lolMiner 0.9.6

lolMiner 0.9.6

Changes (0.9.6 to 0.9.6.1)

• Fixed a bug loading wrong kernel on Polaris / Vega cards for C31
  (Note: On Linux only Polaris cards were affected)

Changes (from 0.9.5)

• Significant improvement on GRIN-C29M performance (+6-7% on 580 and Vega cards, +10% on Navi)
• Significant improvement on GRIN-C31 and GRIN-C32 solver (~ +6% on all AMD cards)
• Added a 16G GRIN-C32 solver (Approx 20% faster on Radeon VII, Vega FE and 570 16G)

Note

Due to the lack of users I was forced to increase the fee on all Grin codes to 2%. This is a step back to the fee model I had before 0.7, i.e. I am taking 2% until the code payed off its development and lower it afterwards. This is kind of my last resort to keep the development of Grin codes for AMD alive.

Expected performances

GRIN-C29M (stock clocks, tested using amdgpu-pro 19.30)

• Radeon VII: 6.4 g/s
• Vega 64: 4.6 g/s
• Vega 56: 4.1 g/s
• RX 5700: 4.4 g/s
• RX 580 (8 G): 2.6 g/s
• RX 570 (8 G): 2.4 g/s

GRIN-C31 (stock clocks, tested using amdgpu-pro 19.30)

Note: Windows speeds on V56 / V64 / VII are ~7% lower due to a driver issue that prevents to use all optimizations

• Radeon VII: 2.05 g/s
• Vega 64: 1.37 g/s
• Vega 56: 1.15 g/s
• RX 5700: 1.12 g/s
• RX 570: 0.67 g/s

GRIN-C32 (stock clocks, tested using amdgpu-pro 19.30)

• Radeon VII: 0.66 g/s
• Vega FE: 0.5 g/s
• Vega 64: 0.42 g/s
• Vega 56: 0.36 g/s
• RX 5700: 0.34 g/s
• RX 570 (8 G): 0.23 g/s
• RX 570 (8 G): 0.18 g/s

lolMiner 0.9.5.1

lolMiner 0.9.5.1

Change to 0.9.5.1 (Hotfix)

• Fixed a potential crash issue in grin straum

Changes (From 0.9.3 (Win) / 0.9.4 (Linux))

• Added support for Grin CuckarooM-29 (hard fork on Jan 16th) on 8G AMD GPUs, use --coin GRIN-C29M to mine with it.
  Hint: The miner will pause until the fork height is reached not to waste energy. Use --coin FLOO-C29M to test the miner in floonet testnet before if you like.
• Improved C31 performance on 8G cards by ~5% (Windows: relatively to 0.9.3, Linux: relatively to 0.9.4)
• Navi on Windows now runs same kernels as in Linux
• Slightly lowered energy use of C31 / C32 solver
• Added aliases GRIN-C31 and GRIN-C32 for the Cuckatoo solvers (the old names still exist and continue to work)
• Driver requirement for all GRIN codes: 18.50 for all pre-Navi cards, 19.30 for Navi (Linux); Adrenaline 18.12 for all pre-Navi, Adrenaline 19.10 for Navi

Further news:

Added wiki article about C31 fade out and when changing to C32 makes sense:
https://github.com/Lolliedieb/lolMiner-releases/wiki/All-you-need-to-know-about-Grin-difficulty,-C29,-C31-fade-out-and-C32

Expected performance

GRIN-C29M (stock clocks, tested using amdgpu-pro 19.30)

Note: Windows speeds may differ slightly (max 3-4%)

• Radeon VII: 6.1 g/s
• Vega 64: 4.4 g/s
• Vega 56: 3.7 g/s
• RX 5700: 4.0 g/s
• RX 570: 2.2 g/s

GRIN-C31 (stock clocks, tested using amdgpu-pro 19.30)

Note: Windows speeds on V56 / V64 / VII are ~7% lower due to a driver issue that prevents to use all optimizations
Note #2: ROCm speeds often 5% quicker then in this table when available

• Radeon VII: 1.9 g/s
• Vega 64: 1.3 g/s
• Vega 56: 1.07 g/s
• RX 5700: 1.07 g/s
• RX 570: 0.63 g/s

GRIN-C32 (stock clocks, tested using amdgpu-pro 19.30)

• Radeon VII: 0.5 g/s
• Vega 64: 0.4 g/s
• Vega 56: 0.32 g/s
• RX 5700: 0.31 g/s
• RX 570: 0.17 g/s

lolMiner 0.9.4 (Currently Linux only)

lolMiner 0.9.4

Changes from 0.9.3

• New GRIN-AT31 performance code for Vega (+7%) and Navi (+12%). Requires amdgpu-pro 18.50 or newer or ROCm 2.10 driver
• Experimental support for Cuckatoo-32 (use --coin GRIN-AT32) on 8G AMD cards (see further notes)
• Windows release postponed due to incompatibilities with the new performance codes.

Expected performance (C31)

Radeon VII (Stock, 19.30): 1.91 g/s
Radeon VII (1330 / 975 / 775mV, ROCm 2.10): 1.55 g/s
Vega 64 (1550 / 1100 / 1000mV, ROCm 2.10): 1.45 g/s
Vega 64 (1350 / 1025 / 875mV, ROCm 2.10): 1.25 g/s
Vega 64 (1350 / 1025 / 875mV, 19.30): 1.22 g/s
Vega 56 (1270 / 900 / 862mV, 19.30): 1.1 g/s
RX 5700 (Power Color Red Dragon, Stock, 19.30): 0.99 g/s

Expected performance (C32)

Radeon VII (Stock): 0.5 g/s
RX 64 (Stock): 0.39 g/s
RX 56 (Stock): 0.35 g/s

About Cuckatoo-32 (Grin)

The primary grin proof of work Cuckatoo 31+ is designed in a way that the original instance - cuckatoo-31 - will fade out beginning mid January 2020, which means its difficulty will increase slowly over a period of 31 weeks until it gets impossible to mine a C31 block on the chain. But as by design also more difficult instances of Cuckatoo, namely Cuckatoo-32 ... Cuckatoo-63 are implemented on the chain that also can make blocks. The difficulties of this higher instances will remain stable, such that at some point it will be more profitable to mine Cuckatoo-32 instead of Cuckatoo-31 (likely from mid of February).

lolMiner 0.8.4 provides the first public implementation on a C32 GPU miner for 8G cards. The code is not yet tuned to the max and thus flagged experimental. When you want to pool mine with it make sure the pool supports Cuckatoo-32. Some pools (e.g. grinmint.com) - especially the pools using the reference pool software - support it on their C31 ports without mentioning it, other pools currently reject C32 shares. Solo mining with C32 will work as usual. Note that a share of C32 has ~2.064 times the weight of a C31 share and will thus rewarded higher / has a higher chance to mine a block by that factor.

Currently there are already two solo mined C32 blocks on the Grin blockchain using an early version of the code published with lolMiner 0.8.4. The two blocks have height 430112 and 430255.

Have fun experimenting with C32.

lolMiner 0.9.3

lolMiner_v093_Win64.zip

lolMiner 0.9.3

Changes from 0.9.2

• Extended GRIN-AT31 compatibility to older drivers (18.x +)
• Improved GRIN-AT31 performance on ROCm (RX 470/480/570/580/Vega/VII)
• Introduced early job cancellation for GRIN-AT31+ (improves hash on pool side, see further release notes)
• Deeply reworked kernel scheduler
• Fixed GRIN-AT31 kernel bugs (improving stability and fidelity)
• Fixed Bug: Vega FE loading 8G instead of 16G GRIN-AT31 solver in Windows
• Fixed Bug: Watchdog did not call right file in Windows
• Added --disablewatchdog 1 parameter to disable the 0 sol/s / 0 g/s detection

About Early Job Cancellation for GRIN

The GRIN proof of work is a compute intense and rather slow algorithm. Especially C31 often needs (at 1 g/s) about one second for each graph processed. In case a new job (for a new chain height) arrives, the current calculation effort is worthless, because the block that was mined on already got found.
lolMiner 0.9.3 now introduces an early cancellation for long lasting works: if a new work arrives that invalidates the current run calculations, they get canceled to start quicker on the fresh work. To demonstrate the effect see the example describing the 0.9.2 and 0.9.3 difference:

Assume you have a card running 1 g/s on GRIN-AT31. Then in 2 minutes (120 seconds) it will be able to process 120 graphs. On average - Grin has a 1 minute block time - two of these graphs are already obsolete when they got finished.

lolMiner 0.9.2 will display the 1 g/s, because of 120 processed graphs per 120 seconds. But two of the graphs are then filtered out in the stratum module, thus the pool can see at most 118 / 120 = 0.983 g/s (minus about 1 graph fee).

lolMiner 0.9.3 will cancel the running work before it gets completed. On average you will be able to process one more graph then 0.9.2 assuming both are canceled about half way through. Thus lolMiner 0.9.3 starts 121 graphs and completes 119 of them. The displayed hash rate is now 119 graphs /120 seconds = 0.991 g/s. This equals what the pool could see (minus one for fee), because the 119 completed are the ones that got not canceled early.

Conclusion: lolMiner 0.9.3 DISPLAYED hash rate may be a little lower then 0.9.2, but what arrives at pool is better. The slower the cards are, the more drastic is this effect, e.g. a 580 8G running at 0.65 g/s will benefit almost 2% on pool side, a VII with 1.75 g/s only half a percent.

Combined with the mostly on GPU cycle finder lolMiner achieves a stale share rate of less then 0.7% and thus gives you best displayed to pool hash rate ratio available for Grin Cuckatoo-31.

lolMiner 0.9.2

lolMiner 0.9.2

Changes from 0.9.1

• Significant performance improvement of GRIN-AT31 on 8 / 16G cards (+5% on Polaris & Vega, +10 on Navi)
• Experimental support for GRIN-AT31 and Polaris, Vega and VII using AMD ROCm drivers
• Added range checks to GRIN-AT31 code (improves better stability)
• Added function to call external watchdog scripts in case a GPU fails during mining (see release notes)

Usage of Watchdog Script

In case the miner detects no action of a GPU for at least a minute it will call the included scripts "reboot.sh" (Linux) or "reboot.bat" (Windows) and display a warning message in red. Afterwards the counters are reset. The scripts can be used to trigger a reboot of the rig or to call any other watchdog actions. The miner itself will take no further action and continue operation on the remaining cards.

Expected Performance:

• Radeon VII (stock): 1.79 g/s
• Radeon VII (1340/1000/0.825v, 130W): 1.37 g/s
• Vega FE (stock): ~1.3 g/s
• Vega 56 (1220/900/0.825v, 130W): 1.0 g/s
• RX 5700 (Red Dragon, stock): 0.91 g/s
• RX 580 8G: 0.56 g/s

lolMiner 0.9.1

lolMiner 0.9.1

Changes from 0.9

• Added GRIN-AT31 solver for 16G AMD cards (Better performance on Radeon Vega FE, Radeon VII and Sapphire RX 570 16G)
• Updated GRIN-AT31 solver for 4G AMD cards (Better performance on Fiji based GPUs, Polaris 10 4G)
• Fixed a bug causing too low pool hash on GRIN-AT31
• Added experimental GRIN-AT31 support for AMD Navi (8G), AMD Fiji (4G) and AMD Hawaii (4G / 8G) GPUs

Edit: support for Fiji was broken, A fix can be found here: https://github.com/Lolliedieb/lolMiner-preview/releases/tag/0.9.1hotfix

Release Notes

• GRIN-AT31 requires amdgpu-pro 19.30 driver or newer in Linux
• GRIN-AT31 requires AMD Adrenalin 19.10.1 WHQL driver or newer in Windows

Expected Performance (GRIN-AT31)git

• Radeon VII (16G): 1.7 g/s
• Vega FE (16G): 1.3 g/s
• Vega 64: 1.15 g/s
• Vega 56: 1.05 g/s
• RX 5700: ~0.85 g/s
• RX 580 8G: 0.55 g/s
• RX 580 4G: 0.43 g/s

lolMiner 0.9

lolMiner 0.9

Changes from 0.8.8

• Significant performance improvement for GRIN-AT31 on 8G AMD cards (+22% on Polaris to 30% on Radeon VII)
• Disabled 16G solver for GRIN-AT31 (the 8G is faster at the moment)
• Reduced Grin stale shares
• In command line lolMiner now accepts --pool address:port pattern
• Fixed a bug with the API crashing when accessed by Chrome based browsers
• Fixed a bug in EXCC stratum not passing number of submitted shares to the API

Release notes:

• The new codes were tested with amdgpu-pro 19.30 driver (Linux)

Known issues:

• Currently AMD Navi support is disabled for the new algorithms.
• Cuckatoo-31 on ROCm falls back to 4G kernel.

Expected performance on Cuckatoo-31 (stock settings)

• Sapphire RX 580 Pulse 0.55 g/s
• Radeon Vega 64: 1.2 g/s
• Power Color Radeon VII 1.53 g/s

lolMiner 0.8.8

lolMiner 0.8.8

Changes from 0.8.7

• Removed BEAM-I (Beam Hash I), since the fork had happened, BEAM-II and BEAM now both point to the new PoW
• Improved driver detection mechanisms and kernel selection
  -> On AMDs: The invalid work group size bug should be gone for Beam*
  -> On Nvidias: Performance of Beam Hash II should be improved (since it now uses the right kernels)
• Added new simple .bat (Windows) and .sh (Linux) files for one liner configuration.
• Beta: Equihash 96/5 (--coin MNX) should now allow to change the personalization string (--overwritepersonal=newPers)

(*) Note: If you see now BeamHash II running in a compatibility kernel on Polaris or Vega cards please report me your driver version (it was tested with 19.20 drivers) - older drivers may in future be unsupported for newer kernels.

lolMiner 0.8.7

lolMiner 0.8.7

Changes from 0.8.7

• Added support for Beam Hash II on older AMD cards (R 9 200 / 300 4 & 8G cards; R9 280(X), HD 79x0 with 3G). Note that the auto-switcher is only working with at least 4G, for the 3G cards please use --coin BEAM-II manually on fork height.
• Fixed an issue with Genesis Network (GENX) missing in lolMiner 0.8.x

lolMiner 0.8.6

lolMiner 0.8.6

Changes from 0.8.6

• Added support the BEAM hard fork on block 321321 (approx Aug15th), read below about the usage
• Fixed a bug with the BEAM stratum back end in case of formatted job descriptions (e.g. new Nicehash platform)

The BEAM hard fork

There are now 3 parameters for mining BEAM, --coin BEAM-I, --coin BEAM-II and --coin BEAM

• Use BEAM-I for mining the current / old BEAM proof of work that is valid until the hard fork.
• Use BEAM-II for mining the new BEAM proof of work that is valid from the hard fork.
• Use BEAM for an automatic switcher between the two PoWs. This mode will detect the height and the fork height and will automatically switch over on the right block. Note that this requires the pool to send at least the height. On startup the miner will detect if the pool is compatible to the automatic switcher and display a message if all is fine.
  The message is colored in command line: green - in this case the new PoW is guarantied to be supported; yellow - the height is send so the miner can switch over, but its uncertain if the new PoW is implemented; red - the pool does not send the required information.
  In the last case the miner will show the message and then quit after 10 seconds. In this case select the right PoW manually with BEAM-I or BEAM-II. When at least the height is send, the miner will start mining in automatic mode.

List of supported pools (status: Wed Aug 14th, 12:30 UTC)

• 2miners (green)
• antpool (green)
• coinmine.pl (green)
• btc.com (green)
• leafpool (green)
• f2pool (green)
• sparkpool (green)
• sunpool (green)
• suprnova (green)