This should be fixed in 1.11

Dear community,

we have gathered here today to regret the imminent end of 4G cards on Ethash.
Latest on DAG epoch 381 all of this precious GPUs will stop mining ETH.

Will they?

No!

Today there is a new release of lolMiner that will allow you to keep going on!

Read this medium article to learn more and how to configure it:
https://lolliedieb.medium.com/rise-of-the-mining-dead-ea8f6298f813

Have fun mining!

Dear community,

we have gathered here today to regret the imminent end of 4G cards on Ethash.
Latest on DAG epoch 381 all of this precious GPUs will stop mining ETH.

Will they?

No!

lolMiner made a release today that will allow to continue mining ETH with a slight time-memory trade of for some more weeks on top!
It even allows you to go on with ETC on Windows today and until they will swap to their new DAG scheme End of November.

Read this medium article to learn more and how to configure it:
https://lolliedieb.medium.com/rise-of-the-mining-dead-ea8f6298f813

Have fun mining!

Thanks, I reported that guy.

Well ... Phoenix and Claymore almost use the same kernel (I would bet one could swap them without causing incompatibility) and if you trace the number of nonces processed you will see they are a bit over-reporting.
But you can also check that with the tool teamredminer devs created without any reverse engineering - that tool is open source and simulates a low difficulty pool, so you see a high number of submitted shares and compare that with what the dashboard proposed. If you do so you will see Phoenix reports about 2% higher then it should (pool is 2.8% lower then dashboard ... some lower is expected due to fee ofc, but not 2%+) ... so well. It is easy to proof they cheat to look better in the miner output then they actually are.

That said: lolMiner on ethash is way from being perfect yet. But I just started and many cool things yet to come. Next release in already 1-2 days approx adding some more stability  

If 512 gb/s is true even with perfect miner we are bound to ~60MH/s
Compared to the increase in CU that somehow feels extremely crippled compared with 5700. Anf 6700 even with less bandwidth then 5700?

In this situation even the size of 16G will not help.

Good catch... The 380 are Tonga chips ... hmmm ... missed them somehow. Which algos would you like to see on them? Can try to build my kernels for them as well





Are the OpenCL drivers installed for your NVidia cards? Often they come with the Nvidia toolkit package. Make sure clinfo (Linux: in 99% of all repositories, Windows: Ships with the AMD driver) lists both types of cards - then it should work in the miner as well
Good question, I do not know from distance. What kind of settings did you apply? I personally only have a 5700 but that is mining Beam stable for days.

Ah ... yea I have to admit I did not test that case xD It sets it to "AUTO" internally for pool switching ... but yes over CMD that does not work. Will fix it with 1.08 - use --coin in the meanwhile for this function.

Not sure here, too what happens in AM, but there are two ways of using this algos.
Either --algo or --coin. If you still can use --coin you can use --coin ZEL to configure ZelHash + right personalization and --coin AUTO144_5 for --algo  EQUI144_5 with --pers AUTO (aka pool selection).
If --pers is not passed forwards that seems to be an issue with awesome miner / a bit too tight template / not giving the opportunity to add custom parameters outside of their template. If so that will be interesting for the next algos I am adding, because they will add quite a lot custom params  Not mentioning yet which it will be 

Jumping in here.

Just to clarify the curiosity about lolMiner configuration:

Pre 1.0 lolMiner had only one parameter to configure the algorithm - this was --coin and did all the needed stuff in one command: selecting right algo + right personalization plus right stratum.
Issue some users of mine had was when a new coin was added with the same algo or one project changed the personalization parameter. Then the shortcut was not right.

Therefore lolMiner 1.0 and newer added a new parameter --algo (or short -a) that can select the algorithm, while --pers takes the personalization for Equihash based algos. So one old parameter was replaced by two new, but its now cleaner separated and when new projects with same PoW come along they should work out of the box.
--coin is still present in lolMiner 1.0 and up, but only for special functions like algorithm switchers and some often used presets. The standard way for future use and additions should be using --algo and maybe - for Equihash - --pers. It might be that some algos in future will be added that need a 3rd parameter to select the right stratum, but that did not happen yet.

I have the feeling that --algo + --pers is much closer to what other miners do and so hope using this pattern clears most of the configuration confusion. 

Yes indeed - if you got a 8G card in the mix always put it to the first slot and do graphics out with that if needed. This is general good advice also for other miners & algos

Hmm ... does the GPU need to do graphics in parallel? If you are on Windows can you check if the system reports memory being flagged as "shared"?
Like e.g. here:


In case that "Shared GPU memory" is greater then 0 the card copies part of the memory to host system (because something is occupying memory) and that causes the miner to get slow.
Have no real solution for this atm... maybe should add a flag that can force the card to step down one more solver (falling to 3G solver in this case ) .... hmmn

I am very sorry about this, but on C29M I just did not have enough users to justify putting much time (and all the cuckoos need a lot) into a new solver. It was not the worst algo I made, but as said ... the ratio between time spend and roi did not fit.
There may be a rework on C32 later this quarter of the year though.


But today lets say hello to lolMiner 1.05

You can download it here.

The change log:

- Improved performance of Beam Hash III for all supported 4G and higher cards by 4-8% depending on model
- Added highly experimental support (untested) for Beam Hash III on blockchain drivers for Polaris 10, Fiji and Hawaii GPUs
- Added highly experimental support (untested) for Beam Hash III on Kepler (SM 3.2 and higher) GPUs, Requires CUDA 9+.
- Windows: Added more search locations for NVML library
- Fixed bug: Mining Cuckarood-29 not starting up
- Fixed bug: Mining on Nvidia GPUs crashed after few seconds when NVML library was not found
- Fixed bug: Mining Beam Hash III on some 8G cards did not start up (tried to allocate more memory in one buffered then allowed on some OS)

Before = 1.02?
I mean did BeamHash III work with 1.02?
May I ask which driver you use? lolMiner has Fiji builds for every driver released in 2018-2020.

So ... very busily spinning last days to make lolMiner 1.03
reality as quickly as possible.

You can download it here.

This time on board (change-log):

• New Beam Hash III 4G solver, replaces the 6G solver on AMD Vega GPUs and earlier (+5 - 8% performance on RX 580 & Vega GPUs)
• New Beam Hash III 6G solver on ROCm and for AMD Navi GPUs (+8-10% performance on Navi cards)
• Fixed a bug with the API (--apiport) crashing the miner in Windows on startup

So overall expect the miner to be more performant now on Beam Hash III - and the crashed the miner had last versions on startup when the API was active should now be history. Happy mining!

Wired ... 20.0.2 should be way new enough.
What is strange is in particular that it seems to be fine with the first VII, but somehow does not like the 2nd.

What happens when you start on device 4 only and skip 11 (and well other way around: when you start on 11 only) ...

Wonder whats going on there...

First of all Nicehash BeamV3 currently only works for few shares ... then it looses connection since there are no buyers for BeamV3 yet
Wait for the fork.


BUT:
My miner had an other bug with Nicehash, which is fixed with new lolMiner 1.01

Please load it here: https://github.com/Lolliedieb/lolMiner-releases/releases/tag/1.01

Changelog:

• Improved performance of Beam Hash III on all 6G+ cards (by about 4-5%)
•     Added 3G* / 4G solver for Beam Hash III
      (* will not work in Windows 10 and some older AMD cards)
•    Fixed an issue with mining Beam Hash III to NiceHash
•    --tls now allowed to be used multiple times to configure it for each pool separately
•     If found the miner will now load the user_config.json file per default again (fixes issues with minerstat)
•    Fixed temperature and power monitoring for VII and Navi on Windows 10 (Fan speed may still be broken ... work in progress)

So ... took a while, but lolMiner 1.0 is ready to go

You can download it here.

Note:

lolMiner changed configuration scheme a bit, since I did a very deep rework on the internal stuff behind it.
That said:

• --port is no longer supported! Please use --pool <addr>:<port> scheme instead
• For using failover pools: You now can use --pool, --user and --pass multiple times in cmd. They will be processed in order of appearance
• You now can select the algorithm directly. For this --algo (-a) is a new parameter. Use --list-algos for getting a list of supported algorithms and if they need / support personalization (--pers).  --coin is still supported, but now used for more special functions and some selected projects
• There is a new, more simple config file format that just uses same parameters as cmd. That said with --json and --profile you still can import the old format if you liked that more

This is only a small set of the changes. In doubt read the new help function (-h) or the growing online manual.

With regard to new algorithms: Support for Cortex and the new Grin-C29M solver was already given with the last alpha versions of 1.0.
New in this release is

• Support for Beam Hash III (use --coin BEAM for auto for routine from Beam Hash II to III or --algo BEAM-III to select it manually)
• I re-enables Beam Hash I (--algo BEAM-I) with support for personalization (--pers). This should allow mining Beam forks like XGM or DEFIs.

For any questions: feel free to ask


-------------------



They have it all - the compute mode is pretty much what the Beta driver introduced. The only problem is that some miner devs did not recompile their miners for newer drivers (one older Monero miner was famous for this) and thus some miners did not work with the more recent versions. But technically the newer drivers can do everything the blockchain driver can

Is the source code of the DEFIS fork available? Can not find it in your github repo.
You said its postponed, so there must be a consensus change to postpone it, but do not see a recent commit changing the fork height. Overall would like to review the changes from Grimm / see what people install there

Problem is that your drivers turn 3 years old now. AMD had a lot of breaking changes with their drivers, so if I want my code to run on older ones I need to install them and do all the build stuff for them as well. Its already plenty of work to go back to 18.10 drivers (RX 580) or 18.50 (Vegas), which are the oldest I currently build for.

Why do you want to stick on that old drivers? The newer ones are really not bad - and what was introduced with the old blockchain drivers is now part of the newer ones. Afaik there is no technical reason to stick on the (very) old ones.

Hi there

Well KawPow is a bit complicated for me - I have not much knowledge about ethash / progpow and co yet - and I am usually only willing to implement, what I understand in full and do not copy codes ...
Yet - I have it on the radar, but need to find some free time for that first


Still I got news: Yesterday evening I released lolMiner 1.0 alpha 2

the download is here:

https://github.com/Lolliedieb/lolMiner-preview/releases/tag/1.0a2

Whats new:
- New GRIN-C29M code: +10% speed on most supported cards
- Improved GRIN-C29M fidelity (gives additionally 2-3% better pool hash)
- Added Beam Hash III test network code (use --coin BEAM-III). A more tuned variant will be released with next version (before Beam fork height on June 28th)
- temperature, fan speed and consumption readings are now also in API
- Some minor bug fixes and internal changes

Expected performance (C29M):
- Radeon VII: 8.05 g/s
- Vega 64 5.8 g/s
- Vega 56 5.0 g/s
- RX 5700 4.2 g/s
- RX 580 2.9 g/s