Expected 1080 Ti performance


#21

Tromp says the miner code is buggy for the 2080: https://www.grin-forum.org/t/possible-cuda-bug/1338

The 1.0.1 is not improving mining only network connectivity (so you might get blocks faster thus improving your chances…). Working on improving mining, but it means changing things around in non-trivial ways.


#22

Thanks for the info, will probably wait until 2080 issues are solved before investing. Also, how are miners supposed to pay for electricity etc. when there is no way to liquidate the coins (or is there)?


#23

Exchanges are coming.


#24

The bug has been fixed. 2080 cards can mine again…


#25

Wow, that was fast! Are the mining performance similar on 1080 Ti and 2080 Ti?


#26

See here for other Cuckoo variants:

Cuckoo Cycle runs at same speed as Cuckatoo Cycle…


#27

Did I understand right:

if I solve for example:
./mean29-generic -h R4nNmL0Xjir09QswGFtqc7wwyuk10OfI23OlwznvsAw=yYJP/pjF9b8= -n 137 -t 483

and push the solution, and get " Failed to meet target"

is it same as I solve, as requested
./mean29-generic -h R4nNmL0Xjir09QswGFtqc7wwyuk10OfI23OlwznvsAw=yYJP/pjF9b8= -n 0 -t 483

or not? As @aemin asks

Because if I start over and try to run over and over:
./mean29-generic -h R4nNmL0Xjir09QswGFtqc7wwyuk10OfI23OlwznvsAw=yYJP/pjF9b8= -n 0 -t 483

I never get a solution, because I have always the same results, hmm or not. What of my thinking is wrong?


#28

I have rewritten some of your code @tromp and it works great on epoch. And parse_generation_result:448 accept the inputs of any solved solution I push. Not yet published here: https://github.com/Zwilla/BitCaine5_aeternity_miner


#29

Thanks for fast replies. Just to double check: since this will be solo-mining, you have to be “lucky” and mine an entire block to get any payouts? In other words, using a single 2080 Ti, can you expect to mine roughly 1 block every 3 days at current difficulty, but also risk mining for many days without any rewards?


#30

extra_args: “-r -1"

will loop for ever


#31

Is this the reason why I always get the same solution??? (push need information)

Solution aae9b d3caeb db7686 168f946 203daac 3e812be 4603264 5320126 6d28244 7014947 776a712 7a6a4a6 8b44de4 8c25dd0 949d487 a2f253d a99edb5 b9e6dc1 ba6db48 cc609fb d745ea9 e218d20 109bf7b2 1103fb2c 1158febb 125c3a3a 12813d98 13cd0508 154e662d 1658ef1d 16fc2357 188b744f 192c2b03 19813277 1a26f3b7 1aa26ce8 1b18e2fa 1ccca739 1dab39dc 1e0b29cf 1ea4fed8 1edee859
Solution aae9b d3caeb db7686 168f946 203daac 3e812be 4603264 5320126 6d28244 7014947 776a712 7a6a4a6 8b44de4 8c25dd0 949d487 a2f253d a99edb5 b9e6dc1 ba6db48 cc609fb d745ea9 e218d20 109bf7b2 1103fb2c 1158febb 125c3a3a 12813d98 13cd0508 154e662d 1658ef1d 16fc2357 188b744f 192c2b03 19813277 1a26f3b7 1aa26ce8 1b18e2fa 1ccca739 1dab39dc 1e0b29cf 1ea4fed8 1edee859
Solution aae9b d3caeb db7686 168f946 203daac 3e812be 4603264 5320126 6d28244 7014947 776a712 7a6a4a6 8b44de4 8c25dd0 949d487 a2f253d a99edb5 b9e6dc1 ba6db48 cc609fb d745ea9 e218d20 109bf7b2 1103fb2c 1158febb 125c3a3a 12813d98 13cd0508 154e662d 1658ef1d 16fc2357 188b744f 192c2b03 19813277 1a26f3b7 1aa26ce8 1b18e2fa 1ccca739 1dab39dc 1e0b29cf 1ea4fed8 1edee859
Solution aae9b d3caeb db7686 168f946 203daac 3e812be 4603264 5320126 6d28244 7014947 776a712 7a6a4a6 8b44de4 8c25dd0 949d487 a2f253d a99edb5 b9e6dc1 ba6db48 cc609fb d745ea9 e218d20 109bf7b2 1103fb2c 1158febb 125c3a3a 12813d98 13cd0508 154e662d 1658ef1d 16fc2357 188b744f 192c2b03 19813277 1a26f3b7 1aa26ce8 1b18e2fa 1ccca739 1dab39dc 1e0b29cf 1ea4fed8 1edee859
Solution aae9b d3caeb db7686 168f946 203daac 3e812be 4603264 5320126 6d28244 7014947 776a712 7a6a4a6 8b44de4 8c25dd0 949d487 a2f253d a99edb5 b9e6dc1 ba6db48 cc609fb d745ea9 e218d20 109bf7b2 1103fb2c 1158febb 125c3a3a 12813d98 13cd0508 154e662d 1658ef1d 16fc2357 188b744f 192c2b03 19813277 1a26f3b7 1aa26ce8 1b18e2fa 1ccca739 1dab39dc 1e0b29cf 1ea4fed8 1edee859

#32

So your software can be mined? Is the built-in development cost?


#33

If you pass a different N to the miner via -n N it means you will solve a different graph.

But any other value than N=0 will not work (nor will any usage of -r) since this will place a nonce at a place where the PoW verification will not expect it. We are working on updating the miner code to extend its nonces to 64 bits and place it in the right spot.


#34

This is disappointing. Why do we really need to destroy and recreate the memory allocation each time and not support -r directly? at least setting the nonce as a starting nonce by using -n # -r Count would be a lot more effective in terms of mining. Seems to me passing the encoder hex header, the nonce and r by hand could return a more suitable response from the miner.

At least 50% of the gpu power is wasted this way. :frowning:


#35

I’ve re-designed the miner(s) to use a 64-bit nonce (as that is what Aeternity uses) and also placing it where the PoW verification expects it. Some more testing is needed, but it should be done fairly soon. This means nonce can be passed using -n and repeated runs can be made using -r.

But, because of the Bitcoin NG where micro blocks arrive every three seconds, the miner can’t be truly long running even in this way 4-5 seconds makes sense. Still, it should improve the mining quite a bit.

A fully demonized miner would need a lot more design work but I am sure it will land some day.


#36

yes. im tesing it now. seems working. but it can use only 1 GPU. multi_gpu version losing about 80% of hashrate because of restarts. in u new version i can run only 1 GPU. and i still cant run multiple nodes on 1 PC )) i


#37

Which “version” do you refer to?

There will be a multi_gpu version of the same kind (basically combining repeats and instances) - it should be a good deal better, but not sure how much… The overhead when starting multiple GPU-instances at the same time seems to be unpredictable…


#38

i cant merge u branch (with repeats) with multi_gpu branch. so i merged it with master branch. it working fine, but only 1 GPU (-d option takes last argument) how can i run my 6xGPU RIG with U version(with repeats enabled)?


#39

No, there are a couple of conflicts merging those branches, that is what I am working on at the moment :wink:


#40

while u working…can u tell how to run several nodes on ONE system. i want run u new version 6 times on different nodes. i tried different users (with different ports in yaml) for every instance. seems its impossible. *** maybe invite those 6 top miners to the team. they seem to know what to do