4th option, the nodes get lazy and allow false solutions in some cases, fedality >1
Don’t mine on such a node. If it’s lazy to accept fake solution then somebody is exploiting this and taking advantage of the honest miners.
?
I don’t think a bug in pow verifcation would only effect honest miners, your ether looking at a fork or dishonest miners getting a percentage boost for knowing a secret
If I was dishonest and I was aware that there is something wrong with the share verification alg I would try to exploit it. If it accepts some garbage then it maybe accepts a lot of garbage. Can you prove that spamming it with random number generator does not give me 1 share per second?
There are only a couple of things verification needs to check. If you are lazy to check one of them then there is a way to exploit it… I guarantee that. Example: If you don’t check that all edges in the cycle are unique then it’s enough to take any edge and repeat it 42 times and you have a 42-cycle with zero work.
What if some future implantation doesn’t stop overflowing when counting the edges of cycle and the rare 2^16+42 cycle block is generated(assuming its a short).
The miner that checks for this case has the clear upperhand and would have >1 fidelity.
This is a clear a present danger to the chain as given enough time this protental attack will happen, everyone panic.
More realistically, bug are just a thing that happen and given how generating test cases in rather hard an off by 1 error could very realistically get into “the java implantation of grin” or whatever shit ends up being made in a few years and the communities get fighty when any issue comes up in a few years and node software is a proxy war for… reasons
or strange math things I don’t understand doing an asicboost type thing when choosing graphs to get an extra percent
I understand that this is likely some bug or some incomplete implementation of the verification (maybe because someone was lazy or tired). But you’re suggesting that this bug may cause invalid shares to be accepted. And not “few”. If you see fidelity 1.2 then 20% of garbage shares are accepted. It’s very probable that this is exploitable and in that case it’s very probable that someone is exploiting this. So my recommendation remains the same: Stay away from such a node…
But to be honest, I don’t know if invalid shares are accepted. Scenario that this is a bug in the way fidelity is computed is also possible…
6% higher numbers with newest BMiner v.15.1.0: https://forum.grin.mw/t/bminer-15-1-0-experimental-support-cuckaroo29-on-amd-cards-rocm-only/2177/185?u=rabinovitch
Using the Sapphire 570 16GB with ePIC Boost release 1.0.2 with ROCm, in C31 getting about 0.46 gps and in C29 getting around 2.1 gps
Ohhh, I have ~6.5 gps on C29 with GeForce GTX 1080 Ti (11GB)
1 Like
i think still the best miner is bminer today
1 Like
I use KBminer, but after look at performance page l will try bminer 
, thanks!
bminer still gives us the best performance from what we saw thus far
Guide over at https://www.coingecko.com/buzz/how-to-mine-grin-coin-with-bminer-step-by-step-guide