Why in hive TEXTFILE tale scan is too slow？

[luhea]

trino version:418
config:
query.max-memory-per-node=32GB
num workes:16
query.max-memory=576GB
jvm:
-Xmx50G
hive table：
table1：format = 'TEXTFILE' ,textfile_field_separator = '|' ,partitioned_by = ARRAY['pt_dt','cluster']
table2: format = 'PARQUET' ,partitioned_by = ARRAY['pt_dt','cluster']
table1 and table have 133342642 rows in pt_dt='2023-06-13'

whren i run in table1 and table the same query:
SELECT a.path ,
sum(a.filesize)/1024/1024/1024/1024*3
FROM
cluster_stats.table1 AS a
WHERE pt_dt='2023-06-01'
GROUP BY a.path
ORDER BY a.path
limit 20

table1 return in：8min table2 return in 14s

and in stage performance：
table1：

table2：

why in hive TEXTFILE tale scan is too slow？