No, not the one in the seat next or behind or ahead of you in the aircraft, but rather that MS AI guy 🙂

Here's the question:

"I run X-Plane 12.3 beta in a desktop with a Ryzen 5600x, 32 GB RAM (DDR5), a Nvidia RTX 3060Ti (8GB), 1 TB nvme2 SSD for the simulator and a 500 GB nvme2 SSD for the system (win 11). I run it with the option to lock FPS at 30, and my display in 1980 x 1024 @ 60 Hz. What should be my best choice of X-Plane Graphics settings to optimize the display and not putting both the CPU and GPU to stress?"

And the answer(s):

You've got a solid mid-range setup that can deliver smooth performance in X-Plane 12.3 beta, especially at 1080p with a 30 FPS lock. The key is to balance CPU-heavy and GPU-heavy settings so neither component becomes a bottleneck. Here's a tailored configuration to help you optimize visuals without stressing your system:

---

🛠️ Recommended Graphics Settings for Your Setup

Setting	Recommended Value	Why It Works
Rendering Resolution	Use Monitor Settings (1920x1024)	Keeps GPU load manageable and avoids unnecessary scaling.
Texture Quality	High	Your 3060Ti's 8GB VRAM can handle detailed textures without bottlenecks.
Antialiasing	2x or FXAA	Smooths edges without overwhelming the GPU.
Anisotropic Filtering	4x or 8x	Improves texture clarity at angles with minimal GPU impact.
Number of World Objects	Medium	Reduces CPU load while maintaining visual richness.
Draw Shadows	Off or Minimal	Shadows are CPU-intensive; disabling them improves performance.
Draw Parked Aircraft	Off	Saves CPU cycles, especially at busy airports.
Reflection Detail	Low	Reflections are CPU-heavy and offer limited visual gain at 1080p.
Draw Distance	Medium	Balances scenery visibility and CPU load.
Cloud Quality	Medium	Keeps GPU load reasonable while maintaining realism.

---

🔧 NVIDIA Control Panel Tweaks

• Power Management Mode: Prefer Maximum Performance
• Threaded Optimization: Auto
• Low Latency Mode: On
• Vertical Sync: Off (unless you notice tearing)

These settings help ensure your GPU runs efficiently and avoids unnecessary throttling.

---

🧠 Performance Tips

• Monitor CPU/GPU Frame Times: Use X-Plane’s built-in performance monitor to check if either component is consistently maxed out.
• Avoid Ultra Settings: Even with good hardware, flight sims are notoriously demanding. Ultra settings often cause diminishing returns.
• Keep Drivers Updated: Especially for beta versions, driver compatibility can make a big difference.

---

If you ever switch to VR or a higher resolution monitor, we’ll need to revisit these settings. Want help benchmarking or testing specific scenarios like heavy weather or dense airports?

Edited September 16, 2025Sep 16 by jcomm

On that "other" forum there is a solid rule that bans posting AI generated content - and for a very  good reason, I believe.

If you do not understand how large language models work, you should not be asking them questions like this. And if you do, you won´t.

The ONE strength that these models have is that they manage to sound convincing even if they have zero understanding of the underlying matter at hand. Just like in this case. They simply say stuff that sounds reasonable - sometimes it is, sometimes it isn´t.

Edited September 16, 2025Sep 16 by Litjan

20 minutes ago, Litjan said:

On that "other" forum there is a solid rule that bans posting AI generated content - and for a very  good reason, I believe.

If you do not understand how large language models work, you should not be asking them questions like this. And if you do, you won´t.

The ONE strength that these models have is that they manage to sound convincing even if they have zero understanding of the underlying matter at hand. Just like in this case. They simply say stuff that sounds reasonable - sometimes it is, sometimes it isn´t.

True! Just recently I asked (for my job) a question where I wasn't 100% sure about the correct answer. (In that case not Copilot but) ChatGPT responded like "You have to check the datasheets of the respective equipment and please look also into the respective norms." Wow!!! That's something I never expected 😂. At least when the question is quite specialized and "narrow", you won't get good (and safe) answers. If you ask, e.g. "What is McDonalds?" you will probably get good answers as there is much data available about that matter. But still, some information bits can be wrong. Is that common sense? I think so ... Or should I ask ChatGPT? 😉 

29 minutes ago, Litjan said:

On that "other" forum there is a solid rule that bans posting AI generated content - and for a very  good reason, I believe.

If you do not understand how large language models work, you should not be asking them questions like this. And if you do, you won´t.

The ONE strength that these models have is that they manage to sound convincing even if they have zero understanding of the underlying matter at hand. Just like in this case. They simply say stuff that sounds reasonable - sometimes it is, sometimes it isn´t.

@Litjan, 

thank you for your comments, which are pertinent OFC...

Regarding AI, well, I guess I'm used to it since the early 90s of the last century when I started teaching and investigating exactly in AI and Natural Language Processing at the University.

I did move from that area to a place where AI is also being followed since the late nineties, and can be used, and is surely being used already, and will even more in the future...

Anyway, in this particular situation as you can see the answers aren't that much detached from what was expected ...

 

Edited September 16, 2025Sep 16 by jcomm

Bjoern.AI's 2025 language model says that a 5600X and a 3060Ti is not a solid mid-range setup anymore, but rather at the lower end.

Edited September 16, 2025Sep 16 by Bjoern

This AI nonsense is getting out of hand. One guy on other forum used it to check his log, which it has no understanding of. People dont trust people, but will use AI that is written by people you dont know. Mad world we live in.

4 hours ago, Litjan said:

If you do not understand how large language models work, you should not be asking them questions like this. And if you do, you won´t.

I know several companies where using AI is mandatory and several where you get FIRED if you don't use AI to complement your job.  It can increase productivity tremendously, saving time and money. Outright banning anything AI is an overreaction IMO from these forums. If you know how to use it properly, it can become an invaluable source.

Claiming that AI is out of control is pure nonsense. As a software developer, I rely on AI tools like Copilot every day—and the boost in productivity and time savings is nothing short of remarkable.

 

3 minutes ago, Cap737Delta said:

Claiming that AI is out of control is pure nonsense. As a software developer, I rely on AI tools like Copilot every day—and the boost in productivity and time savings is nothing short of remarkable.

 

Thats as long as you think you can trust something that cant be questioned. But each to his own. 

Have faith in AI or the 100 percent accuracy of forum posters?  Hmmm 😁😁☺️

1 hour ago, Bjoern said:

Bjoern.AI's 2025 language model says that a 5600X and a 3060Ti is not a solid mid-range setup anymore, but rather at the lower end.

No no no!!!! My rig is GREATZZZZ!!!!

It depends on the AI and especially the data it was trained on. The best AI algorithm can't give you the desired result, because ... I installed Stable Diffusion (for generating pictures) locally and its manual said, not literally but in essence: "If the model is trained with cat content and you ask it something about dogs - expect the unexpected" 😄 (the part about cats and dogs is actually written that way, but I rephrased the second part of the sentence a little for dramaturgical reasons).

* Cat content = sweet 😻

Edited September 16, 2025Sep 16 by uwespeed

10 minutes ago, uwespeed said:

the data it was trained on

And or the trainer. If its anything like our currdnt schools, where done.

9 hours ago, brinx said:

I know several companies where using AI is mandatory and several where you get FIRED if you don't use AI to complement your job.

I totally believe you. I also know several companies where the goal is to sell inferior products at an inflated price to gullible customers and you get FIRED if you don´t play along.

I personally think that is not the correct thing to do.

So just because people do it - it does not mean that it is the right thing to do.

There is a strong tendency to offload work to machines - it makes sense in an economical way and we have done it for millenia. Especially if someone feels that work is tedious - as is "thinking" for many people these days. 

It is natural to look for help in this case, here is an interesting excerpt from an article - from the US, but things aren´t looking much different in other countries, I am sure:

"In reading, the average score in 2024 was the lowest score in the history of the assessment, which began in 1992. Thirty-two percent of high school seniors scored below “basic,” meaning they were not able to find details in a text to help them understand its meaning."

https://www.pbs.org/newshour/education/u-s-high-school-students-continue-to-lose-ground-in-math-and-reading-as-yearslong-decline-persists

So yeah, if you have an employee force that can´t concentrate on content that is longer than the typical Tik-Tok video and has trouble understanding what they read - I would make AI use mandatory, too. It will likely give you better results than those folks coming up with their own stuff.

I think humanity will follow the evolutionary pattern it has shown before. Advances in medicine made people less resilient, advances in machinery made people less physically strong, and advances in AI will make people more dumb.

44 minutes ago, Litjan said:

I totally believe you. I also know several companies where the goal is to sell inferior products at an inflated price to gullible customers and you get FIRED if you don´t play along.

I personally think that is not the correct thing to do.

So just because people do it - it does not mean that it is the right thing to do.

There is a strong tendency to offload work to machines - it makes sense in an economical way and we have done it for millenia. Especially if someone feels that work is tedious - as is "thinking" for many people these days. 

It is natural to look for help in this case, here is an interesting excerpt from an article - from the US, but things aren´t looking much different in other countries, I am sure:

"In reading, the average score in 2024 was the lowest score in the history of the assessment, which began in 1992. Thirty-two percent of high school seniors scored below “basic,” meaning they were not able to find details in a text to help them understand its meaning."

https://www.pbs.org/newshour/education/u-s-high-school-students-continue-to-lose-ground-in-math-and-reading-as-yearslong-decline-persists

So yeah, if you have an employee force that can´t concentrate on content that is longer than the typical Tik-Tok video and has trouble understanding what they read - I would make AI use mandatory, too. It will likely give you better results than those folks coming up with their own stuff.

I think humanity will follow the evolutionary pattern it has shown before. Advances in medicine made people less resilient, advances in machinery made people less physically strong, and advances in AI will make people more dumb.

Or on a different scale at a different time:

“How much is 5*82?”
“425.”
"Wrong."
"No, right because the calculator gave me the result."
"What do you get when you calculate it in your head?"
"I'm not able to do that."
"Ok, then please type it into the calculator four more times and see how often you get 425. Most probably you will get 4 times 410 as result - because you mistyped the first time, but were unable to question that."

If the four additional rolls produce four different results, then you have a bigger problem, of course. 😄 

P.S. Of course I know that we are deviating (a bit) from the original topic - but as long as it stays civil ...

Edited September 17, 2025Sep 17 by uwespeed