GPU #0 COMPUTE ERROR

[CY4NiD3]

windows 11;
nvidia driver 560.81;
cuda 12.4;
xmrig 6.21.3;
xmrig-cuda-6.22.0-cuda12_4-win64 ;

I'm getting "GPU #0 COMPUTE ERROR" / "GPU #1 COMPUTE ERROR" using "--cuda-bfactor-hint=12".
If I lower "--cuda-bfactor-hint=12" to anything bellow 12 it stops, but the hashing speed then gets basically halved.
I've tried different nvidia drivers, different cuda versions, different xmrig cuda plugin releases, nothing worked.
I've also tried adjusting "--cuda-bsleep-hint=" which made no difference no matter the value assigned.
Here is the console output:

xmrig.exe --cpu-no-yield --huge-pages-jit --randomx-mode=fast --cuda --cuda-devices 0,1 --cuda-bfactor-hint=12 --cuda-bsleep-hint=100 --print-time=30 -a rx/0 --coin=XMR -o *pool -u *address

• ABOUT XMRig/6.21.3 MSVC/2022 (built for Windows x86-64, 64 bit)
• LIBS libuv/1.48.0 OpenSSL/3.0.13 hwloc/2.10.0
• HUGE PAGES permission granted
• 1GB PAGES unavailable
• DONATE 1%
• ASSEMBLY auto:intel
• POOL pool
• COMMANDS hashrate, pause, resume, results, connection
• OPENCL disabled
• CUDA 12.4/12.6/6.21.1
• NVML 12.560.81/560.81 press e for health report
• CUDA GPU #0 03:00.0 NVIDIA GeForce RTX 3080 Ti 1665/9501 MHz smx:80 arch:86 mem:11070/12287 MB
• CUDA GPU 1 01:00.0 NVIDIA GeForce RTX 2070 1740/7001 MHz smx:36 arch:75 mem:7144/8191 MB
  [2024-08-11 19:19:19.300] net use pool pool
  [2024-08-11 19:19:19.300] net new job from pool diff 36000 algo rx/0 height 3213088 (8 tx)
  [2024-08-11 19:19:19.301] cpu use argon2 implementation AVX2
  [2024-08-11 19:19:19.301] msr service WinRing0_1_2_0 already exists, but with a different service name
  [2024-08-11 19:19:19.565] msr register values for "intel" preset have been set successfully (264 ms)
  [2024-08-11 19:19:19.565] randomx init dataset algo rx/0 (16 threads) seed []...
  [2024-08-11 19:19:19.565] randomx allocated 2336 MB (2080+256) huge pages 100% 1168/1168 +JIT (0 ms)
  [2024-08-11 19:19:21.646] randomx dataset ready (2080 ms)
  [2024-08-11 19:19:21.646] cpu use profile rx (8 threads) scratchpad 2048 KB
  [2024-08-11 19:19:21.647] nvidia use profile rx (2 threads) scratchpad 2048 KB
  | # | GPU | BUS ID | INTENSITY | THREADS | BLOCKS | BF | BS | MEMORY | NAME
  | 0 | 0 | 03:00.0 | 4384 | 32 | 137 | 12 | 100 | 8768 | NVIDIA GeForce RTX 3080 Ti
  | 1 | 1 | 01:00.0 | 2304 | 32 | 72 | 12 | 100 | 4608 | NVIDIA GeForce RTX 2070
  [2024-08-11 19:19:21.668] cpu READY threads 8/8 (8) huge pages 100% 8/8 memory 16384 KB (21 ms)
  [2024-08-11 19:19:22.487] nvidia READY threads 2/2 (841 ms)
  [2024-08-11 19:19:25.391] cpu accepted (1/0) diff 36000 (189 ms)
  [2024-08-11 19:19:25.602] net new job from pool diff 36000 algo rx/0 height 3213088 (15 tx)
  [2024-08-11 19:19:26.301] cpu accepted (2/0) diff 36000 (189 ms)
  [2024-08-11 19:19:28.233] nvidia GPU #0 COMPUTE ERROR
  [2024-08-11 19:19:31.431] cpu accepted (3/0) diff 36000 (190 ms)
  [2024-08-11 19:19:36.921] cpu accepted (4/0) diff 36000 (190 ms)
  [2024-08-11 19:19:37.557] nvidia GPU 1 COMPUTE ERROR
  [2024-08-11 19:19:39.211] nvidia GPU 1 COMPUTE ERROR
  [2024-08-11 19:19:43.930] nvidia GPU #0 COMPUTE ERROR
  [2024-08-11 19:19:46.675] nvidia GPU 1 COMPUTE ERROR
  [2024-08-11 19:19:47.471] net new job from pool diff 36000 algo rx/0 height 3213088 (19 tx)
  [2024-08-11 19:19:47.819] cpu accepted (5/0) diff 36000 (190 ms)
  [2024-08-11 19:19:48.305] nvidia GPU 1 COMPUTE ERROR
  [2024-08-11 19:19:49.184] nvidia GPU #0 COMPUTE ERROR
  [2024-08-11 19:19:51.761] miner speed 10s/60s/15m 12397.3 n/a n/a H/s max 12863.9 H/s