I run multiple machines with 258 threads. After upgrading to version 14.20, one-third of the machines only have 22 quils per hour. The other normal machines have 100 quils per hour. The CPU threads, RAM and storage of the machines are exactly the same, and they are also under the same network.
The attached picture shows the number of rewards at ten-minute intervals. As you can see, the original quil balance is 1338.9030, and the quil balance after ten minutes is 1339.0568. The calculated 24-hour reward is 22 quils, which is much lower than the 100 quils per day of other normal machines of the same type.
I tried various methods, such as switching network nodes, restarting, and re-git pull to install the code, but none of them could solve the problem.
I have heard that multi-core CPUs will have various problems. Is this a bug?
All servers with less than 32 threads are as stable as v1.4.19.
But all servers that have more than 48 cores become super unstable since v1.4.20, that’s why the reward drops.
My guess is, the more cores the more rewards drop percentage, due to some parallel issue in v1.4.20.
The problem is not solved. The output is still about 22 quils per day. With the same configuration and network, the output is about 100 quils. But I have about 40 machines with 258-thread CPU, 512 RAM, and 1 t storage in the same subnet. The rewards of machines in other subnets seem to have increased. About two-thirds of the machines have increased rewards. Some have reached 110 quils per day, and one-third of the rewards have dropped to about 20 quils per day. I have tried to re-download and switch the VPN exit IP, but I can’t solve this problem. I guess it’s because the same subnet will automatically balance the rewards?
Any information you need can be found by me. I am your loyal supporter.
ps: The picture shows the reward balance at a time interval of ten minutes, and then the number of rewards for one day can be calculated.
