Releases: bugorcka/bugorcka-miner
Release list
v0.2.1
v0.2.1 - BTX and NUMEN both got faster and more efficient
⚡ BTX (btxv4): energy efficiency up 1-11% depending on the card, across Turing, Ampere, Ada and Blackwell - same hashrate for less power, or raise the core clock for up to 1-11% more hashrate at the same power draw.
🆕 NUMEN (numen): speed up 16-19%, accuracy up from 78-84% to 96-100% depending on the card, income (confirmed shares) up 30-47%.
Reliability (both algorithms): fixed a rare share-submission timing issue that could occasionally cost a share.
Full per-card performance tables are in the README.
v0.2.0
Highlights
- 🆕 NUMEN ($NUMN) — new algorithm in the miner.
-a numen, pool ninjaraider
(TCP:44960/ SSL:44961), dev fee 5%. - ⚡ BTX (
btxv4) faster across the whole 2000–5000 series: +5–10% on every card.
Quick start and full docs: see the README.
New algorithm performance
NUMEN (numen)
| Architecture | Card | Power | Hashrate |
|---|---|---|---|
| Turing | RTX 2080 | 100 W | 300 KH/s |
| Ampere | RTX 3080 | 140 W | 660 KH/s |
| Ada | RTX 4060 Ti | 100 W | 550 KH/s |
| Blackwell | RTX 5070 Ti | 123 W | 1036 KH/s |
| Blackwell | RTX 5080 | 150 W | 1100 KH/s |
HiveOS: adding NUMEN
Flight Sheet → Miner → Custom, fill in:
-
Miner name:
bugorcka -
Installation URL — the
_hiveos_ub22.tar.gzasset from this release:https://github.com/bugorcka/bugorcka-miner/releases/download/v0.2.0/bugorcka-v0.2.0_hiveos_ub22.tar.gz -
Pool URL:
numen.ninjaraider.com:44960(addstratum+ssl://for:44961/TLS) -
Wallet and worker template:
%WAL%.%WORKER_NAME% -
Extra config arguments:
-a numen(append other flags on the same line, e.g.-a numen -i 18)
Or import the ready flight sheet as JSON (set your own SS58 address and worker):
{"name":"Ninjaraider_NUMEN_bugorcka","isFavorite":true,"items":[{"coin":"NUMN","pool_ssl":false,"dpool_ssl":false,"miner":"custom","miner_alt":"bugorcka","miner_config":{"url":"numen.ninjaraider.com:44960","miner":"bugorcka","template":"%WAL%.%WORKER_NAME%","install_url":"https://github.com/bugorcka/bugorcka-miner/releases/download/v0.2.0/bugorcka-v0.2.0_hiveos_ub22.tar.gz","user_config":"-a numen "},"pool_geo":[]}]}For SSL: set "pool_ssl":true and "url":"stratum+ssl://numen.ninjaraider.com:44961".
Hashrate and per-GPU stats show up natively in the Hive dashboard (same JSON stats API as BTX).
About NUMEN's ~75% accepted-candidate rate — not a bug
PoScan's reference hash is defined in double precision (asteroid mesh →
spectral features → quantized buckets → SHA256) — that's what the pool
checks against. We compute the hot path in float for speed: on
consumer GPUs the double-precision pipeline runs at 1/64 the rate of
float (Blackwell), so an honest double implementation would be roughly
20x slower (~35 KH/s instead of ~1 MH/s).
The cost is that some candidates (~25%) get filtered: accumulated float
rounding error occasionally pushes a quantized feature across a bucket
boundary, producing a different, invalid hash. This is a property of the
algorithm itself (a hash built from quantized geometric features), not a
defect on our side — any float-based miner for this coin hits the same
ceiling.
Every candidate is re-verified on the host against the exact
double-precision reference before submitting — filtering happens
locally and never produces a rejected share on the pool (confirmed live:
0 rejects). At ~75% accepted, float nets roughly 15x more accepted
shares than exact double at 100% — a deliberate, measured trade-off, not
a loss.
bugorcka v0.1.5
High-performance CUDA GPU miner for BTX (algo btxv4 / MatMul v4). Closed source, binary release.
GPUs: NVIDIA Turing and newer — RTX 20xx / 30xx / 40xx / 50xx (native RTX 50xx cubins, no PTX JIT). Pascal/Volta not supported (needs int8 tensor cores).
Pool: LuckyPool — https://btx.luckypool.io/
Address btx-eu.lproute.com:8666 — works in both SSL/TLS and plain TCP mode (PPLNS).
Dev fee: 2%. See LICENSE and THIRD-PARTY-NOTICES.txt.
What's new in v0.1.5
- GPU overclocking: per-card core/memory clock offset and power limit (
-lc,-lm,-oc,-om,-pl,-rlc), CSV syntax matches-ddevice order. -pl <N>%: set power limit as a percentage of the card's factory default (in addition to absolute watts).-fan <0-100|X>: manual per-card fan control.-tstop <C> / -tstart <C>: automatic thermal pause — mining pauses when a card hits the stop temperature and resumes once it cools to the start temperature. Off by default.--list-devices: print detected GPUs (index, PCI bus, name, VRAM, UUID) and exit — indices match-dand the OC/fan CSV positions.--log-level error|warn|info|debug: control console/log verbosity.
Downloads
| Asset | Platform |
|---|---|
bugorcka-v0.1.5_win64.zip |
Windows 10/11 x64 (contains bugorcka.exe, start.bat, README.txt) |
bugorcka-v0.1.5_linux_ubuntu22.tar.gz |
Ubuntu 22.04+ / glibc 2.35+ |
bugorcka-v0.1.5_hiveos_ub22.tar.gz |
HiveOS (Ubuntu 22 based images) |
Requirements: a recent NVIDIA driver (for RTX 50xx use the newest available). No CUDA Toolkit needed on the rig.
Quick start and HiveOS setup: see the README.
First start on a new card is slower (one-time GEMM auto-tune, ~1 min). Antivirus tools commonly flag GPU miners as PUA — add an exclusion if needed. Overclocking flags apply your own risk profile — start conservative and watch temps.
bugorcka v0.1.4
High-performance CUDA GPU miner for BTX (algo btxv4 / MatMul v4). Closed source, binary release.
GPUs: NVIDIA Turing and newer — RTX 20xx / 30xx / 40xx / 50xx (native RTX 50xx cubins, no PTX JIT). Pascal/Volta not supported (needs int8 tensor cores).
Pool: LuckyPool — https://btx.luckypool.io/
Address btx-eu.lproute.com:8666 — works in both SSL/TLS and plain TCP mode (PPLNS).
Dev fee: 2%. See LICENSE and THIRD-PARTY-NOTICES.txt.
Downloads
| Asset | Platform |
|---|---|
bugorcka-v0.1.4_win64.zip |
Windows 10/11 x64 (contains bugorcka.exe, start.bat, README.txt) |
bugorcka-v0.1.4_linux_ubuntu22.tar.gz |
Ubuntu 22.04+ / glibc 2.35+ |
bugorcka-v0.1.4_hiveos_ub22.tar.gz |
HiveOS (Ubuntu 22 based images) |
Requirements: a recent NVIDIA driver (for RTX 50xx use the newest available). No CUDA Toolkit needed on the rig.
Quick start and HiveOS setup: see the README.
First start on a new card is slower (one-time GEMM auto-tune, ~1 min). Antivirus tools commonly flag GPU miners as PUA — add an exclusion if needed.