Recently I saw a saying about tempdb performance like "…Since you have 8 CPUs, split temp log files on 8 different physical disks to accelerate the write log performance…" This is not correct.
Data file (MDF, NDF) and log file (LDF) have different I/O behavior pattern. For data file, it is "asynchronous random read/write". For log file, it is "synchronous sequential write". Once a data page is modified, SQL Server storage engine will change the page in buffer pool first, and refresh it to disk later. But for log file, SQL Server uses a write-ahead log, which guarantees that no data modifications are written to disk before the associated log record is written to disk. So data file IO pattern is asynchronous, and log file IO pattern is synchronous.
As SQL Server writes log sequentially, even there are more than one log file, SQL Server will use them one by one. The benefit of multiple log files is providing pre-allocated disk place and avoiding single log file size becoming too huge. But for tempdb, the default recover mode is "simple", which means the log space will be used circularly. For those databases with "full" recover mode, backing up log file regularly can also free the log space.
The conclusion is no need to add multiple log files from performance point of view.