Guild Wars 2

boneduste: I believe the reason sPVP got fixed so fast is probably because of scale. What scales well for a few doesn’t always scale well for the masses. So sPVP was easy to resolve because at any time there are only X number of players to render. Let X be 16 or 32 or 64 this is easier on the server / client to render. However now lets talk WvWvW which can have a maximum of 2,000 players world wide and lets say 500 per map. Lets also say that there are now 200 players around which makes the easy calculations more taxing on the Server/Client relationship. (Note: I am not saying your system is rendering 2000 players, but hopefully providing some insight.)

If they are working on it then lets all be patient and let them correct it.

However fixing these types of things take time why? Well its like rebuilding a car engine, you must first tear the engine down fix the problem then rebuild it. Easy for a mechanic, not the average person.

Now lets talk about function points. Everything that touches said fix needs to be looked at, spec’d out, fixed, unit tested and QA’d.

See how this can add up into a larger effort than addressing it for a smaller subset?

“Z”

What I find silly is people complaining about portal bombing… How did they see you to kill you? Aoe nothing? I’ve seen culling issues but mostly minor things. It exists, but I feel people are exaggerating a bit, though I understand different Internet/rigs

Joiry.2504:

I’m a programmer and I certainly understand changing features or adding new functionality can be difficult and takes time. Even though I work on relatively small applications for academic analysis, I’ve gotten my share of requests for features that would take far too much time to implement.

However, Anet has to realize, players are playing the game now. I’m sure you do realize this, but I just want to reinforce the point. While a good long term solution is great, you guys are just leaving the wvw game in a bad state right now, turning off players…ie customers.

Heck, if you need to break out the old GW1 april fool’s stick figure models for culled players, do so. Even if there’s not the badwidth to tell me what those players are specifically doing, I’d at least like to know they are there.

Probably the best post on this I’ve seen in a while, with the actual point being made. Future fixes and long term fixes aren’t the same. There’s a strong need for this to be dealt with now. Like right now.

can most part of the question “when will the culling be fixed” be answered with “when the financial department agrees with server hardware and/ or network bandwith upgrades” ?

Habib Loew.6239:

We currently use server-side report culling to limit the number of characters that any given game client is aware of.

Assuming the server keeps track of which entities it has reported to each client (does it?), couldn’t it simply make the client immune to damage from entities that haven’t been communicated to the client yet?

This isn’t a perfect fix (there really isn’t one, since data spikes can always occur), but would at least make things a bit fairer and reduce the “trampled by a herd of invisible enemies” syndrome.

Habib Loew.6239:

There are also some client-side issues which have contributed to the perception of how our culling system works. Once a character is reported to a given client there’s a non-zero amount of time required to load and initially display the assets associated with that character.

I had a feeling that was the case (and mentioned it in a couple of long threads about the issue).

But why not simply render those players with a default model (which is always cached, loaded as soon as players go into WvW – or even as soon as they start the game) until the full “correct” model is available? Seems like a pretty obvious situation to apply rendering fallback, no? I know this could lead to a bit of model popping, but surely that’s preferable to models popping out of nowhere long after they should have been visible (and long after they’ve been affecting the battle).

On a related note, GW2’s engine seems to have pretty bad aset(*) caching . Often I teleport between two waypoints (within the same city, even), and have to sit through a 20-second loading screen each time (I run the game on a 700MB/s array – I can only imagine if it was running on a normal HDD). All while 85% of my system’s memory remains unused. Same thing when I switch between two characters who happen to be at exactly the same location (ex., both in the LA bank). Surely 99% of the assets required to render the scene with the 2nd character were already loaded, and are being reloaded unnecessarily.

Are the fixes you describe also designed to address this, or likely to improve it?

(*) Had to write it with a single “s” otherwise it gets replaced with “kitten”.

muylaetrix.2096:

can most part of the question “when will the culling be fixed” be answered with “when the financial department agrees with server hardware and/ or network bandwith upgrades” ?

Unless they can also upgrade the hardware and ISP connection of every player, as well as the hardware and bandwidth of every node between every player and the game server, no.

It can never be 100% “fixed” because there’s always the possibility that the amount of data generated by a large battle will exceed the bandwidth and processing available to the server and / or some of the clients and / or some of the intermediate nodes.

But it can be improved and made “fairer” by changing a few things (see my post above this one).

Habib Loew.6239:

Extra load time varies depending on how beefy the client machine is (those with more memory, faster CPUs, more CPU cores, and faster drives experience shorter load times). One of our engine programmers recently completed an optimization pass on the character loading process and so we should be seeing improvements to that part of the issue very soon.

For those of us that do a lot of WvW, (and often nothing else except stand around in Lions Arch queues!) is there a rough baseline machine spec we could look at using which will assist in getting us a better culling rate?

Account.9832:

muylaetrix.2096:

can most part of the question “when will the culling be fixed” be answered with “when the financial department agrees with server hardware and/ or network bandwith upgrades” ?

Unless they can also upgrade the hardware and ISP connection of every player, as well as the hardware and bandwidth of every node between every player and the game server, no.

It can never be 100% “fixed” because there’s always the possibility that the amount of data generated by a large battle will exceed the bandwidth and processing available to the server and / or some of the clients and / or some of the intermediate nodes.

But it can be improved and made “fairer” by changing a few things (see my post above this one).

intermediate bandwith : in 2001, ’02, ’03 intermediate nodes were an issue for a lot of players, true. now this should no longer be an issue for people in Central Europe playing on an European server. these days, if you log in from a banana republic via dial-in, too bad if you expereince some lag/warping, for almost every other user, it is a non issue.

client bandwith : i think it’s safe to asume that 95% of the players no longer play with an analogue modem and that the downstream i received from Anet has never come close to 1 % of my downstream capacity (a modest 15mbit). for almost every user, this is a non issue

Server side bandwith/capacity : the thing Anet will never ever discus.

cllient side rendering : that’s where graphics options are for.

muylaetrix.2096:

can most part of the question “when will the culling be fixed” be answered with “when the financial department agrees with server hardware and/ or network bandwith upgrades” ? […] Anet has never come close to 1 % of my downstream capacity (a modest 15mbit).

So what you are saying is that, if there are any delays in battles with less than 100 players (and there are), then Arena Net’s servers’ connection must have lower bandwidth than yours.

I have a feeling that’s not the case. And I have a feeling you don’t have much experience writing (semi-)real-time multi-client netcode or profiling the impact of data spikes on things like firewall CPU usage, etc..

In short, even a small increase in allowable bandwidth peaks can have a significant impact on performance and network choke (which can require resyncing, and so on), and there’s very little that the “financial department” can do to fix the WvW issues (other than maybe headhunting some programmers with more experience in the field – but it would probably take them a while to find their way around GW2’s current code, so that’s not a short-term solution, either).

It’s mainly a software (netcode / game architecture) issue. Throwing more hardware at it on Arena Net’s end wouldn’t magically fix things.

A good way to fix this would be to change the stat modifiers in WvW.

For instance: Make AOE buffs when your in your party affect party members only. And make them much much stronger, so strong that a good 5 man party can take out a 20 man zerg. If a person Stats for offense exagarate those stats compared to how it is now and reduce his healing and defense from his own stats so that he will be requred to get buffs from other players to survive on his own. And if a person stats for defense or healing buffs let those people have such weak offense they wont be able to kill anything without an offensive player.

This would do several things. It would force coordination, and reward those who take the time to coordinate. It would also give those who coordinate reason to keep playing. And would make for longer lasting, funner battles because if your in a party then you could definately survive for more than a couple seconds.

And it would penalize players who simply participate in a zerg mob. Zerg mobs would still be greater than single players, but grouped up coordinated players would be greater than a Zerg mob. Several groups working together would be greater still.

WvW is about working together with the people of your world. When you do it should be rewarded with success. A commander in charge of several groups should be rewarded with more success than a commander in charge of a group of uncoordinated Zerg.

No, he’s suggesting that the maximum bandwidth of the server allocated to an individual client is nowhere near saturating his somewhat standard broadband connection. Habib even explicitly states that they are limiting bandwidth consumption by BOTH server and client with the sever side culling. To me that sounds like it is both an optimization for the client resources, as well as the server resources.

With the sheer amount of players that participate in WvW battles it might not be technically possible to show everyone on screen at the same time for all battles.

The problem is that in WvW you die, and you die in 2 seconds from a Zerg mob you can’t see.

Empower a group so that Zerg mobs will be destroyed quickly even if you can’t see them you should be able to hit them, and they won’t be able to destroy you in 2 seconds.

This would also spread battles out along the arena instead of require everyone to mob up in one small area thereby overtaxing the ability of the client to show everyone playing.

Visiroth.5914:

No, he’s suggesting that the maximum bandwidth of the server allocated to an individual client is nowhere near saturating his somewhat standard broadband connection.

Looks like you didn’t read what I wrote, or don’t understand the implications of “1%”.

What he said was the the problem would only be fixed “when the financial department agrees with server hardware and / or network bandwith upgrades”.

Then he said the game never uses more than 1% of his bandwidth (which seems a number pulled out of some anal cavity, but let’s ignore that for now and assume he actually measured it accurately, that the packets didn’t get chopped up and resequenced by some node along the way, that the bandwidth wasn’t averaged out over several seconds, etc.).

So, if the server is running at maximum capacity (and only buying extra hardware or a wider upstream pipe could increase its output) and if the current server output ceiling is 1% of his connection’s bandwidth, that means that even his residential connection would be able to handle at least 100 players as well as Arena Net’s servers currently do.

Since there are noticeable delays and culling problems even in battles with less than half that many players, that can only mean (if his statements are correct) that Arena Net’s servers have half the bandwidth of his residential connection (or less).

And that (along with the notion that the issue can be magically fixed “by the financial department”, that presumably refuses to do so due to greed) strikes me as a very silly theory, coming from someone who doesn’t have much (or any) experience in networking code development.

Kaizen.1704:

Empower a group so that Zerg mobs will be destroyed quickly even if you can’t see them you should be able to hit them, and they won’t be able to destroy you in 2 seconds.

Do you realise how little sense that makes? If you can’t see them, where do you know where they are? How are you supposed to select them or aim at them? How are you supposed to know if they have retaliation or projectile reflection on?

And if being in a party somehow makes you more powerful, how long do you think it will take until that “zerg” of 40 players simply forms 8 ad-hoc parties, and destroys your “empowered party” just as it does now?

The way to fix the issues caused by invisible models is to make the models visible. Improve the netcode, enable rendering fallbacks, keep better track of which client has been informed of what (and make sure a player isn’t damaged by things his client hasn’t even been informed of). It’s a technical issue; it should be fixed by technical improvements, not by changes to gameplay mechanics.

Some gameplay mechanics do need changes, but that’s a separate issue.

Is there any estimated or desired conclusion of the culling fixes? Or any designated time frame for when the majority of players will notice significant improvements? I only ask because I’m tired of hearing about it in game while I play.

Do you guys even read the Dev’s replies, or just skip to the last page and post your thoughts?

I still have faith in ANet to fix this, the same way they fixed the BLT loot during Halloween. He said it’s going to take time, and I trust they are working around the clock to fix it for us.

And before somebody accuses me of being some PvE noob who doesn’t know what he’s talking about; I run WvW four to five nights of the week, so yes, I do know what I’m talking about and I hate it just as much as everyone else. But there’s no point in continuing to post about it. Pressuring the Dev’s isn’t going to help, it only gets annoying.

>>can most part of the question “when will the culling be fixed” be answered with “when the financial department agrees with server hardware and/ or network bandwith upgrades” ? […] Anet has never come close to 1 % of my downstream capacity (a modest 15mbit).

<<So what you are saying is that, if there are any delays in battles with less than 100 players (and there are), then Arena Net’s servers’ connection must have lower bandwidth than yours.

>>no, that’s not what i am saying.

<<I have a feeling that’s not the case. And I have a feeling you don’t have much experience writing (semi-)real-time multi-client netcode or profiling the impact of data spikes on things like firewall CPU usage, etc..

>>i admit i have absolutely no experience with programming real-time multi-client netcode. i do have ccna1-4 and many other networking related certifications and have worked for 20 years in IT and IT related bussines. don’t be condecending about people you know nothing about.

<<In short, even a small increase in allowable bandwidth peaks can have a significant impact on performance and network choke (which can require resyncing, and so on), and there’s very little that the “financial department” can do to fix the WvW issues (other than maybe headhunting some programmers with more experience in the field – but it would probably take them a while to find their way around GW2’s current code, so that’s not a short-term solution, either).

>>i all too well realize that the amount of data that the server has to send us when 100 players are fighting eachtother is substantially larger than when you have 10 instances with 10 people each. in the first case, 100 players have to receive data about 100 players. to make it easy, the server has to send out 10 000 pieces of information for those 100 players. now, in the case of the 10 times 10 players, the computer has to send 10 player infos to 10 player, multiplied by 10 instances, for a totall of 1000 pieces of info. so, during the same amount of time the 100 players in w3 will (.. /can at times) consume 10 times the server upload resources than the 100 players in their 10 × 10 instances.

to me it ‘seems’ that the Anet serverpark can cope with the small instances fine and is ‘underpowered’ to cope with large scale events (both pve and w3). to compensate for insufficient upstream from the ‘server’, the server tries to make a selection of ‘relevant’ info to send to each client in order not to exceed it’s ‘quota’ of available upstream, and it’s that ‘selection’ of info creates most of the culling issue… imho

<<It’s mainly a software (netcode / game architecture) issue. Throwing more hardware at it on Arena Net’s end wouldn’t magically fix things.

>>what i do is ask questions and phrase thoughts, being very clear that they are asumptions and educated guesses. you state ‘facts’ with the certitude of revelation.

Visiroth.5914:

No, he’s suggesting that the maximum bandwidth of the server allocated to an individual client is nowhere near saturating his somewhat standard broadband connection. Habib even explicitly states that they are limiting bandwidth consumption by BOTH server and client with the sever side culling. To me that sounds like it is both an optimization for the client resources, as well as the server resources.

cripling every client by not sending the required info to everyone seems like worst possible idea to resolve issues that might affect a few people, asuming ofcourse the server is able to send all that data.

if they send not enough info, 100 % of people are afected.

if they send all the ‘info’ , then, maybe, some clients network bandwith will be flooded and he will experience lag, framerate loss,… , but for most people who don’t rely on manual morse code for their networking needs, this is a non issue.

there are of course the chances that some low spec clients would be unable to process/render all the info that it receives. unfortunately this game has so few graphic options that allow us to degrade the visuals, that tuning the performance from the client side is not as usefull and efficient as it could be. i think loads of people their game experience would increase a lot if there was the possibility to downgrade the quality of the models, turn of spell effects, turn of all ambient things (especially that cursed, loot bag hiding grass !) etc… from a minimum spec, frame rate is life, point of view, if i see the target is A) asuran b) wears cloth c) has two daggers, then i don’t kittening care about the rest ! i don’t need to see the details of his face, i don’t need to see the exact colloring of his skin, i don’t need to see that he is a slightly below average length but fat representative of his race, as long as i can see it’s an asuran. i don’t need to see exactly what skin of armor he is wearing as long as i can see it’s cloth and not plate. i don’t need to see what dungeons he has farmed to get the skin of his weapon, as long as i see it’s a dagger and not a 2h sword. why will Anet not allow low spec users to downgrade their graphics that far ? are they more concerned with bland screenshots and ‘yikes’ comments from low spec user screenshots on the internet than by the player experience they are trying to offer ?

for the record, i have a decent gaming computer, i get rolled by invisible zergs at maxed out settings in glorious 60+ fps. large groups of people who rush in my direction can sometimes have dozens disapear while moving to me and reapear many seconds later over my dead corpse without even dropping a frame, which is even stranger because these are guys that my client was aware of and then loses awareness of while they never went out of my ‘view’.

Account.9832:

What he said was the the problem would only be fixed “when the financial department agrees with server hardware and / or network bandwith upgrades”.

hm

>>can most part of the question “when will the culling be fixed” be answered with “when the financial department agrees with server hardware and/ or network bandwith upgrades” ?

hm

english isn’t my first language, but your representation of my phrase seems a bit ‘off’ in my apreciation. let me try to clarify. i used a rethoric question to state that it is my opinion that the core issue with culling can be found in the inability from the Anet server side of the equation to send out all the requiered data to everyone during events that have loads of players in them, and that fixing it might involve upgrading their network bandwith and/or switching and/or load balancing and/or processing power.

>>>>>Then he said the game never uses more than 1% of his bandwidth (which seems a number pulled out of some anal cavity, but let’s ignore that for now and assume he actually measured it accurately, that the packets didn’t get chopped up and resequenced by some node along the way, that the bandwidth wasn’t averaged out over several seconds, etc.).

even the less tech savvy amongst us can do some simple checks. windows taskmanager tells me that i use around 0.02 % !!! of my network bandwith to play GW2 and it goes to 30-32 % when using a torrent at full capacity (3.3 mbyte/s). do you really want me to measure the amount of kb that my client receives over a fixed period in a way that could withstand your scrutiny ?

>>>>>…Since there are noticeable delays and culling problems even in battles with less than half that many players, that can only mean (if his statements are correct) that Arena Net’s servers have half the bandwidth of his residential connection (or less).

your logic and mathemathics baffle me.

>>>>>>And that (along with the notion that the issue can be magically fixed “by the financial department”, that presumably refuses to do so due to greed) strikes me as a very silly theory, coming from someone who doesn’t have much (or any) experience in networking code development.

having tried to further explain some of my view, i scincerely and humbly await any further sign of that indeed such need for even further clarification exists and can only hope that more people share valuable insights.

Make culling a setting in options so all the toasters and microwave gamers (IE we of the kitten college laptop generation) can still set a lower cull-cap and have some FPS while invisible people blast and hack us to bits.

Lag alone has me dodging zergs already, anyways..

Would it help if the game just displayed a (targetable) name, a HP bar and the buffs/boons if bandwith/CPU needs to be saved?

what i do not understand is why they wouldnt just cut down on rendering by making tension points completely barren… as if THERE WAS A WAR GOING ON. Terrain detail is pointless when there are 80 people tossing fire bombs or what have you.

This is a bigger issue with those of you using min. requirement comps. I was using one as well for the first few months of this games release. Funny thing is that for the first month that comp ran WvW with no issue’s then there was a patch one night, one of the famous no warning WvW is being shut down for a new build routines, and from then on that comp. had rendering issue’s. My guess is that they increaased the population cap per a map in that patch. I went and bought an I7 3.7, 2g Nvidia card, with 16 gigs of ram. Now im back to seeing everything with no rendering issue’s. I even see theifs as normal or so I believe anyhow, Ive seen vids of people with issue’s with rendering theifs and im not experiencing what they are in the vids.

Also im guessing that with alot of you having these issue’s that if you’re running on a minimum reuirement machine then you are also using your ISP’s minimum bandwith option. This most likely is not ment for todays gaming, most are barely enough to surf porn let alone stream information from hundreds of sources instantly. The dev stated there is a maximum amount of information we are recieving, but he didnt say what that was though. Alot of you probably are not even close to maxin that out, which will cause lag and packet loss as the game tries to update what you should see on the screen. Obviously it will get worse with more stuff going on IE: more players.

This game also has a pretty robust sound and graphics engine. For most of us who are used to playing these games as a norm, this will come as a no brainer. For those of you who are new to this type of game, turn your graphics down when in WvW. Also I know back in the beta there was issue’s with the sound causing people to drop from WvW, now we get that kitten Air Raid Horn bug!!! I would still turn the sound options off to only what you need to play the game when in WvW.

I hope some of this helps someone out there. Happy Holidays everyone.

@Tool.3749: It affects everyone. People with machines that match yours have culling problems.

At the end of the day our goal is to dramatically improve the experience of large battles in WvW and provide a substantial increase to the number of players that can be seen by any given client.
[/quote]

I have never really seen any of these large battles that you speak of Habib mostly because there is no incentive to play WvW anymore. Although because there are less and less people I don’t have a problem with culling lol. However making changes to WvW so that people won’t want to play in it is not a fix for culling and something else should be done, please. Thanks.

When you think your fighting 2 players and actualy theres +30 of them.

In all honesty, after 4 months of labor it seems that the culling fix made performance worse for a majority of people playing WvW.
Even people with high end systems are having issues with Low/Low settings.
Why should people have to overclock their computers to run this part of the game at a decent level?

Stormcrow.7513:

In all honesty, after 4 months of labor it seems that the culling fix made performance worse for a majority of people playing WvW.
Even people with high end systems are having issues with Low/Low settings.
Why should people have to overclock their computers to run this part of the game at a decent level?

The lag’s always been there. It’s just that people can’t complain about culling anymore so they look for new things to complain about.

Stormcrow.7513:

In all honesty, after 4 months of labor it seems that the culling fix made performance worse for a majority of people playing WvW.
Even people with high end systems are having issues with Low/Low settings.
Why should people have to overclock their computers to run this part of the game at a decent level?

My friend has a piece of crap computer and couldn’t participate in ANYTHING larger than 30v30. Now with the new culling he can play in 90v90 with MORE FPS and actually be able to do something.

I have a medium-high end PC and I was able to tone down the culling settings and turn up my other graphics to levels where I normally couldn’t have gone to BEFORE this patch.

Please, know what you’re talking about.

Skill Lag has NOTHING to do with Culling or the new Culling change. Skill Lag has ALWAYS been there, but you think it’s due to this change when it’s not.

I think the lag is more of people running in 100man zergs for WXP now, not the culling removal.

Stormcrow.7513:

In all honesty, after 4 months of labor it seems that the culling fix made performance worse for a majority of people playing WvW.
Even people with high end systems are having issues with Low/Low settings.
Why should people have to overclock their computers to run this part of the game at a decent level?

I’m playing with both Culling options on Highest and I’m not having issues… Unless you think that having 20 fps with all graphics on max, and no culling at all is bad.

And no, I don’t have a high-end computer… My video card is pretty much outdated (Nvdia GTX 460) and my processor is a AMD FX 8120 3.1GHz.