Bitcoin Forum

Other => Beginners & Help => Topic started by: Soichiro on July 11, 2011, 04:08:04 AM



Title: Optimizing for efficiency? (RPC-CUDA)
Post by: Soichiro on July 11, 2011, 04:08:04 AM
Okay, so, here's the thing. I'm starting out mining on a pool with my GTX 460. Of course, being a pool, there's a minimum efficiency I have to meet. I get 80% efficiency with the Phoenix miner, and 52 Mh/s average. This is okay. The thing is, with the RPC CUDA miner, I get around 62 Mh/s average, which is almost a 20% improvement. The problem is, this miner drops my efficiency to 10%, which is just awful. I know this is a very noobish question (I guess that's why this is the newbie section), but is there some sort of flag I could set to make the RPC miner more efficient? I really wouldn't like to waste that extra 20% potential from my GPU. (Also I just noticed that Phoenix is constantly using a whole core of my CPU, which RPC didn't do.)


Title: Re: Optimizing for efficiency? (RPC-CUDA)
Post by: Ang3lus on August 03, 2011, 12:05:51 PM
Okay, so, here's the thing. I'm starting out mining on a pool with my GTX 460. Of course, being a pool, there's a minimum efficiency I have to meet. I get 80% efficiency with the Phoenix miner, and 52 Mh/s average. This is okay. The thing is, with the RPC CUDA miner, I get around 62 Mh/s average, which is almost a 20% improvement. The problem is, this miner drops my efficiency to 10%, which is just awful. I know this is a very noobish question (I guess that's why this is the newbie section), but is there some sort of flag I could set to make the RPC miner more efficient? I really wouldn't like to waste that extra 20% potential from my GPU. (Also I just noticed that Phoenix is constantly using a whole core of my CPU, which RPC didn't do.)
Hi there, i don't know if You know what this efficiency really mean so i explain this:
Efficiency is a measure of how good your miner is and that strictly means how many valid hashes your miner finds in single "get work" task. Phoenix miner is searching entire space (2^32 hash combinations) event if he finds one valid hash, so if you get 1 "get work" task and phoenix will find 2 valid hashes by searching entire space then You have 200% eff. The problem is that rpc cuda miner is not searching entire space, while it finds valid hash it asks for another "get work" task. So statistically phoenix will have greater eff that rpc cuda miner.

100% cpu while using OpenCL is known problem.

Also see my rpcminer-mod (maybe it will be worth to add flag to search entire space):
https://bitcointalk.org/index.php?topic=27907.0