Posts on Peter Truchly

Windows 98 SE on Sandy Bridge: the debugging story

Sat, 12 Sep 2026 00:00:00 +0200

After playing with Windows XP on my “new” Optiplex 990 I asked myself a question: “Would Windows 98 SE run on this Sandy Bridge as well?” By running I mean: accelerated 3D graphics and 3D audio (EAX) without any hangs or lockups, being able to play games and run benchmarks.

Component	my Optiplex 990 configuration
CPU	Core i5 2500
Chipset	Q67 (Sandy Bridge, 6-series PCH)
RAM	4GB, 2+2, dual channel DDR3-1333
GPU	Radeon X550, 64bit (most retro part in here)
Sound	Sound Blaster Audigy (SB0090) in the PCI slot
Storage	128GB SATA (BIOS SATA == AHCI)
Other	USB keyboard and mouse, one PS/2 mouse so I could install USB drivers
Turned OFF:	onboard: GPU, NIC, sound; UEFI boot (legacy boot), all power saving/sleep states, virtualization, remote management

Table 1 — the test subject.

Old-School Graphics in C# / .Net 10, Part 3: Hidden Treasures of Old Games

Mon, 03 Aug 2026 00:00:00 +0200

Some of the games I loved on my 386 never got a source release. The companies moved on, the toolchains died, and what remains is a couple of hundred kilobytes of 16-bit machine code on a floppy image. For years I assumed that really understanding one of these games - not just running it in DOSBox, but opening it up, seeing how it ticks, and maybe even improving it - was a job for someone with infinite patience and a decade of spare time.

Why LLM chat feels like an anti-pattern in 2026

Fri, 24 Jul 2026 00:00:00 +0200

A continuation of my previous article, What is better than your AI loop?, which introduced the blackboard architecture this post builds on.

The chat. The message-response chain you produce by typing with an LLM. It could be trivially implemented by a single growing context or a sliding window.

Why does it feel like an anti-pattern to me? Chat as a UI is fine, chat as a world model is not.

What is better than your AI loop?

Mon, 06 Jul 2026 00:00:00 +0200

There is a continuation of this article, Why LLM chat feels like an anti-pattern in 2026?, which shows the blackboard architecture in action.

A blackboard architecture, modernized with event sourcing, content-addressed knowledge source (KS) firing and a few more enhancements.

While working on improvements to various LLM-powered components and subsystems over the past year, I noticed I kept converging on the same architecture — one that, as I later found out, had been described and named four decades ago.

Why Success Favors Action and How This Relates to AI

Mon, 08 Sep 2025 00:00:00 +0200

It’s been over a year since I published my view on the state of AI/LLMs and how it could be significantly improved in my post: “Failure Is Not An Option For AI”. While scrolling through the rStar2-Agent technical report, I couldn’t help but mumble to myself: “I told you so!”.

Consider how quickly the model improved during training, especially given that we are comparing a 14B parameter model to 671B R1.

Fast Memory Copying in C#/.NET (Cache, AVX, Threads, Unsafe, Alternatives)

Fri, 20 Dec 2024 00:00:00 +0100

Numerous copy routine implementations are readily available in .NET. If I were to simply list them alongside a few benchmark numbers and charts, it wouldn’t make for a very interesting article.

⚠️ What if I told you upfront that none of these routines is designed to be the absolute fastest?

If you’re interested in a basic comparison, I recommend checking out this article here on dev.to: What is the best way to copy an array? or a more detailed, older comparison: High performance memcpy gotchas in C#.

Old-School Graphics in C# / .Net 8, Part 2: Fireworks and Advanced Vector Extensions (AVX, SSE)

Wed, 20 Nov 2024 00:00:00 +0100

There’s something uniquely captivating about revisiting the passions of our past. I still remember the countless hours spent tinkering with code on my old 386 and 486 PCs, mesmerized by the magic of pixels coming to life on the screen. The simplicity of those days, where an array of pixels and a spark of imagination were all you needed, still holds a special place in my heart. Inspired by that nostalgia, I decided to breathe new life into one of my early projects - a 2D firework simulation - using modern tools and techniques. I am using Mode-13HX as a basis today, for more details please check my previous part: Old-School Graphics in C# / .Net 8, Part 1.

Old-School Graphics in C# / .Net 8, Part 1: Teaching an Old Dog New Tricks (Introducing Mode 13hx)

Wed, 13 Nov 2024 00:00:00 +0100

I began my journey with computers and programming during the era of 386 PCs, a time when DOS games ruled and sparked endless curiosity. I remember playing those games and wondering how they were created. It wasn’t long before I discovered Mode 13h - a graphics mode that perfectly fit within a 64kB page in 16-bit code, thanks to its 8 bits per pixel, 256-color palette, and 320×200 resolution.

Many years and APIs have passed since, but every now and then, I feel nostalgic for that simpler time. Back then, all we had was an array of pixels and an idea - and that was all we needed. I recall the breathtaking graphical demos that showcased incredible creativity, often categorized by their size. Some were only a few kilobytes but managed to display complex scenes with textures, geometry, animations, and even synchronized music.

Zen5 and C#

Tue, 17 Sep 2024 00:00:00 +0200

Zen 5 landed with a bit of controversy, but anyone Who was paying any attention to the real results wasn’t disappointed. Benchmarks done using more mature OSes (like Linux and Win10) and professional software shown its strengths. But hey, why should we trust just any 3rd party data if we can make a test by ourselves? I used Ryzen 9950x today and a bit of code to get things going.

Failure Is Not An Option For AI (And It Shouldn't Be)

Sun, 16 Jun 2024 00:00:00 +0200

My intent is to write a series of practical posts related to AI and programming. Topics like running local LLMs with agent frameworks, automated source code manipulation, execution under docker and more. All with the use of .Net and C#, Elastic Search, AWS and more.

I realized that I would like to first express my view about more fundamental, theoretical and even philosophical aspects of LLM/AI first. Therefore this post does not require any programming skills.