I have a new fancy pants 24-core 14th gen i9 laptop with a fast 4TB NVME drive. I’m enjoying SIRIL’s speed, with it stacking through about 50 24-megapixel lights a minute.
However, this computer’s CPU is about 2x the speed of my older computer (based on the average of a series of multi-core benchmarks) but SIRIL is “only” about 1.4x faster. I expect it’s because the SSD is the same and it may be disk speed limited. But it got me thinking: My old computer had 64 Gigs of RAM and this has only 16 Gigs as it’s new and I haven’t considered upgrading it yet.
Would adding more RAM to my computer improve SIRIL’s speed even further?
I’m not sure how much SIRIL takes advantage of additional RAM for performance purposes.
I have and it does allocate almost all the RAM but it also did when I had 64 gigs of RAM. So it seems to allocate what’s available, but it’s hard for me to know how much it’s utilizing it and what the performance impact is.
So for me, it’s hard to know the difference between just allocating and utilizing it and whether it provides much of a performance benefit.
in general the limiting equipment is the drive. Siril will use as much memory as it can indeed, you could check what is the CPU usage. If there is not enough memory to handle all images in parallel, some cores may not be used (it’s also written in the console) and if there is much wait time for I/O, CPU usage will be far from 100% too. More memory doesn’t hurt but it depends on many factors to know if it’s useful.
That’s a good thing to check and what I found that during most processing steps, it’s pretty much 100%, so the amount of RAM doesn’t seem to be holding up the CPU. Disk speeds are up to 2.5GB/sec so I expect I’m approaching the limit on that because for sustained writing, many SSDs slow down from their peak to 1-2GB/second range.
calibration and registration are the hardest on disk because of the writes, stacking should always be 100% most of the time, so if it’s near 100% also in those operations too, I’m not sure adding memory will help much, especially with windows. But with 24 cores it’s surprising it’s able to task all that with only 16GB of memory. Are your images monochrome?
They are not, they are 24-megapixel OSC shots, but I’m using a script that sets bit-depth to 16-bit for everything except the stacking stage, where I set it to 32-bit. Perhaps the lower bit depth is helping.