Well, I did some heavy testing over the weekend (took .75 seconds of my Monaco
time in the process!) and have incorporated my findings into the FAQ, to be
posted as soon as I've HTMLized it for the WWW. Here is the section:
---- snip, snip ----
1.2 What sort of performance can I expect on my PC?
---------------------------------------------------
Here is a rough table of machine against performance:
Machine Memory Detail Process Speed
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
P90 8mb 4d T 35% 25fps (DG's tower of power with PCI K64)
486dx266 8mb 4d T 70% 25fps
486dx266 4mb 4d T 60% 25fps
486dx266 16mb 4d NT 35% 25fps (Nigel Bovey's)
486dx266 4d T 70% 25fps (MBP's under OS/2)
486dx250 4d T 95% 25fps (Paul Smyth's with ISA ET4000-W32)
486dx33 4d T 90% 25fps (Graham A's)
486sx33 4mb 4d NT 66% 25fps
486sx25 4mb 4d NT 100% 25fps (a DELL)
486sx25 4d NT 80% 25fps (Nightshade's oldie)
486sx25 4mb 4d NT 100% 23fps (Ben Lester's)
486sx25 2mb 4d NT 100% 21fps
386dx40 4mb 4d NT 100% 20fps
386dx40 2mb 4d NT 100% 20fps (possible optimistic)
386dx33 8mb 4d NT 100% 17fps (DG's old faithful with T8900CL)
386sx25 2mb 4d NT 100% 14fps (Stingray's)
(It appears that as long as you have at least 2mb of RAM, the actual amount
makes absolutely no difference. The difference between the two DX2/66s above
is attributable to graphics card alone.)
The details level is shown by the amount of detail around the track, 1d
being the lowest level and 4d the highest, the other detail option is the
track shading, this is shown by T (track shading on), NT (no track shading).
The process column show the average processor occupancy as you go around ANY
track. This is just a rough estimate, but really shouldn't go above 100%
very much. The final column show the speed in Frames Per Second that this
set-up allows.
Even on similar machines, several things will affect speed. A machine with
some external cache will outperform one without; the actual amount of cache
is probably not going to make much difference. Graphics card performance
also makes a big difference; a local bus card will run much faster that an
ISA card, and some cards have better DOS performance than others (Cirrus
Logic based cards are good, ET4000 and derivatives are even better).
The general consensous seem to be that people would rather have it running
smoother, but with less detail, this shows one of the main advantages of
F1GP over Indycar, in that it runs quickly on a slow machine and smooth
graphics are possible quite easily.
It appears that occupancy has an impact on lap (and race) times. A machine
running at a higher occupancy seems to produce lower lap times than one
running with low occupancy. [DG: My P90 is about 35% occupant and laps I've
timed with a stopwatch are 3-4 seconds longer as measured by the game than
on the watch. I'm awaiting more figures.]
The Amiga version runs at a similar speed regardless of the machine's
capacity, about 3-5fps, depending on circuit, even in the fastest 68060
system.
** Does the performance vary on an ST? Mail me if it does **
1.2.1 So how does this affect lap times?
----------------------------------------
Short answer: it doesn't.
Long answer: it doesn't... directly. DG is in the fortunate position of
having both a P90 and a 386DX/33 on his desk (well, okay, the 386 is under
the desk...), and loaded identical copies of the game up on both machines.
The first and most obvious difference was that the game does not do a good
job of matching "real time" (measured on a stopwatch during laps on
qualifying tyres at Monaco). The first tests were done on the 386. With 100%
to 130% occupancy, the game's timer runs slow, being about three seconds
behind reality. With all the detail turned off and the occupancy down to
about 70% to 110%, it was about three seconds ahead of reality. With the
frame rate reduced and occupancy between 45% and 75%, it was about 4 seconds
behind. Then testing moved to the P90. With maximum detail and about 33% to
44% occupancy, the timer was about 4 seconds fast.
Now, here's the crunch. Despite these differences, the lap times reported by
the game were very close, all in the 1:14.2 to 1:14.6 range. The game was
noticably easier to play at higher frame rates and lower occupancies,
because things move quicker and smoother and make reacting easier. However,
with very high occupancies (more than 200%, such as on the 386 with texture
turned on), the difference from real time becomes very noticable; the whole
game runs in slow motion, and is potentially easier to play as you get much,
much longer to react.
---- snip, snip ----
Chisa
