Off Topic : LFS running slower on faster computer

[DEVIL 007]

Hi,

I already posted that at the end of Patch Z13 thread.I really see some strange behaviour with my new system on LFS. Its around 10 percent in FPS drop. I changed A64 3200+ to Phenom II X3 720. Its actually a change from single core 2Ghz processor to triple core 2.8 Ghz and the new architecutre is even much faster per clock and using just single core.

I got huge boost in every other game but it somehow struggle. Tried 3 fresh installs(actually six because swithing the old mother board with odl CPU and memory and then the new mother board + CPU + memory) and tested various games including LFS and always the same. LFS went down other games so much higher. Not sure what it might be :-(.

Sorry if this wrong thread but I can not ping point to anything then just LFS.

Scawen,
Please could be possible with some test patch give back 0ms Minimum sleep.Just for testing because it might be really interesting. I got idea (might be totally worng ofcourse from my side)that 1ms and anything above significantly holding the CPU and hence possibility of the strange behaviour. Just an idea. It might be something else in LFS ofcourse. Maybe something eveb different but I can not narrow it anywhere else.

[three_jump]

Hi,

I already posted that at the end of Patch Z13 thread.I really see some strange behaviour with my new system on LFS. Its around 10 percent in FPS drop. I changed A64 3200+ to Phenom II X3 720. Its actually a change from single core 2Ghz processor to triple core 2.8 Ghz and the new architecutre is even much faster per clock and using just single core.

I got huge boost in every other game but it somehow struggle. Tried 3 fresh installs and tested various games including LFS and always the same. LFS went down other games so much higher. Not sure what it might be :-(.

Sorry if this wrong thread but I can not ping point to anything then just LFS.

have you tried running without the speed stepping (cool n quiet)? or is there an option to make it game dependent? (i'm not using it, so i can just guess)

[rc10racer]

Did you even try the thing i suggested?

http://www.lfsforum.net/showthread.php?p=1208786#post1208786

[DEVIL 007]

Well I run with CnQ disabled because then my CPU stay at 800Mhz all the time and I get like 30-50 percent drops on LFS. It sucks.

The issue I mentioned above when my CPU it at full speed(no CnQ enabled). The 10 percent drop doesnt make sense to me even if the both CPU would be same fast clock per clock which is not the case. The new CPU is faster per clock and has even +800Mhz more.

Really odd.

[Scawen]

Well I run with CnQ disabled because then my CPU stay at 800Mhz all the time and I get like 30-50 percent drops on LFS. It sucks.

The issue I mentioned above when my CPU it at full speed(no CnQ enabled). The 10 percent drop doesnt make sense to me even if the both CPU would be same fast clock per clock which is not the case. The new CPU is faster per clock and has even +800Mhz more.

Really odd.How about if you have a lot of cars on screen and LFS is paused (so using very little CPU). If you can compare the same scene on the old and new computers (if the old one still works). And what graphics cards are being used in the old and new computers - I think you have only mentioned the CPU.

Also, you say an FPS drop, but what sort of frame rates are you getting?

[DEVIL 007]

Hi Scawen,

thanks for coming back so late.You should get some sleep:-)
Back to topic.I did a lot of tests between these systems several times. I dont have actually 2 standalone systems. I just change mobo+CPU+RAM and always do format and clean install. The graphic card remain the same and its ATI 4870 512MB ram.

The odd is that online I get more FPS but offline I get less when driving alone on Blackwood with XFG( view on wheels).Same on other tracks. In some places same in some I get with better CPU less. As I said the difference is around 10 percent worst maybe a little bit more on the new system and it go really down more then 10 percent(went from around 200 to almost 140FPS) at the place in attached picture. Same behaviour around the whole track. Some places seems same and some worst. Odd and dont know what it could be as the installas of XP are always fresh so it seems not the software related on my side, nor hardware as I mentioend when testing other games there is significant boost.

Well I wanted to report that even the FPS seems really nice and high as I consider this a bit strange behaviour.

The one with CnQ enabled and LFS cant stress the CPU to switch it to higher power state I consider more a AMD driver bug but still other games dont behave like that.

I did not mentioned earlier that I am using FRAPS to see the FPS as the FPS meter in LFS seems to showing not consisten values. It just jump pretty quickly a bit up and down so its not to possile to correctly read the FPS value at any time.

[Trekkerfahrer]

Hi Scawen,

thanks for coming back so late.You should get some sleep:-)
Back to topic.I did a lot of tests between these systems several times. I dont have actually 2 standalone systems. I just change mobo+CPU+RAM and always do format and clean install. The graphic card remain the same and its ATI 4870 512MB ram.

The odd is that online I get more FPS but offline I get less when driving alone on Blackwood with XFG( view on wheels).Same on other tracks. In some places same in some I get with better CPU less. As I said the difference is around 10 percent worst maybe a little bit more on the new system and it go really down more then 10 percent(went from around 200 to almost 140FPS) at the place in attached picture. Same behaviour around the whole track. Some places seems same and some worst. Odd and dont know what it could be as the installas of XP are always fresh so it seems not the software related on my side, nor hardware as I mentioend when testing other games there is significant boost.

Well I wanted to report that even the FPS seems really nice and high as I consider this a bit strange behaviour.

The one with CnQ enabled and LFS cant stress the CPU to switch it to higher power state I consider more a AMD driver bug but still other games dont behave like that.

I did not mentioned earlier that I am using FRAPS to see the FPS as the FPS meter in LFS seems to showing not consisten values. It just jump pretty quickly a bit up and down so its not to possile to correctly read the FPS value at any time.

try 16 bit resolution, mostly it helps with fraps <-> LFS issues

[DEVIL 007]

try 16 bit resolution, mostly it helps with fraps <-> LFS issues
That change nothing.

[Degats]

DEVIL 007:
IIRC the outer loop of the LFS physics engine runs at 100Hz. Whether that has much to do with it I'm not sure, but anything above that and LFS is just drawing duplicate frames and stressing the graphics more.
If you set affinity to a single core, what % of that core is it using? If affinity is set to all cores, what overall % CPU is LFS using? This should help us determine if it's the CPU being the limiting factor, or something else.

[DEVIL 007]

Thats what I have tried as well already and here are the results which I have tested on BL GP track again today with just 1 car(Ai):

When no afinity is set to specific core Task bar monitor is showing around 25 percent CPU usage. Core temp can be more specific showing individual core usage and it showing LFS jumping between the cores and one of the cores is always get basically unused(showing 0-5 percent usage).

Setting the afinity to just one of the core for LFS giving following results and something suprised me here today!!!. Even I set the afinity to 3rd core the following happend:

1st core still get used (why?) ; 2nd core is jumping between 0-5 percent ; 3rd core for which the afinity was choosed for LFS is showing 35-50 percent.
Same strange behaviour is happening when setting the afinity to just 1 of any of the cores. LFS is still using 2 core apparently and 1 is always left unused.

Also regarding the graphic card. There is no change in FPS if I change the AA/AF or resolution down or up. My monitor has possibility to go 1 step but streching the picture just to the middle. I wanted to mentioned about LCD because somebody might asking how is possible on 22"LCD which has native resolution 1680*1050 to go to 1900*1200. I can set for example lower resolution of 4:3 and my monitor is capable not to strech the picture just fill the screen to aspect ratio.

Seems something is holding up LFS on this new system but dunno what. As I described in every other games I got huge boost in FPS.
As I said its really strange the FPS are basically a bit lower. A64 comparing to Phenom II is much slower clock per clock and I have even 800Mhz more now. It looks from looking at the CPU usage that LFS is simply not able to enough use CPU power when the CPU usage never goes above 50-55 on this new system when afinity is set to just one of the cores. Isnt is the cause here the Minimum sleep set to 1ms and not allowing to change it to 0ms? Just an idea as I mentioned already. Sorry if I seems going in circles a but I feel a bit confused by this several strange behaviours and not able to narrow it to anything. Maybe Scawen would be able to find something by my reporting....hopefully. Not a big deal but maybe it would help to clean something in the code(possible bug) or even to dig more performance from LFS.


Also I dont bealive that LFS is drawing multiple frames. How would be possible then that Blackwood would be showing for me 120-230FPS from wheel view and Westhill up to 330. Another thing is why the FPS is jumping so much between both values and FPS behaviour does not change when I set resolution higher or lower. I would expecting as LFS seems not to be graphically limited on my card (I really dont bealive it is) that the FPS would be more steady.

Scawen,
did you start implementing some multithreading(even by something you did and you dont know that it might caused LFS going to multicore - sorry might be stupid questions/idea. Just trying really to find what is going on here) to LFS which might causing some of this strange behaviours.

EDIT:
One more notice. When I move the LFS down to desktop by ALT+TAB then LFS usuage get like max 12 percent only on 1 core and then rest of the cores jumping between 0-3 percents.As I understood the physic calculation still going on even LFS being moved to desktop as the Ai cars simply moving on the track. LFS cant even use single core CPU by more then 12 percent or I am missing something here?

When the LFS in full screen the usage goes much higher but still cant get over 55 percent even when afinity is set to just one of the cores. I am lost to be honest is it does not make sense to me all these things. Sorry if I sound stupid Scawen but I simply does see the reason why LFS can fully use CPU nor graphic card.

1st picture showing LFS usage when running in full screen or SHIFT+F4 and 2nd when I ALT+TAB LFS from full screen.
3rd picture is showing GPU usage on BL GP track with 6 Ai in 3 laps race. When it showing around 10 percent thats when I move LFS to desktop by ALT+TAB.

Just my observation not that it would have to be like that but looking at the pictures my self seems it seems something is not optimized.

[Degats]

AFAIK, LFS does not do any parallel processing. When affinity is set to several cores, it shows usage on more than one core because (I think) lfs.exe calls external modules/system calls which the OS assigns to a different core. lfs.exe can't use more of its core because it is 'paused' when the external calls are running.

I have a feeling that LFS physics will rarely need to use a full core on a Phenom II - on my pc (phenomII X4 955), lfs.exe rarely uses more than about 10-18% total CPU. Strangely, it seems to be at 250.0fps nearly all the time in windowed mode - there's probably a (sensible) software limit somewhere.

[DEVIL 007]

Well I know LFs cant use more then one CPU core, nothing new here but thanks for trying. Still your post does not answer all the above I have mentioned.

[Shadowww]

http://www.lfsforum.net/images/flagpics_tiny/Czech%20Republic.gif DEVIL 007 (http://www.lfsforum.net/member.php?u=22), do you have IE6 running in background? It sometimes takes up to 99% CPU. Get real browser like Fx 3.5 or Opera 10.

[DEVIL 007]

Did you check the pictures above I posted. You can see even IE running but the CPU usage is low. Of course I did most of the comparsion test even without anything then LFS running and switching off the antivirus software/firewall.

I am using same combination of software when I switch the hardware so I am pretty sure I rule out the software issue. Also as I said in other games I get significant boost. This seems to me really LFS related.

[DEVIL 007]

Hi Scawen & all,

You will not bealive but I tried just for curiosity to install Win7 RC1 and guess what. My FPS doubled on new system. So where I had at start on BL track like 160-170 FPS I have now 333 FPS and on some parts of the track even 400+ and some places almost 500!!! :eek:. Its insane but who really care if I can not see in reallity the difference between 60 or 500 but the more important thing my online FPS doubled as well!!!. I was sometime going to 30-40 FPS at start but now I am not moving below 100. I wanted nice smooth frames.

Now the puzzle is what is really causing the issue with better system being slow in LFS with XP. As I said several times other games run fine in XP SP3 and had huge boost in FPS with new system. Only LFS failed to do so. Any idea someone?

I dont think its the system it self but only two options in my view. Graphic driver or LFS itself.
However i tried different drivers and the odd performance drop or basically not gain at all was same using different drivers. I dont want to point to LFS but it looks like that.

Need to find now if my older wheel will work with Win7 RC1.

EDIT:
Found interesting behaviour of LFS in Win7 RC1 with enabling virtual mirror.
For example at BL start line when enabling the FPS drop from 330 to 260. At T1 without mirror the FPS is 427 while with mirror its 500. something weird is happening. Its always reproducable and also on other tracks. Tried westhill and same behaviour with those virtual mirrors.

EDIT 2:
Testing more...
Tried to install newer driver using windows update but got less LFS (10-15 percent) so reverted back to 8.56.1.13 version using rool back function under device manager.

[Renku]

Now the puzzle is what is really causing the issue with better system being slow in LFS with XP. As I said several times other games run fine in XP SP3 and had huge boost in FPS with new system. Only LFS failed to do so. Any idea someone?
I think you are capable of understanding that Scawen has loads more important work to do than figuring out why you have 10% FPS drop! It's not like you can't play LFS @ 140FPS!

[DEVIL 007]

I think you are capable of understanding that Scawen has loads more important work to do than figuring out why you have 10% FPS drop! It's not like you can't play LFS @ 140FPS!

That was not my point or probably I did not make it clear.
Did you read my latest report re the Win 7 RC1 performance. My online FPS doubled at least. I was struggeling even with better computer.

Something looks simply not right to me. I was not complainig or whinigh about my FPS. The main point was that the new computer configuration being superiour to the old should give much better performance and not to leading to 10-15percent drop.

Just trying to help here.

[Shadowww]

DEVIL 007, in XP, go to Control Panel > Power Options > High Performance, also disable Cool 'n' Quiet.

[DEVIL 007]

I isad CnQ is disabled in windows as well in Bios and I never use other the high performance profile. Thats not the issue with windows XP.
As I said I tried to rule out all the possibility of the above odd perrfomance in XP by disabling every power saving features and simillar things.

[quatrolhos]

Did you install on WinXP every specific driver from the MB manafacturer???

Most of the windows just uses general drivers which in some cases can show unusual behaviours in terms of performance and strange crashes.

Try to check the motherboard site for updated drivers for your chipset.

[DEVIL 007]

All,
nice suggestions but I dont how to make it more clean that I ruled out software issues. I am not instllating PC hardware first time, I am actually builduing like every month at least one PC including software as well.

I had all latest drivers for mobo,CPU,graphic card. Also tried older version of graphic card.


EDIT: Scawen, please would you be able to look at this performance "issue" at some time? I fully understand that there are more important things on your list to do and these must be finished first. I think there is something really wrong with LFS and might cause issue with performance on some faster PC. Maybe this investigation could help to gain some more pefromance out of LFS. If you need anything more then let me know. As first idea I got the minimu sleep value to be tested with 0ms.

[hotmail]

Let me Qoute some thing (from a research)

your brain can detect flickering till 60 frames per second

So in other words , why the hell would you worry about it is droping between 350~500, and btw everybody got that, because there is no way that a system can be stable at a that high-end speeds,

This has got nothing to do with lfs, it has got to do with gaming in general,

Look: I am easy touching 550, 580 fps in lfs, and sometimes i am touching 170, But i Dont care , because i know (my brains know) that i cant use more then 60!. And like i said before, i got the same in Counter Strike, CoD 4, etc..

(btw: funny qoute in your Signature)

[SamH]

Just to add to that, whatever framerate is being reported by LFS/FRAPS, the framerate being delivered to your eye is whatever Hz your monitor is running at. If your graphics card is cranking out 500fps but your monitor is running at 85Hz, then ONLY 85fps will pour out of your monitor. Anything FPS, in excess of the number of Hz, is just wasted cycles.

[edit] That also means that, unless you don't have frames locked to Hz, and have that ugly tearing going on (which nobody in their right mind would volunteer for), if you THINK you can see the difference between 400fps and 500fps, you're fooling yourself. BUT, if you're interested, I have some audio cables for sale, with carbo-goldium tipped connectors for £600, which will DOUBLE the quality of the audio out of your stereo.. and you'll definitely hear the difference!...

[MijnWraak]

I dont how to make it more clean that I ruled out software issues.
Obviously not, because an operating system is software and changing it doubled your FPS. :shrug:

[jonny__27]

Let me Qoute some thing (from a research)



So in other words , why the hell would you worry about it is droping between 350~500, and btw everybody got that, because there is no way that a system can be stable at a that high-end speeds,

This has got nothing to do with lfs, it has got to do with gaming in general,

Look: I am easy touching 550, 580 fps in lfs, and sometimes i am touching 170, But i Dont care , because i know (my brains know) that i cant use more then 60!. And like i said before, i got the same in Counter Strike, CoD 4, etc..

(btw: funny qoute in your Signature)

It seems to me you failed to realize the point of DEVIL 007's thread.

Seriously, I don't care if it was 500, 60, 5 or even 64783465453 FPS. Point is, if you buy some new, improved components, you would expect an increase of performance, but in this case it actually decreased.

[hotmail]

It seems to me you failed to realize the point of DEVIL 007's thread.

Seriously, I don't care if it was 500, 60, 5 or even 64783465453 FPS. Point is, if you buy some new, improved components, you would expect an increase of performance, but in this case it actually decreased.
And it seems to me that you only read the title....

[DEVIL 007]

Just to add to that, whatever framerate is being reported by LFS/FRAPS, the framerate being delivered to your eye is whatever Hz your monitor is running at. If your graphics card is cranking out 500fps but your monitor is running at 85Hz, then ONLY 85fps will pour out of your monitor. Anything FPS, in excess of the number of Hz, is just wasted cycles.

[edit] That also means that, unless you don't have frames locked to Hz, and have that ugly tearing going on (which nobody in their right mind would volunteer for), if you THINK you can see the difference between 400fps and 500fps, you're fooling yourself. BUT, if you're interested, I have some audio cables for sale, with carbo-goldium tipped connectors for £600, which will DOUBLE the quality of the audio out of your stereo.. and you'll definitely hear the difference!...

Sam,
from a moderator I would be expecting more friendly attitude - I really did not like the phrase about fooling myself and the cable part. Please stop trying to make an ididot from me. I had always respect for you being neutral all the time yet you had to dissapointed me :(.

I am not such a guys who dont understand things around the PC. Also you probably did not read all my reports/test I did above. As I said I really dont care if its 100 or 500FPS...whatsoever. I have read a lot of as well I know quite well and from my experience as well that I can not see difference between 60 FPS and more. My point was that for me it looks there is something buggy with LFS as my FPS went after upgrade with every game much higher while in LFS I got even a small decrease of perfomance which is suspicious.

Also I said my FPS in online went also with Win7 RC1 much higher(basically doubled at least) and that what really matters here. The problem is that it did not happend in Win XP SP3.

When I made the compersion especially in offline part of the game to rule out also CPU being fluctuated by any network glitches or what so ever.

[DEVIL 007]

Obviously not, because an operating system is software and changing it doubled your FPS. :shrug:

Well in a certain way you are right but why in other games I can simply see the difference while LFS failed to do so?

I would like to say that I rule out LFS to be causing the issue but no matter options I tried it simply did not work and no change with FPS in LFS with new system on Win XP SP3.

[Degats]

It's possible that it's caused by XP not handling the threads very well on a system with more cores or a different architecture than before - particularly as LFS is a fairly CPU intensive multi-threaded (but not parallel threaded) program. LFS is more CPU intensive than most games and is usually CPU limited rather than graphics limited.

Vista and especially W7 are programmed with multi-core and modern architectures in mind and usually perform better with such programs as LFS.


Off topic:
if you're interested, I have some audio cables for sale, with carbo-goldium tipped connectors for £600, which will DOUBLE the quality of the audio out of your stereo.. and you'll definitely hear the difference!...

On that subject, take a look at >> this, << (http://www.amazon.com/Denon-AKDL1-Dedicated-Link-Cable/dp/B000I1X6PM/ref=sr_1_2?ie=UTF8&s=electronics&qid=1247588831&sr=8-2)especially all the comments that praise it ;)

[SamH]

Sam,
from a moderator I would be expecting more friendly attitude - I really did not like the phrase about fooling myself and the cable part. Please stop trying to make an ididot from me. I had always respect for you being neutral all the time yet you had to dissapointed me :(.

I didn't think I was being unfriendly :shrug: and I wasn't trying to make an idiot out of you. I was just pointing out some facts. What I said about FPS is true, and I think that it's important not to obsess about things that don't make a difference. :shrug:

Finding what is causing the drop in reported framerate is a worthy cause but I think it's very important to keep in context what the impact of the drop really is.. and a drop of 100fps from 500 to 400fps is genuinely not pivotal to the visual operation of LFS on a 2009 monitor. ANY monitor. That drop would be proportionate to my Celeron D dropping from 100fps to 80fps. Believe me, it drops a lot more than that sometimes.

Stating what I think and how I feel.. that's what I do, I'm afraid. I'm not neutral.. I'm just me :)

[hotmail]

Whooo, SamH is a person, not a robot:thumb::nod:

[MijnWraak]

Whooo, SamH is a person, not a robot:thumb::nod:
WHAT!?!?!?
this is news to me. Why didn't you tell me that before I slept wi-- nevermind.

[DEVIL 007]

I didn't think I was being unfriendly :shrug: and I wasn't trying to make an idiot out of you. I was just pointing out some facts. What I said about FPS is true, and I think that it's important not to obsess about things that don't make a difference. :shrug:

Finding what is causing the drop in reported framerate is a worthy cause but I think it's very important to keep in context what the impact of the drop really is.. and a drop of 100fps from 500 to 400fps is genuinely not pivotal to the visual operation of LFS on a 2009 monitor. ANY monitor. That drop would be proportionate to my Celeron D dropping from 100fps to 80fps. Believe me, it drops a lot more than that sometimes.

Thanks for clearing that about your attitude. Written things can have sometimes more meaing unfortunately.:shrug:

I think you did not read my last reply carefully also about the online FPS not improving otherwise you would still not repeat the 100-500 FPS thing.I dont want any heated discussion as this would be really pointelss and not hleping the things. I just reported the issue which I bealive might be LFS related. Thats all.

I think really good thing to start with would be to try in some test patch how behave the LFS with minimu sleep set at 0ms under the windows XP SP3 here. Just an idea sure there might be something else in LFS or it can be totally something different however nobody suggested anything what I did not try yet and would fix the issue.


EDIT: some suggestion came to me via PM about sound card and the respective drivers. Tried different drivers as well as to disable the sound card in bios. No change in windows XP SP3. The frame rates are still half compared to Win7 RC1. The offline play is something which really does not bother me but as the online FPS showing low under XP SP3. No gain against my old system.

[Doc44]

Hi,just a quick one,I recently upgraded fron a AMD 6000+X2 to a Phenom II X4 BE and had a HUGE increase in fps and general performance,300+ fps,everything maxed,hi_def etc,etc,magic stuff,LFS loved it!!!!!! Doc.......

[aroX123]

First time i've seen a guy say something nice in his first post :)

[UnknownMaster21]

I can see differense about fps from 0fps to 100fps but over that...man it just wastes my eyes, so I have putted 100fps

[Falcon77]

Just to add to that, whatever framerate is being reported by LFS/FRAPS, the framerate being delivered to your eye is whatever Hz your monitor is running at. If your graphics card is cranking out 500fps but your monitor is running at 85Hz, then ONLY 85fps will pour out of your monitor. Anything FPS, in excess of the number of Hz, is just wasted cycles.

Not strictly true.
Yes, it is meaningless visually above a certain point, but it may make sense regarding the controller polling frequency. Not necessarily talking about LFS, but most sims' controller polling/FFB frequency is linked to fps. That can still mean a tiny difference above the visually meaningful range.

[amp88]

Not strictly true.
Yes, it is meaningless visually above a certain point, but it may make sense regarding the controller polling frequency. Not necessarily talking about LFS, but most sims' controller polling/FFB frequency is linked to fps. That can still mean a tiny difference above the visually meaningful range.

That seems like a really bad way to do things. Have you got any articles or links as proof for that statement?

[Falcon77]

That seems like a really bad way to do things. Have you got any articles or links as proof for that statement?
It does seem like an odd concept, but I don't know of a much better way to do it..

[amp88]

It does seem like an odd concept, but I don't know of a much better way to do it..

Have the graphics and controller inputs on different threads (so they're independent of each other)? That way you can receive controller inputs even if you can't show the controller input in the graphics. The internal game model will still take account of the controller input though.

[Falcon77]

Have the graphics and controller inputs on different threads (so they're independent of each other)? That way you can receive controller inputs even if you can't show the controller input in the graphics. The internal game model will still take account of the controller input though.
Yes, with the current multicore CPUs it does make more sense than before. But you will have to synchronize and make sense of a lot more complicated situation. It *seems* like a better idea but when you try to code it it may turn out to be not really worth it. I'm not saying it cannot be done intelligently, but consider this:

The fps rate will render the sim undrivable if it falls too low. You have a 1 frame lag during which you are driving in the dark, it does not make a too big difference in overall usability if you have multiple samples of the input during that time.
On the other hand, if you have very high fps, it doesn't make sense to poll the controller *less* frequently than the fps rate. In any case, you would want to do it as frequently as possible, why limit it if doesn't harm anything else?

[amp88]

Yes, with the current multicore CPUs it does make more sense than before.

Even with single core CPUs threading is a massively important process. The invention of multicore CPUs has brought the issue more into the public's eye, but threading was used long before that. Even with a single core CPU you can do multiple things at the same time (e.g. listen to an mp3 and drive in LFS).

But you will have to synchronize and make sense of a lot more complicated situation.

Synchronisation isn't really that complicated an issue, especially for game developers. They already need to think about such things as multiplayer racing where the issue of synchronisation is more important.

It *seems* like a better idea but when you try to code it it may turn out to be not really worth it. I'm not saying it cannot be done intelligently, but consider this:

The fps rate will render the sim undrivable if it falls too low. You have a 1 frame lag during which you are driving in the dark, it does not make a too big difference in overall usability if you have multiple samples of the input during that time.
On the other hand, if you have very high fps, it doesn't make sense to poll the controller *less* frequently than the fps rate. In any case, you would want to do it as frequently as possible, why limit it if doesn't harm anything else?

I don't really understand the point you're trying to make. What I'm trying to say is that the controller input and the graphic display don't need to rely on each other. I think what you're trying to say is that if you can't see the result of your inputs (because of too much FPS lag) then those inputs should be ignored until the FPS has improved. Is that right?

[Ped7g]

DEVIL 007: some math:
60 FPS = 16.6666..ms per frame
120 FPS = 8.3333..ms per frame
300 FPS = 3.3333..ms per frame

Most of the internal OS timers/thread switching/interrupts are often running in 1ms to 10ms period, rarely under 1ms, i.e. having application running above 100Hz with MS Windows (or any other ordinary desktop OS) can lead to a bit unpredictable results with serious degree of variation.
For 100+Hz applications it's usually very sane to switch to some real-time OS. Which doesn't make sense for computer game.

So you are measuring something what can be disturbed by so many tiny factors, it's difficult to tell. It may be AMD vs Intel optimizations in gfx driver, it may be AMD vs Intel optimizations inside Win XP threading, it may be some motherboard/bridge/OS communication taking a 1ms longer here and there...

I think you need to run the game engine under 60-100 FPS to measure performance of HW reliably, i.e. your HW is too powerful for this and there's no point for Scawen to solve this. If there's some additional performance needed, it's for that HW which can't run at steady 60-85FPS and I bet Scawen will focus on that, if he will have some spare time to work on optimizations.

[Falcon77]

Sorry for the late reply.

I would like to emphasize that I'm not claiming that the 'one-poll-per-frame' method would be the better solution, only that going more sophisticated may not be worth it for the developer/the issue itself.

Even with single core CPUs threading is a massively important process. The invention of multicore CPUs has brought the issue more into the public's eye, but threading was used long before that. Even with a single core CPU you can do multiple things at the same time (e.g. listen to an mp3 and drive in LFS).

Of course, multitasking makes sense. But there is a cost of a thread switch and as Ped7g wrote threading has a rather coarse granularity when we're looking at the millisec level.
Also it is not irrelevant whether the two threads are dependent on each other or not. In the case of controller polling/physics/frame display, they are not as independent as to have no effect on each other.


Synchronisation isn't really that complicated an issue, especially for game developers. They already need to think about such things as multiplayer racing where the issue of synchronisation is more important.
That game developers need to solve difficult synchronization issues, is true of course. Multiplayer sync is a very good example.
But in general, it is not true that synchronization isn't that complicated, and game developers are not normally über-developers just because they happen to code games (Scavier being obvious exceptions, of course :tilt:).
In general, one tries to tackle the programming tasks with the simplest solution that provides the expected results, that is the idea anyway. LFS is actually a good example as far as I can tell of a cautious approach: there are a lot of quite simple solutions in the engine (that make sense at the same time) and Scawen makes sure he really understands what he's doing when he goes deeper into something.


I don't really understand the point you're trying to make. What I'm trying to say is that the controller input and the graphic display don't need to rely on each other. I think what you're trying to say is that if you can't see the result of your inputs (because of too much FPS lag) then those inputs should be ignored until the FPS has improved. Is that right?No, I'm not saying that they should be ignored. I'm saying there is not much practical difference if you ignore them or not. I'm also saying that there is sense in the polling being connected somewhat to the frame rate, although one or more polls per frame is a different question.
Would have more to say but gotta go now..

[DEVIL 007]

DEVIL 007: some math:
60 FPS = 16.6666..ms per frame
120 FPS = 8.3333..ms per frame
300 FPS = 3.3333..ms per frame

Most of the internal OS timers/thread switching/interrupts are often running in 1ms to 10ms period, rarely under 1ms, i.e. having application running above 100Hz with MS Windows (or any other ordinary desktop OS) can lead to a bit unpredictable results with serious degree of variation.
For 100+Hz applications it's usually very sane to switch to some real-time OS. Which doesn't make sense for computer game.

So you are measuring something what can be disturbed by so many tiny factors, it's difficult to tell. It may be AMD vs Intel optimizations in gfx driver, it may be AMD vs Intel optimizations inside Win XP threading, it may be some motherboard/bridge/OS communication taking a 1ms longer here and there...

I think you need to run the game engine under 60-100 FPS to measure performance of HW reliably, i.e. your HW is too powerful for this and there's no point for Scawen to solve this. If there's some additional performance needed, it's for that HW which can't run at steady 60-85FPS and I bet Scawen will focus on that, if he will have some spare time to work on optimizations.

The issue I mentioned in this thread only happening in LFS to me. I tried untill now to find out many ways to solve it but nothing. Not sure if this some sort of glitch between LFS and Win XP or anything else but Win7 simply solved it.

I did not notice in any other games such a behaviour. Not something really big to look at but maybe,maybe worth to spend some time later on. Thats up to Scawen to decide. My only purpose of this thread was just to report this.

[ozgur]

This problem maybe your gfx card & drivers supports dx10 xp supports dx9 windows 7 supports dx11...
I have same card(q9550 cpu,win 7), getting 320fps (@2xAA & 16xAS & 1680x1050x75hzx32bit) while stopping bl gp pole position in car view then limiting to 85fps i think thats the ideal for both single&multiplayer setup you can try :)
I think all games need %5-%10 gpu power surge for multiplayer fps drops...

[ftop]

Thats true, i was making video with :
Amd Cemprom (or whatever) @ 2800 Processor
1GB Ram
Nvidia FX5200 128mb
And i dont even get 1 fps down than not recording...
now i have
Amd athlon 3200+ @ 2.0Ghz
3.50GB Ram
Nvidia 9600GT 512mb
and i didnt upload video because i cannot make one... to much lag
:shrug::shrug::shrug: