/Tech2h ago

Pieter Levels, PhotoAI founder, says his €6,562 custom PC nearly doubled in value due to surging GPU and RAM prices

Its RTX 5090 GPU can run the Qwen 27B model.

545061041107.8K

#504

Original post

@levelsio@levelsio#627inTech

My gaming PC I built last year to play Flight Simulator 2024 has almost doubled in value now

A year ago I paid €6,562 for it, now it's worth €12,073 (+84%)

The GPU 2x'd, RAM 4x'd, and SSD up a bit

Never expected buying a gaming PC would outperform the S&P500!

@levelsio@levelsio

Best friend @rameerez came to the rescue! That's why it's nice to live in a little village

We had to uninstall the entire CPU fan

It was completely bent out of shape (see pic), like completely crooked, probably by the pillow they put in for transport and they filled it with too much air/gas and that bend the CPU fan heatsink frame

The store didn't wanna help at all, they wanted us to send the whole giant box back to Spain (!) because even though it was sold and assembled in Portugal they can't fix it here 😂 typical experience here btw

So we did it ourselves, we bent it back into somewhat 90 degree normal shape, put thermal paste again, re-installed it, and booted it and it worked!

We already played @MSFSofficial 2024 on it at max really nice!!!

Thanks everyone here for recommending the specs, it's a great machine, RTX 5090 liquid cooled and AMD Ryzen 9 9950X3D air cooled 👌

11:01 AM · Jun 25, 2026 · 64.1K Views

Sentiment

Many users find the doubling value of Flight Simulator gaming PCs fascinating for hardware retention and future local LLM performance on RTX cards, while others call the surge bubble territory with excessive heat and weak local models.

Pos

25.0%

Neg

75.0%

5 comments with sentiment.

Cluster Engagement

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

Posts from X

Most Activity

VIEWS29.1KBOOKMARKS15LIKES155RETWEETS1REPLIES35

@levelsio@levelsio

And yes I've tried local LLMs but with just 1x RTX 5090 it's painfully slow and useless

@levelsio@levelsio

My gaming PC I built last year to play Flight Simulator 2024 has almost doubled in value now

A year ago I paid €6,562 for it, now it's worth €12,073 (+84%)

The GPU 2x'd, RAM 4x'd, and SSD up a bit

Never expected buying a gaming PC would outperform the S&P500!

1h29.1K15515

@levelsio@levelsio

@ifourth My Alibaba stock from 2020 didn't hhaah

2h46851

Marcos Pereira@marcospereeira

@levelsio tinybox green is 75k now, I think used to be 25k?

2h2681

ifourth@ifourth

@levelsio How come everything you buy goes up?! 😂

2h4741

andrew chen@andrewchen

@levelsio 5090 can run qwen 27b dense at good speeds, given a little optimization!

@levelsio@levelsio

And yes I've tried local LLMs but with just 1x RTX 5090 it's painfully slow and useless

1h69330

Gandalf Stormdrain@auroter

Yeah if you want to run local LLMs it's extremely expensive.

If you are thinking about saving money, then you have no business running local LLMs.

But if you value other things like: - Controlling what runs on your computer - Transparency - Open source - Latency - Running models that actually allow you to research and develop AI projects - Running models which built to serve the people

Then you might want to run local AI.

A lot of people are starting to see better performance with models like GLM 5.2, Deepseek V4 and even the smaller Qwen models VS Frontier, especially when it comes to certain categories of software engineering. I have certainly noticed a difference.

One thing to consider regarding cost is that the API rates are still heavily subsidized by the Frontier Labs. Heavily, heavily subsidized. You're paying $200 a month for a Claude Code subscription that's giving you like $8,000 a month worth of API credits that are *still* subsidized. They're probably routing you to more heavily quantized models when you're on the subscription, but still.

AI is expensive. If you want frontier quality AI running in your house, you need to be prepared to spend about as much as you would on a car. And if you happen to have the money and the desire, it is always a good time to invest because those prices are not coming down any time soon on that hardware. I can guarantee you that. I invested $50k in an AI rig a while ago and the same machine today costs $65k at lower profit margins.

1h1101

Javi Lopez ⛩️@javilopen

@levelsio BUBBLE TERRITORY

2h4662

Franz Bruckhoff@franzbruckhoff

@levelsio For some time this also worked well for certain EVs :)

2h4481

Alex 🔔 | updatify.io@_avdept

@levelsio Same shit here. I bought around the same time as you. Can't imagine price of 128gig of ram nowadays

2h2801

Nikita Belokopytov@NikiBelokopytov

@levelsio 120tps on generation with MTP 27B dense, plans and executes well in OpenCode, interactive mode. What have you tried?

1h492

Niklas ⚡️@Niklas_Sikorra

@levelsio Maybe you should sell it and buy a new one, after the prices drop?

2h324

Rui Sousa@heyitzami

@levelsio me looking at the 8gb ram that i lost when changing home

1h501

Renato@renatomoraesp

@levelsio One 5090 running NVFP4 Qwen3.6 27b with MTP can easily get you close to 100 tok/s, likely more with a bit of tweaking, and at the full 262k context size. Isn’t that fast enough? SOTA doesn’t run much faster than that in the popular agents today. 5.5 in Codex fast mode is slower

46m491

Kevin Naughton Jr.@KevinNaughtonJr

@levelsio guess you just have to play more flight sim then :P

1h155

adidshaft@adidshaft

@levelsio we're really far. I'd rather use it for Gaming than running local LLMs. Or, maybe small model tuning.

1h154

klöss@kloss_xyz

@levelsio wild

1h110

EV_Trapper@EV_Trapper

@levelsio I make it happen with 1 radeon VII, but I'm not running full power models honestly

1h331

orlie@sunglassesface

@levelsio What about GTA6 are u gonna play it?

1h96

juzy@juzybits

@levelsio you can't keep getting away with it!!! 😤

1h88

Sandi Slonjšak@sandislonjsak

@levelsio People were dunking on me a couple of days ago because I said that I am not interested in local models because 15k investment in PC seems to much

1h82