As an aside, I was very surprised to read they haven't even started on their server mesh technology, instead they are focusing on the single server performance and optimisation (lol).
That's incorrect, the object containers are the server mesh pretty much, it's one ongoing work for quite a while, the object container streaming (the 3.1 stalled one) is from what I see what localizes the PU game world and will start to allow multiple servers to stream to each other.
Back to form I see. Telling people they're incorrect without having any idea what you are talking about.
Clive Johnson CIG@cjohnson October 14th at 02:35 pm
To give you an update on the specific technologies you asked about :
Server meshing - not started yet. Our plan was always to make the single-server experience better and more optimized first.
Server meshing is going to build on the technologies we're creating for
single servers, so these all need to be in place before we can start.
Also it is going to be challenging and complex work that will need the
focus of the whole network programming team, so once we start work on it
we don't want to be fighting a war on two fronts.
Back to form I see. Telling people they're incorrect without having any idea what you are talking about.
Clive Johnson CIG@cjohnson October 14th at 02:35 pm
To give you an update on the specific technologies you asked about :
Server meshing - not started yet. Our plan was always to make the single-server experience better and more optimized first.
Server meshing is going to build on the technologies we're creating for
single servers, so these all need to be in place before we can start.
Also it is going to be challenging and complex work that will need the
focus of the whole network programming team, so once we start work on it
we don't want to be fighting a war on two fronts.
Oh that, nevermind I confused the containers with binding now on the server mesh itself, on what are the network backbones for it.
I don't know how many times CIG have to explain that the server can't provide better performance for the clients until they implement the new netcode they've been working on for a very long time.
Originally, most of that was supposed to be in 3.0 - but it was postponed for 3.1 I believe.
Meaning, even if there's going to be SOME performance improvement in 3.0 (LIVE version) - it won't be until 3.1 that we're going to see how the planned netcode integration will improve the performance.
Also, it won't be until they have a solid frame-rate that they expand the player count.
This, too, has been made public several times - and yet it seems the detractors are still conveniently forgetting and pretending that this early Evocati build is, somehow, representative of the best CIG can do.
It's the same old bullshit
The Evocati performance is to be taken with grains of salt because they purposely put things as servers in debug mode, to be able to track down issues, it's also pretty much how QA themselves test the game, that's why you see them playing at 10FPS constantly.
A fact often ignored here that's true. We will see how this process flows and what they will achieve once this reaches live.
Sorry Max, I have seen this 'debug mode' used by devs so many times and it has never amounted to more than 1 or 2fps difference if anything at all in my experience!
While the low performance, in this case, probably isn't due to any debug mode - you definitely haven't seen much if that's your experience.
I've played countless games - both in and out of testing - that have had debug/developer modes, and it DOES tend to make a massive difference for performance compared to release/normal modes.
Countless? well for me it could be guessed in the range of 10 - 15 and they all just seemed like poor excuses for bad performance and like i said following the introduction of the 'none debug mode' there was hardly and change in fps. Perhaps I only ever encountered lies? What performance increases have people seen then because I still don't believe it :P
Orinori said: Countless? well for me it could be guessed in the range of 10 - 15 and they all just seemed like poor excuses for bad performance and like i said following the introduction of the 'none debug mode' there was hardly and change in fps. Perhaps I only ever encountered lies? What performance increases have people seen then because I still don't believe it :P
That's because you didn't see how bad the performance of Alpha 2.6 was during the Evocati, when they do they actually do this.
Now when you look back at PTU/Live 2.5 to 2.6, then almost no difference as there was little to no changes of relevance to the FPS in the network front.
I don't know how many times CIG have to explain that the server can't provide better performance for the clients until they implement the new netcode they've been working on for a very long time.
Originally, most of that was supposed to be in 3.0 - but it was postponed for 3.1 I believe.
Meaning, even if there's going to be SOME performance improvement in 3.0 (LIVE version) - it won't be until 3.1 that we're going to see how the planned netcode integration will improve the performance.
Also, it won't be until they have a solid frame-rate that they expand the player count.
This, too, has been made public several times - and yet it seems the detractors are still conveniently forgetting and pretending that this early Evocati build is, somehow, representative of the best CIG can do.
It's the same old bullshit
The Evocati performance is to be taken with grains of salt because they purposely put things as servers in debug mode, to be able to track down issues, it's also pretty much how QA themselves test the game, that's why you see them playing at 10FPS constantly.
A fact often ignored here that's true. We will see how this process flows and what they will achieve once this reaches live.
Sorry Max, I have seen this 'debug mode' used by devs so many times and it has never amounted to more than 1 or 2fps difference if anything at all in my experience!
While the low performance, in this case, probably isn't due to any debug mode - you definitely haven't seen much if that's your experience.
I've played countless games - both in and out of testing - that have had debug/developer modes, and it DOES tend to make a massive difference for performance compared to release/normal modes.
Countless? well for me it could be guessed in the range of 10 - 15 and they all just seemed like poor excuses for bad performance and like i said following the introduction of the 'none debug mode' there was hardly and change in fps. Perhaps I only ever encountered lies? What performance increases have people seen then because I still don't believe it :P
I have no idea how many debug modes I've encountered - but it's certainly enough to have an informed opinion about them.
That said, I'm not terribly engaged by the prospect of trying to convince someone who claims to have tried at least 10 debug modes and none of them made any significant difference to performance.
That sounds like trying to convince Trump that his inauguration wasn't the most impressive of all time
All the same, it's hardly relevant here - as I'm pretty sure the performance issues in the Evocati build are about something else.
I didn't think ultra high details can even be done in an MMO without massive performance hits.
"We all do the best we can based on life experience, point of view, and our ability to believe in ourselves." - Naropa "We don't see things as they are, we see them as we are." SR Covey
I didn't think ultra high details can even be done in an MMO without massive performance hits.
It's not the shiny visuals that hit the MMO factor.
I mean they can, but that is more related to your rendering performance, say go a city in a crowded server in BDO, though the rest of the game always performs fine.
Though even in visuals you can still cut, say when many players gather up in an area, things like it doesn't load textures in high res, or has a limit, limited effects, etc...
If you play GW2 you've come across this on WvW, remember the Zergs of invisible players that you only saw the name tags? xD
Of course ultra high details can be done - it has nothing to do with it being an MMO. Theoretically, anything that can be done in a singleplayer game can be done in an MMO, in a visual sense.
It all comes down to how you economize the visuals and how you scale down when more is happening on screen.
The challenge of Star Citizen and performance is less about visual fidelity - and much more about the fidelity of the interaction and the playfield.
They're going all-in when it comes to detail and interaction - and they want all relevant players to experience that for the sake of immersion.
This is what's putting such a gigantic load on the server and the bandwidth. They're sharing a TON of completely unnecessary information between server and clients and the moment.
Visually, you can be sure it's going to be ahead of the competition.
I believe for a time they were selected based on financial contributions and bug report participation. Regardless, the game is early access alpha. Everyone in alpha testing should have the ability to play on the stable testing branch or the experimental branch for 3.0.
Please explain why everyone should have that ability.
I'm genuinely curious what "right" you think people have?
Here I always thought that a developer could roll out builds any way that they choose, for any specific goals that they set. Or maybe you think that the 3.0 invalidates any other piece of information that they gather from people who don't have that build. If someone doesn't like the way that a developer is rolling out the Alpha, I suppose they could just not participate. No? Everyone's gotta have the exact same thing? If it was a political discussion, I'd call it Liberal Logic at it's best.....
I still don't understand having earlier access to early access.
Because it is not Early Access ?
It is voluntary player testing of a buggy Alpha build. And sending in bug reports.
It is a common misconception that the Evocati are there "to play the game earlier than others". If you think that, you have not yet understood what the Avocados do.
Have fun
I believe for a time they were selected based on financial contributions and bug report participation. Regardless, the game is early access alpha. Everyone in alpha testing should have the ability to play on the stable testing branch or the experimental branch for 3.0.
Then again, It's possible that I don't know what the hell I'm talking about, but the last time I tried to get access to 'earlier access' I couldn't. I enjoy playing SC when major version changes are patched in.
One last thing, lets not kid ourselves. When 3.0 is released to everyone else, it will still be a bug ridden Alpha build no matter how many weird fruit groups tested it first.
No, it's always been based on server activity and dedication to testing.
The reason not everyone should have access to it, is because of people like you. People who don't understand what early testing of unfinished software means - and how it will play.
People who would bitch endlessly about how terrible it is - and who wouldn't contribute anything of value.
It's the same reason a person writing a book doesn't actually release it to the whole world before it's been exposed to an editor.
You want to make the best impression and you want to make people happy.
You don't want 2 million backers hating the game they've been paying for because it's completely broken.
I didn't think ultra high details can even be done in an MMO without massive performance hits.
It's not the shiny visuals that hit the MMO factor.
I mean they can, but that is more related to your rendering performance, say go a city in a crowded server in BDO, though the rest of the game always performs fine.
Though even in visuals you can still cut, say when many players gather up in an area, things like it doesn't load textures in high res, or has a limit, limited effects, etc...
If you play GW2 you've come across this on WvW, remember the Zergs of invisible players that you only saw the name tags? xD
I remember gamers use to crash servers by having everyone go to one location. I think the number of players in game can make a difference in game quality.
Guess I'll believe it when I see it. An EvE type large space battle.
"We all do the best we can based on life experience, point of view, and our ability to believe in ourselves." - Naropa "We don't see things as they are, we see them as we are." SR Covey
I remember gamers use to crash servers by having everyone go to one location. I think the number of players in game can make a difference in game quality.
Yeah but they are already limiting that here because won't allow everyone to feel like to be in the same place at the same time as stated.
Though this still applies with ships, have one Idris with ships docked inside, fight another Idris with ships docked inside and a crew, it'd just 15-20 players in that situation to be enough to force visual cuts based on what your hardware can handle.
It's very noticeable in games like BDO because they do it badly, the popping and the "clay figures" with no textures loading in are too noticeable.
There will never be EvE style battles of hundreds of ships - because the permutations of required shared information would be prohibitive. But it won't be due to visual fidelity.
Also, EvE style battles are already slidehows even on high-end hardware, so....
Realistically, we might see a few dozen ships with multiple players in each of them in close proximity. That might still mean a hundred players, though.
Chris and Erin have both said they want thousands, if not hundreds of thousands in one instance. They honestly believe that their as-yet undeveloped server mesh will be able to do what no one else has done at a graphical quality higher than all other MMOs.
Here's Erin's quote from earlier this year:
So with the next big release a lot of the underlying game is there and
then we can look at transferring people between servers so we can have
hundreds of thousands of people maybe in one instance, but that doesn’t
come online until later.
Weren't you or someone say CIG modified their version of cryengine by over 50%? I wouldn't say that's "some modifications"
They hired the Germans who made the CryEngine and tasked them with modifying it to a great extend, practically rewriting it. So yes, the notion "some modifications" is quite false.
Reading about how a GTX 980 struggles with 3.0, debug mode or not, makes me quite pessimistic about my GTX 970. It is almost certain that a 970 won't be able to handle the game in a decent manner and by decent i mean good visuals plus a frame rate of 40+. This is unexpected and for me unacceptable. Not everyone has the financial potential for a GTX 1070'like GFX.
Reading about how a GTX 980 struggles with 3.0, debug mode or not, makes me quite pessimistic about my GTX 970. It is almost certain that a 970 won't be able to handle the game in a decent manner and by decent i mean good visuals plus a frame rate of 40+. This is unexpected and for me unacceptable. Not everyone has the financial potential for a GTX 1070'like GFX.
A GTX 1070 will struggle as much as the GTX 980 in SC when the servers struggle; again it's not about high-end hardware to achieve performance.
The difference on GPUs is how much FPS you get when the servers are fine, when they are not, everyone will be taken under 20/30 frames.
Well, from the announcement it has been known that SC is for enthusiasts - and that goes for the hardware as well.
While they will optimise it - I certainly wouldn't expect the release version to perform well on a GTX 970.
You have only yourself to thank for that, if you didn't listen when CR specifically said he wanted to push the boundaries and what it meant in terms of requirements.
I think one problem is that people don't seem to understand what "an instance" means. It doesn't mean "in close proximity with no scaling".
A WoW world server is "one instance" with 20K players. That doesn't mean 20K players can stand right next to each other and not be a performance issue.
It simply means you can potentially interact in one way or the other.
If you seriously think it means hundreds of thousands of players in close proximity - visible to each other - then, quite clearly, you have no clue what they're talking about.
While they will optimise it - I certainly wouldn't expect the release version to perform well on a GTX 970.
I disagree.
The GTX 970 already performs over +40FPS in max settings on SC, this without the game being optimized yet, and while they surely will do graphical upgrades here and there, it's crazy to say the card won't work well for the game, unless you are "ultra duper settings or nothing" type of guy.
While they will optimise it - I certainly wouldn't expect the release version to perform well on a GTX 970.
I disagree.
The GTX 970 already performs over +40FPS in max settings on SC, this without the game being optimized yet, and while they surely will do graphical upgrades here and there, it's crazy to say the card won't work well for the game, unless you are "ultra duper settings or nothing" type of guy.
Yeah, the game with a tiny handful of players and no capital ships to speak of - and 90% of the features missing can perform reasonably well right now in Arena Commander.
I'm talking about the game as it's planned - and how hardware expectations will have changed at that time.
Yeah, the game with a tiny handful of players and no capital ships to speak of - and 90% of the features missing can perform reasonably well right now in Arena Commander.
I'm talking about the game as it's planned - and how hardware expectations will have changed at that time.
But, let's see how well 970 handles the MVP
I will say those features will cost more to your CPU than they would to your GPU.
It's also a big reason why people with beefy GPUs complain of performance on games that are heavy in simulation/physics.
rpmcmurphy said: These guys are living in cuckoo land.
Are they? Look at what Dual Universe is doing with theirs. Hm
Well DU are designing the game with the networking technology right from the start instead of building it around single server stuff and then trying to expand it into large mesh servers.
It's not impossible. Load balancing is no longer an ultra exclusive and super expensive solution, especially if they have Amazon's support. The only problem I see them needing to solve is the amount of data sent/received from players.
I agree it's not impossible but I do think they are way outside the reality of what they will be able to handle, making comments like hundreds of thousands in one instance when they don't even have any sort of prototyping in place is just foolhardy.
I think one problem is that people don't seem to understand what "an instance" means. It doesn't mean "in close proximity with no scaling".
A WoW world server is "one instance" with 20K players. That doesn't mean 20K players can stand right next to each other and not be a performance issue.
It simply means you can potentially interact in one way or the other.
If you seriously think it means hundreds of thousands of players in close proximity - visible to each other - then, quite clearly, you have no clue what they're talking about.
Here we go with putting words in people's mouths. It was Erin that used the word instance, really he should have said X number of players on our mesh server. A WoW server is a server, an instance is a spawned area requested by that server, it might be a dungeon, a raid or a quest instance.
This was never about hundreds of thousands in close vicinity, everyone knows that would not be possible with our current tech.
The point you seem to be missing is that if they are struggling to get >10 people on one server how the hell are they going to manage getting hundreds of thousands in one instance.
Can anybody of the experts explain to me please what this "Debug Mode" that is eating away many FPS is doing? And why is it only in the build of the first evocati test group but not in the other Alpha test groups?
@DKLond anything that can be done in a singleplayer game can be done in an MMO, in a visual sense.
While this is true though higher visuals will have an impact on network performance if the client is lagging (Video Lag)
When you have cake, it is not the cake that creates the most magnificent of experiences, but it is the emotions attached to it. The cake is a lie.
Well DU are designing the game with the networking technology right from the start instead of building it around single server stuff and then trying to expand it into large mesh servers.
That's because DU already did the single-server stuff, you don't have multiple servers sustaining the game world and scaling before you have the actual individual servers that will do that worked out for sure.
The point you seem to be missing is that if they are struggling to get >10 people on one server how the hell are they going to manage getting hundreds of thousands in one instance.
Okay I think there's a misunderstanding here of what is meant as one instance, not as a "single-server" as it is now, that instance would already be the mesh of servers. So say this applies better for the word other MMO's use: Regions/Channels (individually run copies of the same world).
Can anybody of the experts explain to me please what this "Debug Mode" that is eating away many FPS is doing? And why is it only in the build of the first evocati test group but not in the other Alpha test groups?
Server performance pretty much ties to the performance of the clients, hence if servers are on debug they are cut on performance that will impact the clients that way.
That's why server memory leaks also cause player FPS issues, noticing how the server degrades over time (they confirmed those are mostly memory-related at least).
And because that's the point of Evocati? The heavy more localized debugging and stress testing, those builds are not meant to be played, yet tested.
Comments
Back to form I see. Telling people they're incorrect without having any idea what you are talking about.
To give you an update on the specific technologies you asked about :
Server meshing - not started yet. Our plan was always to make the single-server experience better and more optimized first.
Server meshing is going to build on the technologies we're creating for single servers, so these all need to be in place before we can start.
Also it is going to be challenging and complex work that will need the focus of the whole network programming team, so once we start work on it we don't want to be fighting a war on two fronts.
Now when you look back at PTU/Live 2.5 to 2.6, then almost no difference as there was little to no changes of relevance to the FPS in the network front.
That said, I'm not terribly engaged by the prospect of trying to convince someone who claims to have tried at least 10 debug modes and none of them made any significant difference to performance.
That sounds like trying to convince Trump that his inauguration wasn't the most impressive of all time
All the same, it's hardly relevant here - as I'm pretty sure the performance issues in the Evocati build are about something else.
"We all do the best we can based on life experience, point of view, and our ability to believe in ourselves." - Naropa "We don't see things as they are, we see them as we are." SR Covey
I mean they can, but that is more related to your rendering performance, say go a city in a crowded server in BDO, though the rest of the game always performs fine.
Though even in visuals you can still cut, say when many players gather up in an area, things like it doesn't load textures in high res, or has a limit, limited effects, etc...
If you play GW2 you've come across this on WvW, remember the Zergs of invisible players that you only saw the name tags? xD
It all comes down to how you economize the visuals and how you scale down when more is happening on screen.
The challenge of Star Citizen and performance is less about visual fidelity - and much more about the fidelity of the interaction and the playfield.
They're going all-in when it comes to detail and interaction - and they want all relevant players to experience that for the sake of immersion.
This is what's putting such a gigantic load on the server and the bandwidth. They're sharing a TON of completely unnecessary information between server and clients and the moment.
Visually, you can be sure it's going to be ahead of the competition.
Why? Well, it's simple. They're deliberately targeting high-end PC-only hardware.
They're the only developer of an AAA game to do that in a very long time. It's actually completely unprecedented in the modern gaming industry.
Squadron 42 will demonstrate that the performance issues are almost entirely network related.
I'm genuinely curious what "right" you think people have?
Here I always thought that a developer could roll out builds any way that they choose, for any specific goals that they set. Or maybe you think that the 3.0 invalidates any other piece of information that they gather from people who don't have that build. If someone doesn't like the way that a developer is rolling out the Alpha, I suppose they could just not participate. No? Everyone's gotta have the exact same thing? If it was a political discussion, I'd call it Liberal Logic at it's best.....
The reason not everyone should have access to it, is because of people like you. People who don't understand what early testing of unfinished software means - and how it will play.
People who would bitch endlessly about how terrible it is - and who wouldn't contribute anything of value.
It's the same reason a person writing a book doesn't actually release it to the whole world before it's been exposed to an editor.
You want to make the best impression and you want to make people happy.
You don't want 2 million backers hating the game they've been paying for because it's completely broken.
Guess I'll believe it when I see it. An EvE type large space battle.
"We all do the best we can based on life experience, point of view, and our ability to believe in ourselves." - Naropa "We don't see things as they are, we see them as we are." SR Covey
Though this still applies with ships, have one Idris with ships docked inside, fight another Idris with ships docked inside and a crew, it'd just 15-20 players in that situation to be enough to force visual cuts based on what your hardware can handle.
It's very noticeable in games like BDO because they do it badly, the popping and the "clay figures" with no textures loading in are too noticeable.
Also, EvE style battles are already slidehows even on high-end hardware, so....
Realistically, we might see a few dozen ships with multiple players in each of them in close proximity. That might still mean a hundred players, though.
Here's Erin's quote from earlier this year:
http://wccftech.com/star-citizen-exclusive-interview-erin-roberts/
These guys are living in cuckoo land.
Reading about how a GTX 980 struggles with 3.0, debug mode or not, makes me quite pessimistic about my GTX 970. It is almost certain that a 970 won't be able to handle the game in a decent manner and by decent i mean good visuals plus a frame rate of 40+. This is unexpected and for me unacceptable. Not everyone has the financial potential for a GTX 1070'like GFX.
A GTX 1070 will struggle as much as the GTX 980 in SC when the servers struggle; again it's not about high-end hardware to achieve performance.
The difference on GPUs is how much FPS you get when the servers are fine, when they are not, everyone will be taken under 20/30 frames.
While they will optimise it - I certainly wouldn't expect the release version to perform well on a GTX 970.
You have only yourself to thank for that, if you didn't listen when CR specifically said he wanted to push the boundaries and what it meant in terms of requirements.
A WoW world server is "one instance" with 20K players. That doesn't mean 20K players can stand right next to each other and not be a performance issue.
It simply means you can potentially interact in one way or the other.
If you seriously think it means hundreds of thousands of players in close proximity - visible to each other - then, quite clearly, you have no clue what they're talking about.
The GTX 970 already performs over +40FPS in max settings on SC, this without the game being optimized yet, and while they surely will do graphical upgrades here and there, it's crazy to say the card won't work well for the game, unless you are "ultra duper settings or nothing" type of guy.
I'm talking about the game as it's planned - and how hardware expectations will have changed at that time.
But, let's see how well 970 handles the MVP
It's also a big reason why people with beefy GPUs complain of performance on games that are heavy in simulation/physics.
I agree it's not impossible but I do think they are way outside the reality of what they will be able to handle, making comments like hundreds of thousands in one instance when they don't even have any sort of prototyping in place is just foolhardy.
Here we go with putting words in people's mouths. It was Erin that used the word instance, really he should have said X number of players on our mesh server.
A WoW server is a server, an instance is a spawned area requested by that server, it might be a dungeon, a raid or a quest instance.
This was never about hundreds of thousands in close vicinity, everyone knows that would not be possible with our current tech.
The point you seem to be missing is that if they are struggling to get >10 people on one server how the hell are they going to manage getting hundreds of thousands in one instance.
anything that can be done in a singleplayer game can be done in an MMO, in a visual sense.
When you have cake, it is not the cake that creates the most magnificent of experiences, but it is the emotions attached to it.
The cake is a lie.
Okay I think there's a misunderstanding here of what is meant as one instance, not as a "single-server" as it is now, that instance would already be the mesh of servers. So say this applies better for the word other MMO's use: Regions/Channels (individually run copies of the same world).
That's why server memory leaks also cause player FPS issues, noticing how the server degrades over time (they confirmed those are mostly memory-related at least).
And because that's the point of Evocati? The heavy more localized debugging and stress testing, those builds are not meant to be played, yet tested.