I have five 7970s. I have been running them in various pools without any problems, they get a high hashrate and tiny reject rate.
However, I would really prefer to be on p2pool, but I consistently get huge reject rates.
My normal cgminer settings include:
intensity: 9
threads: 5
Per the cgminer thread I adjusted intensity to 8 and threads to 1, but still get a massive reject.
I've played with those numbers randomly, but the reject rate is always high.
Is there something stupid I'm missing?
Bounty to the first person to get it fixed.
How many rigs?
I have three rigs with 7970s. One has 4, one has 3, and my workstation has 1. I average between 2-7% reject rate.
I agree with the others, 5 threads is too high. But I've been using 2 threads without a problem for a long time. I just changed mine to 1 to see what changes.
From my experience high rejects can come from a few things:
1 - make sure you have your p2pool port forwarded by your router to your p2pool
2 - don't run p2pool on a machine that is limited in memory or CPU. at least dual core, and you want to have enough memory for the OS and apps so that you aren't swapping. since memory is dirt cheap, I have 4 gigs on all my miners. my main p2pool server has 3 cores.
3 - keep your network latency as minimal as possible. for example, run p2pool and bitcoind on the same box.
here I have p2pool on two boxes, with my main one being the one with 4 gpus. I run p2pool on non standard ports and have my router configured to forward the ports to each, so I have a high number of incoming and outgoing connections on each. I also forcibly connect them to each other with -n.
Here's my cgminer.conf from my main miner:
"intensity" : "9,9,9,9",
"vectors" : "1,1,1,1",
"worksize" : "64,64,64,64",
"kernel" : "poclbm,poclbm,poclbm,poclbm",
"gpu-engine" : "0-1125,0-1125,0-1125,0-1125",
"gpu-fan" : "60,60,60,60",
"temp-cutoff" : "90,90,90,90",
"temp-overheat" : "85,85,85,85",
"temp-target" : "75,75,75,75",
"api-listen" : true,
"api-port" : "4028",
"expiry" : "120",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "1",
"log" : "5",
"queue" : "1",
"retry-pause" : "5",
"scan-time" : "60",
"temp-hysteresis" : "3",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
Note that I've used worksize of 128 with no negligible difference from the recommended 64. Everything below the temp-target cgminer set, I didn't touch.
Also, at first glance, my rig with 3 GPUs that has a single core CPU, my hash rate dropped a bit when I changed the threads to single, so I put it back to two.
Hope this helps.
M
