Build 930 CPU bar maxed out

Discussion in 'Technical & Support' started by Sido Weijer, Feb 18, 2015.

  1. hardtohandle

    hardtohandle Registered

    Joined:
    Jul 15, 2013
    Messages:
    364
    Likes Received:
    54
    I spend a lot of time with this, not really getting anywhere. But I notice one thing and I do not know what it means.

    The pink cpu bars, I have 4 for each core. Say core 1 - there is the main bar, and there is a very thin bar under it between the main bars for cores 1/2.

    This thin bar under the core 1 main bar - sometimes it has this dotted line go up and down. What is that thin dotted line under each core? What does it mean?

    Only thing I can tell you guys is that activity with that dotted line going up on core 1 is connected to the stutters. Sometimes it goes up (even tho core 1 is still only at 60 or 70% usage) and goes away, game runs fine, sometimes though it goes up and then I quickly get the 100% CPU thing and massive race ending stutter. The 2 things go together, activity with that dotted line and the 100% CPU usage they go together.
     
  2. Terence Groening

    Terence Groening Member

    Joined:
    Oct 13, 2010
    Messages:
    169
    Likes Received:
    0
    You might want to try it regardless. But we haven't noticed any issues with relatively newer chips here.
     
  3. MarcG

    MarcG Member

    Joined:
    Oct 4, 2010
    Messages:
    6,715
    Likes Received:
    2,041
    It's just been alerted to me that Green is Graphics (Purple is Physics), so not quite sure what is going on but I just did 3 races at Loch Drummond with the Clios and aside the odd stutter everything was fine. Again it's only these cars I've noticed it with, should I get more conistent stutters at another track I'll just use the "+procmask=3" trick as that worked just fine.
     
  4. Terence Groening

    Terence Groening Member

    Joined:
    Oct 13, 2010
    Messages:
    169
    Likes Received:
    0
    (You call it pink, I call it purple, regardless it starts with 'p' for physics!) The purple bar shows what percentage of the available time is used to calculate the physics/simulation. The graph below it is a histogram. The goal is to run the physics loop 400 times per second, evenly spaced through time. That histogram is showing you how many milliseconds late each loop ran. When the bars are mostly in the left few columns, everything's running about as well as possible under the Windows operating systems. When you see dots scattered along to the right, that means some loops ran late, and the physics is trying to "catch up" to realtime. So yes, the two things are very much associated - when the main purple bar gets near 100%, meaning we don't have enough time for the calculations, then the histogram samples move to the right, because we fall behind. And then you get stuttering because the physics and graphics gets out of sync, etc. Obviously, all of this is bad news, so we try to avoid it.
     
  5. MarcG

    MarcG Member

    Joined:
    Oct 4, 2010
    Messages:
    6,715
    Likes Received:
    2,041
    As I'm getting dots (Nobbly bits sounds better!) on my Green Bar then I've got a slight a slight issue with either the CPU playing catchup to the GPU or vice versa, not too fussed as it's not serious enough to warrant further investigation.

    Only the Clios though, all other cars have been perfectly fine with no stutters at all!
     
  6. Noel Hibbard

    Noel Hibbard Registered

    Joined:
    Oct 5, 2010
    Messages:
    2,744
    Likes Received:
    29
    Is the green bar GPU usage or CPU usage for the GFX thread?
     
  7. Sido Weijer

    Sido Weijer Registered

    Joined:
    Jan 6, 2014
    Messages:
    97
    Likes Received:
    2
    My CPU is an AMD Athlon II x3 445, a triple core of 3.10 ghz. Not exactly a state of the art CPU.
    So far it always ran rF2 well (since I added "prerendered frames=1" in Nvidia settings), but this build seems to take it over the edge.

    Where exactly do I enter that line of code allowing rF2 to only run on 2 cores?

    I also set "Threaded optimization=on" in the Nvidia settings, should I set that back to default (automatic) or doesn't it matter?
    Nvidia's description of this setting:
    Allows applications to take advantage of multiple CPUs.
    Typical usage scenarios:
    - Most newer applications should benefit from the Auto-select or On settings
    - This setting should be turned off for most older applications
     
  8. Terence Groening

    Terence Groening Member

    Joined:
    Oct 13, 2010
    Messages:
    169
    Likes Received:
    0
    To Noel: CPU usage. What we're really measuring for both the green and purple bars is what percentage of the time are we not sleeping.

    In the Launcher, go to Settings. Just above the middle you'll see "Command Line Interface Parameters". Type in "+procmask=3" and then "Save Parameters". You just did it for Single Player. If you want to do it for Multiplayer then you have to click that button above the entry box, then type it again and save.

    Triple core ... now that's something I'm not sure we've tested on recently. Let us know how it goes. (And thanks to those who already have.)
     
  9. Gijs van Elderen

    Gijs van Elderen Member

    Joined:
    May 1, 2014
    Messages:
    1,396
    Likes Received:
    467
    Is there command line option for the dedicatedrFactor2.exe?
    I want to set the dedicated server to 1 core and preferable a predefined CPU core: CPU0 or CPU1...
    Something like "+procmask=1" ???


    We set the CPU affinity manually with Windows task manager after the dedicated server is started. Server 1 is set to "CPU 0" server 2 is set to "CPU 1" etc.
    I don't know why but, this resolved our "Vmod download on a passworded server issue".
    Never had issues with it since. Only once when an admin forgot to set the CPU affinity's for each server after a reboot.

    And maybe a command line to set CPU use to High. This fixed the "user left a 1000 time" bug in build 880.
    A positive side effect was that the server was much more stable,
    - No clients disconnected when advancing to a next session.
    - No little more little freezes when someone leaves or rejoins during the race.
    - Running a 48 car grid on the Nordschleife went without any disconnection or lag.

    I expected some lag issues during the start. But everything was cool.
     
    Last edited by a moderator: Feb 20, 2015
  10. Noel Hibbard

    Noel Hibbard Registered

    Joined:
    Oct 5, 2010
    Messages:
    2,744
    Likes Received:
    29
    The CLI.txt has this:
    Code:
    +procmask=<bitfield>
    Allows user to specify processor affinity mask, using same bitfield as the
    call SetProcessAffinityMask(). (SP,MP,DEV,DS)
    And the documentation for SetProcessAffinityMask() is here:
    https://msdn.microsoft.com/en-us/library/windows/desktop/ms686223(v=vs.85).aspx

    Dec 3 in Bin is 0011 In this case the first two cores will be used.
    Dec 1 in Bin is 0001 so only the first core is used.
    Dec 2 in Bin is 0010 so only the second core would be used.
    Dec 4 in Bin is 0100 so only the third core is used.
    Dec 8 in Bin is 1000 so only the forth core is used.

    And so on.

    Oh and I forgot to answer the second part of your question. To set the process priority to high via command line.. use +highprio
     
  11. Gijs van Elderen

    Gijs van Elderen Member

    Joined:
    May 1, 2014
    Messages:
    1,396
    Likes Received:
    467
  12. Noel Hibbard

    Noel Hibbard Registered

    Joined:
    Oct 5, 2010
    Messages:
    2,744
    Likes Received:
    29
    I edited my post above with more explanation and the +highprio tip.

    Also check out all the command line options here:
    <DataPath>\Support\CLI.txt

    Just don't get too excited with what you see in there. A lot of the options don't really work. +autojoin and +connect for example don't really work. :(
     
  13. hardtohandle

    hardtohandle Registered

    Joined:
    Jul 15, 2013
    Messages:
    364
    Likes Received:
    54
    I think I found something out. I spent a lot of time looking at this now, yesterday I started getting thinking that what it is, is some part of the new tire model code that under certain very specific conditions make the CPU go through the roof. On certain classes of CPU at least, certainly on my Xeon X5450.

    I found it more useful testing this online, purely in order to increase the CPU load to begin with.

    Last night I did some laps at croft in celica gto, I got a bad stutter, I took the same corner in the same way very often and no repeat stutter - but sometimes in that very spot I did see a load of histogram activity no core 1 but no CPU usage increase - in other words it nearly did the same thing but not quite. Of course, I never truly take the corner the exact same way, and the tyres are different every lap, and I start thinking it is some very specific condition, or lines of code, that make it shoot through roof.

    Now I go online to ISR Nords 24h hotlap server, Celica GTO again and I see much clearer evidence what I just said is true.

    Lap 1 exiting karousel corner it stutter real bad - just after flying out the bowl back onto tarmac when the car takes some air and lands with lots of tyres squeeling it did it.
    Lap 2 entering this time not exiting, karousel, again it did it - also the car here is taking air and bouncing around and everything same thing - yet on exit it was fine!
    Lap 3 again exit karousel it did it, entry was ok that time.

    Also at nords there is some strips of temporary tarmac or something, that is all cracked up and you can feel the cracks in the wheel. Again, unusual tyre contact situation and the CPU spiked here again more than once.

    So I think there is some very specific tyre contact scenarios that just kill the CPU dead. Hope this helps.
     
    Last edited by a moderator: Feb 20, 2015
  14. Gijs van Elderen

    Gijs van Elderen Member

    Joined:
    May 1, 2014
    Messages:
    1,396
    Likes Received:
    467

    Thanks! got it.

    Like setting Parc fermé rule. ;)
     
  15. F1Fan07

    F1Fan07 Member

    Joined:
    Jan 11, 2012
    Messages:
    626
    Likes Received:
    46
    I wish I checked CPU but I noted in a URD race today (15 cars or so and my first race with b930) that my GPU usage was quite low (65-80%). Something was bottlenecking elsewhere. No older hardware here though.
     
  16. Sido Weijer

    Sido Weijer Registered

    Joined:
    Jan 6, 2014
    Messages:
    97
    Likes Received:
    2
    Thanks. Yesterday, before I entered the piece of code, I had the same CPU spikes offline with the modern F2 at modern Monza, no AI. I've just tested that same combination with the code and so far so good. I'll have another league race with the URD cars next Wednesday, I'll assume that'll be crowded again which will be a proper test.
     
  17. McFlex

    McFlex Registered

    Joined:
    Feb 23, 2012
    Messages:
    828
    Likes Received:
    240
    So i tried running with "+procmask=3". So far i only tested the SP with no AIs. It seems to be a little bit better. Here are two screenshots i took. First one is without "+procmask=3" and second one is with it enabled

    View attachment 15886
    View attachment 15887
     
  18. DocJones

    DocJones Registered

    Joined:
    Jan 26, 2011
    Messages:
    403
    Likes Received:
    5
    I was a bit sceptical to upgrade to b930 because I too have an older CPU, a Phenom II x3 720 BE (3 core processor, running @ stock). CPU loads might be a tad higher compared with previous builds, but I didn't have the CPU bar maxed out (not even close) at any stage nor did I encounter stuttering. That was while testing the Megane at Bahrain in an online session with another 6-8 human drivers present. The track is quite bumpy in places - noticing that as others reported the issue occured when tires had lots of load changing to cope with.
     
  19. hardtohandle

    hardtohandle Registered

    Joined:
    Jul 15, 2013
    Messages:
    364
    Likes Received:
    54
    I spent a lot of time to limit the risk of this any way I could, I did 2 races over weekend, both were a disaster I cannot race with this build like this :( Stutter city, thankfully I caused no accidents but it's a waste of time trying to race with this build.

    Weird how it only affects a small section of players, I am one of them. But it is hitting us hard. If this cannot be fixed I guess it is the end of the line for me and RF2, my CPU is not red hot but it is not a joke CPU neither, it is 3.0 ghz quad core. I cannot buy a new system just because of this, it run perfect on 910 CPU wise with tonnes to spare.

    The +procmask=3 did very little for me. I hope this can be fixed, the sooner the better. I am thankful for the responses from the devs, hope you can sort this :)
     
  20. jerrymcc

    jerrymcc Registered

    Joined:
    Oct 5, 2010
    Messages:
    160
    Likes Received:
    35
    This build is a huge problem for me as well. Only a q6600 cpu, but 910 and below were manageable in fields up to 15 or so with everything turned down. This build is a problem 2 or 3 ppl on track. Not in a position to upgrade ATM, +procmask=3, though a little different, didn't really help.

    Of course, the problem of upgrading is: the minimum needed is a new MB, CPU, and memory, which add up to about $500 for anything with any longevity. GPU is a 5850, so it is tolerable for a while.

    Any (very low cost) options for us?
     
    Last edited by a moderator: Feb 23, 2015

Share This Page