HadGEM3 Resources and Performance on ARCHER2

Unified Model
1

Hi,

I am writing an ERC proposal and need to do some resource estimates for different resolution versions of HadGEM3. I have seen the information here -http://cms.ncas.ac.uk/wiki/Archer2#Performance. Is there any other information available on costs per model year for different setups?

Thanks, Mat

2

Hi Matt

We don’t have more numbers for jobs on ARCHER2 - there are other sources that can give an estimate good enough for a proposal – what configuration are you interested in?

Grenville

3

Hi Grenville,

Either N512L85 or N216L85 HadGEM3 atmosphere-ocean (i.e. without the UKESM carbon cycle etc.). I can work back from anything you give me really. It is an EU proposal so I need to estimate costs of resources in order to buy ARCHER2 time.

Many thanks for your help, Mat

4

Matt

From ARCHER1 (from a colleague in NOC), 1 year of coupled N512/ORCA12 was consuming around 4.8MAU of charged usage in old money. That translates to around 3.2kCU on ARCHER2.

And here Met Office (Login) for GC4 suites,

and Met Office (Login) for GC3 suites.

There is quite a spread of costs for the N216ORCA025 jobs (one node-hr translates to ~0.36CU)

Grenville

5

Many thanks Grenville. This has opened up a whole new world of computational performance metrics that I didn’t know about!

6

these metrics are CPMIP metrics see GMD - CPMIP: measurements of real computational performance of Earth system models in CMIP6 for details