T-Rex NVIDIA GPU miner (Ethash / Kawpow / Octopus / MTP)
T-Rex is a universal software for mining cryptocurrencies. It supports many algorithms, and we,
as developers, try to do everything possible to make it as fast and convenient to use as possible.
The developer’s fee is
To maximize the hashrate, our software uses all available GPU resources, so it is important that
you check your overclocking settings before starting mining. Our general recommendation is to start with the GPU stock settings
(there is no overclocking, the default power limit), and then, after making sure that it is stable, slowly increases its overclocking
to find where the miner works best and still does not fall.
Watchdog is designed to monitor the state of the miner and restart trexminer if it crashes or freezes for
any reason. In addition, watchdog can additionally perform automatic updates if a newer
version is available. We recommend using a watchdog to avoid downtime in mining and make sure that your
GPUs are busy 24/7. If you really need to disable the watchdog, you can do this with the parameter
- The double buffer function in 0.24.4 is not activated automatically unless –lhr-tune is specified
- The miner now generates a data set for the next ERGO block before it arrives with a small penalty for the reported hashrate, and when the next block does arrive, it immediately starts hashing without losing 1-4 seconds when creating the data set.
- (ethash, autolykos2) Add LHR unlock “auto-tune” functionality (it’s enabled by default, see –lhr-autotune-mode in readme)
- Now the miner will try increasing –lhr-tune if it is stable at the current value, so you’ll find out what the maximum LHR tune your cards can sustain without trying a lot of values manually.
- If you want the miner to stay at the LHR tune value you set yourself no matter what, use –lhr-autotune-mode off.
- Add ETH+ZIL mining mode with an arbitrary ETH pool (not proxying through a ZIL pool)
- See ETH+ZIL .bat/.sh file in miner archive
- New parameter –no-new-block-info to stop “new block” messages appearing in miner console
- (WebUI) Update interface to handle dual mining mode (algo switch button in the top right corner)
- Bug fixes:
- (kawpow) Hashrate degradation in version 0.24.0 on some configurations
- LHR unlock dual mining mode:
- You can now mine ETH (~30% of full speed) and other coins (~70%) simultaneously with LHR cards using their full potential.
- Available combinations along with memory requirements:
- ETH+ERGO (8GB+)
- ETH+RVN (8GB+ on linux, 10GB+ on windows)
- ETH+CFX (10GB+)
- See https://github.com/trexminer/T-Rex/wiki/LHR on how to set it up.
- WebUI is not updated yet to reflect second algo stats in dual mode, but will be in future.
- (ethash) “standard” LHR unlock changes and improvements:
- automatically detect LHR cards even if –lhr-tune is not specified
- new parameter –lhr-low-power reduces power consumption in LHR mode at a cost of a slightly lower hashrate (start from –lhr-tune 68)
- change of –lhr-tune scale compared to the version 0.23.2, now it indicates a percentage of the GPU’s full hashrate the miner tries to achieve. The old values in the [1-8] range are automatically mapped to the new scale, no configuration change required.
- recommended starting LHR tune values: 71 in normal mode, 68 in low power consumption mode (this doesn’t apply to dual mode)
octopus) CPU validation incorrectly detects invalid shares after epoch change
autolykos2) Miner sometimes crashes when
- (WebUI) Multiple fixes and improvements:
added compact mode with the chart hidden
fixed rounding of numbers along chart axes
fixed incorrect hashrate/power labels in pop-up hints
allow authorisation with existing session id (/trex?sid=xxx – for API users)
- NOTE: This version includes several critical changes: changing the GPU order and removing the telnet API. Before updating, make sure that you have removed the OC settings for your GPU order from the files .bat /.sh.
- Automatic updates will not be enabled for this version for the same reason.
- Users of HiveOS and other mining operating systems who forcibly order PCI bus GPUs are not affected by changing the GPU order.
- The new WebUI and API (see https://github.com/trexminer/T-Rex/wiki/WebUI about how to configure it):
- Ability to generate configuration files
- Improved API security: additional password and certificate-based authentication, https
- GPU/Memory clock and exchange difficulty reports
- Log Viewer
- The ability to suspend GPUs
- If you don’t like the new WebUI and would prefer to use the old one for some reason, download webui.zip archive, place it next to the t-rex executable file and restart the miner.
- Order GPUs by the PCI bus ID. You may need to change the settings that depended on the GPU order.
- Remove the telnet API
- Several bug fixes
- CFX mining is broken in 0.22.0
- The CPU share check is broken
- Switching to fault-tolerant pools takes too much time
- Stability fixes
- (Windows) Add optional min/max fan speed constraints in auto-fan mode (see –fan for more details)
- (API) Add timestamp of the last submitted share (last_submit_ts)
- (autolykos2) Improve compatibility with mining pools
- Command line parameter –worker does not take precedence over the one specified in the config file
- (API) Pausing the miner through HTTP POST is not working
- –fork-at functionality is broken
- Cosmetic UI fixes
- NOTE: Telnet API will be removed from the miner in the next release as superseded by HTTP API.
- (Linux) Add GPU power limit management support (see
--plparameter for details)
autolykos2) Duplicate share issue
- (Windows) Fans spin up to 100% in auto-fan mode if GPU temperature is well below the target temperature
- (Web UI)
watchdog-exit-modeparameter is missing in the config
- (autolykos2) Minor performance improvements (mainly Pascal GPUs)
- (Windows) Increase fan speed to 100% in auto-fan mode
- (autolykos2) Invalid shares when mining ERGO at NiceHash
- (mtp) ntime out of range and Low difficulty share errors
autolykos2) Low pool side hashrate caused by invalid shares
--lock-cclockparameter to lock GPU core clock speed
- (Windows) Add
--lock-cvparameter to lock GPU core voltage
- (Windows) Add
--pstateparameter to lock GPU performance/power state (currently
p0is the only supported value)
- (Windows) Allow passing in values greater than 100% to
--plif the GPU supports it
- Reduce miner startup time