Re: [ANN] lolMiner-mnx, Mining MinexCoin(MNX) with GPUs on Windows & Linux
So, as promised it is time for an update:
Here are the binaries for version beta v 0.11:
Linux 64: https://drive.google.com/open?id=1l7TXYEfXbdGgkpQbIQAz9JGaV-5vYVdF
Windows 64: https://drive.google.com/open?id=1remeVpaiVhqKkKZyo966xiVRsP_4OfeZ
Changes:
Added a new command line parameter "--enable-nv-hotfix".
The parameter will reduce the required CPU utilization on Nvidia graphic cards. If you required 100 + x% CPU utilization (so one full core and parts of a second) it is likely that now you will need 10-15% + x. I can not get rid of the x because this is for checking if the solution is below the target. On my 2.1 ghz CPU with a GTX 1080 attached CPU utilization went down from 115% to ~20%. This will also help running multiple instances for large rigs. Unfortunately one will loose about 1-2% of mining speed on Nvidia hardware when enabling the switch. Therefore - and because the loss may be even larger for other vendors - the hotfix has to be enabled by command line for each instance.
Added an automatic reconnect when the connection is broken.
When a share is submitted and there is no acceptance message by the pool for the next 10 seconds the miner will reconnect to the pool in order to get back a stable connection. This helps the miner not to hang.
By the way: I think I found one glitch in the windows version causing the crash that happens after a while. This seem to happen when the pool sends a new job but not with enough parameters (jobId, time, merkle_root....).
The fall back method seems to work fine on Linux, but causes the crash on Windows. Therefore I would recommend testing different pools as long as this is not fixed, because the rate is not equal for all pools. I will try to get this fixed in next version.
By the way beside bugfixing the next features I would like to implement are (in order of their priority):
- The opportunity to run it on multiple GPUs with only one stratum instance.
- A handbrake function for the miner. More concretely a switch that lowers the hashrate but such that the graphic card is not busy 100% over the time. May help for rigs with temperature problems or overambitious graphic drivers that kill tasks that cause too much GPU load.
- An algorithm toggle. The algorithm used can be configured to create different memory pressure on the card. Changing from default behavior may improve the speed graphic cards will less compute cores per shader processor as NV Tesla P100 or AMD Vega series.
I hope that I can make it implementing at least one feature and some of the left known bugs till next week
Ps:
Current hashrate of all lolMiners out there: about 820 ksol/s (1,5% market share)
Here are the binaries for version beta v 0.11:
Linux 64: https://drive.google.com/open?id=1l7TXYEfXbdGgkpQbIQAz9JGaV-5vYVdF
Windows 64: https://drive.google.com/open?id=1remeVpaiVhqKkKZyo966xiVRsP_4OfeZ
Changes:
Added a new command line parameter "--enable-nv-hotfix".
The parameter will reduce the required CPU utilization on Nvidia graphic cards. If you required 100 + x% CPU utilization (so one full core and parts of a second) it is likely that now you will need 10-15% + x. I can not get rid of the x because this is for checking if the solution is below the target. On my 2.1 ghz CPU with a GTX 1080 attached CPU utilization went down from 115% to ~20%. This will also help running multiple instances for large rigs. Unfortunately one will loose about 1-2% of mining speed on Nvidia hardware when enabling the switch. Therefore - and because the loss may be even larger for other vendors - the hotfix has to be enabled by command line for each instance.
Added an automatic reconnect when the connection is broken.
When a share is submitted and there is no acceptance message by the pool for the next 10 seconds the miner will reconnect to the pool in order to get back a stable connection. This helps the miner not to hang.
By the way: I think I found one glitch in the windows version causing the crash that happens after a while. This seem to happen when the pool sends a new job but not with enough parameters (jobId, time, merkle_root....).
The fall back method seems to work fine on Linux, but causes the crash on Windows. Therefore I would recommend testing different pools as long as this is not fixed, because the rate is not equal for all pools. I will try to get this fixed in next version.
By the way beside bugfixing the next features I would like to implement are (in order of their priority):
- The opportunity to run it on multiple GPUs with only one stratum instance.
- A handbrake function for the miner. More concretely a switch that lowers the hashrate but such that the graphic card is not busy 100% over the time. May help for rigs with temperature problems or overambitious graphic drivers that kill tasks that cause too much GPU load.
- An algorithm toggle. The algorithm used can be configured to create different memory pressure on the card. Changing from default behavior may improve the speed graphic cards will less compute cores per shader processor as NV Tesla P100 or AMD Vega series.
I hope that I can make it implementing at least one feature and some of the left known bugs till next week
Ps:
Current hashrate of all lolMiners out there: about 820 ksol/s (1,5% market share)