Sync errors

Abokasee · Post by **Abokasee** » 01 Jan 2007, 11:15

Arm chainsaw's (reapeting long range plasma cannon) are the Number 1 cause of sync errors, because they realease bullet shells wich cause lag, buzzsaw's (core's version) don't realease shells

MadRat · Post by **MadRat** » 02 Jan 2007, 02:50

AF wrote:madrat that sounds like Delayed Sync message and not Desync messages, afterall how cna you desync with yourself if only 1 simulation is running?

I would assume the single player skirmish is a hack of the normal multiplayer. Just because the clients are all one on PC doesn't mean they cannot desync. And it is a desynch message I see, not a delayed sync. I've never seen the latter.

Tobi · Post by **Tobi** » 02 Jan 2007, 10:34

MadRat, no AF is right. If you were running just one spring.exe, then it's theoretically impossible to desync (1 spring.exe = 1 player [can run multiple teams] = 1 sync checksum, and 1 checksum can never be different from other non-existant checksums). In case you still get "Sync error" messages in single player then something is seriously borked, but you're not out of sync

I should have read your post better in the first place btw

MadRat · Post by **MadRat** » 02 Jan 2007, 14:50

In the skirmishes that desync, there is no human player. Does it handle it any differently? Consequently I don't see the error messages unless I look at an individual team during the slowdowns, by using the .cheat and .team commands. When the game suddenly goes choppy its a pretty sure sign of a desync, as plain slowdowns from huge unit counts are much smoother.

Its not looking at the slowdowns that causes it, its the mass of tiny 1x1 units that seems to be the source. When I kill off the units by dropping kamikazi units next to them owned by another team both the slowdown and desync errors disappear.

btw - This is typically when using four AAI ai's on any size map where MAX_GROUP_SIZE is set above 10.

el_matarife · Post by **el_matarife** » 28 Jan 2007, 22:36

Sorry to resurrect this, but are you still analyzing these replays? Some of us are still uploading these and I was just wondering if it was going to waste.

On a different desync track, I've heard that stealth fighters flying off the map are causing desyncs, is that accurate?

Tobi · Post by **Tobi** » 28 Jan 2007, 22:45

Not really, a few weeks ago I statistically analyzed them all together a bit, maybe I do that again, but I'm focussing on resync now.

el_matarife · Post by **el_matarife** » 28 Jan 2007, 22:47

Tobi wrote:Not really, a few weeks ago I statistically analyzed them all together a bit, maybe I do that again, but I'm focussing on resync now.

Nothing turned up? Too bad.

tombom · Post by **tombom** » 28 Jan 2007, 22:54

Is it just me or are most desyncs on maps with water? Also all the desyncs I see are always exactly half the players in the game desync, rounded up. A spec might desync but it's still always the same number of players. I may just be seeing patterns that aren't there.

Tobi · Post by **Tobi** » 28 Jan 2007, 23:33

Only thing that I proved by doing that was that (almost) all games split in exactly two different games on desync, often both with multiple people, which hints it is related to some hardware bug / difference or something, and not to a spring bug.

@tombom: I've seen games where that wasn't true, the above is though (AFAICS).

el_matarife · Post by **el_matarife** » 28 Jan 2007, 23:51

Intel vs AMD perhaps? Some slight difference in FPUs or in general?

LordMatt · Post by **LordMatt** » 29 Jan 2007, 00:16

el_matarife wrote:Intel vs AMD perhaps?

No.

Dragon45 · Post by **Dragon45** » 29 Jan 2007, 00:53

Tobi wrote:Only thing that I proved by doing that was that (almost) all games split in exactly two different games on desync, often both with multiple people, which hints it is related to some hardware bug / difference or something, and not to a spring bug.

@tombom: I've seen games where that wasn't true, the above is though (AFAICS).

That means that there's two main groups of hardware/something that people fall into evenly - this can't be true though, can it?

Intel vs AMD for CPU, or Hyperthreaded vs non-hyperthreaded for CPU, or NVidia vs Radeon for GPU, or America vs Europe for location, or Western Hemisphere vs Eastern Hemisphere for location - all of these are heavily weighted to one side or another in the spring community in general, and even those categories that arent weighted in such a way across the whole community are heavily weighted to one side or another per game.

tombom · Post by **tombom** » 31 Jan 2007, 16:54

I just got a sync error twice in a row. Both times there were the same players, in a 4v4 on metalheck. I get a Sync error for tombom 3101 E both games. We carried on for a few minutes and the sync error didn't change or anything. This is with Balanced Annihilation 4.61. I did pretty much exactly the same thing both time. I have replays and stuff.

el_matarife · Post by **el_matarife** » 01 Feb 2007, 00:55

Dragon45 wrote: That means that there's two main groups of hardware/something that people fall into evenly - this can't be true though, can it?

NVidia vs Radeon for GPU,

According to the Steam hardware survey this may also be a good guess. Either AMD/Intel or Nvidia/ATI are what I'd assume it is. These may not be 100% accurate for Spring players but on here those two as well as SSE2 show up at about a 50/50 or 60/40 split http://www.steampowered.com/status/survey.html

LordMatt · Post by **LordMatt** » 01 Feb 2007, 03:22

Could different GPUs cause sync errors?

silv · Post by **silv** » 01 Feb 2007, 04:20

tombom wrote:Is it just me or are most desyncs on maps with water? Also all the desyncs I see are always exactly half the players in the game desync, rounded up. A spec might desync but it's still always the same number of players. I may just be seeing patterns that aren't there.

No and No.

Btw. for the replays, did you try to get replays from sync and desync players of the same game and compare them against each other? this might be a good or very dumb idea, i dont exactly understand all of spring's mechanics (although, i think the replays should contain the same data, afaik they are like recordings of the commands, and then this method would fail).

LordMatt · Post by **LordMatt** » 01 Feb 2007, 04:24

Tobi said that wouldn't work because on your machine, both replays would run the same.

Tobi · Post by **Tobi** » 01 Feb 2007, 09:21

Indeed. What would probably work is if I could replay one replay on hardware that desyncs from each other here locally, but I don't have enough hardware for that :-D

@LordMatt, no that shouldn't be the cause, all drawing code is unsynced. (unless there's a bug in a driver / drawing code that corrupts synced memory of course)

tombom · Post by **tombom** » 03 Feb 2007, 20:57

Just had another one - two games in comet catcher, both times the same players playing, the same 4 people desynced at the same time. It's getting a little annoying, like 2/5 games every day are desyncing for no obvious reason.

Deathblane · Post by **Deathblane** » 04 Feb 2007, 11:50

Here's an idea. When a game de-syncs a pause and then a quick poll of the players hardware could be very usefull.

Spring RTS Engine

Sync errors