ZWO ASI6200 - Processing 60+ megapixel images.

Robert Q. Kimball

Well-known member
I just purchased a ZWO ASI6200 mono camera. I have a MacBook Pro. I'm scared to death that I won't be able to process my subs in Pixinsight. does anyone one have any experience with these large images?


Screen Shot 2020-07-29 at 5.32.11 PM.png
 
Don't worry, it will work. It will just be long.
I have a ASI6200MC and I use Pixinsight on MBP 15 2019 core i9 32Gb : it rocks. For sure it is sometimes long but no real issue.
Mono camera will take less memory & space than color camera. So, your configuration will run.

But, I want to understand something...
you have the camera...
you have the MBP...
you have Pixinsight...
You don't want to make a try ? you think you will kill your MBP if a 60Mp image jump inside ????

If you don't have yet some "astro" images, take bias or Dark and try to integrate some Masters with maybe 50 of each. You will see how long it takes to integrate.
 
You should have no issues, I use an i5 with 16gb RAM and never had an issue...it just takes a bit longer (more time for whisky) ...

John
 
Don't worry, it will work. It will just be long.
I have a ASI6200MC and I use Pixinsight on MBP 15 2019 core i9 32Gb : it rocks. For sure it is sometimes long but no real issue.
Mono camera will take less memory & space than color camera. So, your configuration will run.

But, I want to understand something...
you have the camera...
you have the MBP...
you have Pixinsight...
You don't want to make a try ? you think you will kill your MBP if a 60Mp image jump inside ????

If you don't have yet some "astro" images, take bias or Dark and try to integrate some Masters with maybe 50 of each. You will see how long it takes to integrate.

My camera is on order but still in China! I started wondering if I'd need to buy a new computer. My "wife" equity is completely spent. Thanks for your answer. Also good questions. If the camera was here, I'd plug it in and watch for smoke and flames.
 
FWIW this is an area I feel like PixInsight could improve, by doing it all in memory like other competitive stacking programs. I really like some PixInsight features especially large scale rejection but CCDStack is just so much faster for the same SNR and FWHM.
 
I am not a Windows fan but my work laptop is an old i5 with 16gb RAM and Windows10 - and things were dragging. I recently got a Desktop PC, again i5 with 16gb RAM and loaded Kubuntu Linux and WOW the performance difference is astounding. StarAlignment, ImageIntergration and DrizzleIntegration really fly....
 
FWIW this is an area I feel like PixInsight could improve, by doing it all in memory like other competitive stacking programs. I really like some PixInsight features especially large scale rejection but CCDStack is just so much faster for the same SNR and FWHM.

?? PI does do it all in memory. ImageIntegration has explicit controls for how much ram to use when stacking. it is impossible to keep everything in memory all at once on most PCs, so either the OS's virtual memory system has to manage memory or the application does.
 
If I am not wrong, Juan implemented the XISF format for efficient access of such large datasets while performing out-of-core computations. The FITS format cannot support that (see for example https://arxiv.org/abs/1502.00996). We
are already in the era of big data amateur astrophotography. How prepared are the other applications to address the challenges of this era?
 
Last edited:
fits does support incremental reading. the problem are other formats like CR2 which have to be entirely loaded into memory in order to be integrated. for the most part ImageIntegration has always been able to manage memory on its own. but when using non-incremental formats like CR2 PI just has to load them all into memory and hope that the OS can handle that efficiently. linux is particularly good at this, macosx less so and windows can sometimes be terrible at this kind of memory management.
 
?? PI does do it all in memory. ImageIntegration has explicit controls for how much ram to use when stacking. it is impossible to keep everything in memory all at once on most PCs, so either the OS's virtual memory system has to manage memory or the application does.
PixInsight only does the final integration in memory. PixInsight requires you write all the calibrated and registered files to disk. This means its several times slower than similar programs as well as a lot of write cycles on the SSD. I have no problem keeping everything in memory. My computers are either 32GB or 64GB of RAM and use 1TB or 2TB Samsung 970 M.2 drives.
 
as we've seen from a lot of threads here, most people do not have enough ram to keep an entire integration in memory at once. even if PI were rewritten to hold everything in ram, the same thing would happen - the OS would have to page stuff out which results in writes to the SSD.

i agree though that when you don't exceed the size of the ram then the disk writes could be avoided. PI's calibration/alignment/integration processes are discrete though and it would be a pretty big rewrite to merge them in such a way that they could communicate with each other entirely in ram.
 
I wonder if it is possible to have a pass of computation where no altered data are stored, besides metadata, like the .xdrz or .xnml files. In this way we would separate the "compute" part from the "apply" part and if possible, perform the "apply" part chunked, during integration.
 
Back
Top