PixInsight 1.8.8-7 Released

Juan Conejero

PixInsight Staff
Sep 2, 2004
8,056
302
57
Valencia, Spain
pixinsight.com
Hi all,

I am glad to announce a new version of the PixInsight core application: 1.8.8-7. This is an important maintenance release providing bug fixes, new features and stability improvements on all platforms.

Installation packages for PixInsight 1.8.8-7 are now available to all licensed users for Linux, macOS and Windows:

https://pixinsight.com/dist/

A FreeBSD version is still not available. We'll try to build it as soon as possible during the next weeks.

As happens with all updates to the PixInsight core application, this version cannot be downloaded as an application update. You have to install version 1.8.8-7 manually:
  • FreeBSD and Linux: The installation packages are xz-compressed tar archives (.tar.xz). To install one of these archives, just extract its full contents on any directory, preferably on one located under your home directory. Then become root and run the installer program from a terminal. Normally you should just type "yes" and press Enter to accept the default settings. Say './installer --help' for detailed information on command line arguments.

  • macOS: If you have a previous version of PixInsight already installed, you must uninstall it by moving the /Applications/PixInsight folder to trash (also known as bin on macOS 10.15 and later). Run the .pkg installation package and follow the instructions. All our macOS installation packages are digitally signed with our corporate Apple developer certificate and have been notarized by Apple.

  • Windows: If you have a previous version of PixInsight already installed, please uninstall it from Control Panel or Add or Remove Programs. Then run the installation package for version 1.8.8-7 and follow the instructions. All executable files in the PixInsight distribution for Windows, as well as our installation packages, have been signed with our corporate Extended Validation (EV) code signing certificate.
If you have version 1.8.6, 1.8.7, or any 1.8.8-x version already installed, this new version does not require a license reactivation. If you are still using a 1.8.5 version or older, see the official 1.8.6 version announcement for a detailed description of our new licensing system. See also FAQ 2.10 for additional information on license reactivations.


Main Improvements

This version is an important maintenance release focused on performance and stability improvements, paving the way towards PixInsight 1.8.9. Version 1.8.8-7 fixes most confirmed bugs since the initial 1.8.8 release and provides some relevant new features. A partial list follows with the most important changes.


New APASS and Gaia Processes with Local XPSD Database Files

XPSD (eXtensible Point Source Database) is a new database format that we have developed for fast and efficient access to large astrometric and photometric star catalogs. We already released a Gaia DR2 local database for the previous version 1.8.8-6 of PixInsight. With version 1.8.8-7 we are releasing four new XPSD local databases, including the AAVSO Photometric All-Sky Survey (APASS) DR9 and DR10 catalogs, plus the Gaia EDR3 and DR2 catalogs, along with two associated modules, APASS and Gaia, which work as XPSD catalog servers for all of our astrometry and photometry tools and scripts.

The new XPSD database files are now available for download on our software distribution system. There are single database files for APASS DR10 and DR9, while the Gaia EDR3 and DR2 databases require 16 files each. Unless you want to perform some comparative analysis between different catalog data releases, you normally should only download the Gaia EDR3 and APASS DR10 databases, that is, the latest data release available for each survey. Please note that the Gaia DR2 files that we released for version 1.8.8-6 are incompatible with the new Gaia process in version 1.8.8-7.


See the announcement post for the old Gaia DR2 database and process, where you'll find detailed information about these processes and instructions to install and use the XPSD local databases. The information given for the old GaiaDR2 process is applicable to the new APASS and Gaia processes in version 1.8.8-7.

High-performance local star databases play now a crucial role in PixInsight, making possible a new generation of astrometry and photometry based tools and scripts, such as astrometric image registration and mosaic construction, photometry-based image normalization, and positional astronomy applications along with our integration of solar system ephemerides, just to name a few relevant possibilities. As you'll see below, the new versions of our ImageSolver and AperturePhotometry scripts can now work with the Gaia and APASS processes to generate astrometric solutions and photometric analyses with unprecedented accuracy, efficiency and flexibility.


ImageSolver Script version 5.5.0

The new version is an important step forward in our astrometry engine. Here are some of the most relevant improvements:
  • Improved distortion models. New surface simplification algorithm for generation of thin plate splines, where an optimal simplifier tolerance in pixels is found iteratively in an automated fashion. This changes the meaning of the simplifier tolerance parameter, which is now an initial value for the iterative search algorithm with a default value of 0.05 pixels (previously 0.25px). This significantly improves accuracy of our spline-based distortion models, which are now completely adaptive.

  • More accurate thin plate splines. The default thin plate spline smoothing is now 0.015. This is in general more appropriate than the previous value of 0.025, given the increased accuracy of our distortion models.

  • Better support of low quality data. New minimum structure size parameter (Advanced Parameters section). This is the size in pixels of the smallest image structure that can be considered as an alignment star. By increasing this parameter one can exclude small, spurious image structures that can cause the initial image registration step to fail under very unfavorable conditions. This mainly includes groups of hot pixels and sensor defects when the field includes very few detectable stars. This may help when working with poor quality data (poor focus, poor tracking, wrong calibration, extremely low SNR, etc).

  • More aggressive noise reduction for improved support of low SNR data. The noise layers parameter is now deprecated. It has been replaced with a new noise reduction parameter (Advanced Parameters section), which is the radius in pixels of a Gaussian convolution filter applied to the working image used for calculation of star positions during the star detection phase. This parameter should only be used for data with very low SNR, or data of very poor quality.

  • Enhanced XPSD local database support. ImageSolver now uses the new unified Gaia XPSD server with support for Gaia EDR3 and DR2 local databases.

  • High accuracy star positions. When one of the Gaia XPSD catalogs is selected (EDR3 or DR2), our astrometric solutions are now computed using astrometric star positions with corrections for space motion (proper motion and parallax) and relativistic deflection of light. This feature makes use of our integration of fundamental solar system ephemerides (currently DE438/LE438).

AperturePhotometry Script version 1.5.0
  • XPSD local database support. New APASS DR9 and DR10 catalogs based on XPSD local databases and the APASS XPSD server process.

  • New automatic limit magnitude parameter (Stars section). When this parameter is true and an XPSD server catalog is selected (such as an APASS or Gaia XPSD catalog), the script finds an optimal limit magnitude automatically using fast XPSD server search operations. In such cases the maximum magnitude parameter will be ignored.

  • Redesigned/refactored script dialog. Fixed all dialog layout problems.

AnnotateImage Script version 2.1.8
  • New VizieR catalog: LBN (Lynds' Catalogue of Bright Nebulae, Lynds 1965), with 1125 objects.

  • New VizieR catalog: LDN (Lynds' Catalogue of Dark Nebulae, Lynds 1962), with 1791 objects.

WeightedBatchPreprocessing Script version 1.5.2
  • Added new image registration parameters to provide more control over the StarAlignment process: minimum structure size, noise reduction, sensitivity, peak response, maximum distortion. This should facilitate working with poor quality data (low SNR, poor tracking, poor focus, wrong calibration, etc).

  • Graceful management of light frames measurement errors. Unmeasured light frames are now discarded.

  • Process logger: readability improved, frame measurement errors reported.

  • Added new global option Separate CFA flat scaling factors. This option is enabled by default, with the same behavior as its counterpart parameter of the ImageCalibration tool.

PhotometricMosaic Script version 3.0

This excellent, comprehensive and remarkably useful script, written by PTeam member John Murphy, is now part of the official PixInsight distribution along with its two auxiliary scripts, SplitMosaicTile and TrimMosaicTile. These scripts come with complete documentation, also written by the same author.


RAW File Format Support Module
  • RAW format now based on the latest LibRaw 0.20.2 Release version.

  • The RAW format can now detect and decode time offset values from EXIF metadata (EXIF tags OffsetTime, OffsetTimeOriginal and OffsetTimeDigitized). This means that for raw digital camera frames that support these EXIF tags the reported DATE-OBS FITS keyword and Observation:Time:Start XISF property are guaranteed to be in the UTC time scale (provided that the camera used has been properly configured).

  • The RAW format can now detect and decode GPS geodetic coordinates, longitude, latitude and height, from available EXIF metadata. Raw frames of digital cameras supporting GPS EXIF tags contain now geodetic coordinates that can be used to compute topocentric positions for accurate astrometry applications and solar system image annotations.

XISF File Format Support Module
  • The XISF format reports now that incremental read/write operations are disabled when compression is enabled. This fixes some critical memory usage problems that have been reported by some users.

PhotometricColorCalibration: XPSD Local Database Support

The PhotometricColorCalibration tool (PCC) makes now use of the new APASS and Gaia XPSD local databases when available. A new automatic catalog photometry parameter will select the XPSD server for the APASS DR10 or DR9 local databases when available.


In addition, when XPSD local databases are available, optimal limit magnitudes are calculated in a completely automated way for astrometry and photometry, which provides optimal color calibration results very easily. With XPSD databases, the PCC tool is basically an automatic process where you just have to execute it when the appropriate metadata are present in the FITS header and/or XISF properties of the target image.

With the availability of APASS DR10, with 128 million stars (compared to 62 million of APASS DR9), PCC now provides much more accurate and consistent white balancing results.


Geometry Processes: Metadata Better Preserved

The Resample, IntegerResample, Rotation, DynamicCrop, Crop and FastRotation processes preserve image scale metadata as much as possible. Existing astrometric solutions are of course destroyed when one of these processes is applied, but the metadata items (FITS keywords and XISF properties) defining image scale and center coordinates are now preserved and/or transformed as possible to facilitate subsequent astrometric solutions. For example, if you resample an image 2:1, its XPIXSZ and YPIXSZ FITS keywords, as well as its Instrument:Sensor:XPixelSize and Instrument:Sensor:YPixelSize XISF properties, will be recalculated with new values consistent with the enlarged image scale.


StarAlignment: New Minimum Structure Size Parameter


This parameter defines the minimum size of a detectable star structure in square pixels. It can be used to prevent detection of small and bright image artifacts as stars. This can be useful to work with uncalibrated or wrongly calibrated data, especially demosaiced CFA frames where hot pixels have generated large bright artifacts that cannot be removed with a median filter (such as the hot pixel removal parameter).

Changing the default zero value of this parameter should not be necessary with correctly acquired and calibrated data. It may help, however, when working with poor quality data such as poorly tracked, poorly focused, wrongly calibrated, low-SNR raw frames, for which our image registration algorithms and tools have not been designed specifically.


Important Changes in Color Management Initialization

In this version we have implemented important changes in the way our color management engine is being initialized upon application startup. A new detect monitor profile parameter is now available and enabled by default. When this parameter is enabled, PixInsight will detect a primary system monitor profile on macOS and Windows each time the core application is executed. This facilitates the automatic adaptation to changes in system color management settings, such as a new monitor profile generated by a screen calibration device. The detect monitor profile parameter can also be disabled to select a primary monitor profile manually, as in previous versions, although this is not recommended on systems with reasonable color management configurations.

Another, probably more important change, is in the way the RGB, gayscale and proofing profiles are now selected by default. In previous versions of PixInsight, all of these working profiles were set equal to the monitor profile by default. This behavior can be (and often is) wrong, since a monitor profile characterizes a physical display device, but does not necessarily have to define a suitable working color space. Since PixInsight core version 1.8.8-7, the RGB, grayscale and proofing profiles are initialized to the standard sRGB color space by default. This is a more consistent and more generally valid behavior.


All users are encouraged to reset their color management settings after installing version 1.8.8-7 of PixInsight. Then the default RGB and grayscale profiles, as well as the proofing profile if necessary, should be redefined manually if sRGB is not the desired working color space.


Preferences: New File I/O Thread Control Settings


The maximum number of threads allowed for concurrent file reading and writing operations can now be controlled globally with dedicated preferences settings. Tasks that perform heavy file I/O operations, such as ImageIntegration, Debayer or StarAlignment for example, can work more efficiently on solid-state drives (SSD) when parallel file access is allowed.


The new maximum number of file reading threads and maximum number of file writing threads parameters are now available in the Parallel Processing and Threads section of the Preferences tool. Not all processes can use these settings, but only those that can be applied to sets of disk files as batch tasks. For now this includes ImageIntegration (for reading only), StarAlignment, LocalNormalization, and Debayer. In a future version ImageCalibration and other tools (some of them new) will also be included in this list.

The default value for these two parameters is 1 thread. This is obviously a conservative value intended to prevent problems with rotational hard disks, which are still being used by many users. The performance improvements that can be achieved with SSD devices can be interesting, but the performance degradation that can result from heavy parallel I/O accesses on rotational disks can be disastrous, which is the reason why parallel disk file operations are disabled by default.

If you use SSD disks to store the data that you process in PixInsight, you can increase the values of these parameters and check if you see some performance improvements. Typical values are in the range from 4 to 16 threads. Whether you'll see some significant improvements or not depends on many factors, including but not limited to the type of disks, the motherboard, the amount of RAM available, the filesystem used, and the operating system.

Please note that these parameters have nothing to do with, and are completely unrelated to, swap storage directories (Directories and Network section of Preferences).


Vectorization Improvements

You'll probably (hopefully) experience that this version of PixInsight is significantly faster than previous ones. The reason is in the hard work we have done to make much more and much better use of SIMD processor instructions. This work has been focused on achieving efficient vectorization in critical parts of our code base. The increased level of parallelism is much more evident on machines with 16 processor cores and more, but significant performance improvements have been verified even on 4-core processors, on all platforms.

Our initial plan was to release two versions of PixInsight 1.8.8-7 on each operating system (hence six different versions): an official version with AVX2 and FMA3 processor instruction support, and a compatibility version with just SSE4.2 requirements. Finally, after much testing and benchmarking work, we have decided to release the compatibility version only, mainly because the increased complexity of having to maintain two versions for each platform is not justified by relatively small speed improvements. So for now we require just SSE4.2. This will probably change in the near future, when we work harder to improve the performance of our code using AVX2/FMA instructions.

Anyway, to get a general idea of the performance improvements we have achieved—although not a precise evaluation—, perform some benchmarks and compare with 1.8.8-6.


GPU Acceleration

It is now official: We are working on the implementation of GPU-accelerated versions of critical tasks and algorithms in our code base, and we are investing significant human, time and hardware resources in this task. For now this includes very basic elements, such as convolutions and multiscale transforms, among others, since these low-level algorithms are critically important for many tools and processes on the PixInsight platform. The performance gains that we are seeing in our initial implementations are promising. I am not going to write numbers here, mainly because we still have no definitive results and anything that I can say here could be misleading, but I'll just say that we are achieving some spectacular results.

If you look carefully you'll see two new options in the Parallel Processing and Threads section of the Preferences tool:


For now we are basing GPU acceleration on the NVIDIA CUDA platform exclusively, for a variety of practical and performance reasons. However, we are designing and building our implementation using abstraction layers that will allow us to extend it to other platforms, such as Metal and OpenCL, relatively easily without costly structural changes.

We expect to have first GPU-accelerated versions of the core PixInsight application and some tools during the first half of 2021. Roberto Sartori, who is now working with us as part of our development staff, is leading the PixInsight GPU acceleration project.


Important Bug Fix on Windows

We have identified a bug in the Windows version of PixInsight that is now fixed in version 1.8.8-7:


Note that this is actually a documentation bug, that is, Qt's documentation was not warning us about a potentially dangerous operation that everybody was considering as something perfectly legal and normal in Qt-based code. Now that we have fixed all instances of this operation in our code (namely, disconnecting all signals connected to an object), we have not been able to reproduce any of the crashes reported in previous versions.


Updated Support Libraries

*
Qt updated to version 5.15.2 LTS on Linux, macOS and Windows.

* Clang version 12.0.0 with the macOS 11.1 SDK used for all macOS development.

* Microsoft Visual Studio 2019 version 16.8.3 used for all Windows development.

* LibCURL updated to version 7.73.0


Recommended Platforms and Supported Operating Systems

The reference implementation of PixInsight, where you may expect the maximum performance and best user experience, is the Linux version. Currently our primary development platforms are:
On FreeBSD and Linux we strongly recommend running PixInsight on the KDE Plasma desktop environment.

PixInsight 1.8.8-7 for Linux requires GLIBC 2.27 or a newer version.

PixInsight 1.8.8-7 for macOS is only compatible with macOS 10.14, 10.15 and 11.x Big Sur. macOS 10.13 and older versions are no longer supported. All of our macOS installation packages are digitally signed with our corporate Apple developer certificates and have been notarized by Apple.

PixInsight 1.8.8-7 for Windows is compatible with Windows 10 exclusively. Windows 8.1, Windows 7 and older versions are not supported. The application might run on Windows 7, but there is no guarantee. All of our Windows installation executables have been digitally signed with our corporate extended validation (EV) code signing certificate.


Known Issues

FreeBSD and Linux
  • Recent versions of the KDE Plasma desktop environment include a morphing popups desktop effect that causes problems with recent PixInsight versions. If this effect is enabled, it slows down some dragging operations in the PixInsight core application, such as dragging icons and view selectors. You should disable this effect, which is very easy from KDE's System Settings panel.

  • On recent versions of the KDE Plasma desktopenvironment, a display tearing prevention option (v-sync) is enabled by default. This option can cause input lag problems (e.g., slow cursor movement over image windows). This option should be disabled for optimal PixInsight performance:
    • System Settings > Display and Monitor > Tearing prevention: set to 'Never'
    • Click Apply
Linux
  • PixInsight is not compatible with the open-source Nouveau graphics driver. If your distribution is using Nouveau, you must remove and replace it with the proprietary Nvidia driver in order to use PixInsight.

macOS
  • On Apple MacBook laptops with dual graphics cards, the integrated Intel HD graphics driver has been causing minor usability issues on macOS 10.14 Mojave. Apparently these problems have been fixed in macOS 10.15 Catalina. However, for maximum performance and stability, we recommend disabling the integrated Intel HD graphics before running PixInsight on MacBook laptops. This can be done very easily:

    System Preferences > Energy Saver > Automatic graphics switching: disable.

Windows
  • Intel HD graphics drivers have been causing serious problems on Windows. Hopefully most of these issues have been resolved once we no longer depend on desktop OpenGL drivers by default.

  • On some laptops with nonstandard screen resolutions, such as the Microsoft Surface, PixInsight cannot find valid screen scaling factors automatically. On these machines, you may have to run the PixInsight core application with the --ui-scaling command-line argument. For example, to apply a UI scaling factor of 2:
C:\Program Files\PixInsight\bin\PixInsight.exe --ui-scaling=2

You only have to do this once, since the scaling factor will be stored in application preferences automatically. Of course, if you already are using version 1.8.6 and the interface looks correct, then you don't need to do this; this is only required for new installations.​

All Platforms
  • The StarNet module uses the TensorFlow library, which requires a processor with support of AVX2 and FMA instructions. This excludes very old machines with processors released before 2013. If your computer uses one of these old processors, PixInsight 1.8.8-7 won't work as it is installed by default. However, this problem is very easy to fix: just remove the StarNet module. Do the following:
Linux
Enter the following command from a terminal:​
sudo rm /opt/PixInsight/bin/StarNet-pxm.so
macOS
Remove the following file:​
/Applications/PixInsight/bin/StarNet-pxm.dylib
Windows
Remove the following file:​
C:\Program Files\PixInsight\bin\StarNet-pxm.dll
In all cases you need administrative privileges to remove the required file, which you also need to install PixInsight. After this simple action, PixInsight will run without problems on your old machine.​

—————————————

Thank you for your attention.
 
  • Like
Reactions: Yannis Doukakis

KenM

Well-known member
Dec 25, 2016
54
0
48
Belgium
Nice, lots of things to try, I did a benchmark before and after, I got a CPU time of 15.26 s to 13.63 s about +12% 'for free' :)
(That is basically equivalent to an upgrade from 16 cores to 18 cores on the old version, or 8 to 9 cores)

Will there be an extra benchmark that is more focused on GPU acceleration or is the current one already covering this?

About the tensorflow lib using GPU instead of AVX2 is that something we have to configure during tensorflow install or is that a choice of PI?

P.S. Ambitious to not only support CUDA (y). I need CUDA for other software so makes no difference for me though.

Edit: my Ubuntu (dual boot) seems to have a bigger jump, I have no before number though.
I guess I will hold on to the top score until the Ryzen 5000 and TR starts to join the benchmark of the new version.
Execution Times
Total time ............. 00:15.21
CPU time ............... 00:12.91
Swap time .............. 00:02.24
Swap transfer rate ..... 7387.634 MiB/s
Performance Indices
Total performance ...... 30926
CPU performance ........ 29313
Swap performance ....... 40917
 
Last edited:

Juan Conejero

PixInsight Staff
Sep 2, 2004
8,056
302
57
Valencia, Spain
pixinsight.com
Nice, lots of things to try, I did a benchmark before and after, I got a CPU time of 15.26 s to 13.63 s about +12% 'for free'
Really nice, but the actual performance improvements are greater than that. The current benchmark script does not make justice to the improvements we have implemented.
 

astrogruff

New member
May 11, 2020
2
0
With the concurrent disk access does this also effect swap directories? Just that I have one swap directory on a HDD while the other is on a PCIe4 M.2 SSD - I'm happy to ditch the swap on the HDD if I get the benefit of the extra thread processing on the SDD.
 

Juan Conejero

PixInsight Staff
Sep 2, 2004
8,056
302
57
Valencia, Spain
pixinsight.com
The new maximum file ... threads options have nothing to do with swap files. They refer to image files that tools performing batch operations read or write, including ImageIntegration, StarAlignment, etc. So you can of course continue using your HDD to store swap files (although you should put it only once in the list of swap directories) and increase the number of file I/O threads, as long as the files that you are processing are in your SDD.
 

rbotero

Well-known member
Jan 19, 2005
57
12
Kent, England
www.astrobin.com
Fantastic release Juan! Thank you and to the PTeam for all your work. I can tell that my processing is faster already and my benchmark results have gone up. I am so glad you included the Photometric Mosaic in this release. As a regular user of this fantastic script since early summer, I really appreciate it being included in the general release.
Thank you again and best wishes for Christmas and 2021!
Roberto
 
  • Like
Reactions: jmurphy

USSFE8

Member
Nov 9, 2020
5
0
Hello,
Not being a programmer, does “NVIDIA CUDA platform exclusively” mean that only NVIDIA GPUs are supported when GPU acceleration goes live? Will AMD GPUs be supported at some point? Will the proprietary divers be required in Ubuntu 20.04? Thanks, and have a wonderful Christmas!

Eric
 

Juan Conejero

PixInsight Staff
Sep 2, 2004
8,056
302
57
Valencia, Spain
pixinsight.com
Not being a programmer, does “NVIDIA CUDA platform exclusively” mean that only NVIDIA GPUs are supported when GPU acceleration goes live?
Yes, our initial GPU acceleration implementation will only support NVIDIA hardware.

Will AMD GPUs be supported at some point?
That's our intention in the medium-long term, but development on the CUDA platform will be prioritized.

Will the proprietary divers be required in Ubuntu 20.04?
The proprietary NVIDIA driver is always required on Linux, irrespective of CUDA support. PixInsight is not compatible with the nouveau open-source driver.
 

USSFE8

Member
Nov 9, 2020
5
0
Yes, our initial GPU acceleration implementation will only support NVIDIA hardware.



That's our intention in the medium-long term, but development on the CUDA platform will be prioritized.



The proprietary NVIDIA driver is always required on Linux, irrespective of CUDA support. PixInsight is not compatible with the nouveau open-source driver.
Thanks for that info Juan. Are you able to approximate the percent performance increase which may result from GPU acceleration should I switch to an NVIDIA GPU? I’m running PI in Ubuntu 20.04 on a newly built 3990x machine, but haven’t decided on a final GPU. I currently have a AMD RX5700XT installed (with the AMDPRO proprietary Linux driver) which is why I asked my initial question. Would a higher end NVIDIA gpu such as a RTX3080 get me significant performance increases? Thanks for you time Juan. It’s greatly appreciated. Here are my current benchmark indices for reference.

Reference BM S/N 52QW1ESV27208F8V60002T7N67JOLCR7

Execution Times
Total time ............. 9.4s
CPU time .............. 8.12s
Swap time .............1.22s
Swap transfer rate .....13576.5 MiB/s
Performance Indices
Total performance ......50027
CPU performance ........46624
Swap performance .......75194
 

KenM

Well-known member
Dec 25, 2016
54
0
48
Belgium
I have 3950X with 2070 Super and with cuda on or off there is (for now) no difference in the benchmark but it's probably not a good test for the currently limited GPU usage.
If I was you I would choose an Nvidia version with CUDA but wait until they release the 16GB/20GB version (unless your computers have a short live cycle)
BTW Don't you think your PC is already fast enough for PI ;)
 

USSFE8

Member
Nov 9, 2020
5
0
Howdy Ken,
Thanks for the info. I’ll definitely wait to make a choice until I hear how folks benchmarks’ change when they turn on the gpu acceleration feature. I’m glad I waited on the gpu purchase. I would have simply went with AMD since they’re more Linux friendly. At least that’s what I’ve read. I’m no Linux expert. This does give a good reason for an NVIDIA though even if the gpu acceleration performance increase is minimal, which isn’t what I expect. They’re slightly better cards anyhow. I’ve just read about a few compatibility issues between them and Linux. Not that AMD is perfect. I had to roll my last Ubuntu version back a kernel in order to get the latest AMD proprietary driver to install correctly. My Linux partition (really a separate SSD) was created simply for PI. 100% worth the effort. The performance increase is astronomical compared to Windows 10, and not just benchmarking. It’s good to know you’ll be marking with a 2070. I’ll keep an eye on 3950x performance indices when gpu acceleration goes live.

Ha! Yes, it’s fast enough for PI. I’m relatively new to AP as a hobby. I’ve been using PI for a number of months now. An amazing tool that demands either extreme patience or number crunching power. I have little patience so I built this PC simply to reduce long processing times to something more reasonable for experimentation. I met my goal to be sure. Pricey, yes, but so is everything else in this hobby. I’m hoping Juan can share some insight, no pun intended, regarding approximately how much performance increase one can expect with gpu acceleration.

Thanks again for the info.
Eric
 

Juan Conejero

PixInsight Staff
Sep 2, 2004
8,056
302
57
Valencia, Spain
pixinsight.com
Execution Times
Total time ............. 9.4s
CPU time .............. 8.12s
Swap time .............1.22s
Swap transfer rate .....13576.5 MiB/s
Performance Indices
Total performance ......50027
CPU performance ........46624
Swap performance .......75194
Aha! So you are the guy with a 50,000 benchmark 😈

Congrats on a truly fast box!

Are you able to approximate the percent performance increase which may result from GPU acceleration
Of course I have numbers, but sorry, I'm not going to make them public for now, at least until we have something real to show. Suffice it to say that the performance improvements that we are achieving are quite important, and that's just with preliminary implementations where we are just exploring how CUDA performs with some basic and relatively easy to transport algorithms. So yes, GPU acceleration will be an important step forward for sure.
 

USSFE8

Member
Nov 9, 2020
5
0
Juan,

Thank you sir. I built it simply because I’m not a patient man ;)

Excellent. Although not a numeric response, that gives me the direction I was asking for. I’m looking forward to the day you make some numbers public.

Regards,
Eric
 

aworonow

Well-known member
Feb 19, 2013
342
12
faintlightphotography.com
Well, I've now spent several days trying to download the EDR3 files. The downloads are very slow, usually around 500KB/s for me. And before a file is down, it aborts (sometimes with 5% or less downloaded, sometimes with 60% downloaded). I have a slow internet, true, but I don't have this problem with other sites. I think it is logging me out before the download completes...any suggestions? I exchange large data sets in DropBox without this problem.
Alex

Got it all done. The longer time-out helped considerably, as did waking up at 4:00 before the internet got swamped. Living in rural America is an internet nightmare!
a
 
Last edited:

aworonow

Well-known member
Feb 19, 2013
342
12
faintlightphotography.com
Thanks, Juan, I will try it when a little later when I don't have to compete with all of Las Angeles on Verizon. Internet is down to 300KB/s tonight, as it often is at this time. Speeds up to 1MB/s or better in a few hours.
 

Phillip Klein

Member
Feb 6, 2015
8
0
Hello Juan,

Thank you for all the work on the recently released PixInsight version 1.8.8-7 which I downloaded. Per your recommendation I also download the Gaia EDR3 and APASS DR10 databases, which I understand are needed for plate solving which is essential for mosaics, something I have recently started. I am unsure if the downloaded files are accessible to the program the way I have listed them. When I right clicked on the downloaded files they would not open so I entered their file path. Please see attached photos. I appreciate your guidance to ensure the plate solving function works as you intended.
Thank you,
Phillip
 

Attachments