Help me decide on a new rig for PI

I would second Alejandro’s comments. If you’re not gaming, the higher end GPU isn’t going to make much difference. Either choice is a BIG step up over a CPU alone with RC Astro tools. A 5060 would be more than adequate.

Re PI, I found that though system RAM is a nice-to-have, jumping from 32 to 64 GB for me wasn’t much of a difference. What matters is having one or more NVMe drives - set them up as swap space for PI, and also process your images from them. The difference in performance is substantial.

And obviously PI scales very linearly with number of CPU cores. So the more cores (rather threads) the better there.

I am getting a good 2TB SSD.

The Ti version has 144 tensor cores vs 120 for the non Ti. Both have 8GB VRAM. RC Astro products use TensorFlow. So was thinking about the Ti card but the non Ti is heavily discounted. There should be ~$80 price difference but the Ti version is $200 more.

I just pulled the trigger on a new PC as well, below were some of my focus points:

1) A CPU with proven performance in PixInsight.

For this, I used the PixInsight benchmark page to compare results across different CPUs. I ended up choosing the AMD Ryzen 9 9950X3D. I went with the 3D model mostly because I like to pretend I have time to game every now and then, otherwise I would have went with the 9950X.

2) 6000MHz DDR5 memory.

I went with 128GB of Crucial Pro (4×32GB). This is probably more then necessary, 64GB is plenty for my workflows, but I wanted to make sure it didn’t become an issue later on. PixInsight does a lot of in-memory processing, and faster RAM definitely helps.

3) Fast, Large NVMe SSD

I focused on getting a fast TLC NVMe drive but at a higher storage rate. TLC drives are faster sustained writes, and this is where I would really like to see some improvement compared to my current setup (I save projects infrequently because of slow write times, and it’s probably bad practice. I’d like to make it better). I went with the Samsung 990pro 4TB.

I already have an RTX 4080, and PixInsight barely touches the GPU outside of the RC-Astro tools; those run fast enough on almost anything Nvidia, so I don't think a high-end GPU is necessary. A modest Nvidia card works fine unless someone is gaming or doing GPU-heavy work outside PI.

Overall, I tried to eliminate the bottlenecks I’m currently feeling: CPU speed, memory bandwidth, and storage performance. File load/save times and certain heavy processes are where the slowdown shows up for me. Aside from the RC-Astro tools, PixInsight leans heavily on CPU, RAM, and disk; so that’s where I put my money.

AstroÅmazer · Nov 26, 2025, 06:38 PM

I am thinking of building a new rig for PI and RC Astro tools.

Comparing Intel 265F based Alienware system. Trying to decide between 5060 + 32GB RAM ($1229) and 5060 Ti + 64GB RAM ($1629)

Do I need 64GB RAM for PI? Price difference is $200.

Also, is a 5060 Ti worth $200 more than a 5060?

1. Do you need 64 GB for PixInsight?

Short answer: No, not strictly. 32 GB works. But 64 GB is the “right” target if this is a dedicated PI box.

• 32 GB is fine for:

  • APS-C / full-frame cameras

  • “Normal” projects (single panels, no crazy drizzle, no huge mosaics)

  • Occasional WBPP runs that you’re willing to let churn a bit

• 64 GB starts to pay off when:

  • You use modern big CMOS (high MP, lots of subs)

  • You do drizzle, LocalNormalization, big mosaics, or keep many images open

  • You want to avoid PI spilling to disk all the time

The PI devs and many power users generally call 64 GB a practical minimum for heavy/production work, with 128 GB as “luxury.”Cloudy Nights+3pixinsight.com+3Reddit+3

So:

• If this is a long-term PI/RC Astro machine and the price bump doesn’t kill you → go 64 GB.

• If your data volume is modest and budget is tight → 32 GB is okay, just expect longer crunch times on big integrations.

2. 5060 vs 5060 Ti for RC Astro tools

RC Astro tools care about:

1. CUDA cores / GPU compute → affects how fast BlurX/NoiseX/StarX run.

2. VRAM → need ≥6 GB VRAM to run the full models comfortably; ≥8 GB is

The RTX 5060 and 5060 Ti both clear the VRAM requirement. Main difference:

• 5060:

  • Fewer CUDA cores

  • Still a big jump over CPU-only PI/RC Astro

• 5060 Ti:

  • ~20% more CUDA cores, slightly higher clocks

  • Expect roughly that kind of gain in RC Astro processing (e.g., 25 s vs 30 s on a frame ballpark numbers, not exact)

The huge speed-up comes from “CPU → any decent RTX”, not “5060 → 5060 Ti”. People going from CPU to something like a 4060/4060 Ti already report going from minutes to tens of seconds per frame.

So:

• If the price difference is only about $200 and that also buys you 64 GB RAM, then:

  • I’d take the 5060 Ti + 64 GB – that’s a nicely balanced, long-lived PI/RC Astro box.

• If the real jump is closer to $400 (your numbers 1229 → 1629), then:

  • For pure PI/RC Astro use, I’d prioritize 64 GB RAM first,

  • And treat the 5060 Ti vs 5060 as a nice-to-have, not a must. The base 5060 will still absolutely destroy CPU-only RC Astro performance.

• PixInsight: 32 GB works, 64 GB is the sweet spot if you can swing it.

• RC Astro: any RTX 50xx with ≥8 GB VRAM is already a big win; 5060 Ti is ~20% faster but not transformational.

• If the config with 5060 Ti also gives you 64 GB for about +$200 total, that’s the one I’d buy. If it’s a much bigger jump, I’d rather have 5060 + 64 GB than 5060 Ti + 32 GB.

we have almost the same set up almost.

Is there a massive diffrence using Linux I have been thinking about getting rid of windows from my workstation.

CPU: Ryzen 9

• GPU: RTX 5080

• RAM: 96 GB DDR 5

• Storage:

  • 2 × 2 TB NVMe (system + fast workspace)

  • 4 × SSDs used as scratch

• Water cooled overclocked

Jure Menart · Nov 27, 2025, 08:38 AM

Read noise Astrophotography · Nov 27, 2025, 08:25 AM

we have almost the same set up almost.

Is there a massive diffrence using Linux I have been thinking about getting rid of windows from my workstation.

CPU: Ryzen 9

• GPU: RTX 5080

• RAM: 96 GB DDR 5

• Storage:

  • 2 × 2 TB NVMe (system + fast workspace)

  • 4 × SSDs used as scratch

• Water cooled overclocked

It would be interesting to see it being executed on exactly the same HW, so if you do it be sure to run benchmark on both OSs :)

PI Benchmark site is usually showing Linux being on top, here is one example (but note it’s really hard to do 1:1 comparison because systems are different, it does show some tendency though):

📷 image.png

wow look as those numbers think I feel like I should be getting more out of my machine
Benchmark version ...... 1.00.08

CPU Identification

CPU vendor ............. AuthenticAMD

CPU model .............. AMD Ryzen 9 7950X 16-Core Processor

System Information

Platform ............... Windows

Operating system .......

Core version ........... PixInsight Core 1.9.3 (x64)

Logical processors ..... 32

Total memory size ...... 95.145 GiB

Execution Times

Total time ............. 00:12.74

CPU time ............... 00:09.40

Swap time .............. 00:03.32

Swap transfer rate ..... 4989.745 MiB/s

Performance Indices

Total performance ...... 36921

CPU performance ........ 40267

Swap performance ....... 27636

Hi,

Interesting topic about PC’s for PI.

Anyone has a data about let me say processing a 4656×3520 pixel big image with the Image calibration process ?

The PC I have has an average processing time of 3s per image eg. I processed the other day 161 lights in 8 minutes 8 seconds. I guess I have a slow PC compared to what has been used here.

After analysing my setup I found out I can improve the benchmark by setting up correctly my SWAP files.

So consider this post as nonsense from my side….

🥴🥴🥴🥴🥴

I have been using Macbook M4 Max 48GB Ram and PI works like horse with my rig of data including upto 300 images with the dark/flats and WPB doesn’t cry at all!

I use this rig and it flies through the entire PI process, from WBPP to exporting:

i7-14700k

4070ti

32GB 5600 DDR5

2TB samsung evo NVMe.