Performance ts_vector fulltext search - Mailing list pgsql-performance

I've configured 2 table like this

        CREATE TABLE "public"."User_Statement_Pivot_2" (

"Email" varchar(50),

"UserId" varchar(50),

"ShortId" varchar(50),

"LastDirectJobMailSentDateTime" int8,

"Where" varchar(4096),

"tsv" tsvector

)

WITH (OIDS=FALSE)

;

ALTER TABLE "public"."User_Statement_Pivot_2" OWNER TO "postgres";

CREATE INDEX "IX_btree_usp2_userIdShortIdEmailLastDJMSDT" ON "public"."User_Statement_Pivot_2" USING btree ("UserId", "ShortId", "Email", "LastDirectJobMailSentDateTime");

CREATE INDEX "textsearch_tsv" ON "public"."User_Statement_Pivot_2" ("tsv");

CREATE TRIGGER "tsvectorupdate" BEFORE INSERT OR UPDATE ON "public"."User_Statement_Pivot_2"

FOR EACH ROW

EXECUTE PROCEDURE "tsvector_update_trigger"('tsv', 'pg_catalog.italian', 'What');

Column "What" (table User_Statement_Pivot  is just a single word or max 2 words separeted by space " " (ex: programmatore .NET), and tsv (table User_Statement_Pivot_2)  is populate by materializing column with a ts_vector of "What".

Now if i perform those 2 queries

    SELECT * FROM "User_Statement_Pivot_2"

where tsv @@ to_tsquery('italian','programmatore|analista')

    

Records on Tables (are same) like 8 milion. 

Execution time of 1st query is 2 seconds (result set like 13.027)

Execution time of 2st query is 3 seconds (result set like 13.027) same records

Those are query analize

Configuration 

If i increase words number in to_tsquery in OR condition those 2 queries are more different (exponentially). I don't understand why a materialized column is more slow than a calculeted one.