VOGONS


First post, by matze79

User metadata
Rank l33t
Rank
l33t

Hi,

anyone knows if Windows 3.x uses the FPU on 286/386 Systems ?

Thx

Matthias

https://dosreloaded.de - The German Retro DOS PC Community
https://www.retroianer.de - under constructing since ever

Co2 - for a endless Summer

Reply 1 of 9, by Jo22

User metadata
Rank l33t
Rank
l33t

Hi Matthias,
Yes - As far as I know, Windows 3.1x is said to do so, but not Windows 3.0.

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 3 of 9, by Jo22

User metadata
Rank l33t
Rank
l33t

It had been tested thoroughly in an article of win magazine, issue 5/'92.

Funny, though, that you mention Excel. The magazine said (by analogy):
"On Windows 3.0, co-processors lived a shadowy existence: Even by Microsoft's own estimation,
they were used for Windows applications such as Excel at best."

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 6 of 9, by Matth79

User metadata
Rank Oldbie
Rank
Oldbie

Windows calculator does - also on the Pentium .... we are Pentium of Borg, you will be approximated!
The thing is, what uses floating point - by the standard Windows function calls, it will be passed to the FPU if present, emulated if not

Reply 7 of 9, by Jo22

User metadata
Rank l33t
Rank
l33t

I forgot to mention - Microsoft once released a patch, WW0548.EXE, for systems with i387 FPUs.
Apparently, it was required for Windows 3.10 on 386 PCs and contained a new WIN87EM.DLL.

That dynamic link library does provide both virtualization and emulation of x87,
so multiple programs could safely make use of the math co-processor.

It is akin to WIN87EM.EXE, which was used in the days of Windows /386.
(Back then, it could be added optionally as load=win87em in win.ini.)

Anyway, 80486DX and 80486DX2 processors were excepted by that WW0548 patch..
I don't know if it was/is required for 286+287XL (i387 core) or 386+287 combos, though. 😕

Since the patch files in that self-extractor are dated mid-92, later releases of Win 3.x,
such as plain 3.11, WfW 3.10 or WfW 3.11, may already have shipped with a new DLL.

https://support.microsoft.com/en-us/help/8686 … for-intel-80387

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 8 of 9, by keropi

User metadata
Rank l33t++
Rank
l33t++

I am resurrecting this thread because I did some quick tests on a 486slc33 system with and without a Cyrix CX83S87-33 fpu.
The rest of the system is 64kb L2 cache, CL5420 vga controller (NOT 5426/5428 stated on winmark) and 16mb of ram running wfw3.11 at 800x600/16colors.

WinG result:
p7Suisll.jpg

WinBench 3.11:
6TyciGth.jpg

so there is some boost in gfx speed - keep in mind that the 5420 does not really offer any gui acceleration, maybe a better vga sees no boost at all.

🎵Link to buy a PCMIDI mpu
🎧Orpheus soundcard project
💻WTB Amstrad PC7486SLC-33 system

Reply 9 of 9, by Jo22

User metadata
Rank l33t
Rank
l33t

Thank you very much for testing, Keropi! 😎

What about the standard VGA drivers (3.1 and 3.0) ?
The 3.1 version does dirty-rectangle from what I remember, maybe it has some FPU code also ?

Anyway, just curious - since I have no physical 286/386 at hand now (all boxed up)! 😅

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//