Discrepancy between spark reduce and sql sum statement

Why there is a discrepancy between spark reduce and sql sum statement. Thanks for your help in advance.

Spark ::
val totalRevenue=orderItemsRDD.map(rec => (rec.split(",")(4).toFloat)).reduce((acc,value)=> acc+ value)
printf("%f",totalRevenue)
Output - 34326256.000000

SQL ::
select sum(order_item_subtotal) from order_items;
Output -
±-------------------------+
| sum(order_item_subtotal) |
±-------------------------+
| 34322620.5984249 |
±-------------------------+