New Ryzen Processor Use in Pixinsight

Oct 22, 2022

Thank you for sharing.
i was expecting that new 7950x to be faster.

My older 5900 is getting
30k cpu
30k swap
Total 29 to 30k

While that 7950x speed is good it will still take over 30 minutes for a full frame imaging stack in WBPP.

I suppose we have to hope CPUs get even faster!

Rouz,

John Hayes

Frequent contributor

Oct 22, 2022

That's pretty sweet Jerry. I haven't measured it in a while but your new machine might be 5x-6x faster than my MacBook Pro M2.

Dang.

John

Phil Wright

Oct 22, 2022

I would upgrade your GPU next to a 3xxx series for faster Starnet++ / StarXterminator processing too.

Oct 22, 2022

The biggest improvement you could get now would be an m.2 drive. The swap speed was quite low.
Im assuming that hard-drive or SSD?

The Samsung evo 980 pro m.2 drive is getting very good results.

Rouz

Topic starter

Oct 22, 2022

Thanks for the comments. Here are some pictures and technical descriptions of the parts, with comments from some gamers:

https://pcpartpicker.com/builds/edit/?userbuild=CQDcCJ

They also comment that my graphics card is low end, but I do not believe that comes into play in PI. StarXterminator is faster by a lot with this system.

Also, note that I have 64GB of RAM that is now being used for 8 4 GB RAMDisks and that doubled the swap speed in PI. Was noted that the swap speed only helps on certain operations that are reversible, not WBPP for example. See the PI thread.

Yes, the Samsung blasts away. I have a second oen on which Linux is installed (Kubuntu- the PI development platform). You can change boot OS relatively easy in the ASUS BIOS.

Linux is faster out of the box- hardware to hardware- than Windows 11, however Windows 11 catches up with the RAM Disks. I am sticking with Windows 11 to keep all in synch with OneDrive... not sure there is a way to use OneDrive on Linux. Also I have installed my Office 365 apps on the install and keep in crisp communication with that directly from this install.

I have retired two old Windows 10 machines now and all my legacy files and in the Mirror drives... fingers crossed as this is not RAID, but RAID is a pain... not sure I mentioned here but Windows 11 does NOT support RAID, at least 1 or 0. I am going to try to keep my registered files there for potential reuse, but it is very space hungry since my ASI6200 has 50MB images.... maybe I will back up on AWS when the time comes.

Oct 22, 2022

I had a gaming system built geared around PI processing and used the Ryzen 9 5950x GPU is a iNivia 3060 12g 96gigs of Viper DDR4 ram. I was hesitant to go with W11 but I did and was surprised at how much faster this unit was. Of course I was coming from a souped up i5 so it didn’t take much to beat that. I originally had 64gigs of ram but after about 4-5 hours of solid processing it would crash on my and added the extra 32g and haven’t had an issue since. I was able to get the GPU speed up with Cuda for StarNet2 & StarXTerminator for my i5 but was unable to get the Ryzen w11 to work so I don’t know what the problem is there.

I considered the RAM disk your talking about but from my understanding when the machine is shut off you loose everything in that disk correct? If thats the case to me it seems like it would be a pain to have to keep loading the RAM disk every time you boot back up.

Is that what you do or am I misunderstanding how the RAM disk works???

Dale

Oct 22, 2022

I don't know about Starnet but StarXTerminator does not use any of my GPU and all CPU.

I just ran it again now to double check.

Rouz

Oct 22, 2022

Rouz Astro:
I don't know about Starnet but StarXTerminator does not use any of my GPU and all CPU.

I just ran it again now to double check.

Rouz

Hello Rouz,
Yes it does work on my i5. I’m trying to find where I posted the results before and after. It was significant. I think its somewhere in this post but its an extremely long post about the acceleration of these 2 processes in PI.

https://www.astrobin.com/forum/c/astrophotography/deep-sky-processing-techniques/starxterminator-vs-starnet2/?page=1

Dale

Oct 22, 2022

Dale Penkala:
Rouz Astro:
I don't know about Starnet but StarXTerminator does not use any of my GPU and all CPU.

I just ran it again now to double check.

Rouz

Hello Rouz,
Yes it does work on my i5. I’m trying to find where I posted the results before and after. It was significant. I think its somewhere in this post but its an extremely long post about the acceleration of these 2 processes in PI.

https://www.astrobin.com/forum/c/astrophotography/deep-sky-processing-techniques/starxterminator-vs-starnet2/?page=1

Dale

*I think you have activated GPU acceleration?
I left mine at default.
https://www.williamliphotos.com/starnet-cuda

I don't know if it work with XTerminator as well but I dont use Starnet.

My main delay comes form the actual stacking which takes 45 to 60 minutes with the IMX455 data.

Rouz

Topic starter

Oct 22, 2022

John Hayes:
That's pretty sweet Jerry. I haven't measured it in a while but your new machine might be 5x-6x faster than my MacBook Pro M2.

Dang.

John

Yes, I used to do all my processing on a Dell mobile workstation with a Xeon processor = I7 and the benchmarks came out at about 3,000... with the new system they come out at 30,000.... suspect your MacBook is in the same category as my Dell... have you run the benchmark in PI?

Topic starter

Oct 22, 2022

Rouz Astro:
Dale Penkala:
Rouz Astro:
I don't know about Starnet but StarXTerminator does not use any of my GPU and all CPU.

I just ran it again now to double check.

Rouz

Hello Rouz,
Yes it does work on my i5. I’m trying to find where I posted the results before and after. It was significant. I think its somewhere in this post but its an extremely long post about the acceleration of these 2 processes in PI.

https://www.astrobin.com/forum/c/astrophotography/deep-sky-processing-techniques/starxterminator-vs-starnet2/?page=1

Dale

*I think you have activated GPU acceleration?
I left mine at default.
https://www.williamliphotos.com/starnet-cuda

I don't know if it work with XTerminator as well but I dont use Starnet.

My main delay comes form the actual stacking which takes 45 to 60 minutes with the IMX455 data.

Rouz

I run StarXterminator that does not have these options, but is very slow with their latest AI implementation (but very good).

If you want to see screen shots of this new processor running stacking (WBPP) they are out at the PI forum:

https://pixinsight.com/forum/index.php?threads/new-computer-system-ryzen-9750-and-benchmarks.19400/

Topic starter

Oct 22, 2022

Dale Penkala:
I had a gaming system built geared around PI processing and used the Ryzen 9 5950x GPU is a iNivia 3060 12g 96gigs of Viper DDR4 ram. I was hesitant to go with W11 but I did and was surprised at how much faster this unit was. Of course I was coming from a souped up i5 so it didn’t take much to beat that. I originally had 64gigs of ram but after about 4-5 hours of solid processing it would crash on my and added the extra 32g and haven’t had an issue since. I was able to get the GPU speed up with Cuda for StarNet2 & StarXTerminator for my i5 but was unable to get the Ryzen w11 to work so I don’t know what the problem is there.

I considered the RAM disk your talking about but from my understanding when the machine is shut off you loose everything in that disk correct? If thats the case to me it seems like it would be a pain to have to keep loading the RAM disk every time you boot back up.

Is that what you do or am I misunderstanding how the RAM disk works???

Dale

Yes RAM disks are a pain... I ran across this old software in the gamer word that seems to keep the disks going... think this persisted after a shutdown... does not help on all PI processes:

Here is what it looks like on PI (with todays image):

Oct 22, 2022 · Edited: Oct 22, 2022

Rouz Astro:
Dale Penkala:
Rouz Astro:
I don't know about Starnet but StarXTerminator does not use any of my GPU and all CPU.

I just ran it again now to double check.

Rouz

Hello Rouz,
Yes it does work on my i5. I’m trying to find where I posted the results before and after. It was significant. I think its somewhere in this post but its an extremely long post about the acceleration of these 2 processes in PI.

https://www.astrobin.com/forum/c/astrophotography/deep-sky-processing-techniques/starxterminator-vs-starnet2/?page=1

Dale

*I think you have activated GPU acceleration?
I left mine at default.
https://www.williamliphotos.com/starnet-cuda

I don't know if it work with XTerminator as well but I dont use Starnet.

My main delay comes form the actual stacking which takes 45 to 60 minutes with the IMX455 data.

Rouz

Yes thats what I was saying. Before I did the acceleration tweak it was all CPU, after it was GPU.

Oct 22, 2022

Dale Penkala:
I had a gaming system built geared around PI processing and used the Ryzen 9 5950x GPU is a iNivia 3060 12g 96gigs of Viper DDR4 ram. I was hesitant to go with W11 but I did and was surprised at how much faster this unit was. Of course I was coming from a souped up i5 so it didn’t take much to beat that. I originally had 64gigs of ram but after about 4-5 hours of solid processing it would crash on my and added the extra 32g and haven’t had an issue since. I was able to get the GPU speed up with Cuda for StarNet2 & StarXTerminator for my i5 but was unable to get the Ryzen w11 to work so I don’t know what the problem is there.

I considered the RAM disk your talking about but from my understanding when the machine is shut off you loose everything in that disk correct? If thats the case to me it seems like it would be a pain to have to keep loading the RAM disk every time you boot back up.

Is that what you do or am I misunderstanding how the RAM disk works???

Dale

Yes RAM disks are a pain... I ran across this old software in the gamer word that seems to keep the disks going... think this persisted after a shutdown... does not help on all PI processes:

Here is what it looks like on PI (with todays image):

Ok so then I wasn’t misunderstanding then, thanks for that confirmation. Ya I seen there was software that was able to easily setup the RAM disk but when I found out more about it I decided to just let it go. My Ryzen 9 is so much more powerful then my i5. I need to check my benchmark again. I don’t remember what it ended up at after I added the extra swap files. I see your in the 30k range, If memory serves me I was mid 20k. Now that I know about yours it will be interesting to see how slow mine is now! LOL

Dale

Topic starter

Oct 22, 2022

Dale Penkala:
Dale Penkala:
I had a gaming system built geared around PI processing and used the Ryzen 9 5950x GPU is a iNivia 3060 12g 96gigs of Viper DDR4 ram. I was hesitant to go with W11 but I did and was surprised at how much faster this unit was. Of course I was coming from a souped up i5 so it didn’t take much to beat that. I originally had 64gigs of ram but after about 4-5 hours of solid processing it would crash on my and added the extra 32g and haven’t had an issue since. I was able to get the GPU speed up with Cuda for StarNet2 & StarXTerminator for my i5 but was unable to get the Ryzen w11 to work so I don’t know what the problem is there.

I considered the RAM disk your talking about but from my understanding when the machine is shut off you loose everything in that disk correct? If thats the case to me it seems like it would be a pain to have to keep loading the RAM disk every time you boot back up.

Is that what you do or am I misunderstanding how the RAM disk works???

Dale

Yes RAM disks are a pain... I ran across this old software in the gamer word that seems to keep the disks going... think this persisted after a shutdown... does not help on all PI processes:

Here is what it looks like on PI (with todays image):

Ok so then I wasn’t misunderstanding then, thanks for that confirmation. Ya I seen there was software that was able to easily setup the RAM disk but when I found out more about it I decided to just let it go. My Ryzen 9 is so much more powerful then my i5. I need to check my benchmark again. I don’t remember what it ended up at after I added the extra swap files. I see your in the 30k range, If memory serves me I was mid 20k. Now that I know about yours it will be interesting to see how slow mine is now! LOL

Dale

Yeah, exactly. Mine was in the mid 20s without the RAM Disk. The Linux was faster. Mine is the 7950. You can check the parts list.

Oct 22, 2022

*I think you have activated GPU acceleration?
I left mine at default.
https://www.williamliphotos.com/starnet-cuda

I don't know if it work with XTerminator as well but I dont use Starnet.

My main delay comes form the actual stacking which takes 45 to 60 minutes with the IMX455 data.

Rouz

Yes thats what I was saying. Before I did the acceleration tweak it was all CPU, after it was GPU.

*
Is your GPU faster?
My CPU is much more powerful than my GPU, maybe that is why.
What numbers are you getting in the benchmark test.

I tried Ramdisks but the EVO980 and swap folders was actually faster.

Oct 22, 2022

Rouz Astro:
*I think you have activated GPU acceleration?
I left mine at default.
https://www.williamliphotos.com/starnet-cuda

I don't know if it work with XTerminator as well but I dont use Starnet.

My main delay comes form the actual stacking which takes 45 to 60 minutes with the IMX455 data.

Rouz

Yes thats what I was saying. Before I did the acceleration tweak it was all CPU, after it was GPU.

*
Is your GPU faster?
My CPU is much more powerful than my GPU, maybe that is why.
What numbers are you getting in the benchmark test.

I tried Ramdisks but the EVO980 and swap folders was actually faster.

Hi Rouz,
I can’t remember what they were, but I’ll try and check it tomorrow afternoon and post. If memory serves me I think I was in the mid 20k’s. Again I just don’t remember. I’m sure its not as fast as what you guys are running but like I said compared to my souped up i5 it was a remarkable improvement.
I never did do the RAM disk so I don’t know how much that would have made a difference. Like @jerryyyyy mentioned it only helps on some processes and its kind of a pain to have to keep loading the RAM disk everytime you reboot the computer.

Dale

Oct 22, 2022

To get almost 2x the speed out of the Samsung Evo you need to install Samsung magician and configure it more max. speed.

Rouz,

Oct 22, 2022

Dale Penkala:

Hi Rouz,
I can’t remember what they were, but I’ll try and check it tomorrow afternoon and post. If memory serves me I think I was in the mid 20k’s. Again I just don’t remember. I’m sure its not as fast as what you guys are running but like I said compared to my souped up i5 it was a remarkable improvement.
I never did do the RAM disk so I don’t know how much that would have made a difference. Like @jerryyyyy mentioned it only helps on some processes and its kind of a pain to have to keep loading the RAM disk everytime you reboot the computer.

Dale

*
That's great Dale,

I'm sure will be a nice bump up from the I5.
Problem sensors a re getting larger, I dread to think what processing the larger than IMX455 files will be like with the IMX461 and 411.
That's one of the reasons I went with the reducer option to get more optical FOV rather than getting a larger sensor.

The camera prices are becoming reasonable but then you have to deal with the massive filters and files!

Rouz,

Oct 22, 2022

Dale Penkala:
Dale Penkala:
I had a gaming system built geared around PI processing and used the Ryzen 9 5950x GPU is a iNivia 3060 12g 96gigs of Viper DDR4 ram. I was hesitant to go with W11 but I did and was surprised at how much faster this unit was. Of course I was coming from a souped up i5 so it didn’t take much to beat that. I originally had 64gigs of ram but after about 4-5 hours of solid processing it would crash on my and added the extra 32g and haven’t had an issue since. I was able to get the GPU speed up with Cuda for StarNet2 & StarXTerminator for my i5 but was unable to get the Ryzen w11 to work so I don’t know what the problem is there.

I considered the RAM disk your talking about but from my understanding when the machine is shut off you loose everything in that disk correct? If thats the case to me it seems like it would be a pain to have to keep loading the RAM disk every time you boot back up.

Is that what you do or am I misunderstanding how the RAM disk works???

Dale

Yes RAM disks are a pain... I ran across this old software in the gamer word that seems to keep the disks going... think this persisted after a shutdown... does not help on all PI processes:

Here is what it looks like on PI (with todays image):

Ok so then I wasn’t misunderstanding then, thanks for that confirmation. Ya I seen there was software that was able to easily setup the RAM disk but when I found out more about it I decided to just let it go. My Ryzen 9 is so much more powerful then my i5. I need to check my benchmark again. I don’t remember what it ended up at after I added the extra swap files. I see your in the 30k range, If memory serves me I was mid 20k. Now that I know about yours it will be interesting to see how slow mine is now! LOL

Dale

Yeah, exactly. Mine was in the mid 20s without the RAM Disk. The Linux was faster. Mine is the 7950. You can check the parts list.

Ya mine is the 5950x so I’m sure mine is slower then yours! I have a 12gig graphics card and 96gigs of ram so I know that is a plus for me. I did think about the “linex inviornment” option for windows too but when I started looking into that I just figured I’m best to stay with one OS. I’m far from a computer genius so the simpler the better for me 😊

Dale

Oct 22, 2022

Rouz Astro:
Dale Penkala:

Hi Rouz,
I can’t remember what they were, but I’ll try and check it tomorrow afternoon and post. If memory serves me I think I was in the mid 20k’s. Again I just don’t remember. I’m sure its not as fast as what you guys are running but like I said compared to my souped up i5 it was a remarkable improvement.
I never did do the RAM disk so I don’t know how much that would have made a difference. Like @jerryyyyy mentioned it only helps on some processes and its kind of a pain to have to keep loading the RAM disk everytime you reboot the computer.

Dale

*
That's great Dale,

I'm sure will be a nice bump up from the I5.
Problem sensors a re getting larger, I dread to think what processing the larger than IMX455 files will be like with the IMX461 and 411.
That's one of the reasons I went with the reducer option to get more optical FOV rather than getting a larger sensor.

The camera prices are becoming reasonable but then you have to deal with the massive filters and files!

Rouz,

Ya you got that right! The bigger these sensors get the more processing power we need in our computers. Especially if your lucky enough to have an abundance of clear skies! Something I do not have unfortunatly.

Dale

Sh2 171 (Various Processing Options)

Topic starter

Oct 22, 2022

Dale Penkala:
Rouz Astro:
Dale Penkala:

Hi Rouz,
I can’t remember what they were, but I’ll try and check it tomorrow afternoon and post. If memory serves me I think I was in the mid 20k’s. Again I just don’t remember. I’m sure its not as fast as what you guys are running but like I said compared to my souped up i5 it was a remarkable improvement.
I never did do the RAM disk so I don’t know how much that would have made a difference. Like @jerryyyyy mentioned it only helps on some processes and its kind of a pain to have to keep loading the RAM disk everytime you reboot the computer.

Dale

*
That's great Dale,

I'm sure will be a nice bump up from the I5.
Problem sensors a re getting larger, I dread to think what processing the larger than IMX455 files will be like with the IMX461 and 411.
That's one of the reasons I went with the reducer option to get more optical FOV rather than getting a larger sensor.

The camera prices are becoming reasonable but then you have to deal with the massive filters and files!

Rouz,

Ya you got that right! The bigger these sensors get the more processing power we need in our computers. Especially if your lucky enough to have an abundance of clear skies! Something I do not have unfortunatly.

Dale

Yes, the sensor size is what pushed me over the edge with 50MG images from my ASI6200. Took forever to process anything, and the AI was a good 2h....

Along these lines, here is the latest 9200x6200 image using all these toys... good colors for Halloween... sorta looks like a skull with blue eyes and a red hat:

Oct 22, 2022

Yes, the sensor size is what pushed me over the edge with 50MG images from my ASI6200. Took forever to process anything, and the AI was a good 2h....

Along these lines, here is the latest 9200x6200 image using all these toys... good colors for Halloween... sorta looks like a skull with blue eyes and a red hat:

Sh2 171 (Various Processing Options)

*Nice work I'm processing the same target right now!

Rouz