Welcome to Keen Software House Forums! Log in or Sign up to interact with the KSH community.
  1. You are currently browsing our forum as a guest. Create your own forum account to access all forum functionality.

Multiplayer Build Feedback Thread: Rock

Discussion in 'Suggestions and Feedback' started by Drui, Nov 24, 2016.

Thread Status:
This last post in this thread was made more than 31 days old.
  1. Gwindalmir Senior Engineer

    You don't understand.
    It is my own local server. It has about 5 mods on it, most of which are mine (or supplement mine). All are already predownloaded except the one in question. The one mod causing a problem is just an old version of essentials I didn't remove. It just doesn't compile.
    I'm testing a mod, and I delete the mod before loading the server, as the server will not download updated mods correctly.

    The only difference for me is that I switched from stable to rock, to test my mod under it.

    EDIT: Deepflame spoke to me, it's possible steam has some 'unreliably reproducable bugs in there'
    Last edited: Nov 28, 2016
  2. JasonG Trainee Engineer

    ok, have you tried going into your users, appdata( you must make invisible files set to visible) then go to roaming, space engineers and delete the Mod folder in there, steam will automatically make a new mod file for you as soon as try to load a world. I had to do this to actually get rid of mods i didnt want to be subbed to any more, becuase steam doesnt acttually delete the mods. hope this helps
  3. Malware Master Engineer

    @JasonG psst... you're talking to someone with quite a lot of experience with SE and modding, and who has been contributing hugely to vanilla code, improving the mod API. He probably knows more about this than you do ;)
  4. Megachill Trainee Engineer

    Server did crash but the game was somewhat ok to play (i'll iterate below). This was tested with 2 / 3 people on a dedicated server running:

    Server: Windows Server R2012 with an i7-5960 and 64GB RAM. SE installed on Intel SSD 256GB.
    Simulation Rate: Whacky from time to time, avg > .85

    While we did not experience a lot of jitter / warping around (syncing / replication issues) the sim rate was quite acceptable. However the server would frequently crash when a bigger grid was causing physics updates. Server crashed within 1h approx 5 times and everytime it did it was caused by someone causing a physics grid update.

    tldr; better than paper but sometimes still having issues with net replication and very frequent crashes on bigger physics updates
  5. Pie Apprentice Engineer

    Perhaps not a very useful comment but after a bit more testing I definitely think Rock would be the one to go with if performance could be improved a bit. After moving the server to my old PC (i7 2600k 16Gb) most of the performance problems I had with 1-2 players and the DS and my client on my main PC went away.
    Everything always appears where it's supposed to be. I've had no inventory issues (like Scissors), no bases and ships randomly moving (but not moving in reality) like you do on Paper and Scissors. I can also move around on ships - I was walking around on a ship at 70m/s with an override in 2 directions on the gyros.

    Things that used to guarantee a visit from Klang seem to be better as well. I've not used rotors and pistons much but piston based 'ship printers' work without exploding although they still ignore parts clearly within welder range at times. Merging and un-merging large grids cause a short performance spike and sometimes need a power cycle (Y) to work once disconnected but landing gears and other sub-grids no longer explode like they used to.

    In quite a large world I've yet to experience a crash.

    Issues so far - but mostly bugs inherited from the old MP with slightly different behaviour:
    - Odd behaviour on the edge of a gravity well. At one point with inertial dampeners off I was actually accelerating upwards in a ship with dampeners and engines off.
    - It really doesn't like beacons. They seem to appear and disappear. Whenever they appear the client suffers a lag spike (although the server seems un-impacted).
    - Even on a LAN connection there seems to be some kind of 'judder smoothing' going on - what used to be a large jump is now smoothed out so I guess this will be much worse on a remote connection. I'll try and get a friend of mine to test remotely later.
    - Ships with small ships docked via landing gears are invulnerable to environmental collision damage - they just bounce off asteroids and planets. Grid - Grid collisions work as expected. This has been around in SE for ages tho and I think still happens in SP.
    • Informative Informative x 1
  6. Armada115 Trainee Engineer

    Could a Nice Person from Keen, please reboot the Rock Survival(Keen Official test - Rock) server please. although it is showing it is up, you get "The Server host has left the game" message trying to join the server.
  7. Gwindalmir Senior Engineer

    Oh, I forgot to mention, I had a really good experience:

    Rock doesn't seem to have that bad prediction teleporting error that is on stable and dev.

    The regular builds have a bug where if you teleport a player (via ModAPI in my case), it seems the prediction gets screwed up (I performed the operation on both the server alone, and server and the client together). This desync shows the player at one location, but physics interactions (block grinding/welding, gravity) behave as if the player is somewhere else. At least until it resyncs some time later.
    This causes the player to repeatedly to teleport back to a specific position after a few seconds. I found toggling the jetpack corrects this error.

    However rock didn't have this problem at all that I could see.
    So props for that!

    NOTE: This was on a LAN hosted DS with an empty world (no voxels, and 4 grids), so latency isn't an issue. It's just a game bug that began a few months ago.
    Last edited: Nov 29, 2016
  8. roophio Trainee Engineer

    created my own local online public game, no one joined. Even with only me in it, it would display connection issues in the top right corner, and even with super fast internet speeds. Wolves would appear out of nowhere with a lag spike.
  9. DerrickMehaffy Trainee Engineer

    So far the testing I have been doing on my Dedicated rock server has been fairly smooth. Few small bugs such as suddenly unable to weld or place blocks but can still move, talk, and put items in cargo. Had a bit of jittering of the ships when near the ground on planets.

    Planning to do some small ship // Large ship testing in the next day or so, and I have noticed about 7-10 people on my server consistently during the day. Seems fairly stable but will continue with more testing.
  10. terribleperson Trainee Engineer

    Okay, I decided to give rock a second shot, thanks to DerrickMehaffy and his servers. I'm pretty impressed so far. Voxels and mining work way better than Scissors (I talked more about that in detail in the scissors thread). It's been a smoother experience than Scissors on an identical server with the same two people, though we haven't done much testing with ships yet. I have experienced one minor bug so far, though. Derrick and I were unable to weld large blocks placed onto the lander-turned-station (we could place and weld blocks off of the station), but relogging fixed that.

    edit: Getting the unable to weld bug again. If it's this common, it's a showstopper.

    edit: Went to creative. Pasted in a small ship, took off. Tried to land... everything went crazy. There's a crater and 3 pieces of a small ship now. So many things went wrong I can't even enumerate them or find a single point of failure. The server started CONSTANTLY complaining about performance. That's.. really bad.

    edit again: Okay, figured it out: Pasting things just.. absolutely drives everything INSANE. There you go.

    edit FOUR: Yeah. Pasting a single ship in (it was running when it got pasted in) broke everything. Server performance went to shit, physics went crazy, me as a player bouncing everywhere, constant performance issues as long as that ship existed. This isn't a design flaw with Rock, this is something being horribly wrong somewhere. Is Rock on an older build?
    Last edited: Nov 29, 2016
  11. Jan.Nekvapil Developer Staff

    Sharing the pasted ship could help identify the issue.
  12. DerrickMehaffy Trainee Engineer

    To elaberate on the issue of unable to place/weld blocks on a grid, I noticed this happened anytime I went to place blocks that connected to the voxels on the planet. Like @terribleperson said, relogging fixed the issue. We could weld and place blocks not on that grid until after relogging.
  13. terribleperson Trainee Engineer

    I'll be on to test it again tomorrow, but I believe it was this.
  14. DerrickMehaffy Trainee Engineer

    Another issue I've noticed is grids owned by myself at a remote distance (I'm guessing outside the server view distance in this case 15k) are appearing and disappearing every few seconds and are unable to maintain a laser/radio link. I am the only player on, so I'm guessing the remote grid is being unloaded on my client side to help stability. While this is nice and all, it makes having a remote base as well as long distance radio/laser pointless in the game.

    Edit #1
    Pictures of console for grid showing and grid disappearing. Takes roughly 10 seconds for it disappear.


    Last edited: Nov 29, 2016
  15. terribleperson Trainee Engineer

    Okay, noticed something that's a problem on both Rock and Scissors. Large ship antennas have a range of 50km, correct? Despite this, at ranges below 50km (I'm usually between 40km and 45km) an antenna on an orbital com station is only intermittently visible. Yes, my signals are set to fully visible. The antenna is usually only briefly visible, less than 10% of the time, and I'm not sure what makes it become visible.
    • Agree Agree x 1
  16. piratep2r Trainee Engineer

    Everything seemed great for the first 10 minutes; mining, moving, lag all superior to scissors. Then I alt-tabbed out and opened my browser and got a "graphics card unavailable" error * may have wording wrong there. SE crashes.

    I then close the browser and rejoin (relaunching SE), and now the experience is terrible. My lander seems like it can't decide where it is, and (every second) switches position up or down about a block. This is horrifying when you are near the ground, and I watched my lander repeatedly slam itself into the ground (while landed) like a hammer for a while. Then I gave up. Will test more later!
  17. terribleperson Trainee Engineer

    I feel like I've seen that behavior somewhere but I can't remember where. I get somewhat similar behaviour on scissors sometimes when trying to land, though...

    edit: Okay, I think I've figured out part of why Rock is often reported to run way worse than scissors. I've been playing on DerrickMehaffy's rock server for like... two days now? It was smooth as silk when fresh, much smoother than a fresh Scissors server. Now, though... it's lagging. It often has server performance issues, and we're getting client side ship... jitter? Bouncing? Ships you are piloting bounce up and down awfully, but it's not visible to other players. I think I know part of the reason it's lagging.

    Collision-generated rocks (strangely, unlike the very nice way mined rocks are handled) are sometimes invisible to players. They're not interactable, but still putting a load on the server. That's one source of lag that likely just gets assigned to Rock being bad. Another more serious problem however is that I've noticed parts of wrecks interpenetrating the ground. That's cool, it looks neat. The problem is they're still acting physically, and some fall further into or through the ground. In fact, I started to scrap part of a wreck and one piece fell deeper into the ground. It's still visible when you go far enough away for the crater rendering to be simplified. Active physical objects inside the ground cannot be good for server performance.

    Near: [​IMG]

    Far: [​IMG]

    edit 2: Yep. Digging down and removing that made a difference. Server's not as good as it was at the start, but doing better.
    Last edited: Nov 30, 2016
    • Informative Informative x 1
  18. X e r o Trainee Engineer

    Rock Test -

    MP Rock Test ~
    - Downloading update 163.5MB updating from Paper.
    - Server started with no issues.
    - Joined server and immediately met with 'Multiplayer connection is unstable' in the top right, this is NOT a fix.
    - Also met with 'Server is having performance issues' despite everything being fine. this tells me nothing, this is not a fix, it is a cop-out....
    - Testing large battle. - test later
    SP Rock Test ~
    - No major issues, rotors/pistons seemed slightly more stable than paper.
    - Also had ship thruster bug where i had to press Y-Y to restart the ship to get thrusters/gyros to move.
    - De-sync from high-ping was apparent and there was plenty of rubber-banding, almost worse than current DEV/Stable versions with my ping.
    Note: I keep hearing 'Rock can take more if you have a better machine'. Better hardware should not be the basis of 'better' in these tests, SE already needs a decent CPU to run, I'm actually leaving rock 3rd in my list because of the fact it's based off of 'better hardware makes it better' That is true for the current build of SE, and is not the answer.
    - Server sim speed dropped just pasting and crashing a 5k block ship, dropped to .8 server sim speed, and fell through ground as well as kept sim speed at .8 just by the ship being present. server sim speed went back to 1.0 after ship was deleted.
    - Rock battle test between same 4-5k block ships as in paper, dropped server sim speed to .4-.5 while battle was going on.
    - aftermath of battle still left server sim speed at .8 and it did not recover to 1.0 once the ships were gone. I restarted server and it was back to 1.0 again. not exactly sure why, but paper and scissors both had no issues recovering.
    - no major crashes or issues, just overall unstable build.
    (Note: I noticed a new network graph, despite how useful this is, it didn't actually help with a higher ping and/or slower connection.)

    Conclusion: Rock seemed stable at first, but after the same testing i put paper and scissors through, it seemed to perform worse than both of them. I was advised 'you don't have good enough hardware to run rock' but the other 2 builds run better... why would I ever choose rock over a less-stressful build?

    Rock is 3rd for me,

    Main issues,
    Server sim speed and local sim speed did not recover after crashing 5k ship, + 4-5k ship battle. sim speed stuck at .6 until i deleted all ships, then it restored to 1.0
    I did test the same ships on idle, and it dropped sim speed to .8-.9, which didn't seem to bother scissors or paper that much.

    positives, are the rotors and pistons seemed to work much smoother than on any other branch, even with server sim speed changes, the piston, rotor contraption i made performed much better on rock. I'm not sure what else they added, so i can't comment on any of that besides the (sort of) useful network graph?
  19. Blazero Trainee Engineer

    painting the while grid (CTRL+Shift+middle mouse button) crashes my game, every, god damn, time.
  20. terribleperson Trainee Engineer

    No. We should not be choosing the netcode that provides adequate results given adequate hardware. I (and I assume many others) want the best POSSIBLE experience for Space Engineers, and to get that you need netcode that gives exceptional results on exceptional hardware, even if it makes the performance on adequate hardware worse. Sacrificing the possibility of great performance for adequate performance in a wider range will prevent large servers from ever taking off. If you really want to host Space Engineers on a home machine, you can. Hosting a server from home anywhere outside South Korea your server is likely to be seriously impaired by your network connection regardless of netcode, but an appropriate CPU for Space Engineers is not that expensive. Considering that extra cores don't really matter, there's a 2015 i3 with appreciable single-thread performance. A small home server for a handful of people does not require the best performance. If you want the best, do what anyone who is serious about running a game server does, and RENT. They are CHEAP, and they'll have a better internet connection than you to boot.

    edit: To cap that off, hardware is improving and will continue to improve (even though we're running up against a wall with regard to CPUs). Choosing netcode that provides adequate performance on adequate hardware at the expense of great will put a ceiling on the performance we will EVER get from Space Engineers servers, regardless of future hardware improvements. The CPU with the best single-thread performance for two and a half years was just dethroned this quarter and the new one (probably, it's not priced for sure yet) gets you more performance per dollar.
    Last edited: Nov 30, 2016
    • Agree Agree x 2
  21. X e r o Trainee Engineer

    The machine I ran this test on was a DS, with what i would call adequate hardware, only because it's a server processor, however it was one of the more high-performance server processors.
    Anyway, in regards to "I (and I assume many others) want the best POSSIBLE experience for Space Engineers, and to get that you need netcode that gives exceptional results on exceptional hardware, even if it makes the performance on adequate hardware worse"....
    While i agree most want the best possible experience for SE, putting the focus on 'exceptional hardware' and basically saying adequate isn't even really adequate... is how you drop a large portion of your player base quickly.
    It's like if you only supported AMD cards and not nvidia cards, you'd have people constantly complaining about support. So, there's no way in hell they're doing anything like that.
    Say you were to say an i7 is exceptional, and an i3 was adequate, lots of servers run on high-grade xeon's because they're from shared machines. While mine wasn't a shared, it's a Xeon 3.40GHz CPU, more things are being set into parallelization, so that's always good.
    anyway, i agree and all of us want the best performance, but saying 'adequate' hardware should suffer at the expense of making the 'exceptional' better, is just wrong.
    I'd wager a couple server owners would even stop renting as well. SE just needs to be optimized, the performance should increase given increased hardware, not degrade.

    I'm too tired to do this :woot:
  22. Pie Apprentice Engineer

    So far so good - found some minor bugs although the last two may not be a bug:
    - Changing the order of ores in an Refinery queue sometimes refuses to update in the GUI if there's Uranium in the queue. Works fine with any other material. A re-log fixes this.
    - Rag dolls are broken - they teleport around. Eventually the bag will appear allowing items to be recovered tho.
    - Weird LCD issue I couldn't re-produce easily - I had a 'blank' LCD update with a copy with what was on the one next to it. Turn away and look back and it's fixed.
    - Small ship drills - they seem far more effective than I remember! I've not played survival in a while but I seem to recall mining in gravity being very difficult - you had to carefully right-click drill your tunnel then mine away slowly at the ores remembering to leave a big enough gap to rotate and get out. Now, right-click on a small ship with 3 drills easily produces a cavern large enough for the entire ship to fit through. Even left click seems far more effective - I managed to fill my ship to the point it was too heavy to fly in less than a minute!
    - Projected batteries aren't constructed with the usual half-charge (although I seem to remember this being an issue in SP Survival).
  23. terribleperson Trainee Engineer

    @Xero: You're treating i7s and i3s like they can be compared interchangeably, and acting like Xeons aren't good. Noting that the most important criteria for Space Engineers are single-thread performance and RAM (which you can neglect entirely because it's dirt cheap and modern RAM doesn't have performance issues), your remarks make no sense. They make less sense when you consider that there are seven year old i7s which are terrible for Space Engineers, and that 9 of the top 20 CPUs for single-threaded performance according to PassMark are Xeons. #20 is a 6th generation i3. The only Xeons that don't score well are either old (5+ years), low power SoC/IoT hardware (Xeon D), or cheap, EOL, AND low-power.

    Any responsible reasonable server host WILL be using Xeons or other server-class processors, and they will not be using low-power or beyond end of life processors for hosting game servers. Also, saying that SE "just needs to be optimized" ignores the complexity of optimization, and the sacrifices that are often made in the name of performance.

    Now for some more productive feedback.
    Note: We have a speed limit mod enabled. That's not officially endorsed, of course, but while physics problems are expected, if it's causing problems like this... something is broken somewhere.
    I've been having some bad client/server desync on Rock, in addition to the usual player enter/exit lag and saving lag. Yesterday, I built a new version of a small ship I'd been working on. I powered it up, disabled the landing gear, double checked everything. Tried to take off, and it didn't respond immediately. It did take off after a few seconds, though. I then put it back down and noticed my landing gear weren't able to lock on. I jumped out, and appeared a good 50 - 100 meters off in the distance, in the air (the ship was build on a platform off of the ground). I flew back, got back in, and tried to land it again and turn it off so I could figure out what's up. After pushing it into the platform for a second... it crashes into the ground a hundred meters in the distance, where I had appeared previously.

    I build a new one, because I've learned by now to save blueprints. I go to fly it out of the atmosphere (there's a funny story there) and notice that the game can't display my speed in a stable or accurate manner. This seems to get worse as I start leaving the atmosphere and engage my ion thrusters. It also stopped responding to control inputs (or maybe it just seemed like it, because the speed gauge may have been lying to me). I eventually stop it with the space master tools and restart all of the ship parts, and start flying it back. Speed display issues continue, as do the problems with control even when I'm back in the atmosphere. Despite this, I manage to pilot it nearly back to the base (within 500 meters) after having been more than 45km away, and then the game crashes. I hop back on and try to find it (or the wreckage) with the space master tools. I eventually find the wreckage.... thirty KILOMETERS from the base.

    I'm going to go play on dev for a bit, because Scissors is ...okay, and Rock was really smooth at first but as we play more has become a buggy mess. I really WANT Rock to be good, because it's clear it has potential. As it is right now though, it would make Space Engineers even worse.

    P.S. The desync might have a warning sign. Before crashing the ship for the first time, I noticed that the small grid projector mounted on the small grid rotor head attached to a large grid rotor had somehow become disconnected, and was sitting halfway into the top of the rotor without any power. I'd seen this an hour before, and had repaired it. The second time (before crashing) I ignored it, because I was busy messing with my ship. When I relogged after the desync induced crashing, it was all back in one piece the way it was supposed to be. It had never broken in the first place.
  24. X e r o Trainee Engineer

    Processor's can be compared, because an i7 will blow an i3 out of the water, surely you agree with this. I personally have an i7 4790k on my computer and i get better performance than the newer 6700k processor, even the passmark scores are beyond the Xeon for single-core performances, but that should be obvious, with these being high-end processors anyway. My server has a Xeon, 3.40Ghz as noted before, so in no-way was I even close to saying that the Xeon's are bad, I am simply stating that they don't have the performance of an i7 when used for running a server...again, obviously, because of the higher single-core performance.

    I wouldn't say they're cheaper, that's more of a personal standpoint, because most server owners rent from a datacenter, as in the end it's cheaper and you've got the benefit of a Dedicated machine + a gigabit connection. That being said... what you said previously makes it appear that there is some magical mid-way that you can optimize a game and through optimization, make it worse somehow. No developer in their life would do something like that. The only case this would be true is where the 'requirements' are high for a game when it releases, because it has massive graphical scales. I'm not even sure what the 'optimal' equipment would be for SE just from a Single Player standpoint, or from a server standpoint, because due to it's single-core performance dependancy... really it doesn't need too much for servers. But, back to what I mentioned, it just needs optimized.

    I do agree with your next note on if something happens with a mod, there's something wrong in-game, like with speed.
    (note: after listening to the Rock team (physics guy) he actually was attempting to optimize the game through stressing it, and improving it on one set of hardware, that wasn't top of the line at all, which I completely agree with!)
Thread Status:
This last post in this thread was made more than 31 days old.