Budget Replacement GPU for Xterminator Processes

20 replies1k views
Jim Raskett avatar
Hi Everyone,

I have a three year old pc that I put together on a budget. Its current configuration is: AMD Ryzen 7 2700X cpu, Gigabyte B450M DS3H main board, 64Gb 3200Mhz DDR4 ram, 2 Gb gen 3 nvme m.2 , and a AMD FX-570 (4Gb) gpu. It does mostly what I need it to except it is very slow when running the Xterminator processes (StarXTerminator AI ver. 11 takes roughly 4 1/2 minutes to run).


I am not very versed in gpu specs, but I have read that nvidia gpu's with cuda enabled greatly decreases processing times when running the XTerminator processes. 
My system is somewhat on the slower end (PI benchmark ~ 13000), but I am fine with it for several more years.
However, if I can find a lower end (very budget friendly ~$200 or less) nvidia gpu with cuda capabilities, I am hoping that will significantly lower my XTerminator processing times.

I am not looking for high performance, just better performance. An older or used gpu will work for me. BTW, I am running a 600w PS.

Thank you for any suggestions/recommendations that you might have!

Jim
Well Written Respectful Engaging
A. Wegdan avatar
Hey Jim,

I have a GTX1650 Super and its 70% boot in performance over my Ryzen7 3700X. I have the Zotac  and PNY variants , i think you can find this model in other brands under 200$. 

Personally I have not tried this brand but should do the trick, ofc do your own research and read the reviews it just popped up in my search so I pasted the link.

I have this one and this one


Ahmed
Jim Raskett avatar
Thank you Ahmed!

Great information. In fact, I looked at similar cards such as the GTX 1660 Ti which is a similar price and a small performance boost.

You gave me a good place to start looking.

Thanks for your help!

Jim
Well Written Respectful Supportive
A. Wegdan avatar
glad i could be of help smile
Joe Linington avatar
Optimal would be a 12gb GTX3060 but above $200. Make sure you get 8gb whatever you get. 4gb and 6gb cards can run out of memory with larger images and multi panel projects.
Jim Raskett avatar
Joe Linington:
Optimal would be a 12gb GTX3060 but above $200. Make sure you get 8gb whatever you get. 4gb and 6gb cards can run out of memory with larger images and multi panel projects.



Your advice is much appreciated!

Thanks Joe!

Jim
Rouven Spaar avatar
I would personally buy smth Like a used rtx 2060 super. I got this card in my own and am very happy with it.
Konrad Krebs avatar
Hi, I use a RTX 3050 and this is much faster (5-8x) than my Ryzen 9 5950x. If you can get a used one this may also be an option.
Jim Raskett avatar
I would personally buy smth Like a used rtx 2060 super. I got this card in my own and am very happy with it.

The rtx2060 super looks like a great card. 
I will look to see what will work in my budget. 
Thanks for sharing your experience!

Jim
Jim Raskett avatar
Konrad Krebs:
Hi, I use a RTX 3050 and this is much faster (5-8x) than my Ryzen 9 5950x. If you can get a used one this may also be an option.

Thanks Konrad!
5-8x faster than the Ryzen is phenomenal!
I will check the markets for used. Might be too much still on the high side, but I’ll see.
Much appreciated!
Adrian Knagg-Baugh avatar
Even the laptop-grade 4GB GeForce 1650 I use is like night and day compared with CPU star removal processing. I use StarNet rather than the Xterminator products, but the principle is the same. I find 24Mpixel images from my Toupcam 2600 take around a minute on the 1650: it's nearly an order of magnitude faster than processing on my 8-core / 16-thread Ryzen 4800H (Zen 2) CPU. I've had no problem with GPU memory capacity when processing full size (but not drizzled) images from the Toupcam, but if you plan to do star removal on huge mosaics after stitching then of course you would need more GPU RAM, as always the more memory the more headroom it gives you for larger images, but somewhere along the line budget constraints come into play for all of us.
Well Written Helpful Insightful Engaging Supportive
A. Wegdan avatar
Adrian Knagg-Baugh:
Even the laptop-grade 4GB GeForce 1650 I use is like night and day compared with CPU star removal processing. I use StarNet rather than the Xterminator products, but the principle is the same. I find 24Mpixel images from my Toupcam 2600 take around a minute on the 1650: it's nearly an order of magnitude faster than processing on my 8-core / 16-thread Ryzen 4800H (Zen 2) CPU. I've had no problem with GPU memory capacity when processing full size (but not drizzled) images from the Toupcam, but if you plan to do star removal on huge mosaics after stitching then of course you would need more GPU RAM, as always the more memory the more headroom it gives you for larger images, but somewhere along the line budget constraints come into play for all of us.

I completely agree, the 1650 is a good start, the higher GPUs are definitely better but this is up to you based on your budget.
starfield avatar
Hey Jim,

My machine is older as well and only has a 450 power supply.   I have a GeoForce GTX 1660 TI.   With CUDA enabled on windows 11,  StarXterminator peformance is great.  Only takes about 35 seconds on one of my big 2X drizzled images from my 2600.   Since this is an older card I think it's pretty cheap now.
Helpful Concise Supportive
Jim Raskett avatar
Adrian Knagg-Baugh:
Even the laptop-grade 4GB GeForce 1650 I use is like night and day compared with CPU star removal processing. I use StarNet rather than the Xterminator products, but the principle is the same. I find 24Mpixel images from my Toupcam 2600 take around a minute on the 1650: it's nearly an order of magnitude faster than processing on my 8-core / 16-thread Ryzen 4800H (Zen 2) CPU. I've had no problem with GPU memory capacity when processing full size (but not drizzled) images from the Toupcam, but if you plan to do star removal on huge mosaics after stitching then of course you would need more GPU RAM, as always the more memory the more headroom it gives you for larger images, but somewhere along the line budget constraints come into play for all of us.

Exactly the information that I am looking for!
I have been looking at the 1650. It is in my price range and the boost in performance compared to the AMD I currently have would really be welcomed!

Thanks Adrian!

Jim
Jim Raskett avatar
Adrian Knagg-Baugh:
Even the laptop-grade 4GB GeForce 1650 I use is like night and day compared with CPU star removal processing. I use StarNet rather than the Xterminator products, but the principle is the same. I find 24Mpixel images from my Toupcam 2600 take around a minute on the 1650: it's nearly an order of magnitude faster than processing on my 8-core / 16-thread Ryzen 4800H (Zen 2) CPU. I've had no problem with GPU memory capacity when processing full size (but not drizzled) images from the Toupcam, but if you plan to do star removal on huge mosaics after stitching then of course you would need more GPU RAM, as always the more memory the more headroom it gives you for larger images, but somewhere along the line budget constraints come into play for all of us.

I completely agree, the 1650 is a good start, the higher GPUs are definitely better but this is up to you based on your budget.

Thank you again for your input Ahmed!

Jim
Jim Raskett avatar
Hey Jim,

My machine is older as well and only has a 450 power supply.   I have a GeoForce GTX 1660 TI.   With CUDA enabled on windows 11,  StarXterminator peformance is great.  Only takes about 35 seconds on one of my big 2X drizzled images from my 2600.   Since this is an older card I think it's pretty cheap now.

Hi Steve and thanks a lot for your help.
I had looked at the 1650 but wasn’t sure that would be it would be much of an advantage compared to my current AMD card.Sounds like it would be a significant improvement! There are a lot of buying options for the 1650 both new and used.

However, I have been looking at the 1660 TI also. I currently have it in my Amazon cart, but it’s a little cheaper at $220 at B&H.
this is really great information and I really appreciate your input!
Thanks!

Jim
Jim Raskett avatar
@Ahmed Wegdan 
@Adrian Knagg-Baugh 
@starfield 
@Joe Linington 
@Konrad Krebs 
@Rouvi05 

Thank you a bunch to everyone that responded. 
I’ve been comparing features and cost both new and used.
I really appreciated everyone’s responses and suggestions. I had some real world experience to use thanks to you.
Since my system is not new and far from cutting edge, I decided on the PNY GTX 1660 Ti. 
It just edged out the 1650 super. 
Compared to my current AMD RX-570, there ought to be a phenomenal improvement in speed.
Now, I need look up on-line resources for enabling Cuda.
Your input was immensely helpful and thanks again!

Jim
Respectful Supportive
A. Wegdan avatar
@Ahmed Wegdan 
@Adrian Knagg-Baugh 
@starfield 
@Joe Linington 
@Konrad Krebs 
@Rouvi05 

Thank you a bunch to everyone that responded. 
I’ve been comparing features and cost both new and used.
I really appreciated everyone’s responses and suggestions. I had some real world experience to use thanks to you.
Since my system is not new and far from cutting edge, I decided on the PNY GTX 1660 Ti. 
It just edged out the 1650 super. 
Compared to my current AMD RX-570, there ought to be a phenomenal improvement in speed.
Now, I need look up on-line resources for enabling Cuda.
Your input was immensely helpful and thanks again!

Jim

Good luck, quick tip look out for the version of cuda you use not all versions work. I believe I use 11.8 so check.

clear skies
Jim Raskett avatar
@Ahmed Wegdan 

Thanks for the tip Ahmed!

I found this on the Pixinsight forum. Looks like it points to 11.8 and has all the resources too.

https://pixinsight.com/forum/index.php?threads/up-to-date-instructions-for-enabling-gpu-acceleration.20275/
Jim Raskett avatar
@Ahmed Wegdan
@Adrian Knagg-Baugh
@starfield
@Joe Linington
@Konrad Krebs
@Rouvi05 

I received my GTX 1660 Ti yesterday and installed it last night.

I ended up following RT's cuda installation instructions in the PI forum thread and it worked like a charm.https://pixinsight.com/forum/index.php?threads/up-to-date-instructions-for-enabling-gpu-acceleration.20275/

I downloaded the nvidia drivers and the cuda installation files the evening before and installed the new card along with setting up cuda last night in about 35 minutes. Slick and smooth!

StarXTerminator times went from +4 minutes to 14-15 seconds even with a 2x drizzle. Wow!
BXT runs under a minute and NXT under 10 seconds also on a 2x drizzled image.

Thanks again everyone for your help and special thanks Steve (@starfield) for your help.

Jim
A. Wegdan avatar
Great to hear, good luck with your processing.

Clear skies
Well Written Respectful