from the can-it-be-hacked-faster-than-it-boots? dept.
The system takes 28 minutes to boot up but it can multitask:
Could one run a modern Windows operating system on a CPU with such a low clock speed? Developer and popular YouTuber NTDEV has proven that you can, booting and using Windows 7, which launched in 2009, on a Pentium-S processor that was downclocked to just 5 MHz. That's a full 995 MHz below the 2009-era OS's 1-GHz minimum requirement. The test system also had just 128MB of RAM, which is way short of Windows 7's 1GB minimum requirement.
In a YouTube video[...], NTDEV shows the system, which is actually a virtual machine running in the 86Box emulator[...], boot up into Windows 7 Ultimate, launch a program which shows its 5.00 MHz clock speed and even run Notepad. By the way, if you follow the sped-up time counter in the video, you'll note that it takes more than 28 minutes for the Windows 7 desktop to appear!
In the video, you can see NTDEV power on his virtual machine which POSTs as a Pentium-S running at 50 MHz with a 128MB of RAM. However, it has been downclocked to 5 MHz, a low speed which NTDEV told us he achieved by editing 86Box's source code.
[...] Now that he's gotten Windows 7 to run at 5 MHz (or even 3 MHz with less to do), NTDEV says he's looking at ways to get Windows 10 or Windows 11 to run on a processor that's slower than 1 GHz. He's already managed to get Windows XP to run at just 1 MHz[...].
(Score: 3, Funny) by Ox0000 on Thursday December 29 2022, @01:43PM (5 children)
First off, that's friggin' amazing!
Second:
Isn't that about the same time it takes a Win11 on the latest intel chip to become functional upon booting and then sending all its telemetry to the mothership, pulling the daily
borkupdates, coercing your into making edge your default browser, and expanding the start menu to serve you ads?I wouldn't know, I don't actively install malware on my devices...
(Score: 2, Informative) by Anonymous Coward on Thursday December 29 2022, @02:45PM (2 children)
The 28 minute boot time reminds me when I was a grad student and new Pentiums (I think) came out. The secretarial staff all got new computers, but nothing for us grad students, and I was given an i486 with Win95 on it. I tried to install Win98 on it, which had recently come out, and though the computer I had met Windows' "minimum requirements," it was really really slow, like almost 28 minute boot time slow. My graduate work was all done on mainframes and all I really wanted was a computer at home from which I could dial in to the university. Because I had what was, to me, essentially a useless machine, I heard about Linux from a friend and installed Red Hat and Star Office on it. It was wonderful.
(Score: 3, Interesting) by RS3 on Thursday December 29 2022, @06:53PM (1 child)
I'm a little bit more of a tweaker, maybe it comes from being an engineer and the philosophy that the human user should always be in control. Yes, it's a losing battle.
Point is, I'm always getting in and tweaking, tuning, trimming, etc. I like lean and mean, fast, efficient.
You may have had swapping happening, btw? I don't know enough about Windows to understand why they seem to always have some disk swapping happening. I've been trying to disable it and it seems every version of Windows will create some kind of swapfile. I know that Bad Things can happen if the system truly runs out of RAM, but I'd prefer a process or app be shut down orderly, rather than swapping.
Linux servers I admin have a few tiny things in swap, but it's like 0.1% of total swap space.
Of course, yours and OP's cases are greatly helped by hibernation. In the AV world that I work in a bit, many soundboards, lighting controllers, and video switchers are (sadly) run by WindowsCE ("embedded" Windows) and every one I've come across uses hibernation. You can force a full boot, which takes much longer, if you feel like something is unstable, but hibernation is the default behavior.
I hate moving to a new Windows version. Each one takes me more and more time to learn, tune, tweak it... For many years, late 90s to maybe 2015, Linux (Slackware) was my main desktop. For several reasons I was forced to use Windows, and I got enough control of it that it's Good Enough. And frankly, in my experience, the maybe too many Linux distros are so wildly different and constantly changing (and systemd is so pervasive) that for me, Win7 has been more stable and just works. I think browser updates and crashes have been the biggest hassle.
(Score: 2) by dry on Sunday January 01 2023, @11:45PM
You have to remember a few things. Memory was stupidly expensive so you were lucky (or using a work related box) to have over 4MB of ram. I never really used Windows after 3.1 but I think with win9.x, especially after the first release of Win95 when they started using a browser for the shell, needed over 4 MB's just for the OS. NT was worse, 16MB's was the bare minimum. So basically impossible to do anything without a swap file. Also the file systems, variations on FAT, were crap, faster to load something at the beginning, swap it out and when needed, swap it back in then loading it from the FAT file system. Of course hopefully your swap file was not fragmented, another reason to create a large enough one at boot rather then growing a fragmented one.
Linux wasn't much better with 4MB's of ram, it didn't take long before swapping happened, and there were some bad choices. I remember a version of KDE that would regularly commit all memory, system would come to a halt as memory was swapped out and back in. I quit using KDE after that.
Unless you had an abundance of ram, having programs shutdown when not enough real memory would have meant not being able to run multiple programs.
(Score: 3, Interesting) by vux984 on Thursday December 29 2022, @08:09PM (1 child)
Interesting coincidence though, my parents found a Toshiba Satellite Pro 4600 laptop from ~2001 in a closet and brought it out over the holidays. We booted it up, and discovered I'd put Ubuntu on it in 2006. It took about 10 minutes to fully finish loading to the desktop (you could see the desktop within 5 but the taskbar icons/utilities/widgets for the clock, wifi etc took another 5 before it was all done. After that it was usable, but not good, and the lack of a trackpad was pretty strange, and the thumbstick was pretty jerky. Worked better with an external mouse.
The sticker was 'built for Windows 2000", and it too had 128MB RAM; google suggests something in the range of a Pentium III would have been the CPU, but i didn't check that particular model. Obviously far and away faster than the 5Mz cpu in the article, but otherwise similar era OS with the same RAM.
I think its interesting that crippling Win7 it with a 5MHz cpu and 128MB left it only a few multiples slower than my similar era Ubuntu install, on what would have only been a 5 year old laptop at the time. It was certainly usable and would have been better than nothing, but it was not remotely 'good'.
"Isn't that about the same time it takes a Win11 on the latest intel chip to become functional upon booting..."
Since you asked, it's about 10 seconds on my desktop from powered off to using it, including the time to type in my password. That's not to say I love windows 11 or anything, but I'd rather criticize it for the things it's actually bad at. (Although I did a wipe and reinstall over the holidays of my 7 upgraded to 8 upgraded to 8.1 upgraded to 10 upgraded to 11 box with a clean install of 11, and that resolved nearly all the "issues" i was actually having with that box.)
Also, edge isn't my default browser, but it did ask. I've mostly tamed the start menu via settings (turn off 'search highlights', and so forth and that gets rid of the worst offenses), but I agree it's *truly obnoxious* that i can't easily limit it to just searching apps/items on my desktop, on the 'pro' edition. But honestly, despite this, I do like it better overall than previous start menus; especially the current set of right click options.
(Score: 2) by toddestan on Saturday December 31 2022, @06:11AM
Looks like the machine was run on an emulator, so it probably had the benefit of virtual hard disk stored on a fast SSD. That can make a pretty big difference.
At work there's some old laptops with 5400 (or maybe 4200?) RPM drives and 4GB of RAM. Between Windows 10, Windows update, the corporate anti-virus/security software, Teams, and other bloat, you're probably most of the way to 28 minutes from power-up until the drive stops thrashing enough that you can actually do something. The CPU's are actually somewhat dated but still decent Core i5/i7's, but that's not the bottleneck.
(Score: 3, Interesting) by bzipitidoo on Thursday December 29 2022, @03:31PM (7 children)
I have a "nettop" with a 166 MHz Pentium, and 96M RAM, the maximum it could have. Came with Windows 98. Last time I used it, about 15 years ago, I put a lightweight Linux distro on it. I tried to start it up again earlier this year, but found it would no longer work. The hardware has gone bad. Don't know why, and it's really not worth spending time on. Last time it did work, Linux came up reasonably quickly, but after that, it was text terminal or forget it. Firefox 3.5 took 30 seconds to come up, with the start page set to blank, never mind trying to load a web page. Stellarium took 5 minutes to start. I didn't even try to play a video.
I found that turning off font anti-aliasing and hinting noticeably sped up the GUI. Basically, that old computer doesn't have hardware accelerated graphics. Modern software really needs that to perform acceptably. Sure, Doom runs fine on that hardware. It's amazing just how weak graphics of that era was, and what programmers managed to do despite that. In the 1980s when Intel sold math coprocessors separately, before the 486, you had similarly huge performance differences. An 80387 was roughly 50x the speed of an 80386 emulating the floating point math. It's mildly interesting to see benchmarks of the Intel integrated graphics circa 2000. The reputation for being dog slow is well deserved. A low end dedicated graphics card from the 2000s might score 200. Those integrated graphics scored just 1.
(Score: 2) by bzipitidoo on Thursday December 29 2022, @03:34PM
Oops, it was a 133 MHz Pentium, if plugged in. When running on battery power, to save power it ran at half the speed, 66 MHz.
(Score: 2) by JoeMerchant on Thursday December 29 2022, @03:37PM
>The hardware has gone bad. Don't know why, and it's really not worth spending time on.
100% concur on the time-value. In my experience, it's usually the power supply (capacitors, or dust-shorts) sometimes cooked chips due to forced air cooling failure.
My early Nettops came with Win98, ran well, then Windows updates would cripple them into unusability, re-install from CD (using an external CD drive) would get them running like new, but eventually you couldn't connect to the internet without being harassed into upgrade. Never tried Linux on those little things, I'm sure they would have done well, but the new NUC stole the show.
🌻🌻🌻 [google.com]
(Score: 2) by RS3 on Thursday December 29 2022, @07:05PM (4 children)
All of what JoeMerchant wrote, plus possibly corrosion in connectors has caused flaky connections to things like RAM, CPU, etc. You might get in and re-seat (exercise) every connector SIMM/DIMM you can, including hard drive, power supply (if there are any connectors).
If you do, be sure you've properly dealt with CPU heatsink grease. Sometimes there's a silicone heat-conductive pad, but sometimes it's grease that will have dried up and you'll need to clean it, preferably with isopropyl alcohol, apply new grease.
Be careful of metal-bearing greases like "Artic Silver". They're often conductive, especially when they start drying out and the metal particles start interconnecting. There have been many major failures and permanent CPU and MB damage when too much of that stuff is applied and it gets on electrically "hot" components, like the little capacitors often found on top of older CPUs, or just accidentally / sloppily gotten on MB parts / traces.
I assume you've checked (replaced) the CMOS battery? Often 3V "coin" cell.
And maybe simply the hard disk is failing. I'm so sick and very annoyed with how unreliable hard drives have become over the years.
(Score: 2) by bzipitidoo on Thursday December 29 2022, @09:42PM (3 children)
I haven't taken the time to try any of that. There's no data of any value on it. I made sure of that the last time I used it, years ago.
Only reason I fooled with it early this year is that it has a 3.5" floppy drive (external, connects to the parallel port), and I wanted to check what was on some old floppies of mine, and possibly save the data if it was anything valuable. I am dismayed how difficult reading old floppies has turned out to be. That computer failed me. I had kept a combo 3.5" and 5.25" floppy drive, but learned that no modern computer has a floppy drive interface any more. Computers from 15 years ago do have a floppy drive interface, but only for 3.5". I had thought floppy drive interfaces were unified, so to speak, in that if there was one, it would of course have the means of handling both 3.5" and 5.25". But it seems not. What did it save them to omit the means of handling a 5.25" floppy drive, 1 kilobyte of ROM? Licensing fees? I bought a USB external 3.5" floppy drive, and was cheated again. It can read 1.44M floppies, but it cannot read 720K floppies.
(Score: 3, Interesting) by RS3 on Friday December 30 2022, @12:05AM (2 children)
It might not take long to try at least removing and re-seating the RAM. That and I've had a lot of RAM failing lately, not sure why. Random machines, no thermal or any other physical abuse, no "power surges".
Yeah, I have a bunch of floppies that I keep meaning to copy into a hard drive. Over the years, especially the past 10 or so, I'm finding most floppy drives won't read most of my floppies. Back in the day, 90s and early 2000s, most of my several drives would read most of my floppies. I don't know what is degrading. I always wanted to get standard alignment floppies and properly align some drives. Maybe some of the floppies could be read if I tweak drive mechanical adjustments.
https://www.minuszerodegrees.net/5150_5160/tandon_tm100/tandon_tm100-2_problems.htm [minuszerodegrees.net]
I wasn't aware that the floppy interface has been dropped, but I guess I'm not surprised. Heck, most computers don't come with optical drives.
That's interesting about 3.5 but not 5.25. IIRC the floppy electrical interface, signals, and code is the same for both drive sizes, so it's just a few parameters that you'd need to have available in BIOS.
So maybe it's just the BIOS? IE, once Linux loads and runs, if it was compiled with 5.25" support (in kernel or module) maybe Linux would see it?
As I'm sitting here writing, I'm remembering more and more. Technically floppies need a "bus terminator resistor". Most, if not all signaling cables, including coax, need termination resistors, one at each end, equal to the characteristic impedance (complex resistance) of the cable. Common are 50Ω, 75Ω, 93Ω, 100Ω, 150Ω... SCSI uses it, really all high-speed analog and digital signalling cables need it.
So maybe one wasn't installed? Or maybe too many were?
https://www.lo-tech.co.uk/wiki/Floppy_Disk_Drive_Cable_Termination_and_Device_Addressing [lo-tech.co.uk]
https://www.minuszerodegrees.net/5150_5160/tandon_tm100/tandon_tm100-2_terminator_1.jpg [minuszerodegrees.net]
(Score: 4, Interesting) by bzipitidoo on Friday December 30 2022, @01:11AM (1 child)
If there may be something on them that you want, don't sit on those floppies for another 5+ years. My floppy disks are all at least 25 years old. The newest ones can still be read, mostly, but the older ones, not so much. In some cases, only a few bytes were corrupted, something like 5 bytes out of every 100k could not be read. Sometimes it's just one bit got flipped. ASCII text files with that kind of damage are fairly easy to repair, can often figure out from the context what the correct character is. Zip files though are all but impossible to repair. If it's just a very few bytes, and you know which ones were corrupted, you could brute force it, trying each byte value in turn until the decompressor is happy and it produces sane results. Worse though is an entire sector going bad. Then there's track alignment problems. Quite a few disks, the last third, beyond approximately track 50, couldn't be read. That could have been the mechanics of the floppy drive, various rubber components having shrunk a tiny bit after all those years of disuse and making it impossible to align across the full diameter. Or maybe it was that the outer edges of the disks, being a bit more exposed, corrupted a little more and sooner. All in all, I'd say 35 years is about the limit for 1.44M floppies. Possibly lower density disks hold data longer.
(Score: 2) by RS3 on Friday December 30 2022, @04:10AM
All too true. Only thing semi-permanent is a hammer, chisel, and stone!
I forget what's on some floppies. A bunch are 20+ years old, backups from some work work. IE, irrelevant now anyway.
Years ago I learned to format floppies at least twice before using them. Even then I'd run some strong disk sector test software. I can't remember all that I used, but I think Spinrite will do floppies. Another is a very old Norton Utilities that does similar to Spinrite- it'll read a track, store the data, reformat that track, write the data, and fully verify the write, then move on. I forget what it's called, but I have it, yup, on a floppy somewhere! On old hard disks too. I'm pretty sure it retries any sectors that seem iffy, but I forget how many retries it'll do.
Yeah, again, I've always intended to get some alignment floppies and make sure a drive is aligned.
Most of the drives I've opened didn't have many rubber things, but I know what you mean. Somehow metal things seem to move over time. "Cold flow" someone once said. Not sure I agree with that, I doubt metal moves like a glacier! But, I will give thermal cycling a vote for cause of problems.
That said, I know I have a couple of Teac drives that still seem to work well. :)
Maybe I'll get to it someday. It's all a matter of priorities.
(Score: 2) by JoeMerchant on Thursday December 29 2022, @03:34PM (3 children)
Wasn't 7 one of the early NT kernel rehashes, with a bunch of the old Win95 accessory ware ported along for the ride?
Is that where Windows still is, in terms of "modern OS tech?" 28 minutes to boot _is_ inkeeping with the Windows experience...
Gotta know: What's the Crysis framerate?
🌻🌻🌻 [google.com]
(Score: 4, Informative) by RS3 on Friday December 30 2022, @12:09AM (2 children)
Not definitive expert, but AFAIK, NT3.5 -> NT4.0 -> 2000 -> XP -> 7 ...
(Score: 2) by JoeMerchant on Friday December 30 2022, @12:29AM
Sounds familiar, I had a 7 notebook new in 2007... Won it in a crazy Intel contest, super expensive thing - same one Daniel Craig used as James Bond in Cuba... Fragile, overheating piece of junk, I am very glad it was free instead of $3800.
🌻🌻🌻 [google.com]
(Score: 2) by dry on Monday January 02 2023, @12:01AM
There were also some server versions in there. Better also to use the kernel versions, which were basically the same until Win2k, which was 5.0, XP was 5.1, 5.2 was some server version, then I think 7 was kernel version 6, 8 might have been 6.1 then I don't know. Still the jump from win2k to XP was actually fairly minor kernel wise.
(Score: 2) by RamiK on Thursday December 29 2022, @04:28PM
Note how typing in notepad causes the whole line to flicker while the terminal doesn't. To be clear, this shouldn't be happening: Even if there's really not enough time to draw the line, there's no justification for a buffer flash.
Worth keeping in mind when trying to figure out what's wrong with emacs / gvim / firefox / xorg etc and why people tolerate* the electron bloat.
* possibly past tense considering contemporary attempts to do away with it: lapce [github.com]
compiling...
(Score: 5, Insightful) by SomeGuy on Thursday December 29 2022, @04:29PM (5 children)
I don't know, you can run anything as slow as you want. You could do each instruction calculation by hand and take until the heat death of the universe before Windows "boots". Using an emulator for that does not seem that huge of a challenge.
I suppose it is interesting to see if there were time-sensitive pieces of code. In the Windows 9x days it was quite common for Windows to fail if a CPU was "too fast" because some developer assumed some piece of code would always take a certain amount of time to run.
It would be a little more interesting to get real hardware down to that speed, as it is certainly not designed for that.
(Score: 2) by RS3 on Friday December 30 2022, @12:16AM (4 children)
Yeah, I remember something about that. Some stupid CPU-cycle-hogging timing loop went too fast and something failed. IIRC it's one of the reasons the very old "clone" systems had "turbo" switches to slow the CPU if you ran into the problem. I don't think I ever experienced it. My first Windows was 3.11, so maybe that problem was in older Windows versions?
(Score: 3, Informative) by SomeGuy on Friday December 30 2022, @12:29AM (3 children)
Those turbo switches were used more for games that relied on the CPU speed and I/O bus speed to get game play timing right. There were few productivity applications that would fail by themselves - however several unfortunately popular forms of floppy disk COPY PROTECTION would fail on anything faster than an IBM AT 286 (6 or 8mhz).
(Score: 2) by RS3 on Friday December 30 2022, @12:47AM (2 children)
How long is statute of limitations on admitting to disassembling and patching around one of those schemes? :)
It was simple food / diet tracking software that would only run with the original floppy in the "A:" drive. That really bugged me- not only did it have to be in the machine, but you could not back it up in case something damaged the floppy (explanation follows...)
It had a crap-ton of do-nothing code, code that patched the single-step interrupt vector and crashed you hard if you were trying to single-step the assy code.
Also had intentionally buggered floppy with certain sectors intentionally marked bad. Can't copy that! Although I believe there were (are) utilities that can do that, I didn't have them at the time (early 90s).
Good old dot matrix printer with long fan-fold paper was my friend. Still have a couple of those printers and a small box of paper (haven't used it in a while, sadly...)
I ended up putting in a jmp around the whole mess, was then able to run from HD, no floppy needed.
(Score: 1, Interesting) by Anonymous Coward on Saturday December 31 2022, @10:40AM (1 child)
(Score: 2) by RS3 on Monday January 02 2023, @01:03AM
Cool story bro! :) Teasing- actually very interesting. Yes, I used DEBUG to do the debugging. IIRC you could pipe commands into it and get it to spit out the code listing that I'd > into a file to be printed.
The single-step vector hijack was pretty obvious when I looked at the code. Trying to use "p" in DEBUG was supposed to single-step, but aforementioned critters caused it to jump to a bunch of looping loops with a bunch of other code that looked like it might do something more than just waste hours of someone's life.
I never heard of "Quaid Analyzer" but that sounds cool, and looks like it's still around. Thanks!
(Score: 2) by tizan on Thursday December 29 2022, @05:55PM
Life in the slow lane eh ?
(Score: 2) by legont on Friday December 30 2022, @04:54AM
Most Russian hackers run - albeit pirated and twiked - Windows 7. They call it poker.
The reasons galore, but one of them, it'd run on any hardware one could reasonably use nowadays.
"Wealth is the relentless enemy of understanding" - John Kenneth Galbraith.