After spending a few hours last night trying to figure out if S3SPDUP
really worked, especially with GP2's log, I thought I would share the
results with everybody, since I have seen lots of different opinions in
the news...
For a start, let's not forget it only applies to S3 based cards. It
won't help you otherwise. I have a Stealth 64 Video 3200. As you know,
it is not a killer under DOS, so a little help from a utility was
welcome.
I am not going to explain the whole test procedure but the main results.
The Log:on results shows a VGA score, nothing else. S3SPDUP by default
works for VGA. So, it will show a big improvement in the log (was 177 -
now 341). However, if you play SVGA, you won't see any improvement in
frame rate... Go out and unload it, reload with the VESA otpion. Now,
the log is back to the poor 177, but yes, the processor occupancy
dropped quite a bit (from 100 to 93%) BINGO. Now to evaluate the gain,
it redid thew whole procedure in VGA mode, knowing that the gain
reported in the log was 100%. The Processor Occupancy then fell from 70%
to 60%.
Two things must be kept in mind after reading these results:
1) the Proc. Occ. is relatively inaccurate as it doesn't
increase/decrease by 1% but rather around 10% chuncks. It is impossible
to estimate properly the gain but I feel it is about the same as in VGA.
2) The video performance does not seem to be as important as initially
stated (10% improvement for double video performance)
S3SPDUP shows great results when run against benchmarking software,
especially in 32 bit mode. I know GP2 runs with the DOS extender which
means 32 bit but is the graphic engine doing 16 or 32 bit access to the
card? This is critical as the performance improvment is 33% in 16bit
against 300% for 32 bit (SVGA).
GP2 wise, I run DOS 6.22, no fancies, a T2, all sounds on, SVGA no
textures and only smoke and track in mirrors. Details are on Medium
setting. Frame rate 16/sec. Although I was a frame rate freak in GP1
(50), I must admit that it is very playable at 16/sec, if the proc.
occup. is kept under 100%.
As a last request and to enlight this subject even more, I am looking
for people with a similar system as mine and with other video cards
(Millenium, ET6000, Dynamite,...) to set exactly the same options as
mine, go to quick race bith Monza and Hockenheim, wait for the Buzz of
the starter and the get the Proc. Occup. before the start of the race.
My system Spec:
ASUS P55T2P4 256 Ko (Async I think-correct me if I am wrong) Triton II
133 - 32 Mb (Non EDO)
Joystick enabled
GP Rating 258
Thanks for your time
JFD
[snip]
: S3SPDUP shows great results when run against benchmarking software,
: especially in 32 bit mode. I know GP2 runs with the DOS extender which
: means 32 bit but is the graphic engine doing 16 or 32 bit access to the
: card? This is critical as the performance improvment is 33% in 16bit
: against 300% for 32 bit (SVGA).
[snip]
I can't guarantee that the game does 32 bit SVGA (since it has code for
varying weather, but we know that's not true). But the following text
is found in GP2.EXE:
Vesa Extensions using 32 bit linear buffiner
with 32 bit banking
--
**************************** Michael E. Carver *************************
Upside out, or inside down...False alarm the only game in town.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<[ /./. [- < ]>=-=-=-=-=-=-=-=-=-=-=-=-=-=
: [snip]
: : S3SPDUP shows great results when run against benchmarking software,
: : especially in 32 bit mode. I know GP2 runs with the DOS extender which
: : means 32 bit but is the graphic engine doing 16 or 32 bit access to the
: : card? This is critical as the performance improvment is 33% in 16bit
: : against 300% for 32 bit (SVGA).
: [snip]
:
: I can't guarantee that the game does 32 bit SVGA (since it has code for
: varying weather, but we know that's not true). But the following text
: is found in GP2.EXE:
: Vesa Extensions using 32 bit linear buffiner
: with 32 bit banking
GP2 documentation specifically notes that GP2 will set vesa mode 0x101h,
640x480x256. 0x101 is a banked, vesa 1.2 mode. The VESA 2.0 640x480x256
Linear Frame Buffer mode # is 0x4101.
Anyone care to hack the exec code to use the vid mode and also remove the
bank switching interrupt calls while you're at it? (it is function (ax)
0x4f05 of interrupt 0x10 (bx contains the bank #)) :) :) :) :) :)
-------
Carlos Ribas
President
MalSoft
http://www.nol.net/~draconis
