That's because in rFactor, the AI's are not actually cars with the same physics as your car. They're not actually driving the car like the AI's do in LFS, they're merely points with specific behavior guidelines that look like cars.
This is part of the reason why the AI in LFS is so slow.
Going slightly off topic I didn't wanna start a thread about it, when we place cones on track in the editor there a number of different ones colours/pointers etc. ,
do the cones have meaning and would the AI follow them ?
I think if you put an object in the middle of the racing line, they will continue to crash into it until they learn and start to drive around it. I have not tested this myself, though. But no, I don't think they won't follow directions given to them by cones.
Using AEGIA means using their physics engine, not a custom one like LFS's. Also, their physics engine is very simple compared to LFS's. AEGIA's engine is designed for basic physical interactions (i.e. balls bouncing), not the kind of complex interactions that take place in racing.
I have also heard that AEGIA's engine is not entirely accurate in it's representation of physics due to many optimizations to make it run faster.
AGEIA will not help. The hardware has been slated, there are no games and multi core CPUs and physics on multi GPU will kill this one off. 3 years ago it might have stood a chance but it came to market far far far to late IMHO.
i have tried lfs on so many pc combinations that i clearly concluded that even if lfs uses more cpu than other games, the graphic card is still the limiting factor, i believe that even if lfs graphics is very "simple" (low poly, no effects, etc.), it has a lot to be optimized.
I disagree, Honey. At one point, with everything else remaining the same, I upgraded from a GF 4600 128MB to a GF 6800GT 128MB and saw no improvement in performance at all with AA and AF off. With it on, the GF 4600 struggled a bit, but the 6800GT could run 4x AA and 16x AF all day long without any performance hit (except maybe on static scenes with no other cars on the track where the CPU isn't being used as much).
The situation is similar on my 7900GT 256MB. On a static scene with no other cars on the track and AA and AF off, I get about 190 fps. With AA and AF on, I get 160 fps. However, the minimum framerate always stays the same.
In fact, when I overclocked my CPU (Conroe) from 2.13 GHz to 2.67 GHz, a 25% overclock, I saw a 24.4% increase in minimum framerate. The scenes that produce a minimum framerate are those with lots of cars on the screen all very close to each other, such as on T1. Minimum framerate is all that really matters, and it's clearly dependent, if not entirely then nearly so, on CPU power.
i had opposite experiences, specifically:
- with two exact systems (same nvidia 6600) but different cpus (one athlon 64 2500, the other a64 4400 x2) i had almost same performance
- with another system with pentium 4 @ 2.8 GHz with crappy unbranded card i had very poor fps (from 20 to maximum 30) as the same with an old athlon 1.6 with geforce 5200 (varying from 30 to 40)
in those cases i used lfs to witness a speed boost with the better cpu...but i did not, so i guessed that maybe cpu is not so relevant.
of course i may be wrong, because i did not scientifically and except the first case, many other factors could intervene, but that was my experience.
maybe the truth is in the middle and for lfs cpu and gpu are equally important...
I assume you meant the Athlon 64 3500+? This particular CPU is clocked at 2.2 GHz. The Athlon 64 X2 4400+ is pretty much identical to the 3500+, except there are 2 of them in a single package. Both are clocked at 2.2 GHz. Since LFS is not multithreaded, it should not surprise you that two otherwise identical CPU's perform the same.
As for the second case, yes, LFS does require a "reasonably" recent card, such as one made within the past 4 or 5 years that supports DX8. For example, a GF3, GF4, or Radeon 8xxx. These cards are ancient by today's standards. In most cases, integrated video is just complete crap. Your Intel system probably had Intel integrated video, which is known to be absolutely horrible for games.
i meant athlon 64 2500...but my memory was incorrect: i checked the bills and it was an athlon 64 3000.
btw as i said i was not making scientific comparisons, but i just want to remark that the difference from aa a64 3000 and an a64 4400 x2 is very noticeable for almost anything, even not-x2-optimized at the time i bought the two systems the 4400 had the higest-end available core architecture, while the 3000 the lowest.
my feeling is that any recent cpu is more than enough for lfs, whilst even a geforce 6600 is still limiting, but as i said i may be plain wrong.
what is clear for me is that devs can optimize lfs still a lot both graphics and phisics and this even only by a common sense consideration: scawen cannot be a geek of everything (math computing, net developer, graphic developer, etc.) especially for math computing there are plenty of numerical algorithms and theorems that can be used to speed up calculations by thousands or million times, i personally experimented that at university time when i was developing some projects...i still remember how was surprised when (in matlab) i reimplemented the computing of LPC coefficients with the levinson-durbin algorithm instead of the "classic" formula.
well my point is: lfs is made by a small team and this mean a much faster and optimized developing time (or at least usually this is the rule), but paid with specialization...honestly i feel that asking scawen to optimize more the phisic engine, rather than graphics is asking something too far from his core knowledge and thus will require a lot of study (time) by him and with the cpu going faster and being quite cheap maybe is less urgent than graphics, which for S3 must be almost rewritten to be ported to dx9 and add effects to simulate wheather...if graphics will not well optimized, then lfs will be a performance killer! btw phisics for S3 don't need to be updated with more effects to compute, so...
The Athlon 64 3000+ is clocked at 1.8 GHz, and as already mentioned, the X2 4400+ is clocked at 2.2 GHz. This is a 22.2% increase in clockspeed. So, if you were seeing a minimum framerate of 20 fps with the 3000+, you would then see a minimum framerate of about 24 fps with the X2 4400+. Some might say this is "pretty much identical" but that's beside the fact that it is a full 22% increase in performance, thereby scaling perfectly with clockspeed.
I suspect that you may have been expecting a very large increase in performance with the X2 4400+ but the truth is the two chips really aren't much different. You're never going to get anything more than a 22% increase in performance over the 3000+, except for multithreaded apps.
ais probably will only become even worse as far as cpu load goes
but lfs could really do with some mass start optimisations ... from what ive gathered the engine does complete physics calculations for all remote cars which is overkill
