I have just commissioned a 2 x Xeon 2.8Ghz 512MB 533Mhz FSB compute node for research work. I have just been told it's much slower (around 25% slower) than a single processor P4 2.8 Ghz 512MB 533Mhz FSB tower PC. The processes run on a single CPU are not multi-threaded, uses very little disk IO (there is no io wait), and little memory. The process simple reads a file and does a number of calculations on that file and then writes out a simple text file, and then repeats for the next file. The calculations are complex and the process often lasts hours and sometimes days. Running two processes in the twin Xeon seems to be around 25%-50% slower than running a single process in the P4 tower PC. Is this what I should expect give that the bus configuration in the Xeon is not any better than the P4 PC? Also I note from the output from top that the two processes each consume 99.9% of CPU, but the total user at the top is 50%, and the idle is 50%. vmstat gives this output, which backs up the top output: procs ------memory-------- ---swap-- -----io---- --system-- ----cpu---- r b swpd free buff cache si so bi bo in cs us sy id wa 2 0 0 238124 134148 46812 0 0 0 0 7 0 50 0 50 0 2 0 0 238124 134148 46812 0 0 0 12 1010 14 50 0 49 1 2 0 0 238124 134148 46812 0 0 0 0 1004 7 50 0 50 0 2 0 0 238124 134148 46812 0 0 0 0 1005 7 50 0 50 0 2 0 0 238124 134148 46812 0 0 0 0 1004 11 50 0 50 0 2 2 0 238124 134156 46804 0 0 0 19 1007 19 50 0 49 1 2 0 0 238124 134156 46804 0 0 0 1 1007 14 50 0 48 2 I know the two processes are not swapping CPU's but can't see what may be causing the 50% idle time, as I do not know what the process is waiting for. On the P4 2.8 tower there is no idle time at all. OS - FC2 patched to everything available beginning of last week! Am I missing something? (Brain?) Anybody have any opinions on this? Thanks, Neil. -- Neil Marjoram. Systems Manager University College London Adastral Park Campus Martlesham Heath Ipswich Suffolk IP5 3RL 01473 663711
