cosmic clarity benchmark

 The mysteries of closed source; Nvidia, CUDA and stuff. It looks as if it's working. No idea why though!
Anyway, what used to take a few minutes, takes a few seconds instead. 

[

    {

        "timestamp": "2025-11-03 20:31:12",

        "system": {

            "os": "Linux 6.16.3-76061603-generic",

            "cpu": "Intel(R) Core(TM) i5-14400F",

            "ram": "62.6 GB",

            "gpu": "NVIDIA GeForce RTX 5070",

            "cuda_available": false,

            "mps_available": false,

            "onnx_providers": [

                "TensorrtExecutionProvider",

                "CUDAExecutionProvider",

                "CPUExecutionProvider"

            ]

        },

        "benchmarks": {

            "cpu": {

                "mad_single_core": {

                    "first_run": 3215.15,

                    "avg_time": 2372.23

                },

                "flat_field_multi_core": {

                    "first_run": 957.46,

                    "avg_time": 730.46

                }

            },

            "gpu": {

                "pytorch_cuda": "Avg: 0.51 ms | Total: 496.03 ms",

                "onnx_cpu": "Avg: 831.44 ms | Total: 808157.56 ms",

                "onnx_cuda": "Avg: 20.58 ms | Total: 20000.70 ms",

                "onnx_openvino": null,

                "onnx_rocm": null

            }

        }

    }

]

setiastrosuitepro desktop icon

To make a clickable and open-able-with entry:

Create a file with this content and save it to:
~/.local/applications

I used the icon supplied by setiastrosuitepro and called it sasp...

#!/usr/bin/env xdg-open
[Desktop Entry]
Comment[en_US]=sasp
Comment=
Exec=cd /home/steve/setiastrosuitepro && . venv/bin/activate && ./venv/bin/python setiastrosuitepro.py %F
GenericName[en_US]=sasp
GenericName=
Icon=/home/steve/setiastrosuitepro/astrosuite.png
MimeType=
Name[en_US]=sasp
Name=sasp
Path=/home/steve/setiastrosuitepro
StartupNotify=true
Terminal=true
TerminalOptions=\s--noclose
Type=Application
X-KDE-SubstituteUID=false
X-KDE-Username=

cuDNN install Pop!_OS 24.04

Sketchy inforrmation and loadsa misleading stuff, but anyway, here goes...

1. Find your CUDA version

2. Find the code you need from here:

https://developer.nvidia.com/cudnn-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=24.04&target_type=deb_local

Click the matrix to get the code:

$ wget https://developer.download.nvidia.com/compute/cudnn/9.14.0/local_installers/cudnn-local-repo-ubuntu2404-9.14.0_1.0-1_amd64.deb
$ sudo dpkg -i cudnn-local-repo-ubuntu2404-9.14.0_1.0-1_amd64.deb
$ sudo cp /var/cudnn-local-repo-ubuntu2404-9.14.0/cudnn-*-keyring.gpg /usr/share/keyrings/
$ sudo apt-get update
$ sudo apt-get -y install cudnn9-cuda-13

Check the installation

I used the venv I already had for setiastrosuitepro:

(venv) steve@pop-os:~/setiastrosuitepro$ python
Python 3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> print(torch.backends.cudnn.version())
91002
>>>

dawn auriga

 Despite having the sv260 filter, managed to pull out the nebulae.

nebulae ic405 estrella llameante -left- with ic410 renacuajos in auriga
veTEC571 on sv550    20x4min @ hcg100

m45

 This season I went with SASPro's hyperbolic stretch. Yeah, I know.

masks and stuff

The final version is, erm, different:

open cluster with reflection nebulae m45 pleiades in taurus
veTEC571 on sv550    72x4min @ hcg100

double cluster

 Again? Well, I'm trying to wean myself away from StarTools -yeah, I know- and onto SetiAstroSuitePro. In fact, not because of the horses and course thing. Whilst the former excels at stars, the latter, not so much. Or certainly not with my attempt at processing. Let's have a look

ngc884 -top- with ngc869 in Perseus.  36x4min frames using a veTEC571 on an sv550
Pre processing and stacking trial between Siril and SetiAstroSuitePro: I couldn't see any difference. 

StarTools 1.9.577b

SetiAstroSuitePro 1.3.21

... a tiny seti tweak 

seti then startools

seti mfdecon stack

wide Ha

 Another all night session, the heart taking the main share with Alnitak only reaching imageable altitude around 04:30 onwards.

alnitak and surrounding emission nebulae in orion
veTEC571 on ts60    24x4min @ hcg100   

Ha luminance

emission nebulae around ic1805 heart in cassiopeia
veTEC571 on ts60    108x4min @ hcg100

SHO:rrg

Ha luminance

Then back to the sv260 trial:

clusters ngc884 above ngc869 in perseus
veTEC571 on ts60    18x4min @ hcg100