said by Hard Harry7:If you reset/reboot the modem and you see 100+ uncorrectables in 10min, thats bad no matter what IMHO.
Here's some real world data. My SB6120 uptime is 26 days. On one channel, the total uncorrectables is 490,358. There are 3,744 10-minute periods in 26 days, so my average uncorrectables per 10 minutes for that channel is 131. The other three channels have 2,285, 1,576, and 12,370 uncorrectables. OTOH, the total unerroreds for each channel is over 100,000,000,000. For the "bad" channel, the uncorrectable/unerrored ratio is thus about 5e-6, and for all of them taken together, about a quarter of that, so about one in a million. (The number of uncorrectables plus correctables is pretty negligible compared to the unerroreds, so I'm taking the number of unerroreds as the total just to make things simpler.)
I've never noticed any hiccup in performance, and without tracking these things day in and day out, or better still, hour in and hour out, or even less, I have no idea when they occur. For all I know, that one channel had a bad burst or two that I never noticed. I don't know what the threshold is for considering these values "bad", but it doesn't seem to be 100 uncorrectables in ten minutes if there were 20 million unerrored ones, which would be the average for my "bad" channel. Again, computing the average is pretty meaningless when you don't know the distribution of the errors. For all I know, most of the uncorrectables happened in that first 10 minutes.
So, I wouldn't worry about any of these numbers if observed performance is consistently what you're paying for. However, if you had uncorrectables hovering at a large percentage over a longish period of time, that could indicate a problem. Unfortunately, there would seem to be no way to determine this since it appears Motorola has eliminated the total unerrored count. I guess you could compare to the other channels, but for my example, one channel having a total uncorrectable 30x the others isn't a problem, I don't know what the threshold would be, and you'd have to track this regularly over small intervals to get an idea of the pattern of it.