lolMiner 1.69

Changes

• Added Nexa support for AMD Vega / VII (1) and RDNA 1-3 GPUs (2)
• The parameter --keepfree can now be used to keep a certain amount of memory free on each GPU on Nexa mining. This might end up in allocating a smaller lookup table, so the GPU memory has space left for other workloads (3)
• The --keepfree parameter is now working for each GPU separately. Use a comma separated list of values if you want to assign different values for each GPU.
• Nexa mining will now allocate memory for each GPU sequentially and delay later cards startup slightly to help systems with small virtual memory.

Notes

(1) AMD Vega and VII are currently only supported on GPU drivers running the Rocm compute OpenCL driver stack. This is e.g. default for Vega GPUs on the latest Linux drivers, but not on pre 2021 drivers. Support for older drivers (and maybe older GPUs) might be included with the upcoming releases.
(2) Attention: we have seen cases where one or multiple RX 6000 GPUs have been mining significantly slower then identical cards on the same rig. This happens only on older 21.x or 20.40 Linux drivers. We recommend to update to 21.50 or 22.20 to fix the issue.
(3) For users of the alternative Windows version of 1.68: Use --keepfree 1024 or higher to get the effect the alternative version had.

General note: The Nexa gpu codes are (and will be in foreseeable future) joint work with Iedoc from BzMiner. Miner fees will be evenly shared regardless of which of these two flavors of the code you prefer.

lolMiner 1.68

Changes

• Significantly improved Nexa mining performance on supported GPUs (e.g. up to +25% on 8G Ampere GPUs)

Fixes

• Fixed a bug that can make the miner crash with error message "Authorization problem on all configured pools 1" during Nexa mining

Note

The alternative Windows version uses a different mechanism to determine the available memory on a Windows system. On most 8G GPUs this will cause the miner to select a smaller dataset table that has only slightly above 2 G. Therefore the alternative version might be a good choice for systems with connected screens or for Nexa / Zil mining using the Zil switching app.

General note: The Nexa gpu codes are (and will be in foreseeable future) joint work with Iedoc from BzMiner. Miner fees will be evenly shared regardless of which of these two flavors of the code you prefer.

Windows note: Try lolMiner_v1.68_alternative_Win64 ONLY if you get the message GPU ?: error in allocation memory 0 .

lolMiner 1.67

Changes

• Improved Nexa performance on Nvidia Turing based gpus by approximately 35%+.
• Improved Nexa performance on 8G Nvidia Ampere gpus by approximately 1.5% when memory is not locked and ~3% on locked memory (5000). 10G and higher gpus got an additional 5% performance increase.
• Improved Nexa performance on Nvidia Ada based gpus by approximately 25% on locked memory clock (5000) and about 8% on unlocked memory.
• Added Nexa echelon mining protocol to support pools using it, e.g. 137pool.io. The needed format will be automatically detected when connecting to a pool using it.
• Updated WebUI for Nexa mining and more pools to support.
• Windows: Updated GUI

Fixes

• Fixed a bug causing scattered defect shared on Nexa mining (all OS)
• Fixed a display bug when setting power limit on Nvidia GPUs saying the value is out of Range, although its fine. (all OS)
• Fixed a bug in Nexa mining of the miner creating only defects (Windows)
• Fixed a bug in Nexa mining of the miner crashing silently after a few minutes (Windows)

Note 1: Every code update - in particular when so massive as here - may require re-tuning your oc & uv settings for ideal results and stability.
Note 2: memory locking on Ada gpus is only recommended if the core clock is locked as well and not maxed out - on high core clock in combination with memory locked to 5000 the performance will struggle due to worse memory timings.

General note: The Nexa gpu codes are (and will be in foreseeable future) joint work with Iedoc from BzMiner. Miner fees will be evenly shared regardless of which of these two flavors of the code you prefer.

lolMiner 1.66

Changes

• Added support for mining Nexa on Nvidia Pascal or newer generation GPUs (1). Use --algo NEXA to mine it. Fee is 2%.
  Note: The pool protocol supported matches the one introduced by rplant pool. That said pools tested by this version are in alphabetical order: acc-pool.pw, rplant.xyz, vipor.net and woolypooly.com. That said this list is not exclusive and every other pool following this protocol should work.
  Note 2: This code is joint work with Iedoc from BzMiner and it is a refined version with approximately 20% higher speed then the code released in BzMiner v13.0.1. An updated version of BzMiner featuring the improvements of this kernel as well as the same fee level will be released soon.

(1) Primary focus in optimization were the low to medium range Nvidia Ampere GPUs. The performance of other generations may vary, but certainly improve the next versions.

lolMiner 1.65

Changes

• Improved Kaspa performance in ETC and ETHW dual mining by about 6-7% on AMD (Big) Navi GPUs and 10-12% on Nvidia Turing and Ampere, measured at the same Ethash speed. The actual improvement depends on the concretely selected dual factor. Re-tuning your setup is highly recommended.
• Improved Kaspa performance in Kaspa only mining mode by 0.6 to 1.2% on Nvidia GPUs and 0.4-0.5% on AMD (Big) Navi GPUs at approximately same power draw.
• Beta feature: Added experimental support for AMD RX7900 series. Supported algorithms: Et(c)hash, kHeavyHash and the corresponding dual mining.

Fixes

• Fixed some minor display bugs

lolMiner 1.65a

Fixes

• Fixed a bug causing sometimes high number of duplicate shares on Kaspa when dual mining with ETC.
• Changed kaspa stratum behavior: Kaspa mining will now start only after the pool did send an authorization message. This prevents rejected shares at startup.
• Slightly changed dual mining tuning behavior to a mode a bit slower, but less likely to hang up the GPUs at startup.

lolMiner 1.64

Changes

• Improved Kaspa only mining performance. Speed increase is about 8-8.5% on Nvidia Pascal GPUs, 4.5-5% on Nvidia Turing and Ampere GPUs and 3-4% on AMD Navi and Big Navi GPUs
• Beta feature: added options to set core clock offset (--coff), memory clock offset (--moff), power limit (--pl) and a fixed fan speed (--fan) on common Nvidia GPUs. Required are admin privileges and Nvidia drivers 520 or higher!
  The syntax is the same as with --cclk and --mclk - if a single value is given then it will be applied to all compatible GPUs, else a coma separated list of values can be given using a * character to skip over GPUs. (1)
• Added a new parameter --no-oc-reset to turn off the reset of overclock settings when ending the miner.
• Windows: Added a beta gui to generate overclock settings strings / .bat files for the miner. Also the tool can apply the chosen settings directly.

(1) Note: No responsibility taken for the values set. Please use with care. If your mining os had build in functions to set these settings we recommend using them instead of the miner settings.

Fixes

• Fixing a bug with ETHV1 (nicehash) stratum mode that may cause the worker name to be appended twice when it was given by --user . (the use of --worker did not have this issue).

lolMiner 1.63

Changes

• Improved the Kaspa only mining performance on Nvidia Turing, Ampere and Ada GPUs by about 3.5%. (1)
• Significantly improved the Kaspa only mining energy efficiency on Nvidia Turing, Ampere and Ada GPUs by 7-11% depending on the actual model.
• Values given to --dualfactor parameter will now be value checked and rounded / capped to working values.

(1) Can be higher in case the card was power limited before.

Fixes

• Fixed a bug causing the miner to show a crash message when lolMiner was ended via ctrl+c
• Fixed a bug causing lolMiner not to start Kaspa mining on Nvidia GPUs when the Nvidia OpenCL installation on the system is broken.

lolMiner 1.62

Changes

• Added basic support for Intel Arc Desktop GPUs on the following algorithms: ETC, ETC+KAS (1), Kaspa, Flux, Ergo, Beam (2) & Equihash 144/5 (2). Fees are equal to Nvidia and AMD GPUs on the corresponding algorithms. Tested Intel OpenCL driver versions are 22.24.23453 and 22.32.23937.
• Added support for mining Alephium in non-dual mode for Nvidia Pascal GPUs and newer. Use --algo ALEPH to mine it. Fee is 0.75%.
• Slight performance improvement for Kaspa non-dual on Nvidia GPUs.
• Added Aeternity Cuckoo 29 & Grin Cuckatoo 32 kernels for RX 6600 and RX 6700 series GPUs.
• Updated web ui for supporting Intel GPUs, more pools and coins.
• --cclk for locking core clock and --mclk for locking memory clock now work for Nvidia Pascal and Turing GPUs (3).
• If --cclk or --mclk are used to lock clocks, these will now be reset on lolMiner exit.
• Added reading of junction temperature and memory temperature for Nvidia GPUs for drivers 515 or newer.
• Re-arranged nonce consumption for Kaspa to allow pools with 3 bytes extra-nonce without (too many) duplicate shares.
• Linux: Added reading of GPU power draw and core clock on Intel Arc GPUs.
• Windows: updated example .bat files to cover most needed cases and error the previous files had.

(1) Only on the 8G+ Intel GPUs: A580, A750 and A770.
(2) Beam and Equihash are experimental on Intel GPUs. Also see known issues when running multiple cards.
(3) Needs admin / sudo privileges. Tested 460.93 driver and higher. --mclk should only be used to mine core intense coins like Kaspa or Aleph in order to reduce power draw by the memory. Recommended value in this case: 810.

Fixes

• Fixed a bug that may cause a segmentation fault on startup.
• Fixed a bug causing LHR detection to sometimes start on Nvidia 520 and newer drivers
• Fixed a bug causing --list-devices not to work.
• Fixed a bug causing DNS over HTTPs to fail often.
• Fixed a bug causing Intel GPUs not to be sorted by PCIE bus and address in device list.
• Fixed a bug causing rejected shares on Equihash not appearing in API (but they did in miner dashboard stats).
• Fixed a bug in web ui to sometimes not refreshing.

Known issues

• When running multiple Intel Arc GPUs on Beam or Equihash 144/5 the cards will slow down. This is likely due to a limitation of the Intel OpenCL implementation and the way lolMiner works with it. We hope to get this fixed in one of the upcoming versions.
• Windows: On Nvidia driver 526.47 mining Beam Hash III is broken. We recommend using a driver from the 522 series at the moment.

lolMiner 1.61

Windows version re-uploaded without changes on 4th of November. Reason: new upload is self signed with the certificate make public in this github.

Changes

• Slightly improved performance of Flux mining on Nvidia Ampere and Turing cards (by approx 1.5-2.5% compared to 1.60)
• Improved system memory footprint when mining Flux or Beam on Nvidia GPUs. This allows running the solvers on more then 8 GPUs with only 4G system memory.
• Speed exposed by API is now smoothed, so especially on all Equihash based algorithms the hashrate figures derived from API values will appear more smooth

Fixes

• Fixed a bug causing Flux mining on Nvidia Pascal GPUs not working
• Fixed a bug causing stale shares being counted as defect shares in API - and not correctly exposed in miner UI as well.