F15 Frame Rate Issues
                                and
                            Hop-Up Tips

There have been concerns raised about the frame rate of F15 Strike Eagle III
and we certainly agree that it would be nice to have a better frame rate.
Unfortunately, the extra detail and spacial awareness offered by the new
textured-style of graphics has a computational price when compared to 
traditional shaded polygon methods.  In designing F15 III, we decided that
this extra detail was worth the cost, but we felt that we were honest with
consumers by recommending a 386/33 or 486/25 machine.  However, we were 
sympathetic to those with slower machines, and spent a good deal of time 
adding detail-reducing options that would make the game run faster albeit 
with somewhat less-appealing graphics.

In some cases, we have heard claims from players of extremely low rates even
on seemingly high-powered machines.  This is baffling, and we have a few 
hints to offer.  First of all, see the page 13 of the Technical Supplement
that is included with the game, for a discussion of how different game
features affect the frame rate.  Remember also that outside views can be
re-sized with the mouse (F15 III manual page 52) to make them run faster.

All machines of a certain speed are not created equal!  Other factors have
a big effect on performance.  One of these is the number of wait states
for memory accesses.  Many machines are set up for one or two wait states
for each memory access.  This is like having a V8 engine in your car, with
two of the spark plugs pulled!  Many times this can be remedied by 
reconfiguring through a program at boot-time.  Alternatively, it may 
require re-jumpering the motherboard.  This only works if the memory is
fast enough to handle it.  All of the machines MicroProse buys are spec'ed
out for zero wait state operation, but many come configured for one or two.
Restting them through the bios's setup program gives a noticeable performance
boost.

Another issue on the motherboard is the actual bus speed.  Many times you can
run at higher rate than the standard 8 mhz.  Most of ours are running around
10-12 mhz.  This is configurable in the same ways as the memory wait states.

Most machines today have some kind of CPU cache.  This is not to be confused
with a disk cache, though.  Since F15 III uses EMS, it is important to verify
that the cache is configured to include more than just the first 640K of
memory in its cacheable area.

Also important is the type of video card you have.  All VGA cards are not
created equal!  By now most people know about the performance difference
between an 8 and 16-bit video card.  If you don't, suffice it to say that
you don't want an 8-bit card, as they are much slower than the 16-bit cards
for doing intense graphics.  Also, make sure that your card is properly
configured for 16-bit mode.  There are many things that make these cards
revert back to 8-bit mode.  One of these is having a seperate monochrome 
video card installed in your machine.  Other cards may cause the same
interference.

The other thing you want in a video card is a recently designed one.  
Anything based on the Tseng 4000 chip set is good.  We use Orchid ProDesigner
IIs cards and Diamond SpeedStar.  There are lots more, too.  This makes a
big difference in graphics performance.

Another issue is local bus video.  This looked to be the great solution to
all of the video card performance bottlenecks.  Unfortunately, every machine
I have yet seen has been a disappointment.  The video section is indeed 
faster than a comparable regular machine, but the regular memory accesses are
now slower!  Overall performance for flight sims is about the same as a 
regular machine.  I suspect that this will be remedied with 2nd and 3rd 
generation local bus designs.

So, you ask, how can I tell if my system has some of these handicaps?  
Basically you need some kind of performance benchmark that simulates the
kinds of things that go on in a flight sim.  The 3DBench program is a good
one for overall graphics performance.  For reference we have seen the 
following 3DBench numbers for some of our machines: 

386/25           7.5
386/33          14
386/40          15.6
486/25          17.5
486/33          23
486/50 DX2      26.3

But we also have at least one 486/33 that only gets 11.1 because of its
configuration (video card and memory wait states and bus speed).

An even better program for checking out specifics is VidSpeed.  It will 
help determine whether the problem is in the bus speed, the regular memory 
wait states, or the video card.  Run it with the "*" and "L" options to
get memory transfer rates.  If the video rate is much slower than the
regular memory, you need to pay some attention to the video card.  If both
are slow, then the bus speed might need some attention.  In all cases, it 
provides a very good metric to use in evaluating changes you make in the 
configuration of your machine.

Beware of making changes to your machine's setup if you are not fairly adept
at these things, however, because it is very easy to mess up.  If you do
understand these things (or have a friend who does) it can be quite
beneficial to experiment a little.


Andy Hollis
Producer, Simulations Group