Astronomy image processing
The current desktop computer that processes all of our astronomy images is nine years old.
It’s specs are:
Intelcore 2 Duo E7400 @ 2.8GHz, 4GB DDR2, 120GB SSD, DVD, Win10Pro64bit, sata 4tb drive, usb2 4tb wdmybook2016, nvidia geforce gt720 gpu
It typically processed 100GB of video files each run, first with PIPP to centre, crop and convert from .ser to .avi files in a batch run.
Next is autostakkert!3 which can also batch process the .avi and after analysis, typically stacks the best 5%, 10% and 25%. It also normalizes the stack and applies a x1.5 drizzle. This can take literally *hours*.
After that, Registax wavelets are applied but they are pretty quick.
All of this led to the conclusion that it was time to upgrade the computer.
We found a higher end, slightly newer system, refurbished for $240Can. Its specs are:
Lenovo ThinkCentre M91P (Refurbished) Desktop Computer, Intel CI5-2400 @3.1GHZ, 8GB DDR3 RAM, 1 TB SATA drive, DVD-ROM, Windows 10 Home.
Faster processor with more cores, more and faster RAM and in general a higher end system… should be great!
The 1TB SATA boot drive was replaced with a 120GB SSD, the GPU transferred across, the internal 4TB spinning drive and the external USB2 4TB drive moved across and the system was up and running.
BENCHMARKS
PIPP Core2Duo 8m15s, Core I5 8m30s
This indicates that PIPP is not CPU bound but rather drive bound. Will run the tests again with all data files on SSD.
AUTOSTAKKERT!3 Core2Duo 16m43s, Core I5d 5m0s
4 cores instead of 2 is a big improvement.. 70% less time.
Many more files were run and the results were within a few percentage points.
A 12 hour run would drop to under 4 hours, a meaningful difference between finishing tomorrow or today.
The only down side.. it is still pretty old itself.. instead of a nine year old system, we now have an eight year old system 🙁