v3.43 (Windows): Out of memory for Equihash125_4 with RX580 or RX5700

[UselessGuru]

GMiner-v3.43\miner.exe --nvml 1 --algo equihash125_4 --cuda 0 --opencl 1 --server zelhash.auto.nicehash.com:9200 --user *************************************.******************** --pass x --api 4001 --watchdog 0 --devices 1

+----------------------------------------------------------------+
|                          GMiner v3.43                          |
+----------------------------------------------------------------+
Algorithm:          Equihash 125,4 "ZelProof"
DevFee:             2 %
Server:
  host:             zelhash.auto.nicehash.com:9200
  user:             ***************************************.******************
  password:         x
Power calculator:   on
Color output:       on
Watchdog:           off
API:                http://127.0.0.1:4001
Log to file:        off
Selected devices:   GPU1
Intensity:          100
Temperature limits: 90/120
------------------------------------------------------------------
10:22:11 AMD Driver: 23.4.3
10:22:11 Failed to initialize miner on GPU1: GIGABYTE AMD Radeon RX 5700 XT 8GB [0000:09:00.0]: Out of memory
10:22:11 No devices for mining

[PrinceOfParallax]

I also get this on a 6800 (16GB) on first boot it seems. If I close and reopen it works.

[Sibonet]

Getting the same here. Equihash 144,5 Rig with 5 GPU (5xRX480 8GB)
OS Windows 10, Drivers 21.10.2 Gminer
Virtual memory: 154,5 GB (Usually it is 138 GB but it is increased due to issues)
Command line: miner.exe --logfile log.txt -d 0 1 2 3 4 -a 144_5 -s equihash144.na.mine.zpool.ca -n 2144 -u cccccccccccccccccccccccccccccccccccccccccccc -p c=BTC --pers auto -w 0 --api 4028

Which card does work and which doesn't is random, sometimes it is only one not working, and sometimes it is only one working or any combination. It doesn't depend on anything I can grasp.

The screenshot below depicts randomises on 3 working and 3 non-working cards