Loading...
  OR  Zero-K Name:    Password:   
Title: TEAMS: All welcome!
Host: CZrankSpringiee
Game version: Zero-K v1.2.11.2
Engine version: 91.0
Battle ID: 309363
Started: 9 years ago
Duration: 32 minutes
Players: 23
Bots: False
Mission: False
Rating: None
Watch Replay Now
Manual download


Show winners



Preview
Filter:    Player:  
sort
Firepluk
desync :(
30% of all big games end in desyncs... sad
Somehow I highly suspect 2 things:
- laggers
- terraform/terrain deformations

but I doubt anyone in dev team interested in fighting laggers in big games
+0 / -0


9 years ago
Since sfire is prolly the one who has experienced the most desyncs, I'd put some weight on his words.

Terraform/deformations would be easy to rule out, by making a ZK stable where those are disabled and playing it for a week or two.

Would be a small price, imo.
+1 / -0


9 years ago
Sfire of all people should run benchmarks.
+0 / -0


9 years ago
Its not terraforming, I can almost guarantee that. One thing I did notice is that on occasion my sim would ''jump'' ahead a frame or so (projectiles keep moving then they revert back a little ways then continue going). If one of these projectiles hits something, it causes a desync. I can only imagine this happening with other people. I brought it up once a long time ago but I forgot the answer I was given as to what this is called and why it happens.
+1 / -0


9 years ago
It's similar to position interpolation in FPS games and it doesn't cause desyncs, at least not by itself.
+0 / -0
Firepluk
mmm... I remember I've read an article about similar problem of desync in another game,

it was something related with simulation goes different on some devices because different precision of floating point operations - at some point it caused desync when error in calculation was big enough(i.e. error accumulates with the flow of time).

Such explanation would fit well in why this almost always happens in big games - more units, more players, bigger device variety -> error has higher chances of appearing. Also I have a feeling that some players almost never experience such kind of errors, even when game desyncs they see the most common version of the game and can even win such games(yes sometimes such game won't be counted but it's not required - I know a lot of desynced games were recorded and caused xp/elo changes).

Probably I said something really stupid but still does it have any chance to occur in zk?
+1 / -0


9 years ago
The desyncs were fixed in later engine versions, right? If so, what exactly was fixed?

Would ofc still be nice if the desyncs in 91.0 would be gotten rid of somehow.
+0 / -0
Firepluk
The problem of all later engine version is slow game, FIranksprang you know right.
We were playing 98 a bit this morning and even in 2v2 and 3v3 games we experienced sudden lag spikes(0.2-0.3 sec) which completely made the game blocked and in general fps was lower, game seemed to be less responsive than with 91...
+0 / -0


9 years ago
Yes, but I was wondering how the desyncs were fixed in the later engines.
+0 / -0
quote:
We were playing 98 a bit this morning and even in 2v2 and 3v3 games we experienced sudden lag spikes(0.2-0.3 sec) which completely made the game blocked and in general fps was lower, game seemed to be less responsive than with 91...

The border-right solid red reminds me what to do before ... have you performed the benchmarking yet?

quote:
it was something related with simulation goes different on some devices because different precision of floating point operations - at some point it caused desync when error in calculation was big enough(i.e. error accumulates with the flow of time).

Wrong guess again: Spring uses streflop.
+0 / -0
Firepluk
9 years ago
well I just supposed something :) good that we're safe from such kind of problems.
+0 / -0


9 years ago
There's one way to find out:


Break everything until something breaks similar to the breaking desyncs.
+1 / -0


9 years ago
I have made a general thread related to this. Link it to any engine complaints. http://zero-k.info/Forum/Thread/10027?postID=110131
+1 / -0