msacco

1 hour ago, DarkAntimatter said:

For what it's worth, some programs do make use of GPUs to do computations, not just for displaying graphics. Many scientific programs do this. This almost always means running Cuda on Nvidia cards. 

The pixinsight web site mentions that the present version does not use Cuda, but it is planned for the future. 

Well yeah but that's not the case 99% of the times

3 minutes ago, celestron8g8 said:

Ok , then just scratch all I said and ignore it and i'll just bow out of the conversation here . Don't mean to cause problems . I made those suggestions cause I use to do a lot of editing but  I haven't messed with PI due to my preference . But I had trouble with some of my editing software and updating the graphic card driver was suggested and that fixed part of my problem . So as I said i'll bow out , have a great day ! 

Thanks, you too! And thanks for trying to help  Didn't mean to be rude or anything like that.

12 minutes ago, celestron8g8 said:

I have basic knowledge , no college training or tech training . What I know is what I learned on my own or from techs updating me . 

I don't have a college or tech training as well(even though I do work in the industry), but what you said is not really related at all.

A graphics card is responsible for processing the video graphics on your machine, which means that everything that you see on your screen, is being processed in your graphics card.
The reason a GPU is extremely important for gaming, is because things needs to be rendered and fast, the faster something is being rendered on the screen, the more FPS you'll have, and the smoother the game will run.

Why a GPU is not related to pixinsight? A GPU is not important for pixinsight(at least not for processing speed), is because these actions work solely on the CPU, RAM and SSD(or HDD), there is the console on the screen writing a few lines of code, which doesn't make any GPU work hard.

You could also see in the console itself, that each image that's being debayered for example, is working on a single CPU thread, so things will run faster.

I could have 100 of the highest end GPU available on the market connected to my PC, and that would still make 0 difference on my processing speed.

BTW, I also stated in my first post that I have a GTX 770 4 GB, which in today terms it's really not all that good, but it's plenty enough for anything related to pixinsight, and I actually use it for gaming as well and it still rocks high end games.

I'm also not 100% sure about everything I said, the general idea is correct, but could be a few small points that are not spot on accurate.

1 minute ago, celestron8g8 said:

Graphic Cards are what runs the processing  of your software . If your problem is taking too much time to process then you need a faster computer and that's not always ram to fix the problem . That's why for gaming computers they have a different graphic card to add speed to their games .

Don't want to be rude or anything, but do you actually have any knowledge about computers?

4 minutes ago, Alien 13 said:

I like a bargain too.

I am sure Pixinsight has its own benchmark software so its worth looking on their site as to what works best at a certain price point..

Alan

Holy....Bless you! Thank you so much for that!

2 minutes ago, Alien 13 said:

The Achilles heal of AMD is stability and compatibility with a wide range of hardware/software, my experience is that Intel just works even if its not as fast sometimes....

Alan

Well it's not really the point of the thread, but I can understand what you're saying, my personal opinion is that it's a waste, but I'm a cheapo guy.
I roam on aliexpress, I always get the best budget for value, and I just like buying cheap stuff. So far everything is working fine for me

4 minutes ago, Alien 13 said:

Unfortunately I am an Intel man combined with Nvidia wouldn't touch AMD with a barge pole I couldn't comment on those chipsets....

Alan

I've never had an AMD before, but I honestly never understood such statements. i9 9900k costs more than 3900x, why would anyone pick intel chipsets nowadays without a specific reason.

1 minute ago, celestron8g8 said:

First check and see if there is an update for your driver for your graphics board. If not see how much a new graphic cards is compared to a new computer. Right now school starting back you could possibly pickup a new computer cheaper . Check with qvc.com and hsn.com . They been running alot of sales lately . 

Hmm, not sure how a graphics card is related here?

1 hour ago, Alien 13 said:

A fast PC or laptop is just as important as any other astro spend, give me a high end gaming machine over a TAK refractor any day...

Alan

Thanks, do you have any recommendations? Would you got with a second hand cheap 2700x or get a 3900x?

24 minutes ago, Icesheet said:

You could always leave the stacking to run overnight then it’s ready for you whenever. Sounds like you’ve made your mind up to upgrade though and it wouldn’t be a bad investment anyway. 

That's the point, I wouldn't mind if I could, but I can't as I need to debayer, align, integrate, calibrate etc. I can't really do them all together(maybe with the batch preprocess, but I didn't really liked it as much, I could possibly give it another shot though, and that might save me some cash.

7 hours ago, Whistlin Bob said:

Hi- I'm running PI on a 2nd hand laptop that's no great shakes in terms of spec. A couple of times I've used the PI stacking and calibration processes and it has taken ages, not just to stack the pictures but to run through all of the processes and set all the parameters etc etc. Despite this investment of time the output was never vastly better than what I can get from DSS.

I umm'ed and arrr'ed about hardware but eventually decided to carry on using DSS and save PI for the post processing (for which I find it fantastic) because life's too short. 

Your mileage may vary, as they say (and I'd be interested to know if others can get something markedly better from doing their initial processing in PI), but if I had a few astro pounds to spare they wouldn't be going on processors and motherboards!

Even though I do agree that I could use DSS, and I could maybe spare that money on astro gear, I do think that PI is processing better than DSS, and I actually think that might be a good investment.

After all, time is something important as well, and eventually having to spend so much time solely on stacking, is a big waste of time I could use on other things.
Obviously the whole processing takes time, but when stacking, I really can't do anything on my machine, it's not a time I can learn and do much meanwhile, so I do think it might be useful.

I'm currently in a point where I'll be spending quite a lot on astro gear, so maybe spending small partion of that in a better CPU could be good as well.

2 minutes ago, fireballxl5 said:

well my 4.5 year old 6-core i7 5930K with 16GB DDR4 is looking pretty old now lol - it's runs PI well enough though🙂

It's 4.5 years old, but think about it, 6 cores 12 threads compared to 4 cores, 4 threads, now that's a huge difference, isn't it?
Probably higher core speed as well, so yeah that makes a difference. But saying that, it actually makes me think that the 2700x would be more than enough for me and I could spare that cash from 3900x on a cooling camera and a guiding camera(yeah, don't have these yet lol).

33 minutes ago, xtreemchaos said:

in that case you could look at 9900k 8core 16threads  hits 5ghz no problems but runs hot nothing a good AIO watercooler wont sort out about 10% faster than a 2700x but £200 more, I tested one just before Christmas last year and I was well impressed got it clocking at 5.2 ghz and a cb score of around 2200 I think or there about  , ive had the 2700x to 4.4 ghz with a cb of 2080 but couldn't find the screen shot to post earlier.

The 9900k is around the same price as the r9 3900x, and I can't really find it second hand at the moment, so if I'll spend more, that will be on the 3900x, otherwise I'd probably get the 2700x.

44 minutes ago, xtreemchaos said:

me id hang out for the 3900x or 3950x for the higher clocks lower tdp but saying that im a game player and benchmarking geek so I try to go for the best I could afford. 

Well I "could" afford much more than that, but I'm not sure if it's worth it, as I can always use this money for better astro gear, so getting something decent is still an option, but I am thinking about the r9.

6 minutes ago, xtreemchaos said:

I don't know how many threads PI utilizes but for future proofing the 2700x would be the better, when I get the 16 core I will prob use the 2700x for processing but the 7700k dos a good job at the mo.

I believe PI could utilize as many threads as you have.

Do you think I should get the second hand 2700x or maybe invest in the 3900x?

4 minutes ago, xtreemchaos said:

horses for courses mate, yes the 2700x is faster but im into solar imaging and there arnt any need for 16 threads yet, the 7700k is running at 5ghz with 32gb @3600 ram and isn't no slouch 1240 in CB with water cooling where as VR benefits from loads of threads.

I see, for my needs the 2700x will be better though, right?

7 minutes ago, xtreemchaos said:

im running a 2700x at the mo heres my cb15 score in case you was wondering how thay do.

ill be having 16 core one in the new year hopefully, this isn't my processing rig "ive a 7700k for that" its my VR portal. 

You use a 7700k over the 2700x for processing? Shouldn't the 2700x be much better processing wise?

23 minutes ago, Pompey Monkey said:

PI really makes use of threads and RAM for a lot of processes. About a year ago, I bought a second-hand dual-Xeon server. It's got two Xeon X5690 hex core multithreading CPUs and 96 GB DDR3 ECC RAM.

I made a 32 GB RAM drive and configured PI to use it as multiple swaps. It flies!!!

Cost was under £300.

Never really looked much into servers, what motherboard do you need? Did you already had it? Or bought that as well as part of the 300?

What do I search for exactly? And can I have my GPU connected with that?

8 minutes ago, Mr Spock said:

Additional processing power is useful, but I've found most graphics programs are held up by memory size and speed. 

That is true of course, and I do plan on getting around 32 GB RAM, but specifically with pixinsight, it seems like the amount of threads is very useful, as stacking process for each individual image is happening on a thread, so I think that should generally be the biggest factor.

1 minute ago, Mr Spock said:

It's not that simple. You can't just replace the CPU, you'd have to change the motherboard and memory too.

"When processing, the amount of threads would incredibly help here, but the price differences are fairly big, and I'll also need to get a new motherboard and RAM slots."

3 minutes ago, xtreemchaos said:

if you could stretch id go for a 2700x ,mobo and 16gb of ddr 4, more power than you could need at the moment, the zen 3000 ser are too pricey at the mo, but if you can wait thay come down after Christmas me thinks. charl.

Well the 2700x is actually the cheaper choice here for only 200$, I also thought of getting 32 GB RAM.

The good thing about that is that even if I get the 2700x, I could later on upgrade to R3 or maybe even R4 I believe, so that could be very useful.
I can wait generally, there's nothing too urgent for me, but meanwhile it does kinda hurt, as when processing, it means I can't use my PC for other things as well, and that's a lot of wasted time.

1 minute ago, Icesheet said:

I don’t use PixInsight but I’m surprised it’s taking you that long. Despite it being an older processor it’s not exactly a slouch and you have a more than decent amount of memory. The graphics card doesn’t sound bad either. I used to process with a lot worse specs than that and it didn’t take that long (DSS and PS) 

 

How many subs are you stacking and is there anything else running in the background?

 

Anyway if you are going to upgrade check how the software handles the number of cores/ threads. For example Photoshop does not necessarily benefit from hyper-threading. For most processes it’s the clock speed of the CPU that is most important and not necessarily the amount of physical cores or the ability to hyper thread. Not sure if thatnis how PixInsight handles things but best make sure whatever upgrade you are making is the best for what you will be using it for. 

Well, DSS is much much faster than the whole process in PixInsight. It really depends, but usually around 50-200 subs, and nothing else is really running in the background.

When processing with PixInsight, you could see how each thread is being used by 1 image that's been debayered/aligned/other, so I believe in this specific scenario, the amount of threads is pretty much in top priority.

Hi guys, first of all I hope it's the correct place to ask, if not please redirect me to the correct place

I started using PixInsight recently, and I'm kinda struggling with the huge amount of time I'm spending when processing an image(mostly the complete "end to end" stacking process).

It usually takes me around 1-3 hours at least before I'm even getting to the actual image processing, now of course, I don't expect it to take 2 minutes, but I believe it simply takes me too long to bear it.

Here are my PC specs:
i5 3470 OC to 3.6 GHZ
12 GB RAM
Average-fast SSD
GTX 770 4 GB(which I don't think is actually really relevant here..)

So this is a 3rd Gen CPU which I have for 7 years already, and each process is taking really really long, on the one hand, buying a CPU *ONLY* for PixInsight, that kinda sucks, but since I'm spending so much time on that(it also limits me from doing anything else on my machine during that time), I might actually invest in that.

So after this long introduction, what CPU would you recommend me? I'm currently thinking of either Ryzen 7 or wondering if it's actually worth investing more in Ryzen 9.

When processing, the amount of threads would incredibly help here, but the price differences are fairly big, and I'll also need to get a new motherboard and RAM slots.

I could get a second hand R7 2700x with cooling for around 200$, which is fairly cheap, and could incredibly boost my processing time using 16 threads without killing my wallet.

On the other hand, 3900x or 3950x costs quite more, around 650/750$, but would probably make the whole processing fly with 24 threads. But the price here is very big, and I'd also need a more expensive motherboard as well.

So after digging so much, what would you guys recommend me? Would you even get a new CPU if you were in my situation?

Thanks!

On 12/08/2019 at 01:54, carastro said:

That's not bad at all for some-one just starting out with imaging.  I think you are expecting too much of yourself, imaging is  a long learning curve that can take some time to start to get reasonably decent results.

This is my process with the image above which I captured after a year of imaging.  If I processed it today I could probably get a much better result.

Thanks! I know that I'm expecting too much, but I simply can't stand the fact that I can do better, but don't manage/don't know how to achieve that. Of course, that pushes me to learn harder.

So hopefully I'll manage to get much better in the future, thanks a lot for the help!

On 23/07/2019 at 13:08, carastro said:

Horsehead Nebula captured with Canon EOS 450D Modified for AP, captured in November 2011 from

LP location (SE London Uk).  This was captured and dithered in APT by the looks of the file name.  I often use this image for processing demonstrations at astrophotography talks, probably the only reason I still have it to hand.  

Post up your result for CC and advice. 

Carole 

 

Horsehead 29.11.11 20 x mons 800 ISO ED120 APT dither.tif 70.13 MB · 8 downloads

So kinda bumping an old thread right now, but these are my results of that:

I don't really like the results honestly, I don't feel like they're good enough, people are able to get some amazing results with my data, but I never manage to get much our of it.

Currently I'm mostly using ABE/DBE, BG naturalization/calibration/photometric calibration, multiscale linear transform, SCNR, HDR multiscale transform if it fits, and of course histogram, curves transformations and screen transfer.

Still looking into learning more tools, but that's pretty much what I know so far.

I also got some M16 data from a friend and got to this:

I believe it was taken with narrowband, once again, feels like I could get some better results, I'm just not really sure how.

So if you or anyone else got some tips on what I could improve from my images, that would be really awesome.

Also, if you got your final version of horse nebula with this data, that would also be very interesting to see!

And of course, thanks for letting me practice with this!

21 hours ago, wimvb said:

It means that you haven't installed the arcsinh stretch module. I can repost the prkcess without it. But you should consider installing arcsinh stretch, it's really great.

https://www.cloudynights.com/topic/569440-pixinsight-arcsinh-stretch/

Or the pixinsight forum

 

Awesome! Weird I searched for that and didn't saw that.

I didn't even knew that we can do such thing with pixinsight! I should've known that before so I could see some different results by people.

This is gonna be so useful for learning, thank you so much for that, really appreciate that!

10 hours ago, wimvb said:

Here is the PixInsight process container

msacco_m31.xpsm 242.28 kB · 1 download

When I'm trying to open it I'm getting this:

What does that mean exactly?