I could use a legend telling what each column means. Is the hashrate 1.561 or 1.934? Does one of these tell what the voltage is?
2 AIfDSo 55 1.561 1.934 109 7 0 0 183 [0:1] 36 5 7 5 6 7 7 8 8 7 7 7 7 7 7 7 7 2 0 2 1 0 0 0 0 1 1 0 0 0 0 0 0
I like to call the 1.934 number the work rate, though most call it the hashrate. The work rate is based on how many work units were processed. It's a function of the voltage and speed setting (55 in this case), and is usually quite stable. The 1.561 number, commonly called the noncerate, is based on how many good nonces are found, and will vary randomly even under perfect conditions (i.e. no errors).
Here's a legend for your line above:
Chip#: 2
Flags: AIfDSo (A = autotune on, check the source code for the others)
Speed: 55
Noncerate: 1.561
Hashrate: 1.934
Good nonces: 109
Nonce errors: 7
SPI errors: 0
MISO errors: 0
Work units: 183
Chip position: [0:1]
Bad cores: 36 (how many of 756 have not yet found a good nonce)
I don't know what the last two sets of numbers are.
The boards don't have digital voltage or temp meters, unfortunately.
In case you're curious about the math...
After the first 5 minutes of chainminer running, the stats are updated every 5 minutes, so the numbers are based on the previous 5 minutes of work. Each core does 4194304 (2
22) hashes per work unit. 183 work units * 756 cores * 4194304 hashes / 300 seconds = 1.934 billion hashes per second.
On average, 1 good nonce (diff=1) should be found per 4294967296 (2
32) hashes. Since 109 good nonces were found, it calcs the noncerate like this: 109 good nonces * 4294967296 average hashes per nonce / 300 seconds = 1.561 billion hashes per second. This was probably just an unlucky 5 minute interval and you'll see the noncerate higher than the hashrate at other times.
The 7 nonce errors are nonces the chip reported as good but chainminer found to be actually bad. This happens because of bad bit flips during the hashing process in the chip and occurs much more often when overclocked too much. Less than 5% average nonce errors is just fine. In this case, we can expect at least 4.8% nonce errors anyway because apparently 36 of the 756 cores are malfunctioning, at least at this clock speed and voltage. The chip's hashing cores are arranged in a 21x36 grid, at least logically, so it is common to find a whole row or column bad. You can see that visually in /tmp/.core.log
I just added two more boards to my rig. I now have SlotsS 0,1 4,5 8 C occupied. Oddly, the board in Slot 1 is grossly under performing:
