Flashed all cards back to original bios....
There's something funny going on with wattman.

I have a 1920 threadripper with 12 cores and hashing with 11 (1 core to keep system responsive)
I have Large Page Supoort Enabled and hash at around 90 per core = ~1000

Wattman settings for the vega56's    945 mem -30 core  - 20 power
(Got it to mine at 1080 mem once with one card at 1.000 volt mem and it hashed nearly 2000 ?!?! then it crashed after 15 secs.)

if i turn on HBCC it does mines at 900+880 = ~ 1780 per card
Stak settings 1522+1512
The card with the monitor attached is set to 900 mem to keep things stable...

All nice and well , rig is hashing around ~7500 with threadripper and 4 x vega 56's @ ~950 watts.

Point is, when i reboot cards hash @ around 550+550 = 1100. And everything in wattman looks the same.
But for some reason HBCC has changed and doesnt show any  changes  in wattman.
Settings look exactly as before rebooting.

If i adjust the slider , (any way) , it starts hashing @ 900 + 880 again.

So wattman does not correctly set HBCC by itself after reboot, and have to set is for all 4 cards again each time.

 i  have 32gb quad channel mem and 70.0000 swap file on fast NVME SSD.

First have to start stak cpu and then stak gpu , otherway around it bugs hangs...

Any tips and tricks ?

I made it work on 5 it seems : http://prntscr.com/grm9ed

upped the virtual ram to 80. Disabled On board graphics , connected Monitor  with HDMI to GPU0.
Without Team Viewer going the hashrate is slightly higher - around 10500.
Only 4 cards accepted some change on the slider. The 5th had HBCC on, and on/off was not doing anything. On the 6th one HBCC was disabled. It was not even possible to enable it.

Sadly ... that lasted only until the first restart.
After it - we are back to 4. And I can't get em to apply any HBCC settings on at least 3 cards.

I have the same situation.  I can't get HBCC to activate on more than 4 cards.  The sliders work for all but there's no effect for anything over 4 cards.

I had the same issue and found the same fix on another forum. My main issue now is getting a good hash to power ratio. Without applying any soft power table mods I can get 1600 H/s at total system power draw of 200w at the wall for one Vega 64 (idle power draw is about 55w). That's with two threads in xmr-stak (2016 and 1600 intensity), -50% power limit, -30% GPU, and 900 MHz on the ram. Can anyone please advise whether this is a good hash to power ratio or whether I can do better?

Guys, don't take this comment the wrong way but right now we have old questions, which were already answered before, being asked, and old discoveries being shared again.

If you really want to help this thread, start by reading what others shared already and come up with new questions/ideas/discoveries.

Maybe it's good for a mod to freeze this thread for a while and put every relevant info on first page

Monitor attached to the GPU  both times? Vega has this quirk that when monitor is attached it works faster

i'm doing 7500 at 650w(all system), 3xV56+1xV64
v56 stable at 1850, can go 1920+ need more voltage, i didnt test again
while v64 at 1970 with lower voltage,
vega truly awesome card

So, at last I grab my new Vega56, flashed it with liquid bios and apply the same patch I used with my pure Liquid Vegas.
But I decided to clock mem for vega56 to 1090 instead of 1100 as it has not so good cooling as liquid.
Screen bellow, vega56 is running first 2 threads in miner, in hwinfo64 vega56 is the last card.
For Liquid Vegas I set a little bit higher voltage as they run xmr-stak with more intensity (to be sure stable).