Quote from Linsen :

@Davo: Wouldn't the physics to be calculated be the same no matter what gfx-settings you use? It's always 20 cars, I thought. Or does LFS change physics calculation with lower (ingame) gfx-settings?

I don't think the physics calculations change apart from having to simulate less cars in low-res physics if the multiplayer draw distance is very low.

Higher graphics detail means more, and more detailed, objects drawn. That means the CPU needs to do more graphics state changes (switch materials, vertex buffers, blending and so on) and also issue more draw calls to the graphics card. This all happens in the graphics driver where it is queued up until the graphics card is ready to handle it. That, of course, increases CPU load (quite significantly).

EDIT: Heh, just noticed this thread was brought back from the dead...

Quote from wien :

I don't think the physics calculations change apart from having to simulate less cars in low-res physics if the multiplayer draw distance is very low.

Higher graphics detail means more, and more detailed, objects drawn. That means the CPU needs to do more graphics state changes (switch materials, vertex buffers, blending and so on) and also issue more draw calls to the graphics card. This all happens in the graphics driver where it is queued up until the graphics card is ready to handle it. That, of course, increases CPU load (quite significantly).

EDIT: Heh, just noticed this thread was brought back from the dead...

vertex buffering is done via GPU in LFS a long time but it seems the 3D engine of LFS has very much calls to API which slows rendering.Something uneffectiove in the 3D LFS engine seems to be....as it shows less details is less CPU internsive but Looking at temp of my GPU comparing to other intensive games shows that it must be very unused in LFS. Thats really odd.As I said with more object/details there is exccesive usuage of calls to API.

Quote from DEVIL 007 :

vertex buffering is done via GPU in LFS a long time

Vertex buffers are stored in the graphics memory, but the CPU still has to switch between these stored buffer depending on what it's drawing. That's not free either.

Quote from DEVIL 007 :

but it seems the 3D engine of LFS has very much calls to API which slows rendering.Something uneffectiove in the 3D LFS engine seems to be....as it shows less details is less CPU internsive but Looking at temp of my GPU comparing to other intensive games shows that it must be very unused in LFS. Thats really odd.As I said with more object/details there is exccesive usuage of calls to API.

It's not odd at all. LFS' relatively low detail graphics means that for each draw call it issues, the graphics card has to draw only a small amount of triangles. This means that newer cards with extremely high triangle throughput only get to spend a short amount of time actually drawing stuff at max performance before they need to empty the pipeline and change state to draw something else. Add to that the relatively high cost of state changes and draw calls in Direct3D (This has been improved a lot in D3D10), and you soon reach a point where the CPU just can't issue enough calls to keep the GPU busy. You're CPU bound.

The only way to "fix" this GPU idling, is for LFS to issue more work per draw call (higher polycount), or batch triangles together according to state as to minimize need to switch beween them. I'm absolutely certain Scawen has batched what can be batched already though.

well I have some doubts that there is no way to blame LFS 3D engine but maybe thats just me.I think Scawen focus more on the other staff and the 3D engine need definatley revision/upgrade at least to DX9. Instancing could help with less calls,maybe there is also too much overdraw which raise calls as wel...and maybe there is more things.

I understand when there is more object there is also more calls but sometimes the drop doesnt prove more calls.I could find some examples...

Quote from DEVIL 007 :

well I have some doubts that there is no way to blame LFS 3D engine but maybe thats just me.

Oh, I'm sure there's room for improvement, but I'm not sure how big of an improvement or whether it would be worth the extra effort to change it around before going to S3 (...and hopefully a new API. OpenGL "Long Peaks" would make me a happy man ).

Quote from DEVIL 007 :

I think Scawen focus more on the other staff and the 3D engine need definatley revision/upgrade at least to DX9. Instancing could help with less calls,maybe there is also too much overdraw which raise calls as wel...and maybe there is more things.

Instancing would most certainly help (especially on large grids), but is not available on D3D8 if I'm not mistaken (EDIT: I see what you meant now, switch to D3D9 and then utilize instancing ). There is probably room for improvement in the visibility calculations and LOD system too, but as I said I'm not convinced it's worth it at this time.

around 20-25 fps at full grid
1280*1024*32
Intel P4 @ 2.8Ghz
1GB ram
Nvidia Geforce FX 5200
no AA or AF

These comparisons are of no use really unless you state every single little adjustment to the settings. Mirror LOD can make a significant difference, as can the other sliders, but most importantly, in-car is a significant hit for the fps compared with a "wheels" view.
I have an Athlon3000 something or other, and a something or other 6600gt128mb and 512mb of ram. I have a few things turned down a bit, and use in-car view at 76degrees fov. No AA or AF and 1024xwhatever it is. It dips into the high thirties at the back of a 25car plus grid, but if I shift-F I get a couple back without the hud showing. So I'm fairly happy.

Quote from sinbad :

These comparisons are of no use really unless you state every single little adjustment to the settings.

I've said the same thing in many threads over the years about frame rates. You have to compare using the same lfs.cfg file (I think that is what it is).

The unofficial LFS benchmark did this with a min and max config file and a standard single player replay. Is the benchmark site still around and updated for recent patches?

Yeah, I get fantastic framerates. I run 60+ frames online and only drop to 25 at the start. And that's with an onboard gfx chip with shared memory. See, that means nothing for comparison unless you have my config file to compare my and your framerates.

even with a full grid my pc usualy stays at 50fps, Athlon 64 4000+ San Diego

1600 MHz sempron
nVidia GeForce 7600 GS
512 Mb ram
1280x1024, 32 bit
at full grid 10 - 15 FPS
if im alone on track 60 - 80 FPS

Tbird 1.4 ghz + radeon 9600 pro 256 mo + 512 mo pc 2100

0.3 FPS

2.2 ghz amd athlon 64 3700+
nvidiea 9300 gs - 512 mb
1 gb memory

around 25 fps with 11 ai drivers at grid and 10 windows open, 26 fps with 11 ai drivers at grid and 1 window open. stays that as long as i'm near them.

1024 x 1280 no aa and no af.;

*playable*

This thread is just people boasting about their specs...


Anyway..

AMD Athlon X2 4400+
512MB Radeon X1950XTX
2GB TwinX DDR400 RAM


Pre patch X I never saw any kind of slowdown at all with 4x AA and 16x AF and all the graphics sliders on full. Now we have 32-car grids I'm starting to see a bit of stutter on the start line if I'm near the back (or near the front looking backwards), but that's it. Once everyone spreads out a bit its back to silky-smooth.

AMD XP 3200+
1GB DDR1 RAM
256MB Radeon 9600
1024x768
EK High res textures
1x AA

28 FPS - 32 cars on grid

My computer is an old model Dell so not that good

Intel Pentium 4 2.4GHz
128MB RDRAM
64MB Nvidia GeForce 4 MX420
1024x768
Low-res textures

8FPS on a 13car full grid

AMD Athlon 64 3200+ (2,21GHz)
ATI Radeon X1300
1GB of RAM
4xAA, 4xAF
High-res textures, skies, etc

60 fps at minimum, normally around 100.

Intel E6600 (@3.6 GHz) and 8800GTS 640 (@600/1000)
Screen resolution 1680x1050 with 16xAF and 8xAA.
Alone on the track +/- 200 FPS and with full grid never under 100 FPS.

Quote from mrodgers :

I've said the same thing in many threads over the years about frame rates. You have to compare using the same lfs.cfg file (I think that is what it is).

The unofficial LFS benchmark did this with a min and max config file and a standard single player replay. Is the benchmark site still around and updated for recent patches?

Yeah, I get fantastic framerates. I run 60+ frames online and only drop to 25 at the start. And that's with an onboard gfx chip with shared memory. See, that means nothing for comparison unless you have my config file to compare my and your framerates.

but do you drop 25 fps with full grid and whats your res if its 400 x 500 or some super low then maybe then, thats true but otherwise i dont buy it.
your not useing a buit in gfx card and not droping fps at the start just like every one else here. there is no onboard gfx that isnt going to drop if a 300-500 agp/pci-e card drops you . one of two things you dont know what you have or your got every thing turned off and the lowest setings

Quote from Zyber :

Just trying to make a point here that maybe it isnt our hardware that is the problem to lfs running at max 40 fps on start with a full grid..

Would be nice if ppl could post hardware and fps at start with a full grid.. just to check if im right about this.. it looks like 40 is max nomatter how fast or slow pc u got


GFX: 7600 GT 256 mb
CPU: AMD 3200+
MEM: 3 GB

FPS: 30-40

Pardon me, I have an old 800MHGZ duron...gigabyte motherboard...256 RAM.

I have held WR's and can get off the grid just fine....I just move aside and pick off drivers one by one..BUT,,if i have pole...only thing you'll see is my assend.

Biggie, Fotch and my team mate Hallen are the only comp for me in the XRR.

Don't break the bank thinking money buys WR's "Grandsurf May" has the best system and set up I've ever seen....he's a great racer..but I can make him blink for $7,000 dollars less.

That's just MHO

3200+ 2.2GHZ
512 ram
7300GT
27-38 fps

+ im overclo... mine procesor.