It seems that there has been lots of hoopla over ’32 bit’ operating systems
and their inherant advantages over ’16 bit’ operating systems. Certainly
things like flat memory management is a good idea.
OS/2 has demonstrated that (after you get the 8+ MB of memory) by using
various 32 bit device drivers, you can gain a 10%-20% speed increase in
operating speed (as quoted from PC Magazine, a magazine I trust for their
benchmarks, even if that’s ALL I trust them for).
Granted, Windows NT provides many more services to the user than Windows 3.1
and this is a good reason for the 12+ MB memory requirements, but are we
going to see a general increase or decrease in operational speed due to
switching to a Politically Correct sounding OS?
As well, how will the efficiency of processes compare. Theoretically, each
process will end up taking more memory due to the non-sharing, protected
method of operation. Will we end up like OS/2 where you have protected
Windows programs but they each take 2 MB of memory overhead before the app
even starts allocating memory?
So, I guess my root question is, why is there such a great noise about 32
bit if when realized it really only nets a performance improvement of less
than one minor hardware improvement? Now that Windows is going NT, is the
difference really in the 32 bits, or is it in the accompanying extra system
services that NT will bring (and MS is now definately dropping from Dos/Win
as noted in PC Week).
- jcb
In article <1992Jul8.235620.29…@raven.alaska.edu> sx…@orca.alaska.edu (Jay C. Beavers) writes:
>So, I guess my root question is, why is there such a great noise about 32
>bit if when realized it really only nets a performance improvement of less
>than one minor hardware improvement? Now that Windows is going NT, is the
>difference really in the 32 bits, or is it in the accompanying extra system
>services that NT will bring (and MS is now definately dropping from Dos/Win
>as noted in PC Week).
> – jcb
Consider the following benchmarks:
[Compaq SysPro 486/25] IBM-PC AMIGA500 CMPQ486/33 MIPS
———- ———- ———- ———
Calculation – 16 Bit : 34.83 17.03 0.64 1.34
Calculation – 32 Bit : 22.75 1.83 0.66 0.06
Reg To Reg - 8 Bit : 34.97 10.30 0.56 0.54
Reg To Reg - 16 Bit : 38.41 12.01 0.62 0.95
Reg To Reg - 32 Bit : 45.08 5.28 0.63 0.23
Reg To Mem - 16 Bit : 28.81 6.39 0.63 1.38
Reg To Mem - 32 Bit : 38.65 5.39 0.66 0.25
Mem To Reg - 16 Bit : 28.81 6.59 0.62 1.36
Mem To Reg - 32 Bit : 36.13 5.83 0.64 0.17
Mem To Mem - 16 Bit : 30.90 4.68 0.62 0.47
Mem To Mem - 32 Bit : 32.60 4.17 0.65 0.18
File Creation : 9.67 8.51 0.83 0.32
File Opens : 7.56 14.37 0.78 0.09
File Deletions : 15.60 9.01 0.60 0.82
File Writes : 10.00 33.11 0.93 0.15
File Reads : 8.47 25.34 0.87 0.08
File Seeks : 12.50 2.49 0.83 0.35
(This was in a windows window, normally its about 0.73 vs the 486/33)
The important thing is the 32 bit benchmarks. The 16 bit test is 17 times
the speed of an Amiga (68000), but the 32 bit test rates only 1.83. And
compare it to the MIPS, its actually faster at 16 bit operations, but only
6% the speed at 32 bit operations! This is probably largely due to the
fact that the MS-DOS versions were compiled with a 16 bit compiler
(albeit with 386 instruction set where possible).
It’s a rude shock coming to the MS-DOS platform and not being able to do
a 256K malloc, something which is taken for granted on most systems!
Of course the hardware here is not at fault, it was the limitations
imposed by the compiler due to the operating system. You might not see
these things every day, but you are unknowingly being held back by it.
In article <1992Jul10.15222…@vax1.umkc.edu> edo…@vax1.umkc.edu writes:
>In article <1992Jul8.235620.29…@raven.alaska.edu>, sx…@orca.alaska.edu (Jay C. Beavers) writes:
>** >(and MS is now definately dropping from Dos/Win as noted in PC Week).****
> HUH?! Not quite following this one, please elaborate.
>E.D.
The latest PC Week has an article on the front cover that says "Microsoft
to DOS Users: The OS Future Lies in NT" or something close to that. The gist
of the article–If I remember correctly–is that many of the things we
expected to find in the next version of DOS and Windows won’t make it to
those two products, and Microsoft is putting the bulk of its efforts in NT.
According to the article the next version of DOS will contain file-compression
but few other improvements. The neato Cairo things we hoped to get in some
future version of Windows are now said to become available in Windows NT only,
etc., etc.
Stephen