|
Some great advice here.
StarTools is an unapologetic memory hog. More RAM can definitely help if it means less swapping to disk.
However, another cheap and great way to speed up your PC is adding an SSD drive (as Andy suggests). StarTools can routinely move multiple GBs of data back and forth between RAM and storage. An SSD will speed this up immensely. Check Ozbargain.com.au for some great SSD deals currently going on eBay.
The absolute fastest is running StarTools from a RAM drive (in Windows) or mounting your /tmp as a RAM drive (using tmpFS on Linux).
The 9010 and 9020 range from Dell are killer bang-for-your buck machines if you pick them up refurbished. Your i5 CPU is pretty quick still. And though it doesn't have hyperthreading, it can still hold its own when it comes to ST's number crunching.
Unfortunately more cores (or using the GPU to offload calculations) doesn't necessarily mean 1:1 improvements in speed. Much depends on the algorithms; not everything can be perfectly parallelised. Often the memory bus is the bottleneck, rather than raw CPU power.
StarTools is especially memory intensive compared to other software due to the Tracking functionality having to consult vasts amounts of data and previous states of pixels.
Lastly here is a no-cost option;
Consider binning your data before processing if your dataset is oversampled - you won't lose detail and will benefit from the noise reduction, allowing you to push your data harder. Reducing your resolution by 50% on the X and Y axis, means a 4x speed increase and a 4x reduction in RAM requirements. That said, you indicated your needed to drizzle your data, so it probably isn't oversampled in your case...
Hope this helps!
|