OMFG - SLI on a laptop??? Wow... :-O
And yes, single-thread performance hasn't advanced very much in the last several years. Worse on AMD CPUs than Intel but even Intel are massively off the curve.
Oh. I thought SLI was "transparent" - that it didn't need app support to work, but that the device drivers just did the sharing under the hood...
Reading some of these recent posts about CPU/GPU utilisation, fps and vsync reminds me that I've been meaning for weeks to post about my recent observations on this, but I've been wanting to do some more testing first.
I'm not at the right PC now so can't provide full details, so I'll just state the basics, which may be relevant to this topic.
I got a big surprise after upgrading my CPU recently. It was coping really well with most tracks and I decided I didn't want to heat up my bedroom needlessly (!!) by having both the CPU and GPU running flat out and getting stupidly high fps (like 300). Thus, I decided to limit the frame rate (I had been running without any limit). I tried various rates, settling on 50ish fps and found I had massively reduced the wall-plug power consumption (IIRC, a ~100W drop!). Yay. I then decided to try vsync to see if I could spot the difference (absence of tearing) and indeed I could so I left vsync on.
A bit later I got the big surprise: I realised that when vsync was ON, the CPU was absolutely flat-out on the LFS thread. Bit of a WTF moment... I checked the power consumption and sure enough, it was stupidly high with vsync at 60 fps, and massively lower with frame rate SET to 62 fps (which gave me an actual fps of 60ish). I'm driving w/o vsync now
My planned-for testing was meant to work out if old versions of LFS did this (I was running H or maybe H2), and get some hard numbers on it, but that's the basic story already. I checked some older threads here, which had some mention of fps and vsync, but nobody observed what I've just mentioned (at least not that I found).
A quick bit of googling gave me some hints that vsync may be implicitly causing a busy wait (either in a game or the video driver), which seems a bit faily really but maybe there's just no better way, which would be tragic. If there is a better way, then perhaps LFS could use it. Or maybe my rig is just odd? (Win XP.)
NB: in the cases above, if thermal throttling does ever rear its head, it's more likely to happen with vsync on I guess (that's throttling on the CPU, not the GPU).
I'm not at the right PC now so can't provide full details, so I'll just state the basics, which may be relevant to this topic.
I got a big surprise after upgrading my CPU recently. It was coping really well with most tracks and I decided I didn't want to heat up my bedroom needlessly (!!) by having both the CPU and GPU running flat out and getting stupidly high fps (like 300). Thus, I decided to limit the frame rate (I had been running without any limit). I tried various rates, settling on 50ish fps and found I had massively reduced the wall-plug power consumption (IIRC, a ~100W drop!). Yay. I then decided to try vsync to see if I could spot the difference (absence of tearing) and indeed I could so I left vsync on.
A bit later I got the big surprise: I realised that when vsync was ON, the CPU was absolutely flat-out on the LFS thread. Bit of a WTF moment... I checked the power consumption and sure enough, it was stupidly high with vsync at 60 fps, and massively lower with frame rate SET to 62 fps (which gave me an actual fps of 60ish). I'm driving w/o vsync now
My planned-for testing was meant to work out if old versions of LFS did this (I was running H or maybe H2), and get some hard numbers on it, but that's the basic story already. I checked some older threads here, which had some mention of fps and vsync, but nobody observed what I've just mentioned (at least not that I found).
A quick bit of googling gave me some hints that vsync may be implicitly causing a busy wait (either in a game or the video driver), which seems a bit faily really but maybe there's just no better way, which would be tragic. If there is a better way, then perhaps LFS could use it. Or maybe my rig is just odd? (Win XP.)
NB: in the cases above, if thermal throttling does ever rear its head, it's more likely to happen with vsync on I guess (that's throttling on the CPU, not the GPU).
Hey, awesome.
Have been racing a lot on cargame recently, as that was just about the only non-empty server (!!) but it has been a bit hit and miss there with crashers and speedhackers.
When a server has no admins or even limads, bad stuff happens...
See you all online!
Have been racing a lot on cargame recently, as that was just about the only non-empty server (!!) but it has been a bit hit and miss there with crashers and speedhackers.
When a server has no admins or even limads, bad stuff happens...
See you all online!
Hmmm
Are you gonna say that LFSLazy can do that? I think I really need to stop procrastinating and install it! LOL...And LOL again. How did you work out the URL syntax for LFSW? (Though one annoying thing if you use a link like that is that it remembers the options unless you open in a private window.)
Anyway, thanks for the nice link, but a big slap with a wet fish for failing to notice that I already know LFSW can filter by demo and don't want to have to use it
Filter out demo hosts in LFS host list...
(There was a thread on something rather like this topic in the past, but it has been mangled by the OP.)
LFSW has a way to filter by Demo/S1/S2.
LFS does NOT.
It's a pretty simple improvement suggestion: I just want to be able to eliminate all Demo hosts from the list I get when I hit "Display list of hosts". I don't really care about filtering out S1 or S2.
(Reason is simple: my experience is that demo servers are full of crashers.)
I can't understand why LFS has such a complex filtering system (it's actually more capable in some ways than the one in LFSW!!) but yet can't hide Demo hosts. This is particularly painful these days when most of the populated servers are Demo.
Please consider making this possible...
LFSW has a way to filter by Demo/S1/S2.
LFS does NOT.
It's a pretty simple improvement suggestion: I just want to be able to eliminate all Demo hosts from the list I get when I hit "Display list of hosts". I don't really care about filtering out S1 or S2.
(Reason is simple: my experience is that demo servers are full of crashers.)
I can't understand why LFS has such a complex filtering system (it's actually more capable in some ways than the one in LFSW!!) but yet can't hide Demo hosts. This is particularly painful these days when most of the populated servers are Demo.
Please consider making this possible...
Intriguing problem.
If I'm reading it right, you still had the problem with the wheel disconnected, yeah?
It's great that you've ruled out LFS as the origin of the problem (by finding another app with the same problem). However, maybe I missed it, but I'm not seeing any reason why you're confident the problem is within Wine. Why could it not be at the Linux level? E.g. rogue process (either CPU activity or more likely IO).
It's been a (very!) long time since I hacked the Linux kernel, but one stupidly basic test I used to perform to debug scheduler issues was to simply hold down the space bar in a shell window, and watch the cursor, looking out for jerky movement happening from time to time.
If I'm reading it right, you still had the problem with the wheel disconnected, yeah?
It's great that you've ruled out LFS as the origin of the problem (by finding another app with the same problem). However, maybe I missed it, but I'm not seeing any reason why you're confident the problem is within Wine. Why could it not be at the Linux level? E.g. rogue process (either CPU activity or more likely IO).
It's been a (very!) long time since I hacked the Linux kernel, but one stupidly basic test I used to perform to debug scheduler issues was to simply hold down the space bar in a shell window, and watch the cursor, looking out for jerky movement happening from time to time.
Aha, most excellent, thanks. Will have to give it a try now
Well I'll give it a shot I guess. The major win for it is that it has an active developer, so features ought to be addable. (Can't recall if it's open source, which would be even nicer.)
One real killer feature of Aonio (for me) is the live delta (time and speed) vs. your PB. (In the cfg file it's called "ShowComparison", and yes, I suspect the feature uses some of the columns in the file you mentioned above
!) Somebody in chat a few weeks back led me to believe that LFSLazy can't do quite the same thing, which would be a bummer.Happy to help
And d'oh, of course, yes, I keep forgetting how cool Google Docs can be (I've got a bunch of stuff in there myself but for some reason I've never tried to make a graph within a sheet...!).
As for updates, well I'm no Insim expert but I know things have changed a wee bit (new packets). I got some errors in the console in previous session (something or other was unrecognised I think) but when I drove some more this afternoon to generate some fresh error text to cut and paste in here, Murphy's Law struck (no errors). But yeah, it's working almost perfectly for me, just a few odd glitches now and then. I'll probly try that new Lazy thing soon, especially if Aonio dies.
PS: it's really weird that Aonio stores that stuff in the pb files - I can understand time, speed & track position, and maybe even X/Y position; elevation seems a little bit pointless, but steering angle??
PPS: I guess most of that info would be useful if you wanted to overlay a ghost car (but Insim can't any more I think?). I still don't get steering angle tho
And d'oh, of course, yes, I keep forgetting how cool Google Docs can be (I've got a bunch of stuff in there myself but for some reason I've never tried to make a graph within a sheet...!).
As for updates, well I'm no Insim expert but I know things have changed a wee bit (new packets). I got some errors in the console in previous session (something or other was unrecognised I think) but when I drove some more this afternoon to generate some fresh error text to cut and paste in here, Murphy's Law struck (no errors
). But yeah, it's working almost perfectly for me, just a few odd glitches now and then. I'll probly try that new Lazy thing soon, especially if Aonio dies.PS: it's really weird that Aonio stores that stuff in the pb files - I can understand time, speed & track position, and maybe even X/Y position; elevation seems a little bit pointless, but steering angle??
PPS: I guess most of that info would be useful if you wanted to overlay a ghost car (but Insim can't any more I think?). I still don't get steering angle tho
Last edited by Neilser, .
Just spotted this thread, while checking if Aonio has been updated yet (yeah, I know, not very likely...).
Did you ever figure out the telemetry?
For what it's worth, the second one you couldn't figure out looks very like a heading to me (or other angle, cos it wraps at 36000). No clue what the first one is, but in a random pbs.txt file I just looked at, the 4th column goes positive sometimes...
PS: what did you use to make those graphs? I'm always on the lookout for nice tools
That does ring a bell.
However, that's a truly lame reason to make rules for skin uploading. The real answer is to fix the flaw that causes the lag in the first place (which isn't gone btw)...
However, that's a truly lame reason to make rules for skin uploading. The real answer is to fix the flaw that causes the lag in the first place (which isn't gone btw)...
Fair point. Shouldn't even use up much dev time (useful since you won't get many people clamouring for this particular fix!).
I had managed to forget that the fetch from LFSW only happens if the skins folder doesn't already have a match
I had managed to forget that the fetch from LFSW only happens if the skins folder doesn't already have a match
Ah...
Help me out here though - if the guy you're racing with in this one-on-one session can't see your skin (cos you haven't uploaded it), then what is the point of having the skin in the first place? (If you're in the cockpit view, you can't even see it yourself.)
Help me out here though - if the guy you're racing with in this one-on-one session can't see your skin (cos you haven't uploaded it), then what is the point of having the skin in the first place? (If you're in the cockpit view, you can't even see it yourself.)
Yup, should be possible, I imagine, but why wouldn't uploading some plain white skins (with matching names) also fix your problem?
Jeez, that's a poor attitude from anyone, never mind a mod.
He may well have told you a bunch of stuff you already knew, but I suspect that in fact he was trying to help you.
And while I'm baffled about why you'd want a custom skin in the first place if you don't want anyone else to see it, quite frankly I don't care.
LOL, yes
But I'd guess they wouldn't be specced until a new attempt was made to fetch the skin - e.g. someone new joining the server. Easy to test tho
Huh? Have I lost the plot entirely? If not, you're confused:
It's incompatible, so if you update now nobody at all will be able to join your server.
Edit: it's moot now, cos 0.6H is out. Damn, I keep checking the TEST patch forum instead of the main forum... D'oh!
Last edited by Neilser, .
OK, cool
Having read the LYT file format now though, I start to realise why it's somewhat non-trivial to make an LYT editor perform the way someone might want...
The interaction between floating and non-floating objects is going to be a pain I guess; e.g. you can't simply translate a set of objects and expect them to maintain their relative positions unless all are floating (or the ground is horizontal).
But at the very least, building structures should be much easier if you write a program to do it rather than do it totally by hand.
Is there a reason btw to keep the LYT file so astonishingly compressed? I would have thought that it's a once-per-track-load kinda thing, so having it much bigger (e.g. hundreds of kB) would be a non-problem, no?
I wonder if the FO8 can do it? (Smallish margin for BF1 though, so maybe not.)
This is exactly why a better editor for LYTs would be a neat thing to have. I haven't checked the file contents but it ought to be pretty easy, no?
Even just as a nasty hack suited best to things like this tunnel, you could draw the first ring or two, and then extend it (virtually "manually") by looking at the file contents and just duplicating the rings with an offset in the direction of the tunnel axis.
Unless, that is, the devs have outlawed such a thing.
This is exactly why a better editor for LYTs would be a neat thing to have. I haven't checked the file contents but it ought to be pretty easy, no?
Even just as a nasty hack suited best to things like this tunnel, you could draw the first ring or two, and then extend it (virtually "manually") by looking at the file contents and just duplicating the rings with an offset in the direction of the tunnel axis.
Unless, that is, the devs have outlawed such a thing.
Yup, that would be useful. But if the layout files are comprehensible, maybe "we" could hack up an external layout editor with that kind of capability. I admit I haven't checked the wiki so maybe something already exists. It's also possible that Scawen wouldn't want such a thing to exist?
..then I think LFS isn't the right "game" for her.
Fantastic news!
Very naughty though to release it the same day I take a 2 week break! :O
Very naughty though to release it the same day I take a 2 week break! :O
Good luck! Hope to see you back here some time...
FGED GREDG RDFGDR GSFDG