Hi,
Have you tried monitoring your PC performance whilst running the process? (Go to the START menu, type in Perfmon, and select the .EXE file that should be offered. Then you might want to consider monitoring your total commited RAM memory, as well as the %age of time used by each processor. There are a few useful YouTube videos out there that can help you with setting up PerMon - I have it running on one of my auxiliary monitors so that I can see 'what's happening' as PixInsight starts to munch its way through available resources!!).
The next thing that can (sometimes) make a difference is to set two parameters in the ImageIntegration process - the BufferSize and the StackSize. These normally default to BufferSize=16MB and StackSize=1024MB. However, the default options can (will ??) usually result in fairly intensive use of swap files. If you examine the ToolTips on these parameters, you will glean the following:
Set BufferSize = W x H x C x 4, and
Set StackSize = BufferSize x N, where
W = image width, in pixels;
H = image height, in pixels;
C = number of colour channels (e.g. 1 for Mono, 3 for RGB);
N = number of images being stacked.
(the multiplying constant of '4' should be changed to '8' for Linear Fit integration)
Doing this, for an image stack that I was recently working on (30 x RGB images, each 3800 x 2500 pixels) pushed my memory usage up to 80% ov my available RAM, and that was without doing linear fits.
Interestingly, the CPU usage for each of my 4 cores, was pushed hard against the 100% end-stop - demonstrating that PixInsight will happily grab all the CPU resources it can get at. And, because of that, you either accept that PixInsight 'has priority' (and so all other PC activity will be compromised), or you - as a mere human - try and 'get in its way' and make the CPU try to do stuff that PixInsight would rather you left alone.
Finally, as a matter of interest, for Biases, I would have thought that simple 'additive averaging' is all that is required - no statistical pixel rejection is required (nor is any noise estimate required). All pixel values, in all the images, have equal wright, and should be processed as such. And, it would be very questionable whether you would see any improvement in your Biases by processing as many as 250 subs (irrespective of how 'easy' it might be to acquire that many Biases in the first place).
What kind of camera are you using? Does it have TEC cooling? Is the TEC set-point reliable? If so, then you may have absolutely no need for Biases whatsoever, you would only need Darks, Flats and Lights.
Anyhow - I hope some of this helps.
Cheers,
Niall
