scripts/MINE/xmr/container_mining_results.md

Mining Performance Tests in Containers

Test Results Summary

Original Performance

• Original xmrig performance: ~12,000 H/s

Container Performance Tests

1. ROCm Container (amd-strix-halo-llama-rocm)

• 28 threads: 14,084.2 H/s
• 32 threads: 14,728.4 H/s ⭐ BEST PERFORMANCE
• 31 threads: 14,233.6 H/s

2. AMD OpenCL Container (amdopencl)

• 28 threads: 14,020.0 H/s

3. Vulkan Containers

• amd-strix-halo-llama-vulkan-amdvlk: No OpenCL support
• amd-strix-halo-llama-vulkan-radv: No OpenCL support

Memory Allocation Tests

Current System Memory

• Total RAM: 30GB
• GPU VRAM allocated: 96GB (only ~1.6GB used)
• Available for CPU: ~21GB

Memory Pool Optimization Tests

• Default (no memory pool): 14,728.4 H/s ⭐ BEST
• --cpu-memory-pool=-1 (auto): 14,679.9 H/s
• --cpu-memory-pool=16: 14,059.0 H/s
• --cpu-memory-pool=32: 14,000.1 H/s
• --randomx-1gb-pages: 14,612.1 H/s (failed to allocate 1GB pages)

GPU Mining Attempts

TeamRedMiner (AMD GPU Miner)

• ROCm container: Failed - "Failed to list OpenCL platforms"
• AMD OpenCL container: Failed - "unknown device name: 'gfx1151'"
• Issue: RDNA 3 GPU (gfx1151) not supported by TeamRedMiner v0.10.21

lolMiner (GPU Miner)

• AMD OpenCL container: Detected GPU but "Unsupported device or driver version"
• Vulkan containers: No OpenCL support
• Issue: RDNA 3 GPU not supported by lolMiner v1.82

Optimal Configuration

Best performance achieved: 14,728.4 H/s (22.7% improvement over original)

Recommended setup:

• Container: amd-strix-halo-llama-rocm
• Threads: 32
• Memory pool: Default (no memory pool parameter)
• Command:

./xmrig -o pool.supportxmr.com:443 -u bc1qjn4m6rmrveuxhk02a5qhe4r6kdcsvvt3vhdn9j -p NUC --tls -t 32

Memory Allocation Analysis

Current GPU Memory Usage

• Allocated: 96GB VRAM
• Actually used: ~1.6GB
• Wasted: ~94GB

Impact on Mining Performance

• Reducing GPU memory allocation: Would free up significant RAM for CPU
• Current CPU memory usage: ~12-13GB during mining
• Potential improvement: Could potentially improve performance by 5-10% with more RAM available

Recommendations for Memory Optimization

1. BIOS/UEFI Settings: Check if GPU memory allocation can be reduced in BIOS
2. Kernel Parameters: Consider adding amdgpu.gttsize=32768 to reduce GPU memory
3. Current kernel params: Already has amdgpu.gttsize=131072 (128GB) - could be reduced

Notes

1. GPU mining not viable: Current GPU miners don't support the AMD Strix Halo RDNA 3 GPU (gfx1151)
2. CPU mining optimized: Container environment provides better performance than native
3. Thread optimization: 32 threads provides the best performance for this CPU
4. Memory usage: ~6.5GB RAM used for mining with 32 threads
5. Memory pool: Default settings work best - manual memory pool configuration reduces performance
6. GPU memory waste: 96GB allocated to GPU but only 1.6GB used - significant optimization opportunity

Future Considerations

• Monitor for GPU miner updates that support RDNA 3 GPUs
• Consider trying newer versions of miners when available
• The ROCm container provides the best environment for CPU mining
• Memory optimization priority: Reduce GPU memory allocation to free up RAM for CPU operations