The holidays are rapidly approaching, and you probably already have a topic or two to argue with your family about. But what about with your hacker friends? We came upon an old favorite the other day: whether it “counts” as retrocomputing if you’re running a simulated version of the system or if it “needs” to run on old iron.
This lovely C64esque laptop sparked the controversy. It’s an absolute looker, with a custom keyboard and a retro-reimagining-period-correct flaptop design, but the beauty is only skin deep: the guts are a Raspberry Pi 5 running VICE. An emulator! Horrors!
We’ll admit to being entirely torn. There’s something about the old computers that’s very nice to lay hands on, and we just don’t get the same feels from an emulator running on our desktop. But a physical reproduction like with many of the modern C64 recreations, or [Oscar Vermeulen]’s PiDP-8/I really floats our boat in a way that an in-the-browser emulation experience simply doesn’t.
Another example was the Voja 4, the Supercon 2022 badge based on a CPU that never existed. It’s not literally retro, because [Voja Antonics] designed it during the COVID quarantines, so there’s no “old iron” at all. Worse, it’s emulated; the whole thing exists as a virtual machine inside the onboard PIC.
But we’d argue that this badge brought more people something very much like the authentic PDP-8 experience, or whatever. We saw people teaching themselves to do something functional in an imaginary 4-bit machine language over a weekend, and we know folks who’ve kept at it in the intervening years. Part of the appeal was that it reflected nearly everything about the machine state in myriad blinking lights. Or rather, it reflected the VM running on the PIC, because remember, it’s all just a trick.
So we’ll fittingly close this newsletter with a holiday message of peace to the two retrocomputing camps: Maybe you’re both right. Maybe the physical device and its human interfaces do matter – emulation sucks – but maybe it’s not entirely relevant what’s on the inside of the box if the outside is convincing enough. After all, if we hadn’t done [Kevin Noki] dirty by showing the insides of his C64 laptop, maybe nobody would ever have known.
Maybe someone will do a tube computer as a badge?
I don’t know about a full-tube computer but a badge with a few sub-miniature tubes as a proof of concept should be doable. My biggest worry would be power. Maybe instead of a “tube badge,” create a “tube vest” that was just something very simple, like a comparator or accumulator or something to make das blinkenlights do something pretty in a “programmed” way. The nice thing about a vest is you can put some beefy batteries in the pockets.
This 2016 paper (doi) had some ideas for not-necessarily-vacuum but through-the-air-electron-flow technology that overcomes some of the limitations of the silicon tech of the last 75 years. I haven’t seen anything come out of it though, so maybe it flopped.
It. would certainly keep you nice and warm. Wouldn’t want to go out in the rain wearing it though.
You could make an x-ray tube badge and walk around giving everyone radiation poisoning. Bonus points if you hide it inside a backpack and use it to irradiate commuters on a bus.
What if you made a vacuum circuit inside a glass hip flask?
Something simple might be possible. After all, battery-powered tube radios were a real thing for a while.
A Nuvistor tube is cute and tiny, and its filament takes just one watt, and I’ll bet you can still make it do something useful and interesting on half that.
Or even more interesting: use a vacuum fluorescent display tube as a computing element. With the right sort of multiplexing you could get 8 or more separate “devices” in a display, and the blinkenlights would be actually meaningful.
I’ve been building a mashup of a system with a STDBus card cage (two 3-slot cages actually) in a 3d printed luggable case and custom boards to simulate Acorn System 1/2/3/4/5, EuroBeeb & EuroCube, as well as Z80 CP/M systems. Video, Audio, and most other I/O is handled by a pair of 20K LUT GoWin FPGAs, with the keyboard connected to the Teensy 4.1 that is also doing in-circuit emulation of the 6502, 6809, or Z80 CPUs.
The 16-bit address and 8-bit data busses are real 5v logic between the STDBus cards, and most 8-bit STDBus cards from the 1981 Pro-Log catalog will work in the luggable.
Running the CPUs in software on the Teensy has allowed breakpoint debugging, dynamic loading of the MOS ROMs for the various systems, and dynamic remapping of memory & peripherals. When talking to peripherals, everything slows down to match the speed the peripheral can respond at, but when executing code from internal SRAM or external PSRAM (gotta have those sideways RAM/ROM slots!) the simulated CPU can zip along unrestricted.
I can see the appeal of the C64 laptop for Commodore fans, but also understand why purists get mad about modern recreations / continuations of the line. (I hear arguments all the time about the CX16 and F256 computers.)
Nice, especially the Gowin cheapo FPGA part, got a link ?
I’m inclined to think that it’s OK to emulate, but it’s increasingly less impressive the greater the Host:Target CPU performance. Emulating a 68K Mac at (mid-range) 68030 speeds on a PowerPC at 66MHz is impressive. Emulating a pdp-8/e (effectively 0.3MHz) on a quad-core PI running at 2-3GHz per core. Not so much.
Running Apple’s file system and network stack on the emulated 68k for years was ‘impressive!’
Going to say it … For the most part, leave the ‘real’ hardware to museums and just use simulators going forward. For example, I’ve found it a lot of fun to just use DOS-X for retro-computing. Running Turbo Pascal and C, Tasm, has been a blast and all running on a RPI-5 500+. I am thinking of turning one of my Lego PCs into a Reto PC now. The PiDP line is another example to experience DEC platforms with small consoles that can sit on a shelf rather than a large building with big power requirements… If you are like me, I don’t have a lot of room to store a collection of ‘retro’ sized computers.
With 3D printing, and availability of making PCBs easily, it sure make it ‘easier’ to simulate the old hardware if desired without touching the originals. Nothing wrong with bringing the old hardware to life though either for those so inclined.
I both collect old systems and get them running again, and emulate. I have a bunch of emulated things running all the time (a Pi 4 running about 10 systems: PDP-11, VAX, S/370, dps8m; also one each of Oscar’s PDP kits; finally a little PC for Alpha emulation), but I only fire up the real iron on special occasions, because (for the computers and workstations, not so much the videogame consoles) it’s hot, it’s noisy, and (for everything) every time I feed it power I’m running the risk that this is the time something difficult to fix blows.
There’s no ‘real deal’. No matter what you’re doing, the modern ecosystem means you do it much differently than you would have 30+ years ago.
For example, my friend gave me an HP Omnibook 486 (it looks like an hp95lx scaled up about twice in every dimension). When it was current, i never could have afforded it, of course. I wanted to know a little about its context so i found an old Computer Shopper magazine that someone had scanned. Which is to say, i downloaded a giant PDF in half a second and paged through it as fast as i could push the buttons — that’s not period correct! From that same issue, i found an ad for a PCMCIA SSD for it, and since the HDD it came with was suffering constant faults, i decided i wanted an SSD for it. To my surprise, i found basically the same SSD from the ad on ebay, but again, i was able to afford it, which is fantastically anachronistic. And to get it booting, I copied a bunch of files over from some modern supercomputer (budget laptop) running Linux.
Same friend left a VIA C3-based thin client with me. I found a plausible use case for it 6 months ago, so i downloaded a period-appropriate Debian install, which took a severaly inappropriate number of seconds to download. I brought it up under QEMU so i would have a good platform for cross-compiling a kernel for it. I compiled a custom kernel in a totally wrong 45 seconds. To find stuff in the kernel source, i used ‘find . -type f | xargs grep blah’, which was not a plausible way to work with source in 2003. I was overall really pleased with how much easier it is to do 2003 tasks in 2025 than it was at the time. I had actually played with this same device many years ago and it was much easier this time.
Everything we do with old computers is only really plausible because new computers are so great, IMO. Every tool we use to bring back an old computer is a new tool. There’s surely some weirdo out there using old tools on old computers but fundamentally you can’t isolate yourself from the modern supply chain. We just don’t do it like we used to.
Is Google AI “modern software” Big Tech attempt to promote 1960s
software technologies into giga/terabyte memories low-watt Nanocomputer
software technologies?
LLVM/CLANG, frameworks, tool chains , another Big Tech BAD IDEA?
freedos has a 64-bit memory manager now. if only they could do a gpu driver. a multithreaded software renderer in dos might do the trick though. might be too much to ask of such a diminutive os.
are the magnificent 7 in a bubble?
AI Overview
Yes, there are significant “bubble”
concerns around the Magnificent
7 (Mag 7) tech stocks, driven by
high valuations and massive AI
spending, but many analysts
debate whether it’s irrational
exuberance like the dot-com
era or justified by transformative
AI technology, with some stocks
even underperforming the broader
market recently, showing a potential
rotation in the market.
Arguments for a Bubble/High Risk:
…
Arguments Against a
Bubble/Justified Growth:
…
Current Situation (as of late 2025):
The market is experiencing a mix
of both, with some Mag 7 stocks
(like Alphabet and Nvidia) still
outperforming the S&P 500, while
others (like Tesla, Apple, Microsoft,
Meta) have lagged, indicating a
complex dynamic rather than
a uniform bubble, according to Inc.com
does bitcoin consume more electricity
than Austin TX?
AI Overview
Yes, Bitcoin mining consumes vast
amounts of electricity, often
exceeding the usage of entire
cities like Austin, Texas, especially
considering the collective power
draw of numerous large-scale
mining operations in Texas alone,
which strain the state’s power
grid and rival the energy
consumption of millions of
homes or smaller cities, with
single facilities sometimes
using as much as 300,000
homes’ worth of power.
Key Comparisons:
…
In essence, while Austin itself is
a large city, the cumulative
power consumed by Bitcoin
mining operations, particularly
within Texas, far surpasses
what is needed to power
Austin, making it a significant
energy consumer.
Take your medication.
i like neoretros better as you can screw around without breaking a priceless artifact (read: something you can sell on ebay for a few bucks).
Nothing inherently wrong with emulation as long as the emulation is provably correct, and even then only if the emulation is actively being promoted in an academic or authoritative sense. Otherwise, “good enough” can be good enough. That said, people settling for a “good enough” solution really need to keep themselves in check and not speak on things authoritatively.
The emulation of Taito’s Bubble Bobble was “good enough”, right up until someone worked out how to dump the protected 68705 microcontroller which, it turned out, had a rather significant effect on gameplay: The timing of when “extend” bubbles spawn, other item spawning, and more.
The emulation of Taito’s Operation Wolf was “good enough”, right up until the copy-protection chip was delayered and dumped as well. Then it turned out that using the best-guess simulation of the copy-protection chip, the levels were in the wrong order, bosses had the wrong amount of health, enemy patterns were wrong, and to a great extent, it wasn’t even like playing the same game.
The emulation of Capcom’s 1942 was “good enough”, then it turned out that there’s a box that displays an expanding-and-contracting animation behind the Capcom logo on the title and ending screens, which was previously shown in all emulators as two separate boxes moving back and forth. The hardware only has enough time to analyze 24 sprites per scanline, but RAM for 32 sprites: It always analyzes the first 16 sprites, but then analyzes the next or last 8 depending on how far down the screen it is. Each box was in a different group of 8 sprites, so they would be cut off at the mid-point of the screen, producing the expanding-and-contracting animation instead.
That last one may seem somewhat innocuous, but here’s the thing: Decades later, when a company went about writing a recreation of 1942 for one of those TV plug-and-play handheld things, the developers must have been using a flawed emulator as a reference, rather than a real arcade board, as the TV plug-and-play thing recreates the behavior of the flawed emulation, not the behavior of an actual arcade board.
Thus my point about not treating “good enough” as authoritative: There is literally the potential for a mistake in that “good enough” implementation to be mistaken for authoritative, perpetuating errors in what is tantamount to a rewriting of history.
Something like VICE, which is as close as possible to 100% hardware-accurate, with its results verified on real hardware? Something like WinUAE, which is the nearest thing to VICE but which covers the Amiga? Those are fine, as they are both intended to be, and verified to be, an accurate model of vintage hardware.