Yes, you cannot simply say, "oh, lets use less CPU power and utilize the GFX card more". While they do have a relation, you must imagine them as completely different entities. The CPU is doing the physics, while the GFX card is handling the viewable representation of what the CPU calculated. As the graphics effects are pretty simple, the GFX card has a relatively low workload. But to reduce CPU load, you'd have to dumb down the physics calculations, making them less accurate.

It depends on the game how much the CPU is responsible for graphics. You can play unreal tournament 2004 without having a dedicated 3D graphics card at all simply by switching over to a software renderer.

Quote from Electrik Kar :

It depends on the game how much the CPU is responsible for graphics. You can play unreal tournament 2004 without having a dedicated 3D graphics card at all simply by switching over to a software renderer.

But you can't load the GPU with Physics calculations now, can you?

Quote from Electrik Kar :

It depends on the game how much the CPU is responsible for graphics. You can play unreal tournament 2004 without having a dedicated 3D graphics card at all simply by switching over to a software renderer.

Yes, of course, but the CPU in LFS is already doing almost nothing for the graphics. The biggest hog is the physics, and besides a few optimizations you cannot do anything to make them take less CPU cycles. You can easily experience yourself what happens when the CPU has to do more graphics stuff, by going to the GFX options and turning off "HVS". Your framerates will not thank you for it, and your GFX card will be even more bored. Another proof is for example, that my framerates do not change one bit if I use no AA/AF or full AA/AF.

Quote :

But you can't load the GPU with Physics calculations now, can you?

From my understanding, no.

And that is again the explanation why you cannot simply shuffle load from the CPU to the GPU, ergo my first post about this issue.

Quote from Flycantbird :

All this talk did make me want to try rFactor for myself.
downloaded the multiplayer test demo, loaded it, hit start.
Nothing. Did the firewall test, no connection, just endless waiting.
Checked the 'test forum' and didn't see specific info.
I'm not using a router, am using zone alarm.

Any suggestions ?

Maybe you should download the new demo.

Quote from axus :

But you can't load the GPU with Physics calculations now, can you?

you can, ATI demonstrated using either a second or third graphics card to handle physics.

The question is, proper self programmable kind of physics (maybe even understanding CPU instructions?), or fudged aegia-kind of physics? Even then, buying a second graphics card to do physics... I dunno if that would be a wise investment. Maybe if you can use it in SLI mode for other games, then yes.

Quote from Electrik Kar :

From my understanding, no.

You can: You just have to write a shader that does physics calculation. Then you save all your input data into a texture, bind this texture and the shader, render into a buffer and read back the values from this buffer... This is called GPGPU (General Purpose programming on GPU) and becoming important in science, since we have GPU's with so much computation power.
There is even a framework from ATI to simplify the whole process and NVidia implemented parts of the Havok physics engine for their GPU's...

Quote from severin_schoepke :

You can: You just have to write a shader that does physics calculation. Then you save all your input data into a texture, bind this texture and the shader, render into a buffer and read back the values from this buffer... This is called GPGPU (General Purpose programming on GPU) and becoming important in science, since we have GPU's with so much computation power.
There is even a framework from ATI to simplify the whole process and NVidia implemented parts of the Havok physics engine for their GPU's...

But you can't do it practically in a racing simulation such as LFS. For one thing not everyone has good graphics cards and varying the amound that the GPU handles would probably be difficult for one thing. For another, it would take a substantial amount of time to program and we should be getting LFS upgraded to DX9 sometime in the not-too-distant future so once that is done, all the efforts will be wasted.

yeah... eric cant have been sitting around for all of this time. he has probably been working hard on stuff. it was confirmed that he is doing new interiors and track updates. i wonder if he is working on updating to dx9?

Quote from Gabkicks :

i wonder if he is working on updating to dx9?

Only Scawen can do that. Eric is "only" modelling the cars/tracks, and for that he's using the tools and editors written by Scawen.

Quote from AndroidXP :

The question is, proper self programmable kind of physics (maybe even understanding CPU instructions?), or fudged aegia-kind of physics? Even then, buying a second graphics card to do physics... I dunno if that would be a wise investment. Maybe if you can use it in SLI mode for other games, then yes.

crossfire for ATi, but im sure nvidia are working on it aswell. and it was havoc physics it was demonstrating, not sure about self programmable physics.

and yeah, in games that dont need the physics, you can use both cards in crossfire/SLi.

Quote from axus :

That "radial street tyre curve" looks like nothing I've seen.

The graph at sasco seemed to be aimed at making a point, not in being accurate. The main point is that biay ply racing slicks are a lot more forgiving, and the grip doesn't fall off. The graph from the smithees web site is based on real world data. In this case I was trying to make the point that you don't see a 30% fall off in grip due to slip angle (within reason). I'm guessing that the people who created the data with a 30% fall off based this on the that there is significant fall off in forwards traction if you spin the tires, but this is a different case than slip angle. Just recently, Hoosier released a new series of radial slick that are very forgiving. One thing I'd like to see in LFS is that it keep up with the latest technology in the tires actually being used on race tracks these days. For one thing, have you ever seen a Hoosier DOT racing tire (for classes that don't allow slicks), it only has two small groves in it (FIA F1 class doesn't allow this and spec's a tire with more groves). Go to this page, click on "specs" (left side menu), then scroll down and click on the "M" to see the tire profile: https://www.hoosiertire.com/rrtire.htm. These tires are popular in classes where the cars are restricted to DOT (street legal) tires, and even hot lappers (track day event drivers) use them as well.

Quote from JeffR :

The graph at sasco seemed to be aimed at making a point, not in being accurate. The main point is that biay ply racing slicks are a lot more forgiving, and the grip doesn't fall off. The graph from the smithees web site is based on real world data. In this case I was trying to make the point that you don't see a 30% fall off in grip due to slip angle (within reason). I'm guessing that the people who created the data with a 30% fall off based this on the that there is significant fall off in forwards traction if you spin the tires, but this is a different case than slip angle. Just recently, Hoosier released a new series of radial slick that are very forgiving. One thing I'd like to see in LFS is that it keep up with the latest technology in the tires actually being used on race tracks these days. For one thing, have you ever seen a Hoosier DOT racing tire (for classes that don't allow slicks), it only has two small groves in it (FIA F1 class doesn't allow this and spec's a tire with more groves). Go to this page, click on "specs" (left side menu), then scroll down and click on the "M" to see the tire profile: https://www.hoosiertire.com/rrtire.htm. These tires are popular in classes where the cars are restricted to DOT (street legal) tires, and even hot lappers (track day event drivers) use them as well.

Fair enough. I actually had a comparison picture posted somewhere on the forum... *digs*

http://www.lfsforum.net/showthread.php?p=181831#post181831

Aha!
There's a curve generated entirely by FTire there, which is a little edgy but the general shape of the curve should be noted. The edgyness is due to "spin" which is an effect from steering the tyre gently upto high slip angles. As the slip angle is increased, imagine what happens to tyre deformation and imagine it carrying through for a whole revolution and then affecting the side force. Despite my sicerest efforts, I have been unable to entirely eliminate this. Then there is a picture of my curve which I made using a piece of C code to match the FTire curve's profile as closely as possible. And then there is the GTR2 Pirelli curve which makes the word "catastrophe" come to mind. The rFactor curves are not much different from the GTR2 ones.

Quote from axus :

I actually had a comparison picture posted somewhere on the forum...

http://www.lfsforum.net/showthread.php?p=181831#post181831

There's a curve generated entirely by FTire there, which is a little edgy but the general shape of the curve should be noted. And then there is the GTR2 Pirelli curve which makes the word "catastrophe" come to mind.

From that curve the GTR2 tires aren't very forgiving. There's also a "alignment reaction" versus slip angle, which is what the driver feels at the steering wheel (in addition to caster effects), and with that curve, it might explain why the FFB goes to near zero on understeer, but I've always thought it was just a canned effect to act as a driver's aid.

Do you have any curve plots for the tires in LFS?

Quote from JeffR :

From that curve the GTR2 tires aren't very forgiving. There's also a "alignment reaction" versus slip angle, which is what the driver feels at the steering wheel (in addition to caster effects), and with that curve, it might explain why the FFB goes to near zero on understeer, but I've always thought it was just a canned effect to act as a driver's aid.

Do you have any curve plots for the tires in LFS?

No, I don't but after putting my tyre stuff in GTR2 it begins to feel very simmilar to LFS in handling... the longitudinal curve, I still can't get sorted (if you have some real life mesasurements there, from slicks specifically, they could be very helpful) - mind you, I haven't had time to mess with that stuff for about a month now. I might be able to get the LFS curves, relatively close anyway, once I have time to add some options to this little telemetry program (monkster's done all the code upto this point but he doesn't have the time to continue development). The idea would be to only add the data to the curve if some other variables are within a certain range. As in, only plot the lateral force values versus slip angle if slip ratio is between -2% and 2% or something like that.

Quote from JeffR :

Do you have any curve plots for the tires in LFS?

Here is an example for road tyres: http://www.lfsforum.net/attach ... tid=4791&d=1136410311 However, it was done before the last physics patch. It looks like lateral grip could drop by at least 15 % after the peak in that older version of LFS. Unfortunately I don't have more current data.

Quote from JeffR :

Do you have any curve plots for the tires in LFS?

somwhere on this forum android xp posted his measurements on lfs slip angle cuves and they look fairly similar to the ones generated by ftyre

edit:
a few of his measurements:

http://www.lfsforum.net/showth ... highlight=slip#post142033
http://www.lfsforum.net/showth ... highlight=slip#post138830
http://www.lfsforum.net/showth ... highlight=slip#post138762

I just must say this...
I downloaded rFactor demo and ran it.
How cute!!! Like a Garfield cartoon!!!

Maybe my settings was messed up, but it looked really cute!
Didnt have time to investigate it more... but my first impression...
SugarSweetCutePie!

/mike

Quote from axus :

But you can't load the GPU with Physics calculations now, can you?

Sure you can. GPU stuff is just shader language. You can fill it with matrix/vector calcs to your heart's content, stuff it into a texture, then read it back out from the exe. I was doing something like 40 billion calcs/sec in a shader when my P4 3.6Ghz HT spu could only do 1.3 billion. This was happening during a simple shader animation and there was 0 loss in frame rate. If they were 3Vectors it could have done about three times that many. There's some bandwidth overhead because you have to transfer the data back and forth between the exe memory and the GPU, but yes, it's absolutely possible to do this kind of thing.

Edit - After playing the demo, I decided I'd get the full game and try the mods, as someone suggested.

Driving Laguna Seca in a McLaren F1 is very addictive.
The karts and sprint cars look like they'd be fun too.

The one thing rFactor has over LFS is the mod aspect, imho. But they are apples and oranges, for the most part.

Quote from jtw62074 :

Sure you can. GPU stuff is just shader language. You can fill it with matrix/vector calcs to your heart's content, stuff it into a texture, then read it back out from the exe. I was doing something like 40 billion calcs/sec in a shader when my P4 3.6Ghz HT spu could only do 1.3 billion. This was happening during a simple shader animation and there was 0 loss in frame rate. If they were 3Vectors it could have done about three times that many. There's some bandwidth overhead because you have to transfer the data back and forth between the exe memory and the GPU, but yes, it's absolutely possible to do this kind of thing.

Impressive stuff that some hardware can do nowadays then.

http://www.sendspace.com/file/307ome

here is video of me running a 1:26 (in 2006 BMW) on the new monza released by ISI. i wasnt running full boost and i am still very green with the car. My setup needs alot of work as well... i was running too low susension or something... and i wasnt running enough boost or using the temp boost. the car's capable of 1:21's with extremely good drivers. IRL the car's done a 1:22 sofar in practice.