During the last 30 days, I have measured the performance of HIVE and STEEM API nodes. Every hour, a new benchmark of all nodes was done and the results of the benchmark have been storing at 
The following plots show a dot for each working fullnode for every hour of the last 30 days. HIVE and STEEM API nodes are marked with different colors (HIVE is shown by a red diamond and STEEM is symbolized by a blue dot).
This plot shows the number of currently responding nodes. The number of API nodes for HIVE is significantly higher than for STEEM.
| HIVE | STEEM | |
|---|---|---|
| median | 11 | 3 |
| HIVE | STEEM | |
|---|---|---|
| median | 0.17 s | 0.48 s |
This test measures the mean duration for receiving information about a vote, a comment and an account.
Almost all API nodes are faster on HIVE.
The median call duration is significantly lower on HIVE.
It can be seen on the data that splinterlands has moved from STEEM to HIVE at 1. June. The call delay on STEEM is higher before the 1. June.
| HIVE | STEEM | |
|---|---|---|
| median | 2059.09 op/s | 740.24 op/s |
All HIVE API nodes are able to serve account history faster than on STEEM. The median account operations per second are 2.8 times higher on HIVE.
It can be clearly seen that splinterlands has moved on the 1.june to HIVE. The amount of served blocks per second is below 10 blocks/s before this date and jumps to around 35 blocks/s on STEEM.
| HIVE | STEEM | |
|---|---|---|
| median | 29.40 blocks/s | 35.14 blocks/s |
| HIVE | STEEM | |
|---|---|---|
| median | 50.60 s | 53.49 s |
This benchmark measures the time delay of the latest received irreversible block to the current time. The median delay is lower on HIVE.
HIVE shows a consistent API node performance. On June 1 splinterlands has moved from STEEM to HIVE.
While the loss is clearly visible in the node performance on STEEM (due to less activity the performance did go up), there is no drop in the node performance on HIVE.
If you like what I do, consider casting a vote for me as witness on Hivesigner or on PeakD
