Jim Raskett:
David Foust:
I've got a new GPU on my list... Still using my old 1660ti!  Glad to see it would be of some benefit to upgrade in my processing workflow.
I also use a 533 MC-Pro and my 1660Ti blazes through the Xterminators. I call blazing 25 sec including initialization and 12 seconds for subsequent runs. For the $$, I can't justify a new GPU until I upgrade my 4 year old desktop.
Of course, I am using a 533. I am sure the difference would be greater with a larger sensor format camera. Yeah same. It's not like my process is particularly slow right now with my 533mcp, but I've debated upgrading to a larger chip and I think the GPU upgrade would pay dividends whenever that happens! Have an i7-13700k and 64gb RAM so I think my bottleneck is my GPU, but it's certainly not bad by any means.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Charles Barker:
So guys I don’t want to spam you but here is a detailed listing from the artificial brain of processes make sense /nonsense on GPU /CPU
and I do agree to this analysis.
No please... by all means!
This was the intent of what I was trying to push continuing to encourage PixInsight to really start digging into improving the user's experience of their product instead of telling us how hard it will be. Don't get me wrong I certainly appreciate the fact that others have made strides in the post-processing side and PixInsight generously invited them in and includes them. Using NoiseXTerminator from several minutes to less than one is happily welcomed. I also admire and benefit everything PixInsight brings to the table! However, if you want to really improve the PixInsight experience and the opportunity is there, make it so I don't have to wait 24 hours to get a result that I may or may not like... It's why I use DSS for stacking, because many times I've waited 36 hours to get a crap image or to find out I did something wrong...
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
I have a 2024 MacbookPro M4 with 24GB of RAM and I can tell you that it is FAST. Faster than my i7-14700K and a 4070Ti.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
I also updated my system this week, from an i7 10900, 16 GB with a GTX 1060 to an i9 14900, 64 GB and an RTX 4090, 16 GB, both in the laptop version.
Here too, the differences (after changing the tensorflow.dll) are significant: Here are the examples with a stack of QYH268m recordings with the new I9 on two 4k monitors in Pixinsight
StarX i9: 30 seconds (without GPU it was still 233 seconds with the i9!)
StarX i7: 92 seconds
StarX i3: 422s ! (without GPU, just for comparison )
BlurX i9: 15 seconds
BlurX i7: 55 seconds
NoiseX i9: 3 seconds
NoiseX i7: 12 seconds.
Cool stuff!
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Freddie Alequin-Ramos:
Charles Barker:
So guys I don’t want to spam you but here is a detailed listing from the artificial brain of processes make sense /nonsense on GPU /CPU
and I do agree to this analysis.
No please... by all means!
This was the intent of what I was trying to push continuing to encourage PixInsight to really start digging into improving the user's experience of their product instead of telling us how hard it will be. Don't get me wrong I certainly appreciate the fact that others have made strides in the post-processing side and PixInsight generously invited them in and includes them. Using NoiseXTerminator from several minutes to less than one is happily welcomed. I also admire and benefit everything PixInsight brings to the table! However, if you want to really improve the PixInsight experience and the opportunity is there, make it so I don't have to wait 24 hours to get a result that I may or may not like...
It's why I use DSS for stacking, because many times I've waited 36 hours to get a crap image or to find out I did something wrong... For you guys the fully automated out of the box 3 stacked different versions solution would be the right choice. It is Pixinsight! For the tinkerers and geeks! Pixinsight gives you many possibilities that you don’t want to make use from?
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Friends of the GPU and Graphical Processing, It took me quite a while to decide whether or not to post this, but I’ve finally convinced myself that I should share my thoughts. Over the past five years of my astrophotography journey, I believe I may have stumbled upon a method for directly observing and examining exoplanets, along with other details that are traditionally thought to be invisible due to the resolution limits of camera sensors. Through extensive experimentation with extreme color grading on my datasets, I’ve noticed what appears to be a kind of lensing effect that can emerge when viewed on an LCD screen or smartphone display. This effect seems to unlock extraordinary levels of detail—details that, in theory, should not be visible. I’ve tested this approach on several photometrically confirmed exoplanet transits and found myself able to observe the exoplanets with what I can only describe as graphical detail. To me, it feels as though astronomy, while grounded in the rigorous work of brilliant scientists with PhDs, also benefits from us “simple” folks—those of us who may not fully understand the complexities of the science but are willing to tinker, hack, and push boundaries with the most straightforward tools we have. Sometimes, unconventional methods lead to surprising results. I’d love to hear your thoughts on this or if anyone else has experimented with similar techniques! this here shows Tres5b with possible exomoons  this here shows a Rocky planet passing in front of white dwarf WD 1145 + 017 orbiting the white dwarf in a orbital period of 4hours in detailed lensing you can see the planet beeing eaten from its host star   this here was a discovery by accident showing a exoplanet undiscovered, while I was observing assumed activity on Mira type star AmCyg which was showing a infrared leaking halo on dual band filters, i found this green blue giant writhin my dataset   |
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Franz:
I also updated my system this week, from an i7 10900, 16 GB with a GTX 1060 to an i9 14900, 64 GB and an RTX 4090, 16 GB, both in the laptop version.
Here too, the differences (after changing the tensorflow.dll) are significant: Here are the examples with a stack of QYH268m recordings with the new I9 on two 4k monitors in Pixinsight
StarX i9: 30 seconds (without GPU it was still 233 seconds with the i9!)
StarX i7: 92 seconds
StarX i3: 422s ! (without GPU, just for comparison )
BlurX i9: 15 seconds
BlurX i7: 55 seconds
NoiseX i9: 3 seconds
NoiseX i7: 12 seconds.
Cool stuff! *** Please be sure to update your BIOS with your motherboard. The current BIOS for an Intel Core i9-14900 processor includes the "Microcode 0x129" update. I have the i9-14900K and it is a beast. But the instability/ degradation issues remain in question even despite the BIOS update but are almost guaranteed without it.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
I had a new computer build for work and to handle Pixinsight. Had the Ryzen 9 7950X with the Geforce 4060 with 128mb ram. After I installed the CUDA accelerator, I went from stacking files over night at 6-8 hours down to 20 minutes and StarX used to take an hour and went down to 20 seconds. Definitely go CUDA if you have to supported components.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
It is an absolute game-changer for most people to use the CUDA acceleration. Now, I just wish WBPP could take advantage of it as well.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Stellar Nomads:
It is an absolute game-changer for most people to use the CUDA acceleration. Now, I just wish WBPP could take advantage of it as well. We discussed this earlier LoL
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
For anyone using RC ASTRO tools in PixInsight who hasn't tried GPU acceleration, and wants to see the massive speed increase when using BlurXTerminator, NoiseXTerminator or StarXTerminator, I'd like to point out that RC made it real easy. All you have to do is this: To proceed, add the following to your PI repository list using Resources -> Updates -> Manage Repositories: https://www.rc-astro.com/TensorFlow/PixInsight/GPUBut first, Get full details including which Operating Systems and GPU cards are applicable by checking the RC ASTRO web site: https://www.rc-astro.com/gpu-acceleration-for-ai-powered-tools/Of course, this information too will become obsolete in time. My hope is, now that AI is so popular, the line between CPU and GPU will narrow and whatever the future brings will benefit all aspects of PixInsight.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Michael Cory:
For anyone using RC ASTRO tools in PixInsight who hasn't tried GPU acceleration, and wants to see the massive speed increase when using BlurXTerminator, NoiseXTerminator or StarXTerminator, I'd like to point out that RC made it real easy.
All you have to do is this:
To proceed, add the following to your PI repository list using Resources -> Updates -> Manage Repositories: https://www.rc-astro.com/TensorFlow/PixInsight/GPU
But first, Get full details including which Operating Systems and GPU cards are applicable by checking the RC ASTRO web site:
https://www.rc-astro.com/gpu-acceleration-for-ai-powered-tools/
Of course, this information too will become obsolete in time. We are many with linux waiting to do the jump.....
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
StarXterminator ran twice as fast on my beafy Dell laptop with an A5500 GPU as it did without the GPU.
Nowhere close to the gain of a desktop with a power-hungry GPU, but better than nothing.
Eric C
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Freddie Alequin-Ramos:
Charles Barker:
However, if you want to really improve the PixInsight experience and the opportunity is there, make it so I don't have to wait 24 hours to get a result that I may or may not like...
It's why I use DSS for stacking, because many times I've waited 36 hours to get a crap image or to find out I did something wrong... Are you using Blink to weed out bad images before WBPP? After 24 or 36 hours in WBPP, how is the image bad? Were the raw images bad (GIGO) or did WBPP make the images bad (i.e., wrong settings), or you just don't like the image?
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Eric C:
Freddie Alequin-Ramos:
Charles Barker:
However, if you want to really improve the PixInsight experience and the opportunity is there, make it so I don't have to wait 24 hours to get a result that I may or may not like...
It's why I use DSS for stacking, because many times I've waited 36 hours to get a crap image or to find out I did something wrong...
Are you using Blink to weed out bad images before WBPP?
After 24 or 36 hours in WBPP, how is the image bad? Were the raw images bad (GIGO) or did WBPP make the images bad (i.e., wrong settings), or you just don't like the image? The image was bad because of wrong settings (flats weren't applied for example or suspected wrong darks on wrong images). The resulting stack is clearly bad.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
I noticed today when running several processes including BlurX and StarX that when I opened windows Task Manager to see how resources were doing, my CPU workload was at 100% but my GPU was at 0%!
I am probably misunderstanding something, but if PI (especially WBPP) puts that much load on the CPU and not the GPU, why would CUDA help with faster processing?
I am running a fast i7 CPU with 16GB of RAM and 6GB of graphics memory, not too shabby a system for one that's over 4 years old.
I'd love to speed up BlurX, StarX and WBPP but not sure downloading all the CUDA related apps and files will help. I wish that PI were not so CPU intensive. I rarely, if ever, see an app use 100% of CPU resources.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Jerry Gerber:
I noticed today when running several processes including BlurX and StarX that when I opened windows Task Manager to see how resources were doing, my CPU workload was at 100% but my GPU was at 0%!
I am probably misunderstanding something, but if PI (especially WBPP) puts that much load on the CPU and not the GPU, why would CUDA help with faster processing?
I am running a fast i7 CPU with 16GB of RAM and 6GB of graphics memory, not too shabby a system for one that's over 4 years old.
I'd love to speed up BlurX, StarX and WBPP but not sure downloading all the CUDA related apps and files will help. I wish that PI were not so CPU intensive. I rarely, if ever, see an app use 100% of CPU resources. Your observation about PixInsight’s (PI) CPU-intensive nature is accurate; many of its processes, including WBPP, primarily utilize the CPU. However, certain tools, especially those from RC Astro like BlurXTerminator, StarXTerminator, and NoiseXTerminator, can leverage NVIDIA GPUs for acceleration. Enabling GPU Acceleration in PixInsight To utilize GPU acceleration for these tools, follow these steps: 1. Verify GPU Compatibility: • Ensure your NVIDIA GPU has a Compute Capability of 3.5 or higher. You can check your GPU’s compatibility on NVIDIA’s CUDA GPUs page. 2. Install NVIDIA CUDA Toolkit: • Download and install the NVIDIA CUDA Toolkit version 11.8.0. This version is compatible with the required TensorFlow libraries. 3. Install cuDNN Libraries: • Download and install NVIDIA cuDNN version 8.7.0 for CUDA 11.x. These libraries are essential for accelerating deep neural network computations. 4. Replace TensorFlow DLL in PixInsight: • Download the GPU-enabled TensorFlow library (version 2.9.0). • Navigate to C:\Program Files\PixInsight\bin. • Rename the existing tensorflow.dll to tensorflow-old.dll. • Copy the downloaded tensorflow.dll into this directory. 5. Set Environment Variables: • Add a system environment variable named TF_FORCE_GPU_ALLOW_GROWTH and set its value to true. • Ensure the system Path variable includes the path to the CUDA bin directory, typically C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin. 6. Install zLib DLL: • Download the zLib DLL (version 1.2.3). • Copy zlibwapi.dll to the CUDA bin directory. Detailed instructions for these steps are available on RC Astro’s GPU Acceleration Guide. Important Considerations • PixInsight Updates: After updating PixInsight, the tensorflow.dll file may revert to the CPU-only version. You’ll need to replace it again with the GPU-enabled version. • Experimental Repository: RCAstro has introduced an experimental repository to simplify enabling GPU acceleration on Windows. This repository automates some of the setup processes. More information can be found on the PixInsight Forum.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Jerry Gerber:
I noticed today when running several processes including BlurX and StarX that when I opened windows Task Manager to see how resources were doing, my CPU workload was at 100% but my GPU was at 0%!
I am probably misunderstanding something, but if PI (especially WBPP) puts that much load on the CPU and not the GPU, why would CUDA help with faster processing?
I am running a fast i7 CPU with 16GB of RAM and 6GB of graphics memory, not too shabby a system for one that's over 4 years old.
I'd love to speed up BlurX, StarX and WBPP but not sure downloading all the CUDA related apps and files will help. I wish that PI were not so CPU intensive. I rarely, if ever, see an app use 100% of CPU resources. Check @Michael Cory 's post above: the RC-Astro repo did it for me and it's pretty straightforward.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
I have a 4070 laptop GPU, and the CUDA makes a considerable difference; although I thought it would be faster than my Macbook M1, it's not impressively faster. I was also just working on a mosaic, so the image was huge, and the speed difference may not have been noticeable since the image was twice the size of a normal image!
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
Freddie Alequin-Ramos:
Charles Barker:
So guys I don’t want to spam you but here is a detailed listing from the artificial brain of processes make sense /nonsense on GPU /CPU
and I do agree to this analysis.
No please... by all means!
This was the intent of what I was trying to push continuing to encourage PixInsight to really start digging into improving the user's experience of their product instead of telling us how hard it will be. Don't get me wrong I certainly appreciate the fact that others have made strides in the post-processing side and PixInsight generously invited them in and includes them. Using NoiseXTerminator from several minutes to less than one is happily welcomed. I also admire and benefit everything PixInsight brings to the table! However, if you want to really improve the PixInsight experience and the opportunity is there, make it so I don't have to wait 24 hours to get a result that I may or may not like...
It's why I use DSS for stacking, because many times I've waited 36 hours to get a crap image or to find out I did something wrong...
For you guys the fully automated out of the box 3 stacked different versions solution would be the right choice. It is Pixinsight! For the tinkerers and geeks! Pixinsight gives you many possibilities that you don’t want to make use from?
I would love to but sometimes what seems to me the exact same settings just doesn't work because reasons. Waiting such a long time and having my workhorse computer down for a bad result and near impossible to track or test why? I truly don't have time for that. I stack with DSS and the rest of the processing I use PI.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Copy the URL below to share a direct link to this post.
This post cannot be edited using the classic forums editor.
To edit this post, please enable the "New forums experience" in your settings.
