• Please review our updated Terms and Rules here

Win95 on a 386?

L

leeb

Veteran Member
Joined
Jan 21, 2009
Messages
709
Location
Palm Springs CA
Has anyone been able to run Windows 95 on a 386?
I thought I had at one time, but now Im not so sure...

I attempted to install onto my recently-acquired-and-decked-out GRID 1755 (386sx/20+387, 1.3gb HD, 8mb) laptop.

It SEEMED to do a clean install, but subsequent boots came up to the splash-screen then shortly cleared the screen and asked for a 3-finger salute...

Starting in safe mode provided the same result.

Im willing to concede that the GRID may be too old/slow/strange to run 95, but it would be 'kinda neat' even tho it would be a slug...
I cannot remember what my first 95 machine had, but it may have been a 486sx/33...

It would possibly do it if I could find a 486sx to put in it... if a 486sx can directly replace a 386sx... That may be another thread...:roll: :mrgreen:
 
My main computer for a while was a 33 MHz 386DX with 12 MB RAM, so it's definitely possible. I was running the "A" release of Windows 95.
 
Yup, I ran 95 on a 386DX once... but it was way after market (first officially used 95 machine was a P-100 after all)... just had a 386DX laying around and figured, hey, let's see what happens...

But TNC's page there: that's AWESOME haha :D
 
Well, according to the KB page it needs a DX... I guess an SX with a 387 doesnt count. :(
(tho the technet page didnt work...)

I guess I'll live. :mrgreen:
 
Well here's what I'd do, assuming the install is a pain.

Install DOS and Win95 on it's HDD while it's hooked up to a different machine with at least a 486. Make sure it's set up to boot to DOS, that way you can get access if Windows crashes easily. Transplant the drive back to the 386.

Try running Windows 95, and if it tosses an error or crashes, reboot to DOS and troubleshoot/tweak. Rinse and repeat, Google as necessary. :D
 
Like I said, the install seemed to go w/o a hitch... it was the actual running that broke.

Once I get the new clock in it (Dallas type) I'll try again.

How can I tell if it is a 303? there is no marking stating that AFAICT....:confused:
 
Last edited:
leeb said:
Well, according to the KB page it needs a DX... I guess an SX with a 387 doesnt count. :(
Click to expand...

Yeah, that nomenclature is 486 specific. 386SX vs DX refers to the size of the external buses, which are narrower on SX and full size on DX. (So a 386DX does not imply coprocessor, nor does adding a copro to a 386SX make it qualify as a DX.) That being said, most likely SX vs DX in and of itself wouldn't PREVENT Windows 95 from working ... they are essentially equally functionally capable, just not equally performance rated. Just like that site which shows old Pentiums running Windows 7, when they specify a minimum sys requirement, it doesn't necessarily specify the minimum hardware required to actually make it run -- it just might specify the minimum hardware that they think will make it run and actually be USEFUL at all! (Read: Win95 on a 386SX is probably going to be crazy slow.)
 
Personally I found the performance of Windows 95 to just barely scrape the bottom of the usability barrel at 486SX-25, with some machines with that CPU running it impressively usably, and some running it like a slug. It seems to depend on RAM and other factors as well when running on a machine that slow. A 386 would be unbearably slow, but possibly not much different than the 486SX-25 if the 386 in question is 25mhz or more, the copro might even make it run better than the 486SX-25s I've run it on.
 
This won't help you get it running, but will make it run a little bit better once installed. I suggest editing setuppp.inf and removing the lines that reference the following files:
* ohare.inf - IE 3 browser
* athena.inf - Microsoft Mail and News (IE 3 e-mail client)
* mos.inf - Microsoft Network and on-line registration.
* msinfo.inf - on line services.

More detailed instructions are available here: http://toastytech.com/evil/lab.html#rem95

I think the original Win95 slims down to 17 meg of disk space when you do this, and the less-cluttered c:\windows directory improves performance dramatically on questionable CPUs. I used this trick to make life with Win95 more tolerable on the 50 MHz 486s where I worked back in 1997-98. But it made it run faster on my 350 MHz AMD K6-2 at home also.

Win95 will be very slow on most 386sx boxes, but of course slow doesn't mean uninteresting. And it could be fun to see just how usable you can make it.
 
leeb said:
...How can I tell if it is a 303? there is no marking stating that AFAICT....:confused:
Click to expand...

That specifically refers to an Intel 386DX B1 step (CPUID 0303h). If your Intel 386DX is faster than 16MHz or is not an Intel brand (i.e. AMD), you won't need to worry. For a 16MHz Intel 386DX you look for the "double-sigma" (two 'E's) marking on the chip's marking to know it should be fine.

"White label" Intel 386DX are also later and should be fine...

It's the same warning for Windows NT and 32-bit OS/2 as well...
 
I know for a fact that Windows 95 (the original release, not OSR2) will run on a 386SX, at least when you have a 387SX co-processor installed -- however, it will be VERY slow and will crash very often.
 
southbird said:
Sure that's not just the Win9x kernel part of this equation? :)
Click to expand...

I never got that many errors when using Win95 on a 386SX; I just frequently ran out of patience waiting for menus, windows, and dialog boxes to pop up and finish drawing the screen -- but keep in mind I was only running it with 4 MB of RAM, which is barely functional for Win95, even on a 486DX.

Just think of what it's like on a modern computer when an application gets stuck in a loop that maxes out the CPU at 100% load, causing everything else to slow down to a crawl until you can terminate it -- that's what using Windows 95 on a 386SX is like, all the time.
 
If the machine is maxxed out to 16 meg (the most a 386sx can handle), it'll help. Frankly I never liked Win95 with any less than 16 meg of RAM, no matter what CPU you were running.
 
*remembers things*

Right, the Intel 386s that were first released weren't properly 32-bit. They had problems, but still far outperformed the 286s, so they released them anyway, disabling the functions that were broken. After a while they fixed the problems and released what people are used to as the 386.

@Dave Farquhar:
Thanks for that tip - will make Win95 on my 486-class machines a bit faster and certainly less huge (huge when compared to the 200MB HDD this machine shipped with).
 
Raven said:
*remembers things*

Right, the Intel 386s that were first released weren't properly 32-bit. They had problems, but still far outperformed the 286s, so they released them anyway, disabling the functions that were broken. After a while they fixed the problems and released what people are used to as the 386.
Click to expand...

I think you're mischaracterizing the situation, Raven. Initially, when the 80386 came out, it was discovered that some, but not all, chips had problems in 32-bit mode. It was basically a yield problem, not a design problem. Intel tarted up its testing process a bit and labeled the chips that passed the 32-bit tests with the double sigma. The ones that failed, but would still run 16-bit code, were labeled "16 bit S/W Only". Both came from the same process and masks and usually, from the same wafer.

Intel had been doing this kind of thing for many years before in their DRAM production. If you bought a 250 nsec. 2117 and a 150 nsec. 2117 DRAM in 1978, they weren't chips that differed in design or came from different fab lines or processes--they were categorized by speed as they came off the line. (Which is also why your chance of discovering a 150 nsec. 2117-2 in a batch of 2117-4 (250 nsec.) parts is pretty minuscule--the cherries have already been picked at the factory).

Making chips is not an exact science.
 
You must log in or register to reply here.
Back
Top