Pings & Gamers

[davej99]

NOTE: This thread has been split off from High Pings to make the discussion of why pings matter to gamers separate from the issues which lead to the original thread - Rik

This has been a long thread with some rather hostile criticism of IDNET. It seems to me IDNET have done what they said they would do and commisioned new capacity, hobbled only by BT's tardiness. If we compare the big LLU/own network ISPs, with those using BTW, including BT Broadband, I think IDNET come out of it pretty well, as do some other smaller players using the same model. Essentially some ISP's deliberately oversell their capcity and throttle, and others don't. Some ISP's have tech support and management that hides in the bushes others do not.

Much of the less than polite invective directed at IDNET concerns response time or ping. It is natural for gamers to get really frustrated when this mission critical parameter rises to unacceptable levels. But what are those levels?

I have always wondered why gamers call for such low latency. Here are some rough numbers and I may be wrong. A ping of 20ms is equivalent to a single video frames at 25Hz.  A blink is 300-400ms. The finger movement reaction time in the very simplest game, the ruler drop test, is typically 200ms, the trigger time if you like. In more complex situations, such as driving and braking, an alert driver can apply the brake in no better than 700ms. A cricket ball traveling at 90mph takes about 500ms stump to stump, so a player has to strike after about 400ms. At best there might be 200ms decision time. In other words playing parameters are of the order of hundreds of milliseconds. So much of the concern that IDNET pings are moving about by tens seems misplaced, especially when you consider that an international hop might be 100ms.

Back to cricket; if you shortened the pitch by 5 metres, or 100ms in time, that's like facing a 120mph ball; if you shortened by twice that, or 200ms in time, it has to be broken teeth. Do either for one team and not the other and it is game over. So gamers have a point, but more about  the differential response time than the absolute and at levels closer to 100ms than 10. However, if your ISP suddenly has a pipe block and response times lenthen by 100ms or more, it is blind rage time.

So what seems to count is not a very low local response, but the near certainty that there will be no excursions above a certain level. I suggest, for example, that a local response time of 40ms that is under 100ms for 99% of the time is better than 20ms that is under 200ms for 90% of the time. In other words much of the preoccupation with a low ping is misplaced and now IDNET have installed more capacity, response time excursions should fall to an acceptable level.

Be wary though; my neigbore (not a spelling mistake) has just spent a fortune to get good Pings. Sadly it has not improved his handicap and he blames his Pings.

[Technical Ben]

Sorry dave, your examples fall apart. I don't care if I miss something when I blink, I am human  . But if my SOFTWARE works at very high speeds, and falls apart if a connection is lost "within the blink of an eye" then my personal preference has no correlation.

I personally don't mind having only 1-4mb download speed. For most things, this is fine. I do however like to have a constant trouble free connection. And for gaming, anything over 150ms gets kicked. And pings of greater than 150ms tend to suffer from glitches caused by the delay in communication.

It's a bot like cars. Some people needs a big truck, or a people carries. Some need a faster, smaller car. Everyone has different requirements.
But as said, IDnet have been rather good getting a fix out, and the problems have effected customer less than other ISPs (O2, Sky, BT etc).

[quandam]

Not a techie by any means. Do not IDNet (and others) have to cater for those that do not require good ping results and those that do? Pings are important to some and to others mean nothing. The average subscriber to, let's say Tiscali, have never heard of pings and just want a 'normal' connection.

I have no interest in pings as it means nothing to my enjoyment of the internet.  I feel that there is a them and us attitude regarding an ISP's responsiibilty. My question is 'do you want a steady overall connection' ? Or do you want to have  a steady overall connection AND  pings to game'?

If you require the latter then there may (may) be problems, you clearly cannot have both (unless your lucky!).

In my non techie opinion IDNet provide a very steady connection that would be more than suffice for the vast majority of customers. Too much emphasis is given to 'gamers' worrying about their 'ping problems', let's just have a good steady connection with no throttling and most customers will be more than happy.

[Simon]

With the new hostlink, IDNet should be able to provide both, but much depends on the quality of users lines, which IDNet have no control over.

[Rik]

Plus the BT netwrok, MSANs, DSLAMs and VPs. There's a long journey to reach IDNet, unfortunately.

[psp83]

At the end of the day, if everything is working the way is should be, your connection should be stable and your pings should be low.

Ping mostly affects gamers. But it also affects VOIP.

Anyways, its always good to have a low ping if your not a gamer aswell.

[davej99]

Sorry dave, your examples fall apart. I don't care if I miss something when I blink, I am human  . But if my SOFTWARE works at very high speeds, and falls apart if a connection is lost "within the blink of an eye" then my personal preference has no correlation.
.........

Software may work at very high speed but your body does not. At best a trigger response in 200ms and a blink is several times that. Surely we all blink. When you add thinking and physiological delays together you are talking several 100ms before we get to the electronics.

A low priority IDNET ping of 100ms, equates to a game UDP packet transit time of less than half that. A total gaming response, including thinking, physiological, visual rendering and internet transit times, probably amounts to at last 500ms, a cricket ball transit time. I suggest therefore that a 100ms ping is not that significant.

What counts is the statistical variation in packet transit time. This is usually caused by congestion. It can be large, transitory and a game stopper. The best way to detect this condition is pipe loading not baseline ping measurement and I think IDNET have a handle on that.

Since response times and download speed tend to go hand in hand, all customers benefit from addressing congestion which IDNET have done.

[Supanova]

I suggest therefore that a 100ms ping is not that significant.

I haven't been taking part in much of this discussion however I really have to correct your suggestion as a hardcore gamer. A ping of 100+ makes a SIGNIFICANT difference, especially when playing in an FPS game. It is obvious to me that you yourself have not played many FPS games seriously for you to have made such a suggestion.

Whatever you might think, a ping of 150+ will cause your sniper bullets to miss and your bunnyhop to be late. If you attempt to snipe at 180+ ping you have to aim ahead of your target which is totally unacceptable from a serious gamers point of view. When you are firing 10 rounds per second from your uzi that 150ms makes a difference between your bullets hitting first and your bunnyhop going off first.

Your ping makes a huge difference in competitive FPS gaming and although I recognise your statement was only a suggestion, it was an utterly incorrect one.

Tried and tested. I have suffered with cr*p pings at the hands if IDnet and BT and it completely ruins your game trying to play with a ping of 180ms. And when I say ruined, I really mean ruined. You cannot hit a damn thing without spraying the area with a high RPM weapon.

[Simon]

So, you're saying that with high pings, your bunny can hop out the way?  

I'm not a gamer, but just to shove my two pennyworth in, I can understand it from the point of view of two (or more) people competing.  If person A fires a shot at precisely the same moment as person B, and person A has a faster ping than person B, person A's shot is going to reach it's target first, even if there's only 10ms of difference.  
I think the human response times Dave is talking about is probably quite correct, but putting it in computer game terms, I can see where the theory falls down a bit.  I'm not agreeing or disagreeing with anyone, just expressing how it appears logically to me.  

[Supanova]

Bunnyhopping is pressing jump just as you enter combat with another player. It throws off his aim and in most games counts as a 'stance change' resulting in a few milliseconds realtime of immunity. If that stops even one bullet tearing into you, it's worth it. In older games like counterstrike that's why you see playing constantly jumping around while shooting. In more modern games they have made 'hops' immediately after the initial 'hop' far less effective so you can't jump more than once effectively, however that makes your first and only hop during combat all the more important.

I also used to think that human reactions are way too slow for a 100ms ping to make any difference, but then I actually started playing FPS games instead of amateur RTS and realised how unbelievably wrong I was.

[Aaron]

Having ping increased by +100ms is also enough to make button abilities in a game such as World of Warcraft feel really unresponsive. I can see how you mean about 100ms being really tiny as its only 1/10th of a second but it funnily makes a ton of difference.

[zappaDPJ]

Having ping increased by +100ms is also enough to make button abilities in a game such as World of Warcraft feel really unresponsive. I can see how you mean about 100ms being really tiny as its only 1/10th of a second but it funnily makes a ton of difference.

I couldn't agree more. That extra 1/10th of a second puts you continually 1/10th of a second behind everyone else and that's more than enough to ruin hours, days, even weeks of preparation for 24 other players. I think some of you are seeing it as a single event, hence the comparison to the time it takes to blink. The reality is that you are continually lagging behind the game and if you are playing any FPS or MMORPG at the highest level that extra 100ms might as well be a life time.

[davej99]

A low priority ICMP echo-reply packet or ping is not the same as a high priority gaming UDP packet. The ping is controlled by both the outward and return times as well as the response priority. The UDP is controled only by the outward time. So at very least the ping is double the UDP time and because of priority no reliable measure of it.

However, when networks get congested both types of packet get delayed. I have in the past correlated download-speed with pings. As a pipe gets congested the download-speed falls off progressively whereas the pings rise sooner and sharper. In particular the variation in ping over baseline becomes wild. We have seen some of that in ping plots earlier in the thread. The average rises slowly but the peaks rise dramatically. When a pipe moves towards congestion it is unplayable. In other words gamers get hit harder and sooner by congestion.

I suggest though when gamers think a 100ms ping is a problem they mean the peaks are a problem. The game dies; check the ping; it is in the 100s; conclusion a 100ms ping is no good. In reality it is the peak delay many times the average that is the killer, not the average and certianly not the baseline.

When the network is not congested the difference between a 20ms ping and a 100ms ping is no measure of UDP packet transit time, and even if it was it is half and small compared with the total gaming response. You get the ping characteristic of you connection, my first hop baseline is 40ms, Rik's is 20ms, it might be a more for others. The new pipe made no difference to my baseline ping but it is now flat.

My point is simple; that baseline ping in an uncongested network is no measure of UDP transit time or game performance and demanding the lowest possible ping is a waste of time.

Seems to me IDNET have fixed congestion and my guess is the gaming problem will go away. A policy of no ISP congestion is all we can ask for and that seems to have been delivered. Well done IDNET.

By the way, when shooting a running bunny, if you aim at it you will will never hit and there is nothing to blame but your skill, or maybe a clever bunny.

If you want to check your reaction times try HERE

If I was a gamer, I would be the weakest link.

[Technical Ben]

Fair enough, pings are a bad mesure of gaming quality, however, it is all we have.
And it's not down to the 10ms of difference in reaction time, it's the game being half a second out of sink with everyone else. Did you get that item? Did you go left? Who knows because everyone else has gone off and the game did something else while waiting for your packet.

From Wiki "In various network multiplayer games, the server notes the time it requires for a game packet to reach a client and a response to be received. This round-trip time is usually reported as the player's 'ping'. It is used as an effective measurement of the player's latency, with lower ping times being desirable. Note that this style of ping typically does not use ICMP packets."
Sorry, the gaming term "ping" is different here, but I usually get the same results from each test, even though one uses UDP and one uses ICMP. 

Latency is the main problem in gaming. But I don't know of any latency tests.

[quandam]

Understand gamers having problems with high pings but to a non gamer (as per my earlier post) there is not a problem. An earlier post said that better pings are an advantage even to non gamers, this could well be true but the end result is not significant to a non gamer. There are two 'problems' here one of download speeds and download speeds & ping speeds required by gamers. Depending on percentages of customers that require a steady 'enjoyable' connection and those that require a steady  'enjoyable' connection PLUS acceptable pings for gaming is a difficult recipe for any ISP to fulfill.

I will now offer a suggestion that we now have a dedicated thread for ping devotees to avail themselves of problems with pings.

[davej99]

Understand gamers having problems with high pings but to a non gamer (as per my earlier post) there is not a problem.

I probably did not express myself very well earlier. I was trying to say when congestion arises it hits both gamers and downloaders, with the distinction that gamers notice the problem first because UDP packet transmission time, ping if you like, is more sensitive than download speed. In reality the gamers are doing everyone else a service by are telling us the pipe is at its limit before the rest of us notice. When that limit is reached the game is unplayable and the rest of see a progressive degradation in download speed.

With a previous ISP we used a script to logon, check the ping and connect if it was low. If it was not, the script would log off and try again in the hope of hitting another central that was not congested.

As a result of BT delays, IDNET has had problems recently and although I noticed it in the variation of pings, the gamers picked it up big time. We are fortunate that IDNET has a policy of no congestion and no throttling, when it gets BT support, and we can all enjoy our time on line, be it gaming or downloading back copies of Rubber News.

Now the problem is fixed, we need be less concerned, but I feel content knowing that if congestion arises again the gamers will let us all know first. We are better in this together.

BTW, has anyone tried their reaction times HERE yet? If you can beat your ping let us know.

[Technical Ben]

I'll see if I can find any figures for you dave. Again, I say it's not down to player reaction time, but the ability of the software to cope with delays or slow responses from clients and servers. Perhaps I could compare it to a crackling phone line. It may not effect 99% of the customers using the phone, and even some people don't notice it. But for the person who has it, it will be very annoying!

[dujas]

Software may work at very high speed but your body does not. At best a trigger response in 200ms and a blink is several times that. Surely we all blink. When you add thinking and physiological delays together you are talking several 100ms before we get to the electronics.

Latency being vital to online gaming is not necessarily due to human reaction time, but because of the relationship between the clients (i.e. the players) and the server they are connected to and playing the game on.

The client can never be trusted due to hacks etc, so the server acts as the referee, what the server determines has happened is final. But the server usually prioritizes the client that sends in its request first (usually the one with the lowest latency for that situation). For example:

Player A's computer says: "Hey server, I've shot and killed Player B at x location".
Player B's computer says (only milliseconds later): "Well actually I've moved 2 steps forward from location x".
Server says: "Too late! I spoke to Player A first, you're dead".

If your latency to the game server is higher than the people you're playing against, you will always be at disadvantage, regardless of player skill. This is one of the reasons why gamers obsess so much about pings

[davej99]

Thanks, Dujas, that is very enlightening to a non-gamer that does not really understand the mediation process. Since you guys pay top money for a quality service with UDP packet priority I think you should get it. Most sportsman though accept that in near dead heat situations you win some you lose some.

If you are playing over a congested link and you experience UDP delay I can see you are disadvantaged and when that happens pings will go out as well. So as an indicator of congestion a high ping is meaningful. But I question its relevance to UDP latency when there is no congestion. I am not convinced that an ICMP ping of 20ms versus 100ms means that a UDP packet will show the same difference. In any event I am arguing that the total gaming latency is many 100s of milliseconds, and therefore pings <100ms in uncongested networks, implying a UDP latency half that, is small in the scheme of things. 

When there is no congestion UDP latency has to be determined by distance, number of switches and repeaters, the speed of light, client and server processing and so on. So for gaming equity players would all have to be in the same room.

But anyway I am really speculating because there is little research, other than knowing that when a central is full its game over. How are you all doing on the reaction test BTW? I have just tried it after a couple of beers and it is rubbish.  There is a lesson there. Meantime, thanks for the chat everyone, I have learned a lot. Cheers.

[Ninny]

Even if it does take 200 miliseconds for a trigger response, people are not taking into account that if a player with a 30 ping shooting at someone with 100 ping can aim at their location in space sooner and fire, the other player will press fire only when the server has updated the other players location 70 milseconds later, plus the latency of human faculties to press the trigger, so you have to add 200 ms to each person. So the 30 ping player will have a 230 response, the other will have a 300 response and hence always die. And this is not even a real indication as a higher ping player has to aim ahead of everything to account for their higher latency so it's entirely unfair, the other player only has to point and click and will probably have a quicker physical trigger response due to this, the second player has to guess.. so even if he pulls the trigger faster, he has to predict.. which is just guessing.. where the player will be when the trigger is pulled. So the human trigger mechanism will be longer due to the brain having to make a judgement where to aim.

And for competitive gaming this is just incredibly unfair. And even worse than this is unstable pings which is what IDnet have been displaying is much worse.. at least with a high latency you can become accustomed to the amount you have to predict ahead, but the pings have been so up and down there is no judge of what extent this will be. It would be like trying to drive a car that just speeds up and down without you having any control of it.

That said, I have asked for my MAC code by the end of the month, this was before the improvement, but it has improved a lot I have to say. So I will reconsider whether to leave, but I have been recommened Aquiss as they patronise gamers specifically and are based not too far from where I live so I may give them a shot. I actually tried to move to them before IDnet but the guy who runs the company never answered the phone! So I actually joined IDnet due to the great customer support, after dealing with BT that's all what most people ask for!

[dujas]

I would read up on the Entanet situation (whose services Aquiss resell) at Thinkbroadband, as download speeds have been dire at peak times for significant number of their customers. They've now implemented permanent traffic management from the 4th of November (it should help improve latency though).

A few weeks ago Entanet informed us that their monitoring systems have recently flagged up a number of potential causes for the slow speeds and they have been investigating this in depth. They have identified that approximately 90% of our customers are seeing their connection adversely affected by approximately 10% of the customer base who are using their connection for heavy downloading via newsgroups and P2P (peer-to-peer) applications.

Although existing management tools are in place, some customers are still seeing significantly slower speeds even during their peak allowance period of the day, and this is the reason as to why action needs to be taken to improve the service.

[shazzy]

Having ping increased by +100ms is also enough to make button abilities in a game such as World of Warcraft feel really unresponsive. I can see how you mean about 100ms being really tiny as its only 1/10th of a second but it funnily makes a ton of difference.

Yeh especially when you're a mage and you rely on getting spells off on time.

Just lately, my ping has fallen below 100ms most of the day and in the evening in wow; it's just over 100ms.  My server seems a little empty though and most of the time it's down to Blizzard's servers if the ping spikes. I know this because if I play Lotro (Lord of the rings online) my ping is always and I mean always below 50ms, which is amazing.  

[dujas]

Have you tried adding the following registry settings to stop Windows adding a time delay when sending very small packets?

Step 1 - TcpAckFrequency

Type "regedit" in windows "run.." dialog to bring up the registry menu.

Then find:
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters\Interfaces\

There will be multiple NIC interfaces listed in there, find the one you use to connect to the internet, there will be several interfaces listed (they have long names like {7DBA6DCA-FFE8-4002-A28F-4D2B57AE8383}. Click each one, the right one will have lots of settings in it and you will see your machines IP address listed there somewhere. Right-click in the right hand pane and add a new DWORD value, name it TcpAckFrequency, then right click the entry and click Modify and assign a value of 1.

You can change it back to 2 (default) at a later stage if it affects your other TCP application performance. It tells windows how many TCP packets to wait before sending ACK. if the value is 1, windows will send ACK every time it receives a TCP package.

Step 2 - TCPNoDelay

Type "regedit" in windows "run.." dialog to bring up the registry menu.

Then find:
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\MSMQ\Parameters

Right-click in the right hand pane and add a new DWORD value, name it TCPNoDelay, then right click the entry and click Modify and assign a value of 1.

For Windows 7 users the TCPNoDelay setting need to be in the same directory as described in "Step 1 - TcpAckFrequency" to work, rather than in \MSMQ\Parameters

After adding the registry settings, you'd need to reboot your PC for them to take effect.

[davej99]

This gaming is clever stuff. Have a couple of questions?

Which is the more common and best gaming protocol, TCP or UDP? Are gamers here using a mixture?

In terms of perceived UDP latency, can you tell the difference between delayed packets and mising packets? Both will lead to faulty positional information.

[dujas]

Well that's really a question for computer game developers. Broadly speaking, slower paced game genres such as Real-Time Strategy and Massively Multiplayer Online Role-Playing Games etc tend to use TCP, faster paced games, like first-person shooters, tend to use UDP.

I can sense the effects of TcpAckFrequency and TCPNoDelay, on a game like World of Warcraft, my rhythm will be off somewhat without the changes described above. Most games will deal with some packet loss until a threshold level, whereupon the client and server will lose sync, noticeable by an in-game warping effect.

It is amazing how quickly your brain's hand to eye co-ordination can sense that average latency is increasing.

[Fox]

Most modern games use a mixture of UDP and TCP, for example Battlefield 2 uses ports 4711, 27901, 28910, 29900, 299 for TCP and 1500-4999, 16567, 27900, 29900 for UDP. Latency does have a major effect on gameplay especially in first person shooters. Most professional leagues and ladders have rules about what ping is allowable in matches, TWL and CAL both have rules on minimum and maximum pins (the minimum rule is to prevent a team competing on a LAN against a team connected via the internet).

The easiest way i can think to describe it is that if you are aiming at a moving target and shoot with high ping, even if you appear to be on target you are actually aiming at an image that has been delayed by its passage over the web. The higher your ping the higher the delay and the higher the chance of missing the target completely (and for a headshot at 300-400 yards the target is only a few pixels wide).

Most good gamers can adjust for high ping (lead the target more) but when latency times become unstable (constantly varying say between 50ms and spikes of 250 ms) the game becomes harder and harder to play.

........and yes, I do consider myself to be a good gamer. I have been number 1 in TWL with several teams and when I played BF2 on a regular basis I was ranked in the top 25 players in the world (out of 1.1 million players at that time).

[Supanova]

Everyone has their gaming claim to fame. I'm not going to bore you with mine.

No offence but....

Dave you really should give up here. All of the gamers know you are wrong and are repeatedly trying to convince you otherwise. If you think it takes 0.7 seconds to react in an online FPS game you are demonstrating that you are a few years of FPS gaming behind the rest of us.

The problem with a higher ping that is the pipelining effect it has on you as a human. Generally to get a kill you have to go through a few actions and any delay between those actions passes to the next action and so on. If you can feel any delay you are scuppered.

Either the UDP ping that the player sees on the screen is merely a representation or we really are that fast. I'm happy with either result being true, while you Dave seem to have already made up your mind which is true.

Maybe this will give you an idea of how fast players are. A quick youtube search will yeald tonnes of vids. Some of those shots are a hell of a lot faster than your cricket ball.

Note most of his kills are headshots, so he is shooting a hitbox that is 1 or 2mm on his screen.
http://www.youtube.com/watch?v=I1FHjrjwuBA

[Rik]

I think the discussion has been useful for all of us who are not gamers and, therefore, don't understand what difference a few ms make. I've certainly found it instructional.

[Fox]

...... are you going to come over to the darkside Rik, and play a few rounds with us?

[Rik]

Not a chance, gaming is the waste of a good computer.

[Fox]

I think we should agree to disagree on that particular point before World War Three erupts on the forum

...but you should try it sometime, it shouldn't interupt your downloading those erm..."art" films too much 

[Rik]

If I'm not here, I'm usually to be found up to my elbows in Photoshop.

[davej99]

..... If you think it takes 0.7 seconds to react in an online FPS game you are demonstrating that you are a few years of FPS gaming behind the rest of us. ........

I believe I said simple human reaction time was the order 200ms. Try for yourself HERE. I also said in complex situations, such as driving and braking, an alert driver can apply the brake in no better than 700ms. These are scientific facts not a personal opinion.

[Rik]

Average 273.8ms.

[davej99]

Most good gamers can adjust for high ping (lead the target more) but when latency times become unstable (constantly varying say between 50ms and spikes of 250 ms) the game becomes harder and harder to play.

Thanks, FOX, for the background on use of TCP and UDP. Makes sense to use TCP for non time sensitive data. I believe I suggested earlier that as a pipe gets congested the average pings rises slowly but the peaks rise dramatically and unpredictably making the game unplayable. I also suggested that when gamers think the ping is a problem they mean the peaks are a problem. This is not too different from the point you are making on stability. So what counts is a stable and uncongested pipe and a baseline ping of is say better than a 100ms, which would equate to a UDP transit time of half that. Let us hope that is what IDNET are providing.

Also interested in your opinion on UDP packet loss %, which will probably track congestion, and if this feels different to unpredictable latency.

[Technical Ben]

On topic, but off subject... Things like Interleaving also have an effect. My pings (whatever the game measures) halved when I had interleaving turned off. And as the games usually have error compensation built in, or just request the occasional lost packed again, I'm better off not having it. I am better off having 99% of packets get through at half the speed compaired to 100% of the packets in double the time. There is only 1% difference in service for me, but twice the speed. I know that interleaving does not use twice the bandwidth, etc, but by the time it's all passed through the exchange it was about twice the time to get a response.

Where as most customers just browsing the web will not notice the difference in speed interleaving causes, but will notice that they get a perfect service with no errors. Apples and oranges really

[Fox]

With regards to interleaving, most gamers prefer it off, the preference seems to be for lower response times as most multiplayer games use quite a low percentage of your available bandwidth so the occasional resent packet has very little effect on the overall game feel. Some players even set up asymmetrical ping to a game server (I wont go into the technicalities of this) to gain an advantage - lag armor - but most well coded games have measures to prevent/reduce this.

[davej99]

... the occasional resent packet has very little effect on the overall game feel. ....

Is that a TCP resend or is there a way of initiating a UDP resend?

[Technical Ben]

Wiki Says (in the main UDP article):
"Reliability and congestion control solutions

Lacking reliability, UDP applications must generally be willing to accept some loss, errors or duplication. Some applications such as TFTP may add rudimentary reliability mechanisms into the application layer as needed. Most often, UDP applications do not require reliability mechanisms and may even be hindered by them. Streaming media, real-time multiplayer games and voice over IP (VoIP) are examples of applications that often use UDP. If an application requires a high degree of reliability, a protocol such as the Transmission Control Protocol or erasure codes may be used instead.

Lacking any congestion avoidance and control mechanisms, network-based mechanisms are required to minimize potential congestion collapse effects of uncontrolled, high rate UDP traffic loads. In other words, since UDP senders cannot detect congestion, network-based elements such as routers using packet queuing and dropping techniques will often be the only tool available to slow down excessive UDP traffic. The Datagram Congestion Control Protocol (DCCP) is being designed as a partial solution to this potential problem by adding end host TCP-friendly congestion control behavior to high-rate UDP streams such as streaming media."

So UDP does not, but the games themselves will have a way to request the missing data to be resent, or just simply smooth out or fill in the gaps. In my own experience though, a large loss will result in a disconnect form the server. Smaller losses result in strange effects in game as the game tries to fill in the gaps automatically, or gets the results wrong (like the player going in the wrong direction or the animation being stuttered).
The article also mentions that UDP gets priority over TCP on most systems. Not because it goes first, but because unlike TCP it does not automatically throttle back, but keeps blundering on regardless or packet loss.

(we seem to be getting tangled over what part of the things we are talking about. Ping/latency/UDP/TCP. When really, most software acts independent of the limits or actions of the protocols. I may be cooking on gas, electric or microwaves, but my water still boils at 100 degrees. )

[davej99]

..... The article also mentions that UDP gets priority over TCP .....

.... which gets priority over ICMP (ping) round trip. So how meaningful is ping as a measure of UDP transit time? Would be interested to know if game servers have a way of initiating UDP resend as a measure of transit time and so gaming equity.

[Technical Ben]

My last post on the subject...

From wiki (hint here )
"In various network multiplayer games, the server notes the time it requires for a game packet to reach a client and a response to be received. This round-trip time is usually reported as the player's 'ping'. It is used as an effective measurement of the player's latency, with lower ping times being desirable. Note that this style of ping typically does not use ICMP packets."

Read the last line of my previous post. I apologise and recognise that at the beginning of this discussion, gamers (me included) and other netters did not realise this difference.

[davej99]

Thx Ben. So server reported player (UDP) ping is not the same as the (ICMP) ping we might assess with tracert. It would be interesting to correlate. But that is for another day.

I have very much enjoyed talking to fellow IDNetters about gaming. After all we are one family and I believe that what is good for gamers is good for us all. I set out to enquire rather than to provoke, but anyway we all have something to say and enlightenment is always welcome, however put. I was impressed by those that clearly have the skill, knowledge and experience, and more importantly the time, patience and courtesy, to share the finer points of gaming and its technology with the rest of us. Hope you all stick with IDNET.