I would also like to mention that FSBench site is still alive and well; however I do have a problem to connect to the database. I hope that is cleared up soon so we can start the FS2002 benchmarks.
I will keep you posted.

Back in 1999 the main conclusions were (1) there is a one-to-one relation to between the speed of FS 
(in fps) and the speed of the cpu and (2) nor cpu-type nor type of graphic card does matter much with respect to the speed of FS, provided you have at least a Voodoo or TNT(2). These results were also corroborated by Marcin Slawicz’s analysis.
Marcin took a different approach, which resulted in the conclusion that for an increase of cpu-speed by 100 MHz the fps were increased by about 3.5.
Again: cpu-type or graphic card type did not really make a difference.

This time I took an approach somewhere between my earlier one and that of Marcin’s, in order to try to forge it all together. For this I took the latest 120 measurements, roughly the ones that were added after my previous analysis. The first conclusion was readily made after just a glance at the data.
With respect to FS2000 the graphic card still doesn’t really matter when it comes to FS speed.
Not even the latest GeForce 3 cards can do anything about that. I’ll come back to this later.

For the analysis once again I only looked at the benchmarks for FS2000, for a resolution of 1064x768,
full screen or window doesn’t matter. Disregarding the graphic card the benchmarks are clustered in
groups with respect to cpu-type and speed. At the bottom selected results from the earlier benchmarks are added. The clustering results in the following table:

The rows in the table probably speak for themselves, but to be sure, I’ll explain them anyway.
The first column shows the processors (speed and type), first the AMD processors then the Intel ones,
ordered according to speed. In the previous analyses of November 1999 the 600 MHz processors were top of the bill, now you can’t even buy them any more! The next column is the number of benchmarks of this type.
The next 2 columns give the spread and average results of the benchmarks in fps,
the last 2 columns the relative results, i.e. the results, divided by the speed of the processor.

There is a considerable spread in the results. It is not possible to relate this to the different graphic cards.
So the obvious reason is that some systems are much better tuned to FS than others.
Or: some people have a "more optimistic view" of the fps they measure in the benchmark than others?
To be honest, I also like to see FS performing better than it does on my P III-700.
Anyway, I think this middles out over the whole lot of benchmarks.

The fourth column show the strong relation between the speed of Flight Simulator (in fps) and the speed of the processor, still showing the one-to-one relation we found in the previous analysis of 2 years ago.
Also there is not much difference between the AMD Athlon and Intel P III processors as far as speed in Flight Simulator. As AMD processors of the same speed are cheaper, you seem to be better of with an AMD processor.

A dramatic thing however happens with the new Intel P IV series.
Of course the number of entries is quite small, but the trend is there.
With respect to Flight Simulator their performance is very bad.
However this is old news, since the same is shown in other benchmarks.
That there is such a small number of benchmarks with the Intel P IV is a reflection of the fact
that flight simulator fans already know and already have chosen for the AMD Athlon.

Looking at the last column, you can clearly see the one-to-one relation.
The relative speed for any processor is about 35 fps per 1 GHz, give or take a few. 
Except for the new Intel P IV series.

If you add the results of the previous analysis, it shows that this strong relation has been so from the 266 MHz Pentium II to the latest 1.4 GHz Athlon. So one other conclusion also still stands: even with the fastest processors of these days Flight Simulator 2000 seems to be compute-bound.

And yes: the choice of the graphic card does of course influence the benchmark result,
but compared to the influence of the processor speed this influence is largely neglectable.
This too is not a new conclusion, the reason probably being that the relatively old-generation
FS2000 does not take any advantage of the features of the (new generation) GeForce series cards.

I guess this concludes the era of FS2000. We’ll all have to move on to the next generation: FS2002.
From what I have seen from the screenshots and read in the reports we are in for a treat.
But will our current systems hold, will we need a 1.4 GHz Athlon?
Will a GeForce 3 card really make a difference and in what respect, speed or graphic quality?

Let’s hope that Pierre Poirier manages to get the new benchmarks for FS2002 ready soon,
so that we can again see if we can hold on to our current systems for a while or that we
have to run to the store to improve upon it, but a new graphic card or whatever.

note: results are in fps; 1600-32 means 1600-1200 32 bit.

note: AA2 means antialiasing x 2

Conclusions:  For FS2000, it is obvious that the 6.50 give the best overall results but they are very close altogether.
For FS2000 and antialiasing, the results are somewhat different. The 12xx series drivers show a significant increase over the 6.50 and 5.33 drivers.

For Fly ll, the 12xx drivers are in a different league: 6 to 14 fps increase... That's huge and the biggest difference is in the antialiasing section.

For X-Plane2.66 and Open GL, the 12.40 lead by a smal margin over the 12.42.

If you use antialiasing and Fly ll at all you know what to do: use the 12xx series drivers. If you use Fs2000 only without antialiasing, the 6.50 will do very well.
Personally, I will use the 12.42 because they give the best overall performance in all the Simulators here.
I understand I may not have answered all the questions you may have, but this report will be updated regularly with new drivers as they appear so you can keep track of their performance loss or gain.

These results may vary with system configurations.

I will soon have an article on the new video drivers and some tests and benchmarks to go with it.
This is mainly due to the fact that Microsoft is updating the DirectX versions, programmers update their software and manufacturers of video card also update their drivers. 
The end result for the user: reduced performance! Indeed, stay tuned...

Basics: There are many types of clouds in FS2K: Cumulus, Stratus, Cirrus 
and Thunderstorms. Each type can cover 1 to 8/8th of the sky coverage
Cirrus clouds are not demanding in as far as processing power is concerned 
but the other types are.

Also a Cumulus or Stratus coverage of 1 or 2/8th isn't very demanding either.
From 3/8 to 7/8 that is another story. The overcast condition is also not
very demanding. It is as if the overcast is treated as one big cloud with 
texture on top and below and whiteout in between.

Another criteria which adds up to the equation is the vertical size of
the cloud layer: the greater the vertical size the more demanding it becomes.

Having said that, one can see that we should aim for the 3/8 to 7/8
cloud coverage. The 3/8 cloud coverage is unreliable: the fps varies too
much and that's because the cloud fragments are relocated everytime
you start up a new cloud situation.

How the test was done: 
location: middle of Lake Michigan, cockpit view with no cockpit bmp at 3855ft alt;
clouds: tests were made with ST (stratus) and CU (cumulus);
everytime a new cloud test was made, I always 
switched back to blue sky to verify fps (68fps at all times);
vertical size: cloud tops at 8000 and base at 1500;
ref system: P3 800 160meg ram, Geforce2mx agp4x, Abit se6, 17gb ata66
The files used in these tests are here:xxx.flt and xxx.wx. You should put them in 
your "Pilots" folder.

Observations (partI): There is a very subtle difference between 8 and 16 bit:
.426 for Stratus and .296 for Cumulus. However, I think I should increase
the number of tests to confirm these readings and that's coming up in part II
later this week.
P Poirier

Here is another graph showing FPS increase for FS98 reports since may 1998.

What does this say ? Basically software is always ahead of hardware: software (typically flight simulator software) is very demanding for the actual CPU and GPU power when it is produced. However, given time (1 or 2 years) CPU and GPU power become available which allow for better performance and quality. And then software producers sit down and develop a "better" product which will be sold because the product has more features, enhancements that the "user" requested and....

It is a never ending cycle.

Last year analysis (dec 99) of CPU and GPU power was done by Marcin Slawicz and Jos Grupping. Here are a few of the remarkable results they had at the time:

1. "Every 100 MHz of your CPU can pump about 3.5 frames more. Remember – the numbers regard the particular FS2000 setup"

This has proven to be true and has been demonstrated with this year report.

2. "What FPS should you expect with the 450 MHz CPU?
TnT or TnT2 15.1
Voodoo3      15.7
Banshee        16.1
It would be about 35 FPS with future 1000 MHz CPU if nothing else will change."

This again is demonstrated with the 1000 mhz cpu.

So, we should very likely get in the 50 fps mark for 2001. Fluidity will remain an issue. The architecture of the PC may be the limiting factor here (hardware vs software). We'll see. 

In conclusion, I'd like to thank everyone who participated in these reports.

Pierre

25Dec00 The Report Form has been updated to reflect the new processor speeds, DirectX and video chipsets. Thanks to Orlando for letting me know.

21Dec00 FSBench is no longer dead. It will gradually be put back in service. For the time being the report system is running and you can update the database with your benchmarks.

I should present myself: my name is Pierre Poirier, I live in the beautiful province of Quebec and enjoy Flight Simming as a hobby. I did give a hand to Bruce in the past with the Results page. 

I felt the site should be revived because we need benchmarks so we can improve on our systems and use the same references when we compare hardware and software add-ons.

However, this is not worth much if you don't use it. So please go ahead, run your benchmarks.

I will make an official announcement later on when I have everything in order at the site.

Thank you