Can You Install Linux On a 1993 PC? (yeokhengmeng.com)

Slashdot reader yeokm1 writes: The oldest x86 CPU that the Lnux kernel supports today is theoretically the 486. However is this theory actually true in practice? I decided to put this theory to the test in my project.
His site describes installing Gentoo Linux on an "ancient" IBM PS/1 Consultant 2133 19C (released in 1993), with 64MB SIMM-72 RAM. (Though to speed things up, he compiled that minimal version of Gentoo on a modern Thinkpad T430 released in 2012.) "Due to the age of the PC, the BIOS only supports booting from the floppy drive or internal HDD," so there was also some disk partitioning and kernel configuration. ("Must disable 64-bit kernel for obvious reasons!") A half-hour video shows that it takes almost 11 minutes just to boot up -- and five and a half minutes to shut down. "Despite the many roadblocks I faced, I was impressed by the level of support Linux has for ancient hardware like this."

And there's one more added bonus. "Given the age of the 486 (1989 technology), it does not support branch prediction... Ironically this makes it safe from the Meltdown and Spectre attacks."

Comments

Interesting project

[93 Escort Wagon]

"Ironically this makes it safe from the Meltdown and Spectre attacks."

No, there's no irony there at all - not even in the manner "irony" gets misused sometimes.

Re:Interesting project

[Anonymous Coward]

It's a bit like rain on your wedding day

Re:Interesting project

[Anonymous Coward]

More like a free ride when you already paid

Re: Interesting project

[Anonymous Coward]

Or like 10,000 spoons

Re:Interesting project

[gravewax]

hmmm I think to turn this into irony you would need for that chip to have been replaced with the new chips due to better security AND you would need the discoverer of the flaw to be the one that made that decision.

Re:Interesting project

[burtosis]

It's like writing a song all wrong about irony, till they change how it's said.

Re:Interesting project

[mrbester]

There is: in attempting to be funny, the statement itself is an example of verbal irony.

Re:Interesting project

[93 Escort Wagon]

Like rain on your wedding day?

Re:Interesting project

[bill_mcgonigle]

No, there's no irony there at all - not even in the manner "irony" gets misused sometimes.

Of course there is - in that aspect the 486 is more secure than the new chips that are billed as having all sorts of security-promoting features.

There's no NX bit on the 486, though, so overall it's not more secure, even with the recent vulnerabilities.

Re: Interesting project

[Anonymous Coward]

Yes I ran slack on my 486dx2

Re:Interesting project

[Anonymous Coward]

Of course there is - in that aspect the 486 is more secure than the new chips that are billed as having all sorts of security-promoting features.

Sure if you rewrite what was said you can find irony. Same way that if you comprised on day of week for the one date you believed the long range forecasts said it wouldn't rain and it did, whilst all the other better days of the week you'd rejected it didn't rain, it would be ironic that you got rain on your wedding day.

Re:Interesting project

[Anonymous Coward]

It's ironic that you say that...

Anyway, I've done something similar with old hardware and the biggest thing that "saves"
the old hardware is a modern-ish GPU (i.e., a nVidia card and supported driver). Even
with Windows XP, having a modern graphics card really makes a huge difference in the
performance feel of some of these old machines. When the graphics task are offloaded
from the main CPU, those old computers actually perform pretty well.

CAP === 'iceberg'

Re:Interesting project

[K. S. Kyosuke]

There's no NX bit on the 486, though, so overall it's not more secure

Why is that? Side channels like those you can't fix easily, but NX is avoided by simply fixing your runtime bugs, or by using safe languages. What is so magical about NX that makes any system without it automatically insecure?

Re:Interesting project

[shess]

No, there's no irony there at all - not even in the manner "irony" gets misused sometimes.

Of course there is - in that aspect the 486 is more secure than the new chips that are billed as having all sorts of security-promoting features.

There's no NX bit on the 486, though, so overall it's not more secure, even with the recent vulnerabilities.

But the exploits are literally predicated on things the new chips have that older chips did not! It's like saying "Ironically, our Christmas vacation in Alaska was cold."

Ironic would be if the old processor was subject to the same exploits in spite of not having out-of-order execution and branch prediction.

Re:Interesting project

[Anonymous Coward]

more like free advice you just didn't take.

Watercooling

[ArtemaOne]

Just need to get refrigerated mineral oil running over that and clock it up to a GHz or two to get it going. Very impressive overall.

If only more old hardware was supported.

[Anonymous Coward]

The big loss is that Firefox and Chromium no longer work on pre SSE2 processors so you can't surf the modern web on old computers anymore.

Re:If only more old hardware was supported.

[Hal_Porter]

Windows 8 and later require SSE2 too

https://support.microsoft.com/...

Windows XP and 7 don't.

Re: If only more old hardware was supported.

[Chewbacon]

Lynx? No?

Re:If only more old hardware was supported.

[Anonymous Coward]

It's a real shame that they don't even care to build a FF version without SSE2 leaving lots of old hadware in use stuck with vulnerable versions.

Re:If only more old hardware was supported.

[Anonymous Coward]

I didn't know that, but couldn't you just recompile from the source and change the compiler flags to support lack of SSE2?

Re:If only more old hardware was supported.

[serviscope_minor]

The big loss is that Firefox and Chromium no longer work on pre SSE2

Firefox works on ARM which has no SSE2.

Is this about the precompiled binaries?

Re:If only more old hardware was supported.

[Anonymous Coward]

Thank god it isn't.

Supporting shit that old is not a good thing. Like it or not, you can't just fire off a few thousand lines of code and forget about it. Either there's an ongoing maintainance cost (resources far better spent elsewhere) or you're in for surprise buttsex.

'sides, it was questionable odds if you could get Linux running on a random PC from 1993 in 1993.

Re:If only more old hardware was supported.

[Anonymous Coward]

Running old code on new hardware is at least somewhat useful at times. Hopefully, nothing serious, but sometimes it's nice to play older games. Games from a time when companies cared at least somewhat about the enjoyment rather than how to best convince players to pay to skip portions of the game.

Re: If only more old hardware was supported.

[Miamicanes]

Before anyone gets *too* nostalgic for old games, remember that in the *really* old days (early 80s), game development went something like this:

1. Discover some cool graphics hack that let you do something novel... reuse sprites, change graphics modes mid-screen, animate by changing the color palette, etc.

2. Come up with some excuse to turn it into a game.

3. Create awful, shitty, pointless, and un-fun ports to every other popular system, regardless of viability.

3a. Don't forget CGA, EGA, and Hercules versions, plus Atari ST. And Apple II (non-GS).

Had it not been for Atari's early-80s implosion, we probably would have seen abominations like "Yars Revenge for CGA" (shudder), ignoring the fact that the game's entire reason for EXISTENCE was the "color static" effect.

Re:If only more old hardware was supported.

[arth1]

The big loss is that Firefox and Chromium no longer work on pre SSE2 processors so you can't surf the modern web on old computers anymore.

This is simply not true. Firefox builds just fine on a PIII here, using gentoo. You just need an ffmpeg that's built without SSE2.

Chromium won't build on a PIII, but that's not because of SSE2, but because you need at least 2 GB RAM to link it.

Re:If only more old hardware was supported.

[Anonymous Coward]

> I didn't know that, but couldn't you just recompile from the source and change the compiler flags to support lack of SSE2?

I may be wrong, but I guess that may not be a trivial task for everyone.

We can all recompile the kernel, but Linux got waaay easier when people stopped doing it -- though it seems some compiling does happen automatically at times.

In the end, there's so many versions which can be offered by distros. SSE2 makes old computers which have them go faster, so there is some merit on using it -- though it bricks some computers.

But -- as I do own some old/weaker computers -- there's also the problem of the size of recent applications. Firefox, Chromium etc. all became too heavy. Old computers simply cannot run such mammoths -- due to lack of memory and maybe exactly because of lack of speculative execution.

Alas, the problem is not the OS -- though I recommend Slackware (or an easier option like Salix) in the present case. But the problem are the applications. Firefox/Chromium, Libreoffice, Gimp... all are now massive, sophisticated programs... very useful with enough resources but too taxing on limited hardware.

Re:If only more old hardware was supported.

[jandrese]

One might get old and die before Firefox finishes building on a 33Mhz 486.

Re:If only more old hardware was supported.

[toddestan]

It's actually worse than that, since Windows 8 and later require the NX bit, which came out around the same time as 64-bit, so you can't install Windows 8/10 on anything older than about 2005 or so. I don't think there are any chips that have the NX bit but lack SSE2, but plenty the other way around, like any Socket 478 P4.

Re:If only more old hardware was supported.

[Bert64]

Can you not compile them to run on a non SSE2 system?
I've used firefox successfully on non x86 processors which clearly don't have SSE2 support.

Re:If only more old hardware was supported.

[Anonymous Coward]

Shut up

Re: If only more old hardware was supported.

[Anonymous Coward]

Depends how old you mean. There was some pretty amazing stuff like that in the '80s, but mostly for consoles. I was more thinking of the '90s when that software is too recent to be able to be perfectly emulated like BSNES does for SNES, but old enough that it doesn't necessarily run properly with new operating systems.

And yes, the number of games worth playing from that era is a small portion of the total sold. But, it is a large enough number to make having a system that's compatible with them potentially worth while.

That being said, I do feel a certain amount of nostalgia for games like that Muppets game for the Apple OS.

Re:If only more old hardware was supported.

[Anonymous Coward]

I don't think you get to blame firefox for that in your case. Looking that that id, I'd say already "hello world" could be problematic. :>

Re: If only more old hardware was supported.

[Anonymous Coward]

This brings back memories. Prince of Persia was one of the later titles that still supported Hercules. Playing it on a 4.77Mhz 8068 in green tint was much of a slide show. I guess the timer to complete the game was slower too.

Re: If only more old hardware was supported.

[loufoque]

Don't know about Firefox, but there are significant differences if you use the x87 unit or the sse one to do floating-point computation (especially on single-precision).
It's quite possible that some applications cannot deal with the x87 extra accuracy.

Re:If only more old hardware was supported.

[Hal_Porter]

It needs cmpxchg16b in 64 bit mode too

https://answers.microsoft.com/...

Which means it doesn't have a 8TB address space limit. On the other hand it also means it won't run on the original AMD Opterons

https://blogs.msdn.microsoft.c...

https://blogs.msdn.microsoft.c...

Re: If only more old hardware was supported.

[tepples]

we probably would have seen abominations like "Yars Revenge for CGA" (shudder), ignoring the fact that the game's entire reason for EXISTENCE was the "color static" effect.

You mean that trench just to the left of the screen's halfway mark? There are plenty of other ways to draw a trench that the player can duck into, other than just filling a background stripe with game code.

Re: If only more old hardware was supported.

[CronoCloud]

3a. Don't forget CGA, EGA, and Hercules versions,

You forgot Tandy graphics. ;-)

Re:If only more old hardware was supported.

[Anonymous Coward]

He's probably cross compiling.

Re:If only more old hardware was supported.

[Anonymous Coward]

FTFY

Which means it doesn't have a 8TB address space limit. Ironically it also means it won't run on the original AMD Opterons

Re:If only more old hardware was supported.

[Hal_Porter]

Careful. The Irony Police will give you a Fixed Penalty Irony notice for falsely claiming things are ironic when they're not. Ironically they often give Fixed Penalty Irony notices to people for falsely claiming things are ironic when they actually are ironic.

Re: If only more old hardware was supported.

[Anonymous Coward]

Probably has nothing to do with floating point. Newer compilers use a newer CRT where stuff like memcpy is optimized with SSE2

Re:If only more old hardware was supported.

[Anonymous Coward]

First, someone complained some software not running because of the lack of a certain SSE instruction. I thought, poor fellow, a shame to be left suddenly without support. But what could I realistically do?

Then some kernels wouldn't run for some other guys, because it required PAE and their processors didn't have such capability. "That sucks", I thought, but then again I felt powerless as things happened, and wondered what were those guys feeling?

Then it was about SSE2 -- they actually said Pentium 4 or superior was accepted, but in truth they wanted to get rid of CPUs without SSE2. It's just a few users and their old machines, they said. I shivered but felt lucky that my CPU had the instruction while those poor fellas were left in the cold.

Afterwards they got fond of 64-bit only software. 32-bit PCs owners protested, some said they didn't need a wider word -- but, to no avail. 32-bit ISOs were brought into public places and there they would be discontinued, to our surprise and horror. I have a 64-bit PC but that was the first time I felt myself great discomfort, as if a shadow were nearing me. I agreed with 32-bit folks, but then again was powerless to do anything. Or so I thought back then.

When they came for those without the CMPXCHG16b opcode, I felt somewhat numb, as if reality had become somewhat out-of-focus. It was all too fast, and again that left a sensation of an unavoidable thing, like an accident or tragedy. Luckily I had a CPU from a brand which had such instruction, so I could only feel sorry for those without it.

Now they say my "good brand" CPU is flawed. How so? Didn't I purchase the same as the majority of buyers? They say it cannot be patched. Soon they'll come for my CPU, too. "No longer supported", "insecure", they'll pronounce it. But now, all those guys I've talked about are gone, and there's no one left to defend me.

Re:If only more old hardware was supported.

[Agripa]

It's actually worse than that, since Windows 8 and later require the NX bit, which came out around the same time as 64-bit, so you can't install Windows 8/10 on anything older than about 2005 or so. I don't think there are any chips that have the NX bit but lack SSE2, but plenty the other way around, like any Socket 478 P4.

I have a number of old systems now which would be fast enough for various things but Windows will no longer operate on because they lack SSE2.

Re: If only more old hardware was supported.

[Miamicanes]

IMHO, Jumpman was the first game that really transcended the "Playability" barrier and broke the "move the joystick one pixel too soon or too late and die" norm that made earlier games so frustrating. I remember seeing Jumpman for the first time & being totally unimpressed by its graphics (to put it mildly), but after agreeing to play a round to appease my best friend, it ended up instantly becoming the favorite and most-enduring game of my childhood. It was the first game I'd ever played for hours because it was genuinely FUN, as opposed to "had awesome graphics and looked cool", or "had a really cool crack screen" (yeah, I had plenty of games in THAT category... of my most favorite Amiga games, probably half were cherished more for the crack screen than the game itself. Amiga had great graphics).

Out of the entire Atari 2600 collection of games that actually existed circa 1982, just about the only ones I still enjoy playing occasionally are Circus Atari (getting to see the stick figures splatter headfirst into the ground almost made up for the frustration of losing a life) and Warlords (the original "party game" if you had an extra pair of paddle controllers). IMHO, those two games were fantastic... I'd KILL for a hypothetical "Atari Flashback 9" that shipped with reimagined paddle controllers built around optical rotary encoders with 16-bit resolution. Such paddles wouldn't be TIA-compatible, of course... but the Flashback systems from 3 onwards have all been running on emulated hardware anyway, so it could just read the optical encoders by modern means, then stuff the result values into the proper phantom TIA registers to make them LOOK (to the game code) like a discharged resistor...

Of course Linux of the same vintage would work

[Anonymous Coward]

For example download Slackware 2.0 or 3.0 and run them. They are much smaller all around, and designed for systems at the time. The article title should be running todays Linux on machines from 1993 when 8 mb of memory was a big system. I remember back then you had a floppy to boot Linux etc.

Re: why does this matter?

[Anonymous Coward]

Itâ(TM)s better than going out and talking to people or getting laid!

well, to some people maybe

Re

[pele]

64MB RAM? Eh??? Back then 64MB of RAM cost £60k. And I don't think PCs supported more than 8 or 16MB. I had 4x1MB 30pin and had 2 72pin slots free. Later added 2x2MB in 95 or thereabouts. X would fly with 8 megs. Anyway I forgot more about Linux then I know right now but I think it did run on DX (i.e. 32-bit) machines only, SX was 16-bit, right?

Re: Re

[pele]

One thing I noticed he has no external (L3 I think) cache installed. He should try with that first. It looked like a sim if I remember correctly. Also there are some PCI messages during boot. No PCI, just ISA and Vesa Local Bus support required. No PNP either. He could strip that down further.
Mine would boot in a minute and a half or so to prompt. Another maybe 2 to X with olwm or windowmaker.

Re:Re

[Anonymous Coward]

my 486slc (essentially a 386sx that understood 486 instructions) i had back then had a whopping 4 megabytes of ram, for which i spent $120 on two used modules to get up to that. i didn't have more than that in a pc until xp on a k6-2.

64? that's totally unrealistic for this 'study'. if you're going to play around with 'typical' configurations of a past era, use a fucking typical configuration.

Re: Re

[Anonymous Coward]

64mb is to smaller to even compile on. CC requires at least 128mb.

Was working on an IOpener from 1999 with 128mb and gentoo. Had a VM on another machine to create image and rsynvcâ(TM)d and puppy 7.2 to copy image from VM to 14gb partition. I have 4 partition 1 boot, 2 win me & puppy, 3 swap, and 4 gentoo. X let alone Firefox would not work with gentoo. But puppy works great!

Re: Re

[jshipp]

No. Intel's i486SX was a modified Intel 486DX microprocessor with its floating-point unit (FPU) disabled. It was intended as a lower-cost CPU for use in low-end systems.

Re:Re

[jellomizer]

My 486 motherboard supported up to 32megs. But 16 meg system were for high end use. Pc came with 4 megs which was considered good. Then I spent $650 for 16 more megs to get X11 to run smoothly that gave me a total of 20 megs with my gigabyte hard drive I was really rocking.

Re:Re

[Anonymous Coward]

SX was 16-bit, right

No, SX was 32-bit too. But the SX didn't support floating point extensions unless you installed a 487 co-processor, whereas the 486 DX has floating point support built-in.

Re: Re

[pele]

I was referring to 386sx. Didn't realise there was a 486sx...I di remember a dx4 though..lots of MHz..80 or 120?

Re: Re

[pele]

My memory is starting to come back now... thanks for that I forgot about fp coprocessors completely!

Re: Re

[mrbester]

I had a DX4 100 with 16MB RAM. Ran BeOS nicely.

Re: Re

[pele]

BeBOX was a dream never to come true sadly...

Re:Re

[Somebody Is Using My]

Actually, in the early '90s, the price of RAM was ~$40/ MB ($33/MB according to this list, so 64MB would cost you around $2500.

Finding a board with 64MB could be tricky, but I seem to remember at least one that allowed it (it supported SMP (dual) processors and was supremely expensive).

Both the 486DX and 486SX were 32-bit processors; the 486SX lacked a float-point processor (you may be thinking of the 386SX, which - although it was a 32-bit processor - only had a 16-bit bus).

Re: Re

[pele]

Sorry to burst your bubble my fried but that list is wrong.
What the author fails to mention is that 4 sims of 1mb each would cost £240 yet a single 4mb simm was priced WAY more that those 4 1mb simms. And then a 72-pin simm would again be slightly more expensive than a 30-pin one. To get up to 64MB you would need 4x16MB simms. Probably 72-pin ones as I don't seem to recall anything bigger than 4MB in 30-pin guise. My unix lab (where I'd later come to work at) bought a 64MB simm (or whatever SparcServer 20 modules were called back then) and paid between 60 and 90k for it, can't remember the exact figure. SunSite at src.doc.ic.ac.uk had 128MB and 2GB disk and was regarded as one of the beefier machines in the country, sponsored by Sun. So no $2400 wouldn't have covered it.

Re: Re

[Anonymous Coward]

The 386SX was a 32-bit processor internally, but it had a 16-pin data bus externally (as a cost-saving exercise). It also only had a 24-pin address bus, limiting it to 16MB RAM - the 386DX had a 32-pin address bus giving a theoretical maximum of 4GB RAM. Although there was a performance hit, any code that would run on a 386DX would also run on a 386SX.

Re: Re

[Anonymous Coward]

My memory is starting to come back now... thanks for that I forgot about fp coprocessors completely!

Was the 487 a coprocessor or really a 486DX that took over from the SX? 387 was the last actual coprocessor? Recollections fuzzy here too. :-)

Re:Re

[gravewax]

that was for low capacity sticks, 512k and 1MB being quite common, that went up exponentially for larger capacities, you would have been paying many times your $2500.

Re: Re

[Somebody Is Using My]

Interesting. I wasn't in the market for 16MB memory in 1993; I purchased 1MB modules because just having 4MB was an impressive upgrade. So I have no direct experience with prices for modules of that size. You position makes sense; the larger modules should cost more, as they would have smaller production yields. Still, to achieve the $60,000 price-point the RAM would have to cost $900+ per megabyte. But 16MB modules would be reserved for servers and business applications, and those always are more expensive than consumer hardware, so maybe?

Still, looking at the advertisements in Byte Magazine of that era (September 1992) this doesn't seem to be the case. The ads don't specify whether or not the memory is 30- or 72-pin, but regardless the memory does seem to stay within the $40MB range (with some increase as the modules get larger; up to 4MB modules stay within the $40/MB range but it jumps up once you start pricing 16 or 32MB modules). 8MB modules go for around $300-$400 (with one outlier being $695). 16 MB modules range from $700 to $900, and the exceptionally rare 32MB modules range from $1800 to $2700. Nowhere do I see prices close to those you suggest. Even the priciest 32MB module is available for a "mere" $85 per MB, a far cry from the $900 you remember. Doubtless all this RAM was of the non-parity variety; adding ECC would increase the price but even so not by an order of magnitude.

Still, it's fun to look at those prices and compare them to the hardware of today. The 32GB RAM I have in my current computer - were such a thing been available in 1993 - would have cost over a million dollars at those prices.

Re: Re

[u801e]

I was referring to 386sx. Didn't realise there was a 486sx...I di remember a dx4 though..lots of MHz..80 or 120?

IIRC, the 486 sx processors were clocked at either 25 MHz or 33 MHz (the wikipedia page mentions 16 MHz and 20 MHz versions, but I never had either of those). The math co-processors that you could add on came in either a DX2 or DX4 variety. The DX2 doubled the clock speed (50 MHz or 66 MHz) and the DX4 tripled the clock speed (75 MHz or 100 MHz).

Re:Re

[sconeu]

1992... Had a 486/33 (no bloody DX or SX). Paid $3200 for 32MB. We were running a departmental server (yeah, wasn't Linux back then... was *gasp* SCO Open Desktop -- before they were evil)

Re: Re

[Anonymous Coward]

386 boards were the last ones that still occasionally had a socket for a math coprocessor. But the 487 DID EXIST, and there was probably.....like ONE motherboard out there that no one ever saw.....

https://micro.magnet.fsu.edu/optics/olympusmicd/galleries/chips/intel487sxb.html

Like the failed precursor to the more successful Overdrive stuff. Remember when you could "upgrade" your 486 to a Pentium and keep the old junk?

Re: Re

[Anonymous Coward]

Yes, the but the 487 was more or less a 486DX that disabled the original 486SX that came with your board.

Re: Re

[raynet]

You can use this thing called swap, it will be slow (though perhaps not so slow if using ssd on that 486), but make compilation possible.

Re:Re

[Anonymous Coward]

I had an original 486/33 in late 91... the ram cost indeed $100/MB for a 30-pin SIMM. I had 8MB which was considered extravagant for a home PC. I think the cache had to be upgraded to 128kb, but I don't remember. The board had 8 30-pin SIMM slots for up to 32MB. The best part of the system was the Tseng Labs ET4000 Turbo with 1MB VRAM. Wow, that was a great card. SBPro, 120MB HD and 2400 baud modem rounded out the initial system build. I kind of miss 5 1/4" floppies.

I ran Linux on it from floppy disks in the very early days. I read Linux original post on comp.os.minix and drooled over the possibility of running UNIX at home. I forgot when I finally got it to work, probably mid 92. Eventually I could run x-windows via the Slackware distribution (probably in 93).

The system ran just fine as a daily driver up until I replaced it with a Pentium II in 1998.

Re: Re

[Anonymous Coward]

They made an SX2-50 as well, that was what was in my old packard bell.

Re:Re

[Reziac]

Yeah, I had a 486 that came with 4mb (pretty good in 1994) and a year or so later a local shop advertised 4mb sticks for $100 each... drove clear across L.A. to get there, ran in with the cash, grabbed four sticks, and ran out the door before they could realise their mistake. :P Machine was still in use in 2001, when a keyboard short killed it.

That 486's sound card is still in my DOS gaming box (now a P4), tho just this week it's finally trying to die... got my money's worth, eh?? :)

Re:Re

[Agripa]

64MB RAM? Eh??? Back then 64MB of RAM cost £60k. And I don't think PCs supported more than 8 or 16MB. I had 4x1MB 30pin and had 2 72pin slots free. Later added 2x2MB in 95 or thereabouts. X would fly with 8 megs. Anyway I forgot more about Linux then I know right now but I think it did run on DX (i.e. 32-bit) machines only, SX was 16-bit, right?

I had an early MIcronics 486 motherboard which had 8 30-pin SIMM sockets. I originally had 8MB but had planned to update it to 4MB SIMMs for 32MB which the motherboard was suppose to support. But when the board was designed they did not have 4MB SIMMs to test it with and as it ends up, it would not work with them even though it was suppose to.

Re:Re

[Matheus]

I had 16MB of ram on my P90 Laptop in (late) 1995 and *that was HUGE! (and expensive...) On desktops it was more realistic to have more than that (I had 72MB on my 486DX4-100 desktop) vintage 1994.. that replaced a 386DX so I have no reference for how much memory I would have had in my theoretical 486-25 ;)

You could run Linux just fine on all of that. The article would be better if it was worded "Modern Linux".

I also would like to see if Slackware has a better experience as historically they were much better with keeping the bloat down...

Re: No ...

[Anonymous Coward]

I ran Linux on a 1993 computer in 1993.

My recently resurrected 40MHz SPARC boots faster

[Anonymous Coward]

I currently have similar projects going on, I'm surprised by his boot time, my 40MHz SPARCstation 2 boots in just a minute or two: https://www.instagram.com/p/BdKbXxlnW_i/ video and details coming: http://youtube.com/renerebe

I Run Linux On A Commodore 64

[dryriver]

I just type in LOAD "LINUX" , 8 , 1 and off the C64 goes. Booting requires some "disc swapping" of course, and is sometimes hard on the Datasettte unit, but it works. Most impressive is that is that I've managed to solder a current Nvidia Titan GPU to the underside of the Commodore as well. This lets me run C64 games like The Last Ninja, International Karate II and Infernal Runner at 8K UHD 144Hz. I can also run all Playstation Pro 4, Xbox One and Nintendo Switch games on the unit. Even games that haven't been developed and released yet, like Grand Theft Auto 6 run great on this souped up Linux 64 unit. Oh, and the unit can time travel as well. I had coffee with Leonardo Da Vinci just this morning. He told me that he was painting a portrait of an Italian lady who hit her head recently and has a strange smile frozen on her visage. Amazing what a few beers before going on Slashdot can achieve, right? =) (The idiocy in this post is released under the GNU GPL 3.0 License)

Re:I Run Linux On A Commodore 64

[Anonymous Coward]

You joke, but the Contiki operating system, which is now marketed as a modern OS for the Internet of Things, started out as a multitasking, networked operating system for the Commodore 64 and other 6502-based systems. They seem to have scrubbed almost all references to that off their web site, though.

Re:I Run Linux On A Commodore 64

[Anonymous Coward]

The message is bullshit but it's framed so well. I'd read more of it.

Re: I Run Linux On A Commodore 64

[Anonymous Coward]

Yes, of course, go run Lunix on your C64.

Back in the day

[tbuskey]

We ran the current versions on our 486, pentium 90s, pentium 200s, pentium 2s and 3s.

For a server type system, they're ok except for power consumption and they can't keep up with gigabit speeds.

For a desktop, internet wasn't something you use extensively. In the pentium 2/3 days AJAX was just starting. Javascript was not used heavily and most people had dialup. Today's internet will be glacially slow if it will even run on older CPUs.

Most of the stuff I do personally and professionally uses web pages with heavy javascript. I think the baseline would be a RasPi or other ARM based system with ~ 1GB RAM, I've found that 4GB+ on x86 is vastly more responsive.

Re:Back in the day

[Anonymous Coward]

Yes.. interpreted garbage that makes today's cpus run like native code did on yesterday's..

Re:why does this matter?

[jellomizer]

In 1994 I was using Linux on a 486 DX 50mhz originally with 4 megs of RAM. I had upgrade to 20 megs a few months later, so I could use X efficiently.
What can you do with a 486 Linux system? Probably more then you think. Just not as many things at once. You can run a web server, a database probably not both at the same time. However if you maxed the RAM you could get a lot done on slow CPU. If you checked you fast Computer most of the time your CPU is idle. On a 486 you can do nearly anything you can do on any other 32bit computer.

40MHz SPARCstation 2, 16 MB RAM

[ReneR]

I have similar projects ongoing, e.g. 40MHz SPARCstation 2, 16 MB RAM: https://www.instagram.com/p/Bd... I wonder about the 486 boot time, the vintage SPARC needs a minute or two to text login. maybe theirs include X, ... or systemd, ... Video and such next: https://youtube.com/renerebe

that machine is a beast

[Anonymous Coward]

64MB ram? that thing must have cost a fortune, At the time most machines capped out at 16MB and that was a bloody expensive config.

Well, this tells me modern software is shit

[lamer01]

The software that existed at the same time this 486 was prevalent ran just fine and rather quickly on the 486. This modern (last 20 years) with layering software with countless abstraction layers has produced utter crap software. I would like to see how fast the software of that era runs on modern PCs versus the crap software we put on them today.

Re: Well, this tells me modern software is shit

[pele]

We used to say (back in the 486-era) how software of today is shit and how everything was flying on 286-es in assembler. And 8085s...

Re:Well, this tells me modern software is shit

[gravewax]

The software of that era had complete and utter shit for security, hardware support, ease of use, stability and graphics etc etc. sure it ran fast, your car would go faster too if you took out all the windows, airbags, seatbelts, the doors and panels, stripped out the seats, air con, reduced fuel tank size to 10% of current capacity, not many people though would say that the car was better and today's cars are shit because of everything they come with.

Re:Well, this tells me modern software is shit

[AvitarX]

Did it?

My 486 struggled with much smaller displays, hung for ages when my turn changed in civ2, took forever to load a map (both bandwidth and rendering), rendering even basic 3d shapes took forever.

Re: Well, this tells me modern software is shit

[pele]

That is one well-substantiated statement right there. How old are you, 16?

Re: Well, this tells me modern software is shit

[pele]

IBM PCs weren't meant for graphics or 3d. Amigas of the era would fly if I remember correctly. So would NeXT boxen. And SGIs of course. And the little sonic the hedgehog wouldn't even twitch on a sega.
There must be a point in your comment but I am struggling to pinpoint it.

Re:Well, this tells me modern software is shit

[Anonymous Coward]

Actually, such a car would go slower.

Re: Well, this tells me modern software is shit

[gravewax]

What are you 12? perhaps get out of your mothers basement before commenting. that was a well-substantiated ad hominin you posted.

Re: Well, this tells me modern software is shit

[Anonymous Coward]

It's not the best analogy but he has a point. Things like hardware detection, filesystem journaling, new functionality like modern encryption systems for sshd, etc. All of those things are made with a modern cpu in mind.

The 486 guy knew what he was doing, and every slowdown was a sacrifice done in purpose just to get a modern filesystem with modern functionality running on a 486.

Re: Well, this tells me modern software is shit

[Shinobi]

Amiga's weren't really good for real-time 3D graphics, such as shaded structures etc, due to using planar graphics, as opposed to chunky graphics. However 3D creation software on Amigas was often quite a bit faster than on Mac's with better hardware, or roughly equivalent PC's, due to more responsive UI etc. However, Commodore senior management fucked up a lot, so the Amiga stagnated, and PC just steamrolled ahead.

Re: Well, this tells me modern software is shit

[AvitarX]

Would they really fly displaying a few million pixels at 24 bit color?

The last Amiga I used was before then (286 era I think), but could only display 64 colors at a time out of a pallette of 1024.

Not sure the resolution.

Re:Well, this tells me modern software is shit

[religionofpeas]

I think you forgot how slow everything was. I remember kernel builds taking half a day on my 486.

Re: Well, this tells me modern software is shit

[lamer01]

Yeah, I wrote some of that as well :)

Re:Well, this tells me modern software is shit

[lamer01]

Well, of course some things just weren't there at the time but to me these sound like straw man arguments. I still dare a comparison on modern hardware running Old time Software vs todays's versions.

Re:Well, this tells me modern software is shit

[Thelasko]

...your car would go faster too if you took out all the windows, airbags, seatbelts, the doors and panels, stripped out the seats, air con, reduced fuel tank size to 10% of current capacity, not many people though would say that the car was better and today's cars are shit because of everything they come with.

So the PC equivalent of this?

Re:Well, this tells me modern software is shit

[Anonymous Coward]

WTF? Yes, you can run Linux on a 1993 PC. Choose from the first versions of Slackware, Red Hat, Suse, Debian and some forgotten distros.

Re:Well, this tells me modern software is shit

[Anonymous Coward]

I'd say it's pretty remarkable that in 20+ years the boot process only blew up by ~10x. Especially when hardware performance increased thousands of times (more than compensating for the additional code).

Re:Well, this tells me modern software is shit

[samwichse]

Your comment reminds me of this classic:

Getting a 5.8s 0-60 time from a 2001 Nissan Sentra:

http://www.rcramer.com/fun/eco...

Completely true for the same reasons.

Re:Well, this tells me modern software is shit

[thegarbz]

You only say that because you don't understand what you are actually doing with modern software.

Re: why does this matter?

[pele]

I ran a web server with cgi off of postgres95 back then would you believe it. And sendmail.

Re:why does this matter?

[Lunix Nutcase]

So you can decode 1080p H.264 in real-time on a 486? Yeah, didn't think so.

Re: why does this matter?

[pele]

Actually decoding mpeg layer3 was rather difficult.

Developers can't be bothered

[FrankHaynes]

Modern developers scoff at the idea of anyone running their software on anything less than the very latest hardware. 32-bit processors?? PFFFT!!

MOAR LIBRARIES!! MOAR DEPENDENCIES!! MOAR, MOAR, MOAR!!! CONSUME ALL AVAILABLE MEMORY AND CPU CYCLES!!!

Re: Developers can't be bothered

[pele]

I don't think they are developers. They are typists that today's universities (and ms courseware) churn out by the truck-load. On their way to extinction...

Re:Developers can't be bothered

[Anonymous Coward]

Yep, efficiency only matters to embedded hardware programmers.

Re: Developers can't be bothered

[pele]

True. But even NASA has started slipping up nowdays, sadly. Because of the aforementioned issues I fear.

Re:Developers can't be bothered

[Anonymous Coward]

It's cheaper to buy faster hardware than it is to write more efficient code. Developers relying on heavier frameworks is just the natural evolution of modern development practices reacting to market forces.

Re:Developers can't be bothered

[AHuxley]

The PAE wall that stops people going back too far into computer history.

Memories

[jgotts]

I switched to Linux in May of 1994. That computer had a 486DX2 66 with a whopping 12 MB of RAM. Slackware was pretty much your only choice, and I installed Slackware 2.0 from 3 1/2 inch floppies.

It took me days and days to get on the Internet with PPP from my dorm room at the university, and from that experience I wrote a mini-HOWTO.

That's where I'd get started if I wanted an authentic 1993 Linux experience. Be prepared for nothing working as you would expect out of the box. Out of necessity I immediately became a Linux developer and author. I even wrote one patch for the kernel and at one time maintained two kernel modules.

Now I pretty much don't do any Linux development except for work, but I've been doing it for 24 years now.

Re: Memories

[pele]

24 floppies. My 12th and 14th didn't copy properly so had to run back to the uni to re-do them.
I guess I was one of your first devoted readers then? But I don't remeber a mini howto on slackware, just the regular one..

Re:Memories

[dow]

My introduction to Linux was on an Amiga with a 50mhz 68030 cpu, fpu, and 16mb ram expansion for a total 18mb. It could run an X server quite fine. Later moved over to the PC, but it was only my first steps with the Amiga version that confirmed Linux as the OS choice when I did. I remember the PPP How-To, and really appreciated that one in particular. Thankyou!

Re: Memories

[pele]

Oh PPP minihowto! Yes, remember that one. Settled for SLIP first to get things going and then went on to tackle PPP. Minihowto saved the day! Thanks from me too.

Re:Memories

[Anonymous Coward]

I had very similar experiences. I also had a 486/66 DX2 (albeit with only 8MB of RAM). It took me a lot longer than 3 days to figure out PPP with Linux. Once I figured it out, I wrote a script it bash (and later in PERL) to help automate the process.

I was able to get more than slackware (i had Red Hat 4.1~5.0, Debian, slackware, and later SuSE installed at various points), but your argument still stands. Linux was a big pain in the ass to get set up back then!

Re:Memories

[Brostenen]

I remember that Slackware all too well.... It had something that have gone lost in modern distro's. Namely a report, on how many failed logins on your login, that have been attempted since you last logged in succesfully. That feature is not active by default anymore, in all mainstream distro's. Anyway.... I tried Unix, back in 1995, and one month later I discovered that Slackware distro. I tried to install it on my 486dx2-66 with 8mb Ram, and failed misserable in getting stuff working. Through the years, I have kept an close eye on the Linux devellopment and tried many distro's from time to time. It was only in Feb. 2017 that I finally went all in and installed Xubuntu as my main OS on an Thinkpad R61. I was tired of lagging Win7 after all them updates, and the software avaliable for me to use Linux 100% was finally ready. I am happy to have "tested" Linux all these years, wich have resulted in me not having to go through any learning curve, from that point when I have finally made the big change. Yeah.... a couple of months of combined Linux training every year for the last 22 years did pay off. It is as natural to me, as Dos, Windows 3.11 to Win7 and obscure operating systems. It's just.... Linux rocks for me personally. And I can still not find my way around Win8, 8.1 and 10. They are so annoying to use.

Re:Memories

[Bert64]

Mine too, but with a 68040... I still have a couple of amigas which can dual boot into linux, and even have a gentoo install running on a 50mhz 68060.

Re:Memories

[Dr. Evil]

I was installing from CD on a 386 w RLL. The CDR was proprietary, and I had a Que book on UNIX to help me... I upgraded from 2M to 5M just to be able to run it... $200+ in RAM right there.

It took a month to get the install working... then everyone told me to spend more money on upgrades...

I wish I had more money as a kid. 12M on a DX2-66 would have been a dream for me. I was on a 12” paperwhite VGA display. And that was 1995...

Re:Memories

[Billly Gates]

So in other words it's just like Debian stable today?

Re:Memories

[dgallard]

I first brought up oceanpark.com on Slackware Linux in 1993. I seem to recall it was running on an 80386 but I may very well be mistaken. It was using a 14.4 modem running gopher and FTP 24x7 and later of course served web sites. I recall when I later converted from acoustic model to DSL -- I remember doing that in stages where at first inbound packets came in through the acoustic modem and outbound went out through the faster DSL connection (I don't recall the routing and DSL shenanigans involved just now). I do recall that all that was configured without having to reboot the machine (something unimaginable for a Windows server). Those were the days man.

Re: why does this matter?

[Anonymous Coward]

486sx-25 w/ 12MB of ram. It is backup firewall with two 10base-T cards.

Works great!

Get off my lawn

[PhunkySchtuff]

Linux, back in the day, originally ran on 486 processors, and ran well. You could boot the system off 1.44 MB floppy disks and it booted in well under 11 minutes.
Why reinvent the wheel and compile a modern linux when OG distros are still available - like Slack 1.01 from Feb 1995.
Download all 13 floppy disk images (less than 20 MB!) from here: https://mirrors.slackware.com/....

Re:Get off my lawn

[Anonymous Coward]

I think the point was to see if it still worked after 25 years of development. I'm sure you could find an old copy of DOS or Windows 3.11 to run on it but Windows 10 isn't going to work.

Re:Get off my lawn

[Anonymous Coward]

It probablt took 11 minutes too bot because he added nginx, udev, SELinux to the init and even compiled everything into the kernel.
He also used ext4 as the main filesystem. Using a journaled fs takes some cpu power compared to ext2/fat.

Re:Get off my lawn

[Chris Mattern]

Linux, back in the day, originally ran on 486 processors

Linux originally ran on 386 processors; support for the 386 wasn't withdrawn until Version 3.8 in 2012.

I bet that runs great

[Anonymous Coward]

Just because you can, doesn't mean you should. Can you imagine how slow a modern browser would work on such old hardware? Maybe use a text only browser. Not worth the time or effort on this one.

They do, but you need to compile manually.

[Anonymous Coward]

Furthermore the browser itself can't fit into that small a profile of memory anymore.

It is like 350 megabytes just to start now. Firefox 3.6 might still be small enough, and the newer javascript engines could be bolted on, although much of the DOM changes and similiar for HTML5 would be lost.

Having said that, netsurf will run on systems that small, although the gtk port has botched text layout and has for at least 4-5 versions. If somebody could come in and fix the font layout stuff, it would make an acceptable low feature browser for sites like slashdot, although it has few of the privacy protections you can get with noscript, adblock, umatrix, ublock, or other addons for pre-webextensions Firefox.

Also AMRISC20000

[Anonymous Coward]

Was a 486SLC-133 Cyrix core on an SoC used on a number of pieces of networkable hardware in the mid '00s.

The MGB100(110?) being an example with 32 megs of RAM, USB2, SATA/IDE, 100 megabit ethernet, and optional serial/jtag/mini-pci slots.

Gentoo ran fine chrooted on it, although the stock firmware used a linux 2.4 kernel and samba for NAS support, making it horribly slow and memory inefficient. Was a fun little device to run a copy of synchronet off of though!

Systemd

[ArchieBunker]

Systemd is probably the cause of the slow boot time. I'd love to see a light weight modern OS like NetBSD tested. Probably boots 10x faster.

Re:Systemd

[DaMattster]

I an sure Net- or OpenBSD might run half way decently but there would not be a whole lot you could do with it. I wonder if either of the OSes could even detect a network card.

Re:Systemd

[greenwow]

And you won't have the problem of systemd swallowing stdout and stderr messages. It just sucks how systemd has made it so much harder to troubleshoot booting and service starting problems.

Re:Systemd

[Anonymous Coward]

> Systemd...

Gentoo Linux uses OpenRC by default. You have to go out of your way to install systemd.

Check 102 seconds into the video. You can clearly see the string "OpenRC 0.34.11 is starting up Gentoo Linux (i486)"

Prior to that we can see that it takes the kernel nearly 14 seconds to pass control to init.

Actually, watch the video. You get a really good sense of which services take an unreasonable amount of time to start. (Under ordinary circumstances, OpenRC doesn't need to regenerate its service dependence cache, so his next boot will shave a couple of minutes off of the start time we see in this video.)

Re:Systemd

[Anonymous Coward]

Chances are any of those NE2000 cards still work fine, which is exactly the kind of networking equipment you would find in a 486. If it has PCI (and it seems to), there are many possible adapters such as Intel / 3com.

Re:Systemd

[Billly Gates]

Lol

SystemD has its faults and deserved hate for sure, but boot speeds isn't one of them. Initd would outright crash as on a 486 loading dependencies of dependences with long if/Fi for every possible outcome and loading everything in parallel.

The problem is code bloat and glibc bloat and many many things being loaded. I am curious about a FreeBSD or netbsd boot as they are very minimalistic barebones out of the box

Re:Systemd

[toejam13]

I can't speak for OpenBSD, but both FreeBSD and NetBSD support numerous ISA network cards, including the NE1000/2000, EtherExpress Pro/10, and DEC 3C50x series cards. You should still be able to use those OSes on early pre-VLB 486 systems.

Re:Systemd

[thegarbz]

Systemd is probably the cause of the slow boot time.

So to be clear, systemd whose single useless feature people advertise as faster boot time is the result of the slow boot time despite the fact that Gentoo doesn't even use systemd?

What next? Systemd kicked your dog and slept with your wife while you were debugging a sysvinit script?

Re: Systemd

[Anonymous Coward]

You could do everything with it because Unix and hence Linux is all about the command line and the tools that are there already to support it.

Re:Systemd

[TyFoN]

I have OpenBSD running on a dual ppro 200 mhz with 96 gb ram and a quantum fireball(!) drive. The drive actually impresses me most as it's 20 years old.

The system is running sshd and irssi, and I've even compiled the kernel and some userland updates a few times. There might be some compile errors if you disable things like ACPI, however it is easy to fix with a few variable guards in the kernel.

Not tried X yet though, but I could see if it could drive my pci matrox mga g550 card.

Re:Systemd

[Anonymous Coward]

(Under ordinary circumstances, OpenRC doesn't need to regenerate its service dependence cache, so his next boot will shave a couple of minutes off of the start time we see in this video.)

Given he has some songs and a python script on that machine this wasn't his first boot. So, this 11 minute boot time is the one with the shaved off minutes already...

Re:Systemd

[Anonymous Coward]

Systemd is probably the cause of the slow boot time.

No, it isn't, because the machine doesn't use systemd. I know it's popular to bash systemd, but at least do it when it's actually being used.

I'd love to see a light weight modern OS like NetBSD tested.

That would be interesting, yes.

Probably boots 10x faster.

Unlikely, but 2 times faster, maybe.

Re:Systemd

[Anonymous Coward]

> So, this 11 minute boot time is the one with the shaved off minutes already...

No, it's not. An ordinary boot doesn't need to regenerate the service dependency cache. His next boot will be able to skip that step. Modifying the config files or init scripts _directly_ consumed by OpenRC causes it to regen the service dependency cache (as changes to any of those files might change the service dep graph). (Changes to config files consumed only by daemons (such as sshd.conf or apache.conf, or...) won't change the OpenRC service dep graph, so it won't invalidate the service dep cache.)

Pointless support...is pointless.

[geekmux]

"...it takes almost 11 minutes just to boot up -- and five and a half minutes to shut down. "Despite the many roadblocks I faced, I was impressed by the level of support Linux has for ancient hardware like this."

This is like Ford advertising their latest F-150 truck can also be retrofitted with a Model T engine.

Ancient hardware is ancient, and pointless support for it, is pointless.

Re: Pointless support...is pointless.

[pele]

Well, I hope you never have to retrieve a quarter-megapixel digital photo from your graduation ceremony off an ATA disk. Or an original LaTeX of your final year project, for example. Don't worry soon enough you won't have a cd reader anywhere around you and loads of burned cds...

Re:Pointless support...is pointless.

[walterbyrd]

I spite of user name you are using, you don't seem like much of a geek.

The point is to experiment and learn.

Re:Pointless support...is pointless.

[toejam13]

For desktop or server use, sure. But the 486 series was quite popular for embedded, industrial, and aerospace hardware use. Intel didn't halt production of the 486 until 2007. I remember using a number of 486 industrial devices running BSD or DOS well into the mid-2000s.

And while early ISA-based 486 systems were incredibly slow, later PCI-based 486 systems were much better, especially when paired with processors like the AMD 5x86/133.

Re:Pointless support...is pointless.

[geekmux]

I spite of user name you are using, you don't seem like much of a geek.

The point is to experiment and learn.

Technical experimentation should have a point and produce value in order to learn from it. This little experiment provides next to nothing, and I guess my time is more valuable to me these days. To each their own I suppose.

Re: Pointless support...is pointless.

[geekmux]

Well, I hope you never have to retrieve a quarter-megapixel digital photo from your graduation ceremony off an ATA disk. Or an original LaTeX of your final year project, for example. Don't worry soon enough you won't have a cd reader anywhere around you and loads of burned cds...

My Apple IIc still reads old games on 5.25" floppies just fine, and my USB CD/DVD burner and USB 3.5" disk reader will likely be useful for some time. Rather ironic that I worry the least about my oldest media (vinyl), but it's still archived as well.

CD media suffers from physical decay over time, so archiving into the cloud serves a purpose, unlike installing Linux on 25-year old hardware.

Re: Pointless support...is pointless.

[eclectro]

Ikr?? What self respecting nerd doesn't futz around with ancient hardware/electronics/tubes/machines/cars to make and see the thing work if not for nostalgia purposes alone?

Re:Pointless support...is pointless.

[Reziac]

Friend who worked at JPL told of NASA scrounging 486s long into the Pentium era, for use in anything that went into space -- because the 486 CPU was by then thoroughly understood, and patching bugs at 300 baud is at best tedious. So they wanted absolutely known hardware.

Re:why does this matter?

[iggymanz]

if you want to watch a movie at the time you didn't need nor use a personal computer, you used an appliance that cost 15% of what a computer did.

Re: why does this matter?

[sg_oneill]

I still have fond memories of giving a talk in the 90s in "the $70 web server" to a Linux conf about dumpster diving an old 486 with a minescule amount of ram and repurposing it as a webserver , IRC host and mail server for a bunch of clubs at the uni I was at. (The $75 was for a hard drive and coax network card). I remember being approached by some IBM drones afterwards offering us a license for OS/2 to replace the Linux of the machine. I think my response was something to the effect of "haha.... god no"

TinyCore?

[Anonymous Coward]

I wonder how this would compare to using a minimal distribution such as TinyCore or SliTaz, distributions which are (in theory at least) designed to run on such hardware. Or, for that matter, how OpenBSD or NetBSD might compare.

Re:TinyCore?

[Anonymous Coward]

TinyCore on a 486 would still boot faster than Windows 10 on today's Intels.

reason for gentoo

[Anonymous Coward]

Why not use an older version of Linux?

4MB of SIPP RAM was $600+

[Anonymous Coward]

8MB of SIPP/SIMMs RAM was $650-ish. 4MB wasn't quite enough for v2.1 to run well on my 486DX/33, but with 8MB, it flew. So did Linux with X11.

I probably still have that RAM somewhere. Couldn't part with something that cost that much! Probably still have the credit card receipt too.

All your homework ...

[perpenso]

What could you possibly do with this?

Probably all your programming assignments in a Computer Science degree program. :-)

Re:All your homework ...

[Anonymous Coward]

Maybe; but you probably couldn't test the program... We use SSE instructions in our assembly courses.

Re: All your homework ...

[loufoque]

Presumably you use amd64 not 486, unless your program sucks.

Re:All your homework ...

[Chatterton]

Most of your programming assignments: yes. All, certainly not! I have coded some assignments that put my i7-6700HQ with 16GB of ram on its knees (salesman problem, perceptron...), running them on a 486 would be madness :( Right now, my current project is a program that could play othello reversi. I am doing it with a neural network without tree search. The trainning of the neural network is done in a similar way of Alphago Zero (self taught with reinforcement learning) and is running since 2 weeks on my GPU (Nvidia 960M). The programm I will deliver to the teacher will only use the CPU, but for trainning I use my GPU with OpenCL to get an >100x speedup for the computations. The IA right now is good enough to beat me on each game, but not yet at the level of programs like Wzebra.

Re:All your homework ...

[Anonymous Coward]

Or challenge yourself to rewrite your algorithms so you don't need OpenCL.

Re:All your homework ...

[Chatterton]

I have actually rewritten my algorithms (written in python) to run them in OpenCL to gain that 100x speedup. And I really need that speedup for the learning process but not for the 'playing' process.

More than 97% of the time is used in the computation of matrix product, matrix sums, applying an activation function on a vector and do some max-pooling on these same vectors. All these functions have some kind best algorithms or implementation that take advantage of cache and instruction sets. For the matrix product, I don't think I will find by myself a better algorithm than winograd. Any advancement I could put on the algorithmic side would be in the 3% left coded in python. Even if I double the speed of the rest of my program the speedup will be less than 2% in the end.

The quote I will get from the teacher is mostly dependant on how well I followed the specifications, the clarity of my code, and how my bot fare against the other student bots. Having a more complex algorithm or implementation to shave some CPU cycles here and there will only make the source code of the bot less readable for my teacher and make him raise more concern and thus more scrutinity on my code. This is why my code for the bot will not use my OpenCL implementation who is around 600 lines of OpenCL code. It use about 30 lines of numpy code and is running under 0.5s (ie: mostly instant for the end-user and in the specs for the project). This is good enough for the part of my project that I will present to the teacher.

Wait a minute

[skoskav]

Why did he put his computers and bicycle in the bathroom?

Re: Wait a minute

[greenwow]

Yes, because being in the bathroom closet versus just the bathroom makes it entirely different. /s

Re: why does this matter?

[Anonymous Coward]

I did the same on a 386 running FreeBSD. Worked fine. The primary server at a place I worked at was also a 386 but running Netware. That also worked fine.

CD-ROM and Plug-and-Play in '93

[perpenso]

I switched to Linux in May of 1994. That computer had a 486DX2 66 with a whopping 12 MB of RAM. Slackware was pretty much your only choice, and I installed Slackware 2.0 from 3 1/2 inch floppies.

On my 486DX2-66 I installed Yggdrasil Plug-and-Play Linux from a cd-rom. Graphics, audio, networking, etc all just worked automatically, it really was plug and play, as easy as a MS Windows install. Only later did I try slackware and learn the more typical cluster-f that was Linux installation, entering various technical parameters for your monitor in order to get graphics to work. To be fair my video was a popular ATI, my audio a popular Soundblaster, my networking a popular ...

Of course you can.

[mark-t]

The more interesting question is can you install a *MODERN* Linux on a 1993 PC?

I was using Linux in '93, so I can state without any doubt that you can *definitely* install it on a system from that period.

Re:Of course you can.

[nasch]

You didn't bother reading the summary before commenting, did you?

Re:Of course you can.

[mark-t]

Heck no.... this is slashdot. :)

Re: why does this matter?

[Anonymous Coward]

If that's how you prioritize your time why are you on slashdot? Go hang out on failbook or some normie forum.

Re:why does this matter?

[Anonymous Coward]

Why is that even relevant? No one claimed you could. No one suggested watching video of any type.

Re: why does this matter?

[Anonymous Coward]

I remember winamp being the first player that could decode mp3 without stuttering on my AMD 5x86 and I was amazed

RAM Prices

[Bangback]

In August 1992 I purchased a 486-66 machine with 4MB of RAM for $3000. It was pretty top of the line for consumer use. In May 1993 I purchased 4MB of RAM for $400. In spring 1995 I purchased 8MB of RAM for $300 for the machine. Ran Linux in each configuration (first install SLS in April 1993 -- May purchase was to get X somewhat functional). Each sum was a lot of money at the time for me as a college student.

In 1987, the company I worked for spent $2000 for 4MB RAM for a 386 Novell server. In 1982, the company I worked for spent $2000 for a 512K RAM disk for CP/M (trade name of "Semidisk" -- it had an external power supply and would maintain state across the reboots and power cycles of the host machine)

CD-ROM and Plug-and-Play in '93-Dry instillation.

[Anonymous Coward]

Yggdrasil was the one that came with an instillation video tape that was so dry, it would put one to sleep.

Re: No ...

[vivian]

I had a quad boot setup on an experimental rig I was using to trial various OS's at the company I was working for in Japan. it had Japanese windows 3.1 running in DosV, English winows 3.1, SCO Unix (which I think cost about $700 or so) , and NeXTSTEP - the intel based release of NextOS, that had just been released and ran like a speed demon on the super hot 486-100 that I'd built Hardest thing was finding hardware that was compatible within the narrow range of hardware supported by both NExtSTEP and SCO.

Re: why does this matter?

[Anonymous Coward]

Stick a pci gigabit nic in there and set up a shared 256mb ram drive on a modern computer to use as swap for the 486.

Sure

[Anonymous Coward]

I installed Gentoo on an SGI Indy R5000.

160MHz MIPS R5000... I stuffed it with 128MB of memory because GCC kept crashing with out-of-memory when trying to complete a number of links.

Ended up having to comment out some MTRR definitions/references that didn't exist any more or something to get the Newport gfx card video driver to compile. Takes a few minutes to start up for sure... XFCE runs on it. s l o w l y. The 10M network card is... slow by modern expectations.

Got Webkit+Midori to compile so it even has a modernish browser... page refreshes are a slideshow.

Difficulty level: R5000 requires that double loads be aligned on an 8-byte boundary. Something somewhere in webkit or GCC doesn't know this (my guess, some structure gets reordered), so after not too long it crashes with nothing more verbose than "Bus Error."

Most remarkable thing about it? The BIOS... A machine from 1993, whose preboot system has SVGA resolution, color GUI with mouse support at 60fps, and sound and networking.

Bringing back fond memories

[Antique Geekmeister]

I well remember testing out operating systems on 486 based hardware. I actually did tests with Windows, with early Linux releases, and with HURD on the same host. HURD was unusable. Linux became a critical part of the environment very quickly, since genuine UNIX systems were much more expensive than our limited development budget could support.

I had a user on a 386 in 2007

[Anonymous Coward]

I actually had a user on a 386 nearly 11 years ago ask for support running my open source program on his ancient computer. That was just over a decade ago (they were running NetBSD, not Linux), so I could see people still using 486s on home networks and as Internet routers, even here in 2018. Heck, the number of people in my extended group of friends still using Windows XP online is a bit shocking (no, you do not use Windows XP online in 2018. Use Windows 7, use Windows 10, use Linux, use NetBSD. But not Windows XP.)

Out of your way

[JBMcB]

eselect profile set 7 (or whatever systemd/desktop profile you want)
emerge -avDN @world
systemd-machine-id-setup

Uncomment the systemd line in /etc/default grub and run grub-mkconfig -o /boot/grub/grub.cfg

Reboot - that's about it. It'll usually run the enabled openrc stuff by default. If not just enable it with systemctl.

Re:Out of your way

[Anonymous Coward]

> eselect profile set 7 (or whatever systemd/desktop profile you want)
> emerge -avDN @world
> systemd-machine-id-setup

Right. That's out of your way. You also have to go out of your way to enable SELinux, or use Gentoo Hardened or...

I did it for years

[Anonymous Coward]

I installed linux on a 1993 PC ... a 66 MHz '486 with 16 MB of RAM. I went through University with that machine. I still have the disks right here: "Official Linux Slackware version 3.0 2 Disc set with ELF binaries... The internet's favorite 32 bit multitasking operating system Ready-To-Run Includes kernel 1.2.13 & 1.3.18. It was hellish to install, you needed boot floppies, and I had to include kernel parameters to tell it that I wanted the cd rom interrupt to be different from the default with a command line kernel parameter. Before recompiling the kernel to have it support Hannu Savulainens soundblaster card support, I had to change the mitsumi cdrom interrupt so that I wouldn't have to keep with the boot parameters. I had never recompiled a kernel before,
  and haven't stopped (I just recompiled a kernel a few minutes ago to isolate the kernel page table... at one time Intel f00f bugs were annoying, but allowing ring 3 applications to steal ring 0 data is really bad).

Mac

[JBMcB]

The Mac Quadra 840av was out in 1993 and supported 128MB of RAM. Heck the old IIci supported that much.

Re: Mac

[Anonymous Coward]

Even the 1988 Macintosh IIx supported 128MB out of the box. Maybe the Motorola 68k platform made it easier to address large (for the time) amounts of memory.

486s were Amazoning

[rsilvergun]

I had a DX100 because when the Pentium hit you couldn't give them away. Heck, when I wanted a Vesa local bus card I drove down to a computer shop to ask for one and they just handed me one out of the junk pile.

But I got 90% of the performance of a $2000 Pentium for about $300 bucks and most of that was hard drive & ram. I played near arcade perfect ports of X-Men: Children of the Atom & Primal Rage on it not to mention Rise of the Triad and Doom.

Re:486s were Amazoning

[hal2814]

I was the dummy who got a Pentium. Mine was probably slower than a DX100 because I opted to find one with VESA local bus. I didn't understand the technical hoops a Pentium board had to jump through to support VLB. It was a major step up from my 386SX though and at least I could enjoy my ludicrous gibs at a decent speed and screen size.

Re:CD-ROM and Plug-and-Play in '93-Dry instillatio

[perpenso]

Yggdrasil was the one that came with an instillation video tape that was so dry, it would put one to sleep.

Not that I ever saw. I bought a CD-ROM, that was it. It booted from the CD-ROM, it asked for very basic things like where to install on the hard drive, username, password, that was about it. I guess it would be boring to sit and watch it install but you could mostly walk away and it would install and configure hardware without much if any interaction from the user. It was years ahead of its time with respect to a simple Linux installation. Again, I benefited from popular motherboards and cards so auto detection worked fine.

If your recollection is correct it wouldn't matter. You could take a nap and awake later to find Yggdrasil installed, rebooted and waiting for you to login.

Re:why does this matter?

[Anonymous Coward]

What could you possibly do with this? The idea makes me cringe..

"Scratch an itch." Isn't this how Linux was born in the first place?

DSL

[Anonymous Coward]

Damn Small Linux will run much better on that PC if you want to actually do anything with it

Re:why does this matter?

[Billly Gates]

This shows how much Xorg sucks and why the kids today can't see the good on Wayland

CGA on NTSC

[Anonymous Coward]

CGA at the time could output to an NTSC TV, although rare. White text would give a color rainbowing effect on its own.
Various combinations of the 4 colors in checkerboard or other patterns would definitely result in some interesting artifact colors.

Well, this tells me modern computers are shit.

[Anonymous Coward]

And the Amiga gets the last laugh by not being subject to Meltdown or Spectre. Took a couple decades but that's OK.

Re:why does this matter?

[SurenEnfiajyan]

You mean the memory consumption for that era?

64 MB?

[m.alessandrini]

This is too much RAM to be realistic in a 1993's 486.

Re:64 MB?

[Anonymous Coward]

It is possible that the system came with 8mb or 16mb in 1993 and was later upgraded. I had 8mb in 1991 (8 x 1mb 30 pin simms). My board supported officially 32mb. 4mb 72pin were common after 93. 8x8mb would be rare then, but not unheard of in mid 90s (after Pentiums came out).

Re:CD-ROM and Plug-and-Play in '93-Dry instillatio

[Anne Thwacks]

If your recollection is correct it wouldn't matter. You could take a nap and awake later to find Yggdrasil installed, rebooted and waiting for you to login.

^^^ This.

Yggdrasil worked on all the kit I tried it on. Most other distros failed in some incomprehensible manner.

OTOH the various BSDs all worked just the same way they had on the VAX at work, although I seem to recall OpenBSD needing rather less RAM than the others, but lacking some tool I wanted, so I switched to FreeBSD, and have been using it ever since (about 1988, I would think).

How is that different from today?

[Anonymous Coward]

There was probably more attempts at gameplay innovation in those old days than there is now, relatively speaking.

While there still is, especially amongst independent studios (many of whom are making their games in 1980s throwback styles, but 2010s memory and cpu footprints), the standard today is 'big licenses' 'big graphics' and formulaic gameplay JUST addictive enough to retain players, timing them to grow bored with it for around the time your next game is coming out. Don't forget the lootboxes or other unlockables every game forces on you today. Rather than making unlocks something you only gave to players with skill, for major in-game accomplishments, they have now become something you buy with money to show how financially superior you are to others rather than what a better player you are (although back in the day that often meant the same thing, since only upper middle class/wealthy shut-ins could usually play games often enough to become masters and unlock any specials a game might have.) In comparison to that, even the horror many had about the commercialization of games in the 80s and 90s pales in comparison, because even the big companies of that day couldn't turn stinkers into big hits, which comparatively they can today, if only enough to meet costs.

Re: why does this matter?

[greenfruitsalad]

just format floppy disks to weird sizes like 1.722MB so they would fit tomsrtbt that you would never use afterwards. and remember to set up the scsi emulation on your ata CD-Rom drive so you can burn your 650MB CDs at slowest speed possible to ensure they were readable by that discman in your pocket (and remember, only TEAC and Toshiba drives are worth buying).

Re: CD-ROM and Plug-and-Play in '93-Dry instillati

[Anonymous Coward]

Tbh a few of the BSDs would still run fine because they have changed so little. In fact, the whole question is pretty much silly because all it bouls down to is software selection. It is possible to have a fullspeed and competent 486 installation if someone were to spend the time resource to sift through, update, and compile the kernels and software that runs on them.

Re: why does this matter?

[loufoque]

On the list of "interesting things to do", getting laid is quite far from the top.
If you're not a horny kid that is.

Re:why does this matter?

[vtcodger]

What could you possibly do with this? The idea makes me cringe

most things text based?

• email (alpine,mutt)
• text editing (emacs, vi, etc)
• get news and weather (I generally read Google News with w3m as it is much faster and somewhat more likely to actually get news content from diverse sites than any of the five GUI browsers I have installed)
• play mp3s?
• read slashdot (Tried that via konsole on this decade old 5x86. I'm able to get to slashdot.org and read articles and replies with no problems. Could probably log in and possibly post a reply if I really wanted to).

Re:why does this matter?

[jellomizer]

I expect that the guy was trying to argue my assertion that “On a 486 you can do nearly anything you can do on any other 32bit computer.”
Except for the fact that I stated “nearly” a 486 can decode h264. It just can’t do it in real time or even close to it.
Yes there are exceptions, and some thing will be slow. But given enough ram and time it can be processed.
I can say the same about the 386 too as it was a 32bit cpu.

486-dx5@160MHz with 32MB ram

[higuita]

I had a 486 since 1993 that i kept upgrading as much as i could, using other people cheap parts, when everyone started to use pentiums

i finally upgraded to a amd 486-dx5@133MHz, with a vesa local bus card and performed similar to a pentium 75 (except in FPU, where the pentium was more powerfull).
I then manage to grab some 8MB EDO SIMMS and upgraded to 16MB, one year later, to 32MB RAM. The MB only supported max of 16MB of ram and testing i found that with 16MB, the L2 cache helped a little, but with 32MB, it would only cache the first 16MB and then much slower to the second 16MB range. If i disable the L2, the first 16MB were just a little slower, but the remaining 16MB would be still the same speed... as the L2 gain was too small compared to the lost on the second 16MB range, i kept it disable. Then i manage to grab some SCSI HDs and CD-RW burner from a friend with a Macintosh and got myself a SCSI card and replaced the old IDE disks.
I also overclocked the bus from 33MHz to 40MHz, that put the 486Dx5@160MHz, a performance similar to a pentium 100Mhz (again, minus the FPU). i could boot with a bus of 50MHz, having a 200MHz cpu, but the sound card and SCSI, using PCI, started to act weird (as it was way overclocked from the spec 33MHz), but if i used the IDE drivers and the vesa local bus graphic card, it was stable... as i wanted the sound and scsi card, i kept the bus to 40MHz and they worked fine.

I run slackware on that machine and kept upgrading the OS until about 2005, where i finally got a new opteron64.
It was my main machine, i use fluxbox, claws-mail and mutt, lynx, dillo and only when really needed (as in the end took 2 minutes to startup), firefox. I compiled the latest kernel during the night and used it for everything, it was my server, my desktop, my learning machine... and of course, i also played some games... not windows games, but the linux ones, so most of they were light enough to run.
that 486 did run windows 3.11 and even windows 95 for a few months, but then i switch 100% to linux.

12 years using the same machine... that is a long time, specially for a old machine as that... now its easier to run a 12 year old machine, they are way more powerful.

Slackware was one of the last distros capable of booting old CPUs and only in 2015 if was forced to drop support for those old CPUs, but it requires now still i586 or above, so still be able to boot old machines. Check the slackware changelog at that time:

# Some more notes, Mon Aug 3 19:49:51 UTC 2015:
#
# Changing to -march=i586 for 32-bit x86 as several things (Mesa being one of
# them) no longer work if constrained to -march=i486. We're not going to use
# -march=i686 since the only additional opcode is CMOV, which is actually less
# efficient on modern CPUs running in 32-bit mode than the alternate i586
# instructions. No need to throw i586 CPUs under the bus (yet).

If you still pick up the last slackware capable of booting the 486 and then compile the kernel, you may still boot a recent kernel... and if you ignore mesa and some other programs, still be able to upgrade userspace too :)

Re:486-dx5@160MHz with 32MB ram

[higuita]

forgot to mention... it still boots... but i stop really using in around 2017, when i also got a cheap HP microserver to replace it as a server

Re:486-dx5@160MHz with 32MB ram

[Anonymous Coward]

Nice post! I love hearing these old 486 stories. What was the original processor on that board?

I had an original 486/33 on an AMI board with 8mb which ran as my only desktop until late 1998. I thought that was scratching it's longevity. Nicely done. I had a friend with an AMD 486/100 that he kept going as a gaming machine even longer.

My wife's P166MMX was retired from being her only office PC in 2012. I turned it into my nostalgia gaming machine and finally recycled it this month, Jan 2018. It did a solid 20 years of service and still worked flawlessly. It was useful as an office PC from 1997 through 2012, 15 years running word, excel, internet browsing and email... and the 3 1/2" floppy drive still worked just fine. I upgraded the HD to a compact flash card and ran DOS 6.22/win98 for gaming. Amazing to me that it still works. Sad that I don't have any space to store it and no one wanted it.

The 486 is still in my parent's garage. Maybe it still works.

Re:486-dx5@160MHz with 32MB ram

[Reziac]

That sounds like my kind of frankenputer :)

I still have a DX4-100 board here somewhere that takes 72pin SIMMs (can also do 30pin), and plenty of old RAM... when I find it again I should see what it can really do. It was my parts-and-RAM tester for a long time because it booted fast and wasn't fussy; pretty much anything that could physically connect to it worked.

Now I'm wondering what's the absolute minimum hardware that the oldest Puppy can run on -- anyone tried that?

Re:486-dx5@160MHz with 32MB ram

[higuita]

the ubuntu based puppy are the 686, so pentium pro or pentium 2
the slackware based one, as its from 2015, it should still have the 486 support, so that should be the one...

the plain 386 support was removed from the kernel and glibc some years ago, IIRC, to a 486 SX should be the minimum

Re:486-dx5@160MHz with 32MB ram

[Reziac]

I was thinking of the very old ISOs that are still floating around... [looking in my archive]

Puppy_Lighthouse215SeaM_Beta5.iso (2007)
Puppy_4.2retro-k2.6.21.7-seamonkey.iso (2009)

are the oldest that I have stashed here.

Here's one from 2003:
https://archive.org/details/Pu...

I have an early P75 ready to hand (well, I could bring it up from the basement) which should be close enough to a 486 for quick test purposes... stay tooned...

Re:486-dx5@160MHz with 32MB ram

[higuita]

In that case, slackware also have almost all the old distro versions available since release

https://mirrors.slackware.com/...

notice that the 1.0.1 have the 2009 date because it was re-uploaded at that time... the release notes are from 1993-08-04 08:33:56 PST

being one of the oldest distros and the oldest still alive... it will be hard to beat that :D

My first linux pc

[sad_]

was an AMD 40Mhz 486, i think it maybe already had a 120MB hard drive and 4MB ram, it kicked ass.
i needed to take my zip drive to work to download all the updates, because i couldn't afford internet back then (it was really expensive back in those days).

The '93 ps/1s were easy

[d3vi1]

The post '92 PS/1s were easy because they behaved like an AT system, but the '92 ones were a bit more difficult. I've done the same thing on a 2133-W13. It was a complicated PS/1 because linux's setup.s couldn't detect the IDE drives. It incorrectly assumed that the FDPT is at 0x41 and 0x46 and the HDD type is at 0x19 in CMOS. While that is true for the AT systems, the PS/1 systems were not AT. IBM released a unixboot.com binary that can solve this for a single boot. With a bit of hexediting to kill the final reboot you can put it as a syslinux .com executable to use as a preload to the Linux Kernel.

You can obviously solve this by adding ide0=0x1f0,0x3f6,14 ide1=0x170,0x376,15 hda=3884,16,63 hda=noprobe hdc=cdrom, but there are still some issues.
This is the boot log of a Red Hat Linux 6.2:

Loading initrd.img................
Loading vmlinuz............
Uncompressing Linux... Ok, booting the kernel.
Linux version 2.2.14-5.0BOOT (root@porky.devel.redhat.com) (gcc version ecgs-2.91.66 19990314/Linux (ecgs-1.1.2 release)) #1 Tue Mar 7 20:31:32 EST 2000
ide_setup: ide0=0x1f0,0x3f6,14
ide_setup: ide1=0x170,0x376,15
ide_setup: hda=3884,16,63
ide_setup: hda=noprobe
Console: colour VGA+ 80x25
Calibrating delay loop... 3.12 BogoMIPS
Memory: 13496k/16256k available (1000k kernel code, 408k reserved, 456k data, 60k init, 0k bigmem)
Checking if this processor honours the WP bit even in supervisor mode... No.
Dentry hash table entries: 262144 (order 9, 2048k)
Buffer cache hash table entries: 16384 (order 4, 64k)
Page cache hash table entries: 4096 (order 2, 16k)
CPU: 386
Checking 386/387 coupling... OK, FPU using old IRQ 13 error reporting
Checking 'hlt' instruction... OK.
POSIX conformance testing by UNIFIX
PCI: No PCI bus detected
Linux NET4.0 for Linux 2.2
Based upon Swansea University Computer Society NET3.039
NET4: Unix domain sockets 1.0 for Linux NET4.0.
NET4: Linux TCP/IP 1.0 for NET4.0.
IP Protocols: ICMP, UDP, TCP
TCP: Hash tables configured (ehash 16384 bhash 16384)
Starting kswapd v 1.5
Detected PS/2 Mouse Port.
Serial driver version 4.27 with no serial options enabled
ttyS00 at 0x03f8 (irq = 4) is a 16450
ttyS01 at 0x02f8 (irq=3) is a 8250
ttyS02 at 0x03e8 (irq=3) is a 8250
pty: 256 Unix98 ptys configured
RAM disk driver initialized: 16 RAM disks of 4096K size
loop: registered device at major 7
hdc: , ATAPI cdrom
ide2: ports already in use, skipping probe
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
ide1 at 0x170-0x177,0x376 on irq 15
Floppy drive(s): fd0 is 1.44M
FDC 0 is a post-1991 82077
md driver 0.90.0 MAX_MD_DEVS=256, MAX_REAL=12
raid5: measuring checksuming speed
8regs : 3.048 MB/sec
32regs : 1.524 MB/sec
using fastest function: 8regs (3.048 MB/sec)
scsi : 0 hosts.
scsi : detected total.
md.c: sizeof(mdp_super_t) = 4096
Partition check:
hda: hda1
RAMDISK: Compressed image found at block 0
EXT2-fs warning: checktime reached, running e2fsck is recommended
VFS: Mounted root (ext2 filesystem)

You can make it boot in less than 1 minute

[d3vi1]

All you need is to compile your own kernel without useless stuff such as ACPI, PCI, USB, SCSI, MD. This is a config that should work like a charm for that system in at most 3 seconds (instead of 14) on a 386sx PS/1 and with a lot less RAM based on 2.4.37.11. You only need SB32, VESA, EL3 (3COM), TTY, ISA, ISAPNP, PARPORT on the hardware side. It also has support for SMBFS. It can further be trimmed without SMBFS and NLS to around 600kb (loading and decompressing are slow on a 386.
https://pastebin.com/Mj0cudLF

Re:You can make it boot in less than 1 minute

[Katatsumuri]

Great point. I disabled and tuned every kernel compile option when installing Linux on a 486 or 386. With modern kernels, for sure there is even more stuff to configure. Most likely, there are also branches optimized for older hardware. It is a valid test to boot with minimum effort, but then don't chuckle at the boot times or other performance metrics.

You're 'old' like me, lol... apk

[Anonymous Coward]

Took a 486sx/25 in 1992 out to a 486 Dx/4 133mhz & 32mb RAM w/ a 'windows accelerator' Diamond Stealth 24 vidcard + a caching disk controller (IDE 16mb) & dual WD 212mb disks - was FAST for its day & competed well w/ Pentium I's in that setup!

* Ran Windows NT 3.51 + OS/2 2.11 - Warp 3.0 @ what was THEN "warp-speed", lol...

APK

P.S.=> Good to see there's "OLD FUCKS" (lol) like us STILL "roaming 'cyberspace'" too - why? YOU "KNOW THE FEELING" of PRE-Plug & PRAY + flipping dipswitches (remember soundcards + network cards 'fighting' for IRQ's?)... apk

When JS JIT emits SSE2 instructions

[tepples]

Firefox works on ARM which has no SSE2.

Is this about the precompiled binaries?

It's also about the JavaScript JIT code generator. If the x86 and x86-64 versions of Firefox are hardcoded to emit SSE2 instructions, the browser can't be so easily recompiled not to require SSE2.

That was the last time your work was relevant

[Anonymous Coward]

That would have been around the last time your work was relevant too.
Unlike you technology has moved on.

TLDR: Yes.

[Anonymous Coward]

See Subject

Re: why does this matter?

[Lord Kano]

I remember being approached by some IBM drones afterwards offering us a license for OS/2 to replace the Linux of the machine. I think my response was something to the effect of "haha.... god no"

They couldn't even give it away, huh? The only person I knew who was using OS/2 in that era was running a BBS and wanted something more robust than desqview.

LK

Linux on 1997 pc

[gordona]

We ported our build environment for an embedded OS from Windoze 98 to a linux distro. Boot up and shutdown was much shorter than what was reported here, but I donâ(TM)t remember the timing. What I do remember is that build times were reduced from 2 hours to about 5 minutes. This similar to builds about 15 years later

U were never relevant & /.ers disagree

[Anonymous Coward]

I'm going to continue using the Host File Engine. Your software is well written, functional. The Host File Engine performs exactly as promised by mmell February 16 2017

(APK's work), I've flat out said it's good by BronsCon February 11 2016

his hosts program is actually pretty good by xenotransplant August 10 2015

his hosts tool is actually useful for those cases in which one does indeed want to locally block stuff outright while consuming minimum system resources by alexgieg September 25 2015

I like your host file system by Karmashock September 09 2015

I do use APK's host file on all my systems at home by OrangeTide December 01 2017

I personally use a HOSTS file blocker produced from a genius called APK by 110010001000 October 27 2017

* Recommended/hosted by Malwarebytes!

(U wish u were me - "ur kind" in unidentifiable ac trolls w/ tech = waste!)

APK

P.S.=> You've done better? LOL, hell no... apk

KPTI penalty (Meltdown fix)

[DeVilla]

Was this with or without the meltdown patch? I would be interesting to try both and get a real measure of the impact of the patch on a system that doesn't need it.

Buildroot

[Anonymous Coward]

https://buildroot.org/

You can easily make a custom Linux system with buildroot to run on an 486.
Just keep the bloat low and use e.g. busybox for the userland.

What kind of question is this?

[luis_a_espinal]

Can You Install Linux On a 1993 PC

Seriously? Linux used to be run back in the day in 386 computers. I did that on a laptop (not the best experience, except for running a single app, like a database or web server.)

My main mode of running it back in the day was on a 486SX with 2M of RAM (later 4M, what's when shit was flying fast man!)

I ran X, postgress and a web server (to the exclusion of everything else.) Later I turned it into a dev system (gcc/gnat) complete with a whole bunch of other goodies.

Pretty primitive by today's standards, but useful and secure nonetheless.

I wouldn't mind having a bunch of spare 486SX's or first gen Pentiums just to slap Linux (or BSD) on them and have them run mundane jobs, like tracking stocks or something.

Re:CD-ROM and Plug-and-Play in '93-Dry instillatio

[perpenso]

Regarding BSD ... At the computer swap meet in 1993 I picked up two dirt cheap CDs, Yggdrasil and FreeBSD. Given a BSD background from university days I tried FreeBSD first, it crashed during install. Then I tried Yggdrasil, to borrow from Apple, it "just worked". And so the Linux vs BSD decision was made for me with respect to PC desktop *nix. Today its back to BSD via macOS mostly, a little Linux via VMs or old headless PCs in the closet.

i bought a pentium 150MHz

[Anonymous Coward]

i discovered my 60MHz bus was food enough for FreeCraft using Blizzard's Starcraft data. i Bought the roadblocked Intel processor that didn't scale to 66MHz bus, and neither it an MMX enabled processor. i wasted a hundred for a total of 128MB sdram, but i must say that was great until i became a multi-host whorethat sunk my soul into a Dual Pentium Pro 200MHz beast and 600$ for buffered edo dimm drams.It had faster fpu's but my soul yearned for more excitement than what my 3DFx Doodoo3 pci 2000 could musturd: So, i sold all the hosts i owned, abandones theVIA Cyrix swarm I hoped to train, and grabbed a DEC ruffian Alpha 21264 633MHz and stuffed a quad tulip nic adapter with a isa Cardbus adapter and sniped a pci Radeon 9100 pci off eGay. i was surprised i sold one unneaded Elsa Gloria-8 vga adaptor from this that netted twice as much what i payed for the host and that bankrolled the modern hardware.

Then the family pushed me onto the porche, looted my equipment, abandoned my cnc welding dreams, and have been homeless ever since. at-least i payed 10$ each for my 4 androids cluster and i am happy it runs the same shitty emu and apps that i used 13 years ago.

Yes, version specific

[Anonymous Coward]

I installed Slackware on a 486SX33 in 1993, so yes.

Re:why does this matter?

[Anonymous Coward]

I wonder if anyone went back and looked at those old video cards with MPEG2 specific silicon ( Matrox G200/400 etc ) could be adapted to at least partially accelerate H.264. I don't know the algorithm ( or MPEG2 ) well enough to know.

Re: why does this matter?

[Anonymous Coward]

I remember downloading and compiling an MPEG audio encoder/decoder ( pre MP3 ) and trying to decode and play a song in realtime on my (quite old at the time ) FreeBSD box with 386sx ( a pretty old system even then ... 20Mhz, 8MB of ram and 200MB Quantum harddrive). I managed to get it to play if I decoded in mono. It was a happy day. I couldn't play MP3 songs at all.
Hard to believe that in just a couple more years AAC and Ogg etc would come out.

Re: Re i486SX

[xanthos]

For some reason I want to say that the SX was a DX with a defective FPU that was then disabled. The FPU was new to the 486 and since a lot of software didn't support it, selling neutered chips instead of scrapping them was viable.

Mine wasn't until 97 or 98...

[Anonymous Coward]

First on a TELNET shell to a remote system a friend adminned at a high school computer program, then a year or two later on my home system when I was finally able to convince my parents I needed a book on this thing called linux, which conveniently included a slakware cd, having tried and failed a few times in previous years (as early as '93 or '94 in fact!) to download the kernel and started images from a local BBS that required subscriptions for more than 1 download a day and didn't allow resuming downloads :(

Never did get to development. People kept telling me you needed to focus on school and get out fast for more school, then you'd get a job and the rest would fall into place. Needless to say I was getting shitty for quite a while there.

Re: why does this matter?

[ncc74656]

Actually decoding mpeg layer3 was rather difficult.

The audio files he was playing weren't even MP3s, but plain old WAV files. He even took the opportunity to rickroll the audience, but it started stuttering when he ssh'd in.

Why new version?

[Anonymous Coward]

Why didn't he just download build 3.0 of slackware? It would have booted up a lot faster.

Exactly

[Anonymous Coward]

That's why this project was so perfect for a Linux wanker.

Hey, I like the idea of supporting older hardware. Within reason! Supporting a 80486 system is just technical wanking though. That system is 25 years old; think about that.

You want to run 25 year old hardware, then run 25 year old software on it. That makes a tiny bit of sense under very particular and constrained circumstances.

The tech compiled his distro on a different system just to make that part of the build more practical. There's a lesson there somewhere, I just cannot imagine what it might be...

I know you're trolling

[rsilvergun]

but my God, the effort. The craft that went into it. It's amazing. The number of period correct references here is staggering. Who even remembers the Elsa Gloria line? How long have you been storing this up?

That's nothing. I'll one up you big time!

[DarthVain]

For "fun" many years ago, probably in 1999 if I recall correctly I decided to see if I could install Linux on my grandmothers old laptop her got rid of. If I recall correctly is was a 386 SX and ran at something like 4Mhz or something like that. It may have had 2MB of RAM but I can't really remember. I believe it originally came with GRID OS installed on it, but had since been installed with Dos/Windows perhaps though I vaguely recall using a DOS/Shell on it so perhaps the Windows install died after awhile.

In any case the first problem was that the smallest Linux distro I could find at the time was Damn Small Linux at 50MB, but the internal HD of the laptop was only 20MB so that was out of the question. However I did have an old parallel port ZipDrive collecting dust, and the laptop had a parallel port... So I installed DSL on a 100MB Zip Drive across a parallel port. I booted. And waited. and waited. and waited. and got tired of waiting and went and did something else for the rest of the day. When I checked it the next day, it had successfully booted, and even very helpfully gave a boot time at the end of it, which was measured in like 20+ hours or something ridiculous like that. Probably easier to communicate with the moon.

Anyway it worked however, and once it had booted it could do Linux type things, although each command no matter how simple was literally like communicating with the moon in that there was a lag of several seconds between entering it and anything actually happening on screen. It had good uptime though :) in that it took so long to boot that I just kept it on, however it really was pretty useless other than a proof of concept, and after being left on the one time for an extended period, I eventually just turned it off and never tried booting Linux on it again except for the one time for obvious reasons.

Re: Re i486SX

[Reziac]

Yeah, that's how I remember it too.

Also remember separate FPUs on 286 and 386 boards.

Lordy, we're a bunch of greybeards...

Re:why does this matter?

[Reziac]

My 486DX2-66 couldn't play MP3s (MIDIs and MODs were fine). My P90 could, but it was a near thing (if WinAmp was going, that was about all it could do). Both with the same Win3.1 setup, so wasn't the OS hogging the system. Decoding 'MP3s simply chewed too much CPU, so the 486 stuttered and gagged, and the P90 was pegged at 90%.

However, the 486 was just fine for the office and internet software of its day. It ran Microsoft and WordPerfect Office, Netscape 3, and assorted utilities of the era, often all at the same time.

Slackware

[ebvwfbw]

That's what I used to use back in those days.

For your reading pleasure:

[Anonymous Coward]

https://ask.slashdot.org/story/09/10/30/2126252/installing-linux-on-old-hardware

Yes you can!

[Anonymous Coward]

But it's still a waste of time and electricity. Just use Raspberry Pi's! :)

Re:why does this matter?

[Agripa]

What can you do with a 486 Linux system? Probably more then you think. Just not as many things at once. You can run a web server, a database probably not both at the same time. However if you maxed the RAM you could get a lot done on slow CPU. If you checked you fast Computer most of the time your CPU is idle. On a 486 you can do nearly anything you can do on any other 32bit computer.

I used a 486 as a firewall and router for a number of years. For even longer I used a Pentium 60 which was not much faster.

I'm calling BS...

[Anonymous Coward]

His site describes installing Gentoo Linux on an "ancient" IBM PS/1 Consultant 2133 19C (released in 1993), with 64MB SIMM-72 RAM.

Unless he actually started compiling in 1993, there's no way that thing would have finished a stage3 yet...