based on the histogram data, i changed the delay(10000) in esp.c to delay(1000)
with the following results:

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
1.4.X      10   315 98.9  1878 99.3  1012 92.0   339 99.5  1965 91.4  17.3 42.8
1.5        10   118 97.6   146 98.5    84 99.6   143 99.4   186 97.0   4.8 96.5
delay=1000 10   366 99.5  1726 85.0   744 54.0   492 99.7  2106 96.9  76.8 38.6

[the 1.4.x results are from my q800, which is fairly similar to the q700.]

i noticed that the delay(10000) crept in somewhere around revision 1.22 of
esp.c when i believe allen added 'correct' support for the dual bus on the 950. 
esp_dualbus_intr also has a delay(10000) that should probably be 1000 but 
then again, maybe not.  i am guessing the delay should be related to the 
clock speed of the scsi controller.  any thoughts?