15th Feb 2024: Astro Pixel Processor 2.0.0-beta29 released - macOS native File Chooser, macOS CMD-Q fixed, read-only Fits on network fixed and other bug fixes
7th December 2023: added payment option Alipay to purchase Astro Pixel Processor from China, Hong Kong, Macau, Taiwan, Korea, Japan and other countries where Alipay is used.
What's the current guidance about using RAM disks to speed up APP processing? I have 128GB of DDR5 memory on a Ryzen 7950x computer that I just built. I'm integrating a 21 panel mosaic from a QHY600 camera, with each panel already calibrated and processed in APP. These 21 individual fits files take APP (latest version) 50 minutes to register, normalize and integrate to come up with the final mosaic. Not bad, but I would like to halve that time. The majority of that 128GB seems to be free during this process. What if I did this processing in RAM instead of using the SSD? Would that speed this process up substantially?
I'd like to experiment with this. Is there a way to tell APP to use a RAM disk instead of the SSD for processing or is APP only processing in RAM anyway? APP does say that x amount GB is required for processing so that makes me think that a RAM disk would be faster. I know that the finished file is saved to an SSD, but don't know how to direct APP to use a RAM disk for the integration of the mosaic or even if that's possible.
My SSD write speed is 6000 MB/s while the transfer speed of my DDR5 memory runs at 4800MHz. Not sure how that translated to MB/s. Just found this though it's old info: "Modern SSD may achieve real-world write speed of 456 MB/s (mine is 6000 MB/s). Whereas, the speed of RAM is determined by the PC number and a module of PC3-12800 memory can transfer data at the speed of 12,800 MB/s." So maybe RAM is twice as fast as my modern SSD? If so, it seems there is potential for a RAM disk to speed up processing.
Thanks in advance for your thoughts about this.
Rick
I just downloaded Beta 13 and I can confirm that Integration time has been sped up greatly. Also, now, 80 GB of my 128 GB of DDR5 memory are being used instead of less than 40GB with Beta 12. My processing time for my 21 panel mosaic is 28 mins, down from 50 minutes. So Integration time is 2x faster than before! Thanks for this improvement Mabula! Looking forward to your future releases.
Hi @rixon,
Thank you for your question. Yes, if there is enough RAM available APP should be able to do the data integration with a RAM disk instead of read/write on your HDD which is definitely slower.
I have it on my ToDo list that APP will perfrom data integration through RAM directly if enough RAM is available. I have already tested this and I know it is quite a bit faster indeed, so for computers with plenty of RAM like yours, this will increase APP speed big time !
I just downloaded Beta 13 and I can confirm that Integration time has been sped up greatly. Also, now, 80 GB of my 128 GB of DDR5 memory are being used instead of less than 40GB with Beta 12. My processing time for my 21 panel mosaic is 28 mins, down from 50 minutes. So Integration time is 2x faster than before! Thanks for this improvement Mabula! Looking forward to your future releases.
Great, thanks for the feedback !
Mabula
@mabula-admin - is this implemented now (2beta17) ?
Ah, thanks. Just FYI, I tried it with a 32GB RAM disk on a 120 image (4000ish x 2500ish) integration, and it only made a nominal (maybe 15 seconds over a 10 minute run) difference even though the RAM drive was 5x as fast as the SSD hard drive.
Hi, something must be off in how you configured things then @heylers, I know from testing myself that integration through a ramdisk with new code that I am developing is much faster provided you have enough RAM for all integration data 😉
Please note, this speed increase only applies for stacking the calibration masters and the light master. Steps 3) analyse stars to 5) normalize will not be affected.
Mabula
