 |
May 06, 2026, 07:12:53 PM |
|
There have been many updates in the last several months for TNN Miner.
With recent priorities being more impactful, though, I've decided to resume updating this forum topic. Notables are GPU support, improved hiveOS wrappers, hashrate uplifts, and the addition of multiple algorithms.
Below is the full changelog list for releases that have come out since the last post to this topic was made
TNN Miner v0.8.2
-----------------------------------
+Improved Xelis v3 CPU performance, especially on Zen4/Zen5 CPUs
+Fixed Xelis v3 SIMD selection so the fastest available paths are used
+Reduced Xelis live mining overhead and hashrate variance
+Repaired NVIDIA GPU mining on Linux
+Fixed NVIDIA CUDA/NVRTC runtime loading and bundling issues
+Fixed NVIDIA KawPow kernel compilation and live mining issues
+Improved Linux GPU package portability with bundled/patched runtime libraries
+Improved Linux huge page reporting and 1GB huge page allocation behavior
+Added general KawPow stability and performance improvements
TNN Miner v0.8.0
-----------------------------------
+Improved Xelis v3 CPU hashrate by roughly 10%
+Removed per-hash bignum overhead from Xelis difficulty checks
+Added BMI2 gating for AVX2/AVX512 Xelis tiers
+Applied faster raw byte difficulty checks to Dero/AstroBWT
+Cleaned up unused Xelis v3 GPU kernel code
+Added initial KawPow support
+Fixed --test-xelis in GPU builds
TNN Miner v0.7.8
-----------------------------------
+Improved GPU hashrate on AMD and NVIDIA
+Added multi-strategy GPU kernel autotune
+Fixed rejects after fresh autotune
+Bundled required Windows runtime DLLs for GPU builds
+Added HIPRTC preflight checks with platform-aware errors
+Added Orochi support for HIP 6.x and 7.x runtimes
+Added ROCm 6.0-6.3.3 shims for older drivers
+Added GPU hwmon backend
+Added debug build flag
TNN Miner v0.7.7
-----------------------------------
+Significant Xelis v3 GPU hashrate improvements
+Improved GPU autotuner with memory bandwidth cliff detection
+Reduced Xelis GPU VRAM reserve from 128MB to 32MB
+Drastically reduced CPU usage during GPU mining
+Reclaimed roughly 500-800MB VRAM previously wasted during compilation
+Added NVIDIA Linux disk cache for compiled kernels
+Added automatic retune when cached settings become stale
+Improved low-VRAM handling when loading cached tunes
TNN Miner v0.7.6
-----------------------------------
+Added unified AMD + NVIDIA GPU build via Orochi
+Improved Xelis v3 GPU performance with new kernel math and autotune changes
+Added parallel GPU initialization and autotune
+Fixed HiveOS/mmpOS GPU wrapper support
+Bundled AMD ROCm and NVIDIA CUDA runtime libraries for wrappers
+Added default tunes for RTX A2000, RX 7900 XTX, and CMP 170HX
+Improved RandomX JIT behavior
+Improved Xelis v3 CPU SIMD dispatch
+Reduced AstroBWT startup memory overhead for non-Dero users
TNN Miner v0.7.5b
-----------------------------------
+Added runtime GPU overclocking flags
+Added per-device GPU core, memory, offset, and power limit controls
+Added real-time GPU wattage reporting
+Added hashes-per-watt efficiency reporting
+Added ADL, ROCm SMI, and NVML power monitoring backends
+Improved multi-GPU tune cache behavior on mixed/free-VRAM rigs
+Bundled default tune files in Linux, Windows, macOS, Docker, HiveOS, and mmpOS builds
TNN Miner v0.7.4b
-----------------------------------
+Added per-device share tracking
+Added per-device accepted/rejected counters to broadcast API
+Improved mmpOS endpoint with per-device shares
+Improved HiveOS stats reporting for GPU wrappers
+Fixed HiveOS/mmpOS wrapper pool URL handling
+Fixed GPU wrapper folder structure for HiveOS extraction
TNN Miner v0.7.3b
-----------------------------------
+Fixed NVIDIA Windows startup crash when connecting to pools
+Improved NVIDIA autotune error reporting
+Bundled required OpenSSL DLLs with NVIDIA Windows builds
+Removed boost::thread dependency in favor of std::thread
+Fixed AMD and NVIDIA Windows dependency cache separation
TNN Miner v0.7.2b
-----------------------------------
+Added LuckyPepe / YescryptR32 support
+Fixed yespower dev pool ports
+Added v0.5 yescrypt support through --yespower ver parameter
+Added --test-yespower with known-vector validation
+Added HiveOS/mmpOS launch shim
+Fixed NVIDIA Xelis v3 register strategy by architecture
TNN Miner v0.7.1b
-----------------------------------
+Added Xelis v3 startup auto-tune
+Added Xelis v3 MSR profile
+Added --xelis-simd override
+Improved Xelis v3 AES/SIMD dispatch handling
+Improved Xelis v3 memory prefetch strategy
+Added built-in presets for multiple yespower/yescrypt variants
+Added --priority and --no-msr flags
+Moved Xelis benchmark to --bench-xelis
+Finished initial Xelis v3 GPU stratum submission pipeline
+Re-enabled LTO for CPU builds
TNN Miner v0.7.0
-----------------------------------
+Upgraded default Linux/devops build to Clang 20
+Fixed Xelis v3 divide-by-zero edge case
+Improved Xelis v3 CPU performance
TNN Miner v0.6.8
-----------------------------------
+Improved Xelis Hash v3 division math
+Improved Xelis Hash v3 hashrate across supported CPUs
TNN Miner v0.6.7
-----------------------------------
+Added modest Xelis Hash v3 CPU performance improvements
TNN Miner v0.6.6
-----------------------------------
+Fixed wallet autodetection and versioned algo routing
+Fixed Xelis v3 Stratum routing
+Improved Xelis v3 Stratum connection behavior
+Added guidance for --xel, --xel-v3, and --ignore-wallet usage
TNN Miner v0.6.5
-----------------------------------
+Moved Linux builds to Clang 20
+Fixed Xelis Hash v3 Stratum bug
TNN Miner v0.6.1
-----------------------------------
+Fixed Stratum watchdog timeout issue after reconnects
+Cleared job cache on new Stratum sessions
+Prevented immediate timeout behavior from stale job receipt logs
TNN Miner v0.6.0
-----------------------------------
+Major netcode refactor for improved long-session stability
+Fixed dangling pointer issues that could cause slowdown or segfaults
+Reworked submission threading to use queue-based logic
+Improved Xelis Hash v2 performance and consistency
+Added Xelis Hash v3 implementation
+Added versioned algorithm scaffolding
+Improved RandomX initialization feedback and stability
|
