Help! my install of ONLINE 4.00 is 30% slower than Standard Engine

I have a 4 column table. Each column is a char. I do this query select count(*) from invent where description matches '*MEM*' and time the result on each of the engines. The table has about 115,000 rows and the description column has length 33. The total row length is 57. The table doesn't have any indices. Each engine resides on the same model disk running from the same SCSI controller. Each query runs isolated from the other. The machine is a single user system during the benchmarking. There is 2MB of shared memory using 800 buffers, and a raw device (50MB) is used for ONLINE. SE uses 36 seconds, ONLINE uses 47 seconds. This ratio is very consistent after running the test multiple times (>10). Using 'set explain on' showed identical results (sequential search plan) *except* that the SE cost was 50% *greater* than ONLINE. I've called Informix tech support, but so far, the 1st level of help available hasn't helped solve my problem. Somebody's looking deeper into this problem. In the meantime, I need to make progress (pun intended) :-). Are there any comments from anybody as to why these results, assuming ONLINE really is faster than SE, are valid? Maybe ONLINE isn't faster. Thanks to all. -- "The Battle of the Gulf was won on the playing fields of Atari" - J. Beaudin My preferable .signature is awaiting apropriate display technology.