compression: 34 images (~70.7MB/file)