Timings for ocean only

Timings from Dave Storkey

Dave has given me some numbers, which can be found on p124 of my UKESM General II.

• He's using 480 cores
• One month seems to take on average about 37.5 mins. This suggests one model year will take 27,000s or 0.3125 days.
• This is 3.2 model years/day
• Speed on IBM for 450 cores was expected to be 4.371 model years/day, which is 0.2288 days for one model year.
• This suggests that Cray is 0.3125/0.2288 = 1.37 times slower than IBM

Richard's times for ORCA025

Richard has some times in OceanIceBio page below UKESM ticket 100. Looks like each of his runs use 1 node (32 cores) for XIOS.

Number of cores	Model run length	Duration (s)	Expected duration for one model year (days)	Expected duration on IBM (Cray/IBM scaling)	Speed (model years per day)
392	10 days	983	0.410	0.274 (*1.50)	2.44
512	10 days	781	0.325	0.218 (*1.49)	3.08
608	10 days	700	0.292	0.192 (*1.52)	3.42