>The same job, with the same data, etc. should show pretty close to the same >TCB time, regardless of system load.
This hasn't been the case since 1981, when SP1.1 came out. Some system functions would then run non-disabled and 'billed' back to the last active TCB. This was apparent specically when the system was lightly loaded and the OS would 'rattle its chains' just looking for work and to keep busy. This caused a great variance in TCB time, and we were still doing benchmarks. This was before the time of LSPR and Joe Major's method of evaluating processors. The busier the processor the less the variance. The introduction PR/SM also caused variance to show up. This was mainly due to cache pollution. More modern methods of cache management reduced this, but it's still there. - Too busy driving to stop for gas! ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
