I’ve been mining with three 8TB drives for about a week now and the highest I’ve seen Effective Capacity is around 19TB. Since the actual capacity of the 3 drives is around 21.82TB, I’d like to at least see that number over 20. While writing this, it’s actually down to a little over 16TB with 308 Valid Deadlines in the last 360 blocks.
I know that my ODROID-XU4 setup will read all 3 drives in under 72 seconds, so I decided to take a look at the last 360 blocks to see how I was doing in relation to the actual blocks being generated. Luckily the wallet stores all of that information for me to go and grab from phpMyAdmin. To get the latest 360 blocks, I used the following SQL statement:
SELECT `height`, `block_timestamp` FROM transaction GROUP BY `height` ORDER BY `height` DESC LIMIT 360
This gives the data set to work with. I went through and calculated the block times of each block and plotted them on a graph. Maybe there’s a way to grab this in the Explorer, but I don’t know of one at the moment. This method worked pretty well anyway.
The average round time is 240 seconds, as maintained by the system and I was told on the forums & Discord that anything under that is fine for read times, but I feel like in my previous tests with the Laptop vs. ODROID that the only real difference was the Valid Deadlines, which may have been from missed opportunities.
In the chart below, you can see how the block generation times worked out. The GREEN is anything above 240 seconds. BLUE is anything longer than my 72 second read time, so I’m still safe there. The ORANGE and RED are in the “danger zone” of not submitting any valid deadlines before the time runs out. That’s not to say that I won’t. It’s possible that in a 5 second round, I could still find and submit something in the first 3 seconds.. but that’s pretty unlikely.
The breakdown is:
- Safe (green & blue): 307
- Maybe (orange): 38
- Probably Not (red/orange): 5
- Very Unlikely (red): 9
So with an actual Valid Deadlines of 308, that means only ONE of the rounds below my round time actually counted. That’s pretty awful. At this point it doesn’t make me all that enthusiastic to add more drives to the ODROID. I may call 3 drives it’s max and move on to a second setup.
only ONE of the rounds below my round time actually counted
On the other hand, it’s spurred me to do some research into ARM-NEON Intrinsics for upgrading the creepMiner setup. I’ve begun a little testing on this, but I’m not sure if it’ll give me any improvement over the SSE2 implementation. If anyone has some data on this, I’d love to hear about it. Feel free to post a comment below. I also recently discovered a Linux port of the Blago Miner, so I may try that soon.
Conclusions:
- The ODROID-XU4 may be topped out with 3 drives without further optimization.
- There’s a strong correlation to your read time and Valid Deadlines. Enough that any blocks forged in under this time are almost never counted.
Leave a Reply