Back to forums

Acceptable laptop for image processing...

Pleiades Astrophoto PixInsight Processing Buying Advice
AstroÅmazerPeter
•25 replies•755 views
Terrance avatar
Terrance avatar
I'm in need of a new laptop computer for image processing while I'm on the road.  My budget is $1,200 - $1,500.  I saw this:

 

https://www.bestbuy....wE&gclsrc=aw.ds

 

I'd like to run the latest version of PixInsight with add ons such as Seti-Astro, and RC-Astro etc...

Are the specs for this computer adequate for my needs?  What are some other options?

 

Thanks in advance!

 

Terrance
Dean Jacobsen avatar
Dean Jacobsen avatar
What cameras do you now use or anticipate that you will be using during the lifespan of the computer?  Less processing power is required for smaller format cameras vs. full frame cameras.
Nick Grundy avatar
Nick Grundy avatar
Hi Terrance, If you can get something with an Nvidia GPU, RC-astro processes will be much faster. (not sure if someone else has tried them with the Iris GPU processor they are using now)

The Thunderbolt 4 ports on there will be important and help using an external drive. (if you need that)

I assume you'll stack and keep calibration files elsewhere? So you'll just be working with master files.

check out the ASUS Rog Zephyrus G14 series. They are intended for gaming, but will benefit your Pixinsight use.
Helpful
Tony Gondola avatar
Tony Gondola avatar
Terrance:
I'm in need of a new laptop computer for image processing while I'm on the road.  My budget is $1,200 - $1,500.  I saw this:

 

https://www.bestbuy....wE&gclsrc=aw.ds

 

I'd like to run the latest version of PixInsight with add ons such as Seti-Astro, and RC-Astro etc...

Are the specs for this computer adequate for my needs?  What are some other options?

 

Thanks in advance!

 

Terrance

That's much more powerful than my 2.3 GHz i7 machine and it works just fine.
Well Written Respectful
Rick Krejci avatar
Rick Krejci avatar
Do you prioritize getting through a WBPP stack faster, or having interactive processes (RC tools) be faster?    Having even an older Nvidia graphics card would speed up the interactive RC tools very greatly.    I definitely wouldn't drop below 32GB memory that the one you posted has.     But I'd personally rather have an I7 with a base Nvidia graphics card than an I9 without one.    I'm not sitting waiting while WBPP is running.    But if I have to wait a minute or more each time I experiment with BXT, NXT or SXT, I'd get very frustrated.

And, if it's not up-gradable with another SSD, I'd consider going with 2TB rather than 1.
Helpful Insightful Respectful Engaging Supportive
Michael Jarvis avatar
Michael Jarvis avatar
Swimming up-stream on this one.  I have an M1 MacBook Pro with 32 GB memory.  Runs both PI and PS simultaneously with no issues.  WBPP times vary with the number of images/session but no complaints.  For Pixinsight, I have all the plug-ins and have upgraded multiple PI versions without problems.  May be worth a look as the new M4 processors are even faster.
Helpful
Oscar avatar
Oscar avatar
I would get a gaming laptop with as much RAM as I can afford, and that uses an Nvidia GPU
AstroÅmazer avatar
AstroÅmazer avatar
This spec has more everyday use features like touch screen etc. which you do not need for processing. It does not have a dedicated graphics card. The Processor is rated 99 TOPS but the Intel Arc Graphics processor on it is rated 77 TOPS. Look for a laptop which has a NVidia graphics card and less frills.

Now a deep dive into components. The processor, Intel Core Ultra 9 285H is a general purpose processor with 6 performance cores, 8 efficient cores and 2 low power cores. This means only 6 will be running at max turbo speeds. The efficient cores will be close to that but the low power one will not. It is good for everyday use.

When you process images, more threads means more work done in parallel. So I think you will end up with 14 threads max for image processing (Don't think this processor has hyper threading). For comparison, my 5 yr old laptop has 12 threads running at about the same speed as the efficient cores and has a graphics card which helps. Architecture has improved but raw speed and parallelism is still the king.

As for memory, it has pretty fast LPDDR5 but 7467MHz is less than what the processor can handle (8400MHz). And storage wise, it has a PCIe4 NVMe SSD. PCIe5s which the processor supports, are in the market but maybe not at this price point. Everything else looks good. (Wi-Fi 7, Thunderbolt 4 support). So long story short, if you plan to use it for more than image processing, go for it. I predict a 6 year usage life before it starts struggling to keep up. If you want to buy something for just image processing, look for something without convenience features like touch screens, better audio systems etc.
Helpful Insightful Engaging Supportive
Keith Mombourquette avatar
Keith Mombourquette avatar
I have a Dell G16 with i9 cpu and 32gb ram, and it works very well with PI.
Pedro A. Sampaio avatar
Pedro A. Sampaio avatar
I have an ASI2600MC pro and shoot 5 to 10 min subs, so take that into account.

By that price range I would suggest the basic model Macbook Pro M4 (not the M4 pro) with 512gb storage and 16gb RAM. It gols for around 1500 USD. Alternatively you could go with a MBA, for a lower price and lighter weight.

I just bought my MBP M4 and was completely amazed by the performance using PI. It completely obliterated my 2022 Alienware M15 R6 with an 12th generation I7, Nvidia 3070 and 32gb RAM [edit] - despite it having only slightly lower theoretical performance on benchmarks.

The difference in speed is huge on all processes, including the RC Astro ones. For instance, whereas NXT, BXT or StarX would take a couple of minutes on my Alienware, on my MBP it would take less than 30s (sometimes less than 20 or so).

Also, battery life efficiency. My last image was processed on my MBP, for around 3-4 hours, and the battery used was less than 30%.

Also, display quality, which on the MBP is probably the best in the business by this price range.

That said, I haven't used WBPP this time,  so take that into account - I ran it on my Alienware thinking it would be too much for my MBP, but now I think I'll run WBPP on my MBP just to see how fast it goes.

So there you go, my recommendation, from someone who was a Windows user for the past 20 years or so. If you don't have an aversion to MacOS, that would be a great notebook.
Helpful Insightful Respectful Engaging
AstroÅmazer avatar
AstroÅmazer avatar
I still use my old laptop with a 10th gen i7, 16GB RAM, 6GB NVidia GeForce RTX 2060 for non-PI processing (Siril+Gimp+GraXpert+StarNet). It takes a while to do stuff on it.

My main workhorse is a desktop with 14th gen i9, 32GB RAM, 8GB NVidia Quadro T1000. I'd say it is 10x to 25x faster depending on the task. However, I do not run PI.

And I use the desktop for more than astro… CAD/general purpose programming, mostly Python based with some Pytorch/TensorFlow. Also some FPGA programming.

The GPU in the desktop, unlike the laptop is not gaming focused. It is good at CAD tasks and I think that will translate to PI. A more recent model would be one of the NVidia Ada architecture GPUs. 

To give you a comparison of the amount of data processing, I was running an OpenCV based homography task (think astro  registration with LancZos4 based interpolation) on data from a Nikon D7500 (APS-C) and a Nikon Z5 (FF) for a non-astro experiment and the desktop was fine with it. The Z5 is astro-modified and I use it half the time for astro.

OS wise, the laptop is still on Windows 10, the desktop has Windows 11 with Ubuntu 22.04 on WSL for non-astro stuff.
Helpful
Terrance avatar
Terrance avatar
I'd like to thank everyone for their input!  I neglected  to mention I'd only be using the laptop for editing my master files while I am on the road.  I'll use something like this:

https://www.cpusolutions.com/store/pc/High-Performance-AMD-Ryzen-9-9950X-PC-16-Cores-32-Threads-5-7GHz-1TB-NVMe-SSD-32GB-DDR5-Windows-11-190p6750.htm


with an NVIDIA card, 64GB of DDR5 RAM, and 2TB SSD when using WBPP and processing from home.
Dean Jacobsen avatar
Dean Jacobsen avatar
Then my suggestion would be to use the desktop instead with a good designer or photo editing display.  smile
AstroÅmazer avatar
AstroÅmazer avatar
Terrance:
I'd like to thank everyone for their input!  I neglected  to mention I'd only be using the laptop for editing my master files while I am on the road.  I'll use something like this:

https://www.cpusolutions.com/store/pc/High-Performance-AMD-Ryzen-9-9950X-PC-16-Cores-32-Threads-5-7GHz-1TB-NVMe-SSD-32GB-DDR5-Windows-11-190p6750.htm


with an NVIDIA card, 64GB of DDR5 RAM, and 2TB SSD when using WBPP and processing from home.

Your desktop is probably better than mine.

If your choice of laptop is for general use with the occasional astro processing, this spec is more than enough.
Rick Krejci avatar
Rick Krejci avatar
Terrance:
I'd like to thank everyone for their input!  I neglected  to mention I'd only be using the laptop for editing my master files while I am on the road.  I'll use something like this:

https://www.cpusolutions.com/store/pc/High-Performance-AMD-Ryzen-9-9950X-PC-16-Cores-32-Threads-5-7GHz-1TB-NVMe-SSD-32GB-DDR5-Windows-11-190p6750.htm


with an NVIDIA card, 64GB of DDR5 RAM, and 2TB SSD when using WBPP and processing from home.

If that's the case, then getting a lightweight basic laptop with a good (OLED) and large enough display, much like the one you linked, is your best bet.    Or a macbook as was mentioned if that's your preferred flavor.    You'll be a bit slower for RC tools with integrated graphics, but hauling a gaming laptop with their heat and weight and short battery life likely isn't worth it.  I have a similarly powered desktop as you linked for most processing, and I often just remote into it for major editing (RC tools and such).
donsinger avatar
donsinger avatar
I have an M4 Mac Mini with 32g of RAM and PI is very fast……..WBPP is 50% faster than my i9 windows desktop…RC Astro tools finish in less than 30 seconds, and most astronomy software has a MacOS flavor these days.
AstroStew avatar
AstroStew avatar
Dean Jacobsen:
What cameras do you now use or anticipate that you will be using during the lifespan of the computer?  Less processing power is required for smaller format cameras vs. full frame cameras.

Actually this is not strictly true, as with full frame cameras you will be taking long exposure images which require no real processing power, but you will need a lot of storage space as they are big files, with smaller format cameras you may be doing lucky imaging or even planetary imaging by taking AVI video’s, which will require more processing power and USB 3 for sure, but take up less space on your hard drive.
James Peirce avatar
James Peirce avatar
It would probably be hard to beat one of the Apple silicon MacBook Pros, if you’re comfortable with macOS. The ARM architecture chips run extremely cool and are subject to far less throttling than traditional desktop CPUs on other machines, and acceleration for tools like the RC Astro suite is baked in. And you’re generally going to end up with a much higher quality, longer-lived, quieter, and more reliable computer.

On the desktop side, where comfortable with Windows or Linux, the best bang for the buck is easily in building your own PC to pay to Pix’s favorite specifications.
Helpful
Peter avatar
Peter avatar
I would be careful with the 9950x build. I did one about 2 months ago and watched in horror as the 8 CCD0 and 7 CCD1 cores switched off  one by one leaving the last running at 800MHz while running WBPP. You also need a fresh install of Windows so you get the Xbox stuff installed due to core parking. 
It could have been a quirk of the motherboard CPU combination I had, there are as many combinations as grains of sand on the beach, though running basic stuff it was fine.
The CCD1 cores run between 300MHz and 400MHz slower than CCD0 cores. It's how the membrane between the CCDs is configured. In effect its P and E cores like the Intel.
In benchmarks the 9950x rules but we live in the real world so take them as a guide not a definitive reason to go that route

I've gone to a i9 285 with 192Gb memory. Running all 24 ( 8 P and 16 E ) cores synced at 5.2GHz and all 4 memory at 5400MHz. The graphics card can be any size Nvidea you have if you want the AI for RC tools. I ran a GTX 1080 until I bought a RTX 5070 which to be honest is overkill for a Pixinsight build.

A decent build based on a i9 285 a Z890 MB with 96Gb memory, Case, 360mm cooler and a RTX 3060  850W PSU will be around £1564 or less if you already have a graphics card, in which case go for 192Gb memory. Nvme drives and SSD 2.5 drives can be added later or transferred.  The plus side to doing your own build is that you can expand as much as you want to over time when finances allow and not locked in as with a laptop.

I can't say anything about Macs as I have never used one, The good people on here who do will be more than happy to give you excellent advice.
Helpful
Sean Boon avatar
Sean Boon avatar
I bought a Lenovo Legion Pro 7i and if I could do it over again, I would buy a Macbook Pro with an M3 Pro chip and 48GB of RAM.  I've already had my motherboard overheat on the Lenovo and while it might stack faster than the Macbook Pro, I'd trade that ability for a Macbook that doesn't require a 5 lb power brick, weighs a ton and has a tendency to overheat.
Médéric Hébert avatar
Médéric Hébert avatar
Peter:
I would be careful with the 9950x build. I did one about 2 months ago and watched in horror as the 8 CCD0 and 7 CCD1 cores switched off  one by one leaving the last running at 800MHz while running WBPP. You also need a fresh install of Windows so you get the Xbox stuff installed due to core parking. 
It could have been a quirk of the motherboard CPU combination I had, there are as many combinations as grains of sand on the beach, though running basic stuff it was fine.
The CCD1 cores run between 300MHz and 400MHz slower than CCD0 cores. It's how the membrane between the CCDs is configured. In effect its P and E cores like the Intel.
In benchmarks the 9950x rules but we live in the real world so take them as a guide not a definitive reason to go that route

I've gone to a i9 285 with 192Gb memory. Running all 24 ( 8 P and 16 E ) cores synced at 5.2GHz and all 4 memory at 5400MHz. The graphics card can be any size Nvidea you have if you want the AI for RC tools. I ran a GTX 1080 until I bought a RTX 5070 which to be honest is overkill for a Pixinsight build.

A decent build based on a i9 285 a Z890 MB with 96Gb memory, Case, 360mm cooler and a RTX 3060  850W PSU will be around £1564 or less if you already have a graphics card, in which case go for 192Gb memory. Nvme drives and SSD 2.5 drives can be added later or transferred.  The plus side to doing your own build is that you can expand as much as you want to over time when finances allow and not locked in as with a laptop.

I can't say anything about Macs as I have never used one, The good people on here who do will be more than happy to give you excellent advice.

Sounds to me that you used a 9950x3d, those have different cores for speed and memory. The 9950x should only have speed cores and don't need the core parking. Also I've found that process lasso does a better job at controlling which core should be active.
Dark Matters Astrophotography avatar
Dark Matters Astrophotography avatar
Pre processing I would get a i7 or i9 mini PC that you can just cycle projects through. Run it headless and just feed it data to munge for you. Max out the ram on it for faster times. Consider Linux for more gains.

Post processing is great on MacBook. All of the RC PI modules support native Apple Silicon so they run in seconds.
Peter avatar
Peter avatar
Médéric Hébert:
Peter:
I would be careful with the 9950x build. I did one about 2 months ago and watched in horror as the 8 CCD0 and 7 CCD1 cores switched off  one by one leaving the last running at 800MHz while running WBPP. You also need a fresh install of Windows so you get the Xbox stuff installed due to core parking. 
It could have been a quirk of the motherboard CPU combination I had, there are as many combinations as grains of sand on the beach, though running basic stuff it was fine.
The CCD1 cores run between 300MHz and 400MHz slower than CCD0 cores. It's how the membrane between the CCDs is configured. In effect its P and E cores like the Intel.
In benchmarks the 9950x rules but we live in the real world so take them as a guide not a definitive reason to go that route

I've gone to a i9 285 with 192Gb memory. Running all 24 ( 8 P and 16 E ) cores synced at 5.2GHz and all 4 memory at 5400MHz. The graphics card can be any size Nvidea you have if you want the AI for RC tools. I ran a GTX 1080 until I bought a RTX 5070 which to be honest is overkill for a Pixinsight build.

A decent build based on a i9 285 a Z890 MB with 96Gb memory, Case, 360mm cooler and a RTX 3060  850W PSU will be around £1564 or less if you already have a graphics card, in which case go for 192Gb memory. Nvme drives and SSD 2.5 drives can be added later or transferred.  The plus side to doing your own build is that you can expand as much as you want to over time when finances allow and not locked in as with a laptop.

I can't say anything about Macs as I have never used one, The good people on here who do will be more than happy to give you excellent advice.

Sounds to me that you used a 9950x3d, those have different cores for speed and memory. The 9950x should only have speed cores and don't need the core parking. Also I've found that process lasso does a better job at controlling which core should be active.

The X3d hadn't been released. I think it came out 3 days after I bought the 9950x on March 3rd. Like I said. It could be the Mboard combination I used but MSI are a good make. Running Ryzen Master showed the speed difference and the cores dropping off 1 by 1 in real time. It's what happened to me hence my caveat. I'm sure the i9 has its quirks as well, as the 13 and 14 series has shown no one is perfect. Though coming from an 18 core Xeon which is old tech. I was surprised and disappointed with what happened plus the lack of pcie lanes. My very old 6 core i7 6850k has 40 same as the 18. You would think that would be normal by now. Shows you how long its been since I needed to upgrade.
AstroÅmazer avatar
AstroÅmazer avatar
Peter:
Médéric Hébert:
Peter:
I would be careful with the 9950x build. I did one about 2 months ago and watched in horror as the 8 CCD0 and 7 CCD1 cores switched off  one by one leaving the last running at 800MHz while running WBPP. You also need a fresh install of Windows so you get the Xbox stuff installed due to core parking. 
It could have been a quirk of the motherboard CPU combination I had, there are as many combinations as grains of sand on the beach, though running basic stuff it was fine.
The CCD1 cores run between 300MHz and 400MHz slower than CCD0 cores. It's how the membrane between the CCDs is configured. In effect its P and E cores like the Intel.
In benchmarks the 9950x rules but we live in the real world so take them as a guide not a definitive reason to go that route

I've gone to a i9 285 with 192Gb memory. Running all 24 ( 8 P and 16 E ) cores synced at 5.2GHz and all 4 memory at 5400MHz. The graphics card can be any size Nvidea you have if you want the AI for RC tools. I ran a GTX 1080 until I bought a RTX 5070 which to be honest is overkill for a Pixinsight build.

A decent build based on a i9 285 a Z890 MB with 96Gb memory, Case, 360mm cooler and a RTX 3060  850W PSU will be around £1564 or less if you already have a graphics card, in which case go for 192Gb memory. Nvme drives and SSD 2.5 drives can be added later or transferred.  The plus side to doing your own build is that you can expand as much as you want to over time when finances allow and not locked in as with a laptop.

I can't say anything about Macs as I have never used one, The good people on here who do will be more than happy to give you excellent advice.

Sounds to me that you used a 9950x3d, those have different cores for speed and memory. The 9950x should only have speed cores and don't need the core parking. Also I've found that process lasso does a better job at controlling which core should be active.

The X3d hadn't been released. I think it came out 3 days after I bought the 9950x on March 3rd. Like I said. It could be the Mboard combination I used but MSI are a good make. Running Ryzen Master showed the speed difference and the cores dropping off 1 by 1 in real time. It's what happened to me hence my caveat. I'm sure the i9 has its quirks as well, as the 13 and 14 series has shown no one is perfect. Though coming from an 18 core Xeon which is old tech. I was surprised and disappointed with what happened plus the lack of pcie lanes. My very old 6 core i7 6850k has 40 same as the 18. You would think that would be normal by now. Shows you how long its been since I needed to upgrade.

Intel and AMD are all about power saving right now. I think a 14th Gen i9 or comparable Xeon is more consistent with CPU frequency scaling and keeping cores running full steam.

I hate these new chiplet designs and performance+efficient cores. You never know how many cores are running and at what frequency.
Peter avatar
Peter avatar
AstroÅmazer:
Peter:
Médéric Hébert:
Peter:
I would be careful with the 9950x build. I did one about 2 months ago and watched in horror as the 8 CCD0 and 7 CCD1 cores switched off  one by one leaving the last running at 800MHz while running WBPP. You also need a fresh install of Windows so you get the Xbox stuff installed due to core parking. 
It could have been a quirk of the motherboard CPU combination I had, there are as many combinations as grains of sand on the beach, though running basic stuff it was fine.
The CCD1 cores run between 300MHz and 400MHz slower than CCD0 cores. It's how the membrane between the CCDs is configured. In effect its P and E cores like the Intel.
In benchmarks the 9950x rules but we live in the real world so take them as a guide not a definitive reason to go that route

I've gone to a i9 285 with 192Gb memory. Running all 24 ( 8 P and 16 E ) cores synced at 5.2GHz and all 4 memory at 5400MHz. The graphics card can be any size Nvidea you have if you want the AI for RC tools. I ran a GTX 1080 until I bought a RTX 5070 which to be honest is overkill for a Pixinsight build.

A decent build based on a i9 285 a Z890 MB with 96Gb memory, Case, 360mm cooler and a RTX 3060  850W PSU will be around £1564 or less if you already have a graphics card, in which case go for 192Gb memory. Nvme drives and SSD 2.5 drives can be added later or transferred.  The plus side to doing your own build is that you can expand as much as you want to over time when finances allow and not locked in as with a laptop.

I can't say anything about Macs as I have never used one, The good people on here who do will be more than happy to give you excellent advice.

Sounds to me that you used a 9950x3d, those have different cores for speed and memory. The 9950x should only have speed cores and don't need the core parking. Also I've found that process lasso does a better job at controlling which core should be active.

The X3d hadn't been released. I think it came out 3 days after I bought the 9950x on March 3rd. Like I said. It could be the Mboard combination I used but MSI are a good make. Running Ryzen Master showed the speed difference and the cores dropping off 1 by 1 in real time. It's what happened to me hence my caveat. I'm sure the i9 has its quirks as well, as the 13 and 14 series has shown no one is perfect. Though coming from an 18 core Xeon which is old tech. I was surprised and disappointed with what happened plus the lack of pcie lanes. My very old 6 core i7 6850k has 40 same as the 18. You would think that would be normal by now. Shows you how long its been since I needed to upgrade.

Intel and AMD are all about power saving right now. I think a 14th Gen i9 or comparable Xeon is more consistent with CPU frequency scaling and keeping cores running full steam.

I hate these new chiplet designs and performance+efficient cores. You never know how many cores are running and at what frequency.

Not wanting to ruin Terrances post. This is my last reply. That's very true. I was running all cores at max without having to do anything.