Miner Benchmarks - What is currently the best Grin miner?


#1

With new miners and new versions of miners constantly being updated and saying 30% faster etc, it would be good if people could share their results so we don’t have to test all the new miners and versions all the time.

Some miners:

GMiner
NBminer
AceMiner
Bminer
Miinerbabe (KBminer)
CryptoDredge
GrinPro.io (updateed version of GrinGoldMiner)
GrinGoldMiner

Please share if you can: OS, CPU, Miner including Version, Video Card, C29/C31, and your hashrate. Also include if your overclocking/underclocking and those settings if you can but of course it varies card to card.

I’m currently running:

Hive OS (linux) OS version 0.6-19@19022402, Celeron G3930 2.9 GHz, Bminer 15.0.0, (4) Nvidia GTX 1060 6GB, C29, hash 3.160-3.230.

I am undervolting and overclocking - hive settings fan 100, core 150, mem 1400, pl 100 and the cards run around 100w (usually under).

My gaming rig is also mining:

Hive OS (linux) OS version 0.6-19@19022402, Intel i5-8600K at 3.6GHz, Bminer 15.0.0, (4) Nvidia GTX 1080 8GB, C29, hash 4.950-5.0

I am running at stock for my gpu.

Vega Rig - 4X Sapphire 64

Minerbabe OS V4.15 (linux), Celeron G3930 2.9 GHz, KBminer v0.7.0 (4) Vega 64, C31, .41-.43

Power usage is way too high and I the power settings don’t seem to bring usage down.

Undervolt/OC settings, core clock 1401, core vol 850mv, mem clock 1100, mem vol 850mv, fan 100

Typical power 150w-220w, its all over the place.


#2

I’m using the Sapphire RX570 16GB with ePIC-Boost-Miner and the ROCm driver under Ubuntu.

I’m using about 200W on average, and getting 0.47gps with C31.


#3

You’ll want fidelity as well; preferably something like fidelity_24_42 for increased accuracy. See


#4

Tromp great point. Some earlier upgrades of miners did have some not so good fidelity.

FYI I’m trying Grinpro 1.2 now on HiveOS. Getting better results than last time I tried gringoldminer and more importantly power control works well - 114w-125w per vega.

Hive OS (linux) OS version 0.6-19@19022402, Celeron G3930 @2.9GHz, GrinPro 1.2, (4) Vega 64, C29, hash 3.2-3.416

Overlock/Undervolt settings

Fan 100, Core 1408, DPM 2, V 925, Mem 1100, MDPM 905

Also CPU load average .86 1 min, .83 5 min, .48 15 min

Also I am now using C29 instead of C31 for my Vegas. According to the calculator here, if I am mining 3.42 C29, I would need to mine .65 C31 in order to make using C31 the more profitable choice. I can only do .41 so far with Minerbabe so it looks like C31 has really gone down profitability wise.


#5

Hey, sorry for the noob question, but what exactly “fidelity” refers to ?


#6

Fidelity is a probability of finding a cycle in a graph that has one. I.e. assuming you have a good nonce how probable it is you will turn it into a solution(share).

Fidelity of 1 means that you always find solution if the graph has one. Lower fidelities generally mean that you have to look through more graphs to find a solution.

Example:
Algorithm1 processes 1.7GPS with 0.7 fidelity. It produces 1.7 * 0.7/42 = 0.028 shares per second.
Algorithm2 processes 1.5GPS with 0.9 fidelity. It produces 1.5 * 0.9/42 = 0.032 shares per second.
Algorithm2 is superior to Algorithm1 even though it has lower GPS.


#7

Well, thanks a lot for this explanation.

So theoretically, a fidelity above value 1 is equal to value 1, right ? You can’t find more shares than you can afford to ?

Another question, if you allow me :

Using the latest Bminer, I usually get 1.1 fidelity rate @ 1.4G/s with a single 1080Ti. The pool reports an average hashrate between 0.8 and 1.3 G/s, over a period >24h. If I understand there are miner’s fee and pool fee to take in consideration, I’m still amazed at how the average hashrate drops significantly between what the miner and what the pool displays.

How can I optimize this ? Is it a matter of choosing the right pool, as a small one will have more “share wastes” ? I kind of don’t want to switch to a bigger one, in order to keep the network balanced [like if my single GPU could make a difference, though…]


#8

Fidelity above 1 is not possible. The probability of something happening can be 100% (i.e. it certainly happens) but cannot be more.

What can happen though is that:

  • You are lucky and in a given period you find more shares than you should on average. But you can’t consistently outperform your odds so this smoothes over time.
  • Your miner may overestimate its GPS and/or it’s fidelity to trick you into using that miner over some other miner.
  • You can get less shares than expected if the pool has high latency (i.e. network delays may cause you to mine on previous block for some time).

On top of all that there are fees that come to play… It’s complicated. One way to figure this out is to switch pools for a day and see what are your rewards. Then report them on this thread so people can compare :slight_smile:


#9

Of course he means the empirical or approximated fidelity, which can exceed 1.


#10

We are seeing some pretty good results with Bminer and written a guide for it


#11

Latest result with bminer-v15.0.2. -> C31. (Pool grinmint EU)

4 GPUS 2080 TI INNO3D.

[GPU 0] Speed: 2.08 G/s Fidelity 1.006 Temp: 73C Fan: 54% Power: 234W 0.01 G/J
[GPU 1] Speed: 2.05 G/s Fidelity 1.058 Temp: 79C Fan: 67% Power: 246W 0.01 G/J
[GPU 2] Speed: 2.05 G/s Fidelity 0.913 Temp: 75C Fan: 59% Power: 239W 0.01 G/J
[GPU 3] Speed: 2.03 G/s Fidelity 0.946 Temp: 81C Fan: 71% Power: 251W 0.01 G/J


#12

Of course I was not talking about approximation but about the probability as defined in my previous post. Example: If you hold an election and you count votes and one party gets 110% of votes that does not mean 110% of voters voted for this party. It may be an empirical approximation of the election outcome but it does not make it possible that 110% of the people voted this way.

Nakhom says that he usually gets this fidelity. He did not say he gets 1.1. half of the time and 0.9 other half. He usually gets 1.1. That is not possible and likely an error on the measurement side of the software.

No! Fidelity above one is not one. It’s a bug. I.e. If you measure fidelity correctly and your measurement yields X then the fidelity is close to X. If you measure it incorrectly to be X then it can be anything. If you measured 1.1 then there is a chance it’s one, but it may well be anything else.


#13


#14

Three options:

  • the number over time converges to 1
  • the number over time fluctuates around 1 being 1.2 for some time and being 0.8 for a while
  • the number is bull*hit

#15

I’ll try and post updated numbers/stats for my rigs tonight.

The vega on GrinPro is doing real well 3.4 with 0 bad over 24 hours, i’ll grab fidelity as well.

I did have an issue with GrinPro on my Nvidia 1060 rig, it kept crashing about every 10 minutes but I am guessing its due to my really weak cpu. I switched to Bminer for my 1060 and no crashes but its like 95% - a fare amount of shares are bad.


#16

My Vega rig with GrinPro

10:25:06 Global solvers fidelity: 0.979

GPU 0 : Vega ONLINE Mining at 3.41 gps Solutions: 3453
Last Message: 10:25:06: GPU AMD0: Trimmed in 292ms to 74112 edges
GPU 1 : Vega ONLINE Mining at 3.36 gps Solutions: 3485
Last Message: 10:25:05: GPU AMD1: Trimmed in 295ms to 69295 edges
GPU 2 : Vega ONLINE Mining at 3.42 gps Solutions: 3627
Last Message: 10:25:05: GPU AMD2: Trimmed in 290ms to 80612 edges
GPU 3 : Vega ONLINE Mining at 3.43 gps Solutions: 3479
Last Message: 10:25:06: GPU AMD3: Trimmed in 290ms to 75292 edges


#17

bminer 15.0.0 running C31:


#18

BMiner is the best!
Bminer ist das beste!
Bminer je najlepší, kurva!


#19

I have just updated bMiner to 15.0.2 to see if there any difference comparing to 15.0.0. No difference:

3 PALIT GeForce RTX 2080Ti , PA-RTX2080Ti DUAL 11G, [ne6208t020lc-150a]. MWGrinPool, C31, SSl, Ubuntu 18.10.


#20

Wait for 16.0 to feel the difference mate. :drooling_face: