EPTStarlink vs fiber optics for better latency
| Forum Index > BW General |
|
QuadroX
386 Posts
| ||
|
WombaT
Northern Ireland24255 Posts
In that you can absolutely take a slower route from wherever you are to Korea to another, but there’s a hard limit of the physical distance that will always lead to pretty chunky latency over large distances. IIRC Starlink’s main selling point is in a high-speed solution for remote areas or regions with poor internet infrastructure, and from what I’ve heard it does it well. But if full fibre is available where you are it’s going to be the better option, especially for latency. Somebody more knowledgeable than moi by all means correct me if I’ve got any of that wrong! Which isn’t all that unlikely | ||
|
iPlaY.NettleS
Australia4327 Posts
Starlink does not have the consistent ping that fibre does, you will get lag spikes. | ||
|
tankgirl
351 Posts
2 .play over ethernet not wifi 3. portforward UDP 6112 and check the setting in starcraft ingame options->network->prefer port 6112  Port Forwarding4. try using a vpn like Lagofast or wtfast (each has free trial) and connect to its South Korea server. https://www.lagofast.com/ https://www.wtfast.com/ https://www.vyprvpn.com/ | ||
|
MeSaber
Sweden1235 Posts
Starlink tests seems to range between 20-100ms which for an RTS game would be ok. | ||
|
MeSaber
Sweden1235 Posts
On April 15 2024 18:48 tankgirl wrote: 1. fibre > starlink 2 .play over ethernet not wifi 3. portforward UDP 6112 and check the setting in starcraft ingame options->network->prefer port 6112 Port Forwarding4. try using a vpn like Lagofast or wtfast (each has free trial) and connect to its South Korea server. https://www.lagofast.com/ https://www.wtfast.com/ https://www.vyprvpn.com/ You dont need portforwarding anymore. Fibre isnt necessarily better if you live far away. Starlink routing is very straight forward while fibre is a spiders web. I myself from Sweden with Fibre got 300-350ms to Seoul. | ||
|
tec27
United States3694 Posts
Blizzard's netcode is not particularly resilient to packet loss, as it only resends data once it receives packets from another player and can see that that data has not been received. This is something we've changed/improved in ShieldBattery, but you won't get those benefits playing over bnet. In any case, you always want to reduce packet loss as much as possible for the best experience (which is why other people have also suggested not playing over WiFi, for example). | ||
|
jinjin5000
United States1396 Posts
On April 16 2024 09:00 tec27 wrote: If you have the option of both, the better choice is fiber, no question. The latency of a single packet isn't really the problem here, it's packet loss that will screw you over. Starlink theoretically can improve upon latency (although last I heard, the things that would allow for these theoretical improvements are not even in use) but it will be worse for packet loss. Blizzard's netcode is not particularly resilient to packet loss, as it only resends data once it receives packets from another player and can see that that data has not been received. This is something we've changed/improved in ShieldBattery, but you won't get those benefits playing over bnet. In any case, you always want to reduce packet loss as much as possible for the best experience (which is why other people have also suggested not playing over WiFi, for example). I'm curious why battle.net is so consistently bad.... | ||
|
Smorrie
Netherlands2921 Posts
On April 16 2024 09:00 tec27 wrote: If you have the option of both, the better choice is fiber, no question. The latency of a single packet isn't really the problem here, it's packet loss that will screw you over. Starlink theoretically can improve upon latency (although last I heard, the things that would allow for these theoretical improvements are not even in use) but it will be worse for packet loss. Blizzard's netcode is not particularly resilient to packet loss, as it only resends data once it receives packets from another player and can see that that data has not been received. This is something we've changed/improved in ShieldBattery, but you won't get those benefits playing over bnet. In any case, you always want to reduce packet loss as much as possible for the best experience (which is why other people have also suggested not playing over WiFi, for example). Is port forwarding still beneficial? I remember reading somewhere that whatever improvements were put in place weren't even working properly and it is still recommended to forward ports. I've never had any issues but forwarded ports regardless just in case. Also, when I host replays it consistently takes at least one of my friends exceptionally long to join my lobby. Could this be related with him not having his ports forwarded properly? | ||
|
castleeMg
Canada758 Posts
| ||
|
Bonyth
Poland537 Posts
![[image loading]](https://upload.wikimedia.org/wikipedia/commons/thumb/8/89/Submarine_cable_map_umap.png/640px-Submarine_cable_map_umap.png) Starlink is supposed to connect you with South Korea via straight line, connecting from satelite to satelite, but i don't think they have this technology yet, so the signal ends up going from your home --> satelite --> ground station --> the same route from picture above Unverified info. | ||
|
tec27
United States3694 Posts
On April 16 2024 10:48 Smorrie wrote: Is port forwarding still beneficial? I remember reading somewhere that whatever improvements were put in place weren't even working properly and it is still recommended to forward ports. I've never had any issues but forwarded ports regardless just in case. Also, when I host replays it consistently takes at least one of my friends exceptionally long to join my lobby. Could this be related with him not having his ports forwarded properly? The answer is really "it depends". Modern bnet uses a combination of STUN (to identify what your various IP addresses might be and possible ports, as well as what type of NAT you might be behind) and TURN (to relay packets between users who can't connect otherwise). STUN will often allow holepunching to work for a lot of home routers, but with NATs that are more strict it will fail (especially if the 2 players that are trying to connect both have more strict NATs). In that case TURN should still allow them to connect, but I have no idea where Blizzard hosts their servers for this, how reliable they are, how well the location choice works, etc. So, to be safe from that ever occurring, forwarding ports (or enabling UPNP on your router to make this happen automatically) is still a reasonable thing to do. Unfortunately, people that most need to forward ports are also generally the least able to do it (either because they lack the knowledge or they are on some network with infrastructure they do not control). Generally the game will send packets to both the player's IP and the TURN server until it verifies it can reach the player directly, so I doubt it should really add super noticeable time to joining a lobby. More likely it's the thing Blizzard changed in a recent patch randomly to scan people's maps/replays for duplicates when joining lobbies, and that person has a large map+replay collection that takes a while to scan. | ||
|
tankgirl
351 Posts
On April 16 2024 06:11 MeSaber wrote: You dont need portforwarding anymore. Fibre isnt necessarily better if you live far away. Starlink routing is very straight forward while fibre is a spiders web. I myself from Sweden with Fibre got 300-350ms to Seoul. ok maybe it not work for you sorry :/ my games against koreans went from TR14 low~TR20high to consistent TR16low w/ occasional TR20low (edit- canada) | ||
|
sophisticated
58 Posts
On April 16 2024 14:47 Bonyth wrote: fiber connects you with South Korea through these lines Starlink is supposed to connect you with South Korea via straight line, connecting from satelite to satelite Just to make it clear: the signal doesn't go through earth, lol. It has to go round the planet too. If someone wants to connect from sweden there will be multiple hops among the satellites including "ceremony" for each just like when using cables (may be more hops too, cause at least you can curve the cable around the earth's surface) | ||
2Pacalypse-
Croatia9488 Posts
On April 16 2024 14:47 Bonyth wrote: fiber connects you with South Korea through these lines (and has to go through some sort of procedure on every dot?): Starlink is supposed to connect you with South Korea via straight line, connecting from satelite to satelite, but i don't think they have this technology yet, so the signal ends up going from your home --> satelite --> ground station --> the same route from picture above Unverified info. That map only shows the undersea internet cables. There are a lot of cables over land as well :d But yeah, the path that packets take to go from one place to another on internet is pretty complex and not consistent. You can sort of get a sense of the direction the packet takes with tracert command. For example, I just ran the command for a random Korean IP address, and the packet went from Croatia -> Germany -> Kansas (America) -> Seoul. It's wild. Starlink could theoritcally improve on latency of the packets (making them take shorter paths) with their satellite-to-satellite lasers, but it remains to be seen how this actually works in practice (see this video for an animation of how this could work). They do have this technology already; IIRC they added satellite-to-satellite lasers to their v1.5 and v2 mini satellites, which might even be the majority of their currently active satellites at this point. However, there's still a lot of variables which might impact the latency. For example, the signal still needs to go from/to the space/ground station which makes the path longer by ~1000km which land-based cables don't have. Although, that might be offset by the speed of light being higher in the vacuum of space than in fiber glass. In the end, it remains to be seen how it works actually in practice by doing some real-life tests with people who have Starlink. For now I would do what tec said and go with the fiber, since it's much more important to have a stable connection than potentially slightly lower latency on Battle.net. | ||
|
RJBTV
194 Posts
In that case starlink would be wayfaster. | ||
|
WombaT
Northern Ireland24255 Posts
| ||
|
MeSaber
Sweden1235 Posts
On April 16 2024 15:14 tankgirl wrote: ok maybe it not work for you sorry :/ my games against koreans went from TR14 low~TR20high to consistent TR16low w/ occasional TR20low (edit- canada) Dynamic Turn Rate is very sensitive from what i can tell so you should ping a Korean server instead of trusting these numbers. | ||
|
RJBTVYOUTUBE
Netherlands696 Posts
On April 16 2024 20:55 MeSaber wrote: Dynamic Turn Rate is very sensitive from what i can tell so you should ping a Korean server instead of trusting these numbers. Dynamic turnrate is often one off where it has to be. Its incorrectly calibrated. | ||
|
Smorrie
Netherlands2921 Posts
On April 16 2024 15:01 tec27 wrote: The answer is really "it depends". Modern bnet uses a combination of STUN (to identify what your various IP addresses might be and possible ports, as well as what type of NAT you might be behind) and TURN (to relay packets between users who can't connect otherwise). STUN will often allow holepunching to work for a lot of home routers, but with NATs that are more strict it will fail (especially if the 2 players that are trying to connect both have more strict NATs). In that case TURN should still allow them to connect, but I have no idea where Blizzard hosts their servers for this, how reliable they are, how well the location choice works, etc. So, to be safe from that ever occurring, forwarding ports (or enabling UPNP on your router to make this happen automatically) is still a reasonable thing to do. Unfortunately, people that most need to forward ports are also generally the least able to do it (either because they lack the knowledge or they are on some network with infrastructure they do not control). Generally the game will send packets to both the player's IP and the TURN server until it verifies it can reach the player directly, so I doubt it should really add super noticeable time to joining a lobby. More likely it's the thing Blizzard changed in a recent patch randomly to scan people's maps/replays for duplicates when joining lobbies, and that person has a large map+replay collection that takes a while to scan. Right, that makes sense. Thanks for the insights. I never heard about the folder scanning before - I'll ask my friend to archive his replay folder and test it out. I'll report back if it actually made a difference. | ||
|
rusty23456
United States107 Posts
| ||
|
MeSaber
Sweden1235 Posts
On April 16 2024 21:50 RJBTVYOUTUBE wrote: Dynamic turnrate is often one off where it has to be. Its incorrectly calibrated. Or it sets TR8 because micro stutter that isnt even noticeable. | ||
|
WGT-Baal
France3341 Posts
On April 17 2024 07:11 Smorrie wrote: Right, that makes sense. Thanks for the insights. I never heard about the folder scanning before - I'll ask my friend to archive his replay folder and test it out. I'll report back if it actually made a difference. I actually didn't know it was scanning the replay folder too. Let us know your result. If it s better i may archive my replays elsewhere | ||
|
iFU.pauline
France1469 Posts
 | ||
|
RJBTVYOUTUBE
Netherlands696 Posts
On April 18 2024 00:20 iFU.pauline wrote: I highly doubt a non wired connection outperform internet fiber in term of latency, then it also depends on the paths and nodes you are traversing to reach your destination but regardless, non wired connection is heavily prone to interference, and you can forget about constant rate in term of bandwidth. Also, satellites are orbiting, therefore paths are constantly changing depending of plenty of constraints such as nodes saturation, network load balance etc... so latency may fluctuate a lot which is the worst imo when you're playing. Starlink was not made for low latency gaming so I would avoid such solution unless you have no choice in the present day there's no options to really improve inter-continental connectivity. in regards to moving replays, it works. I have 25GB+ of replays and I store them elsewhere otherwise my starcraft instantly crashes when joining a lobby for which I don't have the map or the replay. | ||
|
WGT-Baal
France3341 Posts
On April 18 2024 06:51 RJBTVYOUTUBE wrote: in the present day there's no options to really improve inter-continental connectivity. in regards to moving replays, it works. I have 25GB+ of replays and I store them elsewhere otherwise my starcraft instantly crashes when joining a lobby for which I don't have the map or the replay. possibly a stupid question but what exactly do you mean my moving? do you archive them and remove them entirely from SC (and thus you cant open them at all from the game) or do you just take them out of the main /maps folder (and if so, where and how are you able to link them to the game so you can watch them) | ||
|
RJBTV
194 Posts
On April 19 2024 23:21 WGT-Baal wrote: possibly a stupid question but what exactly do you mean my moving? do you archive them and remove them entirely from SC (and thus you cant open them at all from the game) or do you just take them out of the main /maps folder (and if so, where and how are you able to link them to the game so you can watch them) Take the replay files out of the starcraft folder and place them in a folder outside of the starcraft directory. So yes, you cant open them from the starcraftclient anymore. You would have to put them back in the starcrafy replays folder. | ||
|
WGT-Baal
France3341 Posts
On April 20 2024 01:26 RJBTV wrote: Take the replay files out of the starcraft folder and place them in a folder outside of the starcraft directory. So yes, you cant open them from the starcraftclient anymore. You would have to put them back in the starcrafy replays folder. ok yeah. I guess i need to sort which i need to keep and which I dont. probably 95% of them i dont | ||
|
namkraft
424 Posts
Btw sometimes when I play I see this red text in a game: Due to your network configuration, your connection to other players must be routed via a proxy server ![[image loading]](https://i.ibb.co/dL7Lq24/Proxy-Server.png) Why is this happening? Must I find a way to remove this text? | ||
|
MeSaber
Sweden1235 Posts
On April 20 2024 04:26 namkraft wrote: Very cool insights. I've played Starcraft for 10+ years from US, Europe, Japan and a bunch of others, and I've ALWAYS experienced lag. Btw sometimes when I play I see this red text in a game: Due to your network configuration, your connection to other players must be routed via a proxy server Why is this happening? Must I find a way to remove this text? If you feel its working without lag for chosen TR then no. | ||
|
rusty23456
United States107 Posts
On April 20 2024 04:26 namkraft wrote: Very cool insights. I've played Starcraft for 10+ years from US, Europe, Japan and a bunch of others, and I've ALWAYS experienced lag. Btw sometimes when I play I see this red text in a game: Due to your network configuration, your connection to other players must be routed via a proxy server Why is this happening? Must I find a way to remove this text? you are probably connected to the interent with wifi on a router and not opening ports. two options. one is to connect directly to the internet with ethernet cable without wifi router. or you need to go into router settings and open up UDP and TCP ports. you can google this to see exactly how to open ports. | ||
|
Smorrie
Netherlands2921 Posts
On April 17 2024 21:42 WGT-Baal wrote: I actually didn't know it was scanning the replay folder too. Let us know your result. If it s better i may archive my replays elsewhere It definitely helps! The situation still isn't perfect but it has improved a lot, from having to wait 30-40sec to 10-15sec. If you have a large library I'd definitely recommend trying this yourself. | ||
|
iopq
United States876 Posts
On April 15 2024 18:48 tankgirl wrote: 2 .play over ethernet not wifi let me ping my router --- 192.168.2.1 ping statistics --- 628 packets transmitted, 628 received, 0% packet loss, time 631884ms rtt min/avg/max/mdev = 0.570/1.400/15.957/1.193 ms I only got one big spike of 15ms and most of the time it's within 3ms, I don't know that 15ms will make a difference I usually get TR24 low vs. Koreans and I still don't know if these tiny spikes are noticeable | ||
|
tec27
United States3694 Posts
On April 24 2024 01:14 iopq wrote: let me ping my router I only got one big spike of 15ms and most of the time it's within 3ms, I don't know that 15ms will make a difference I usually get TR24 low vs. Koreans and I still don't know if these tiny spikes are noticeable Ping is not really a realistic test of this as BW will be sending packets much more quickly (and also larger packets). Wifi is worse because of potential interference (which can happen from many things in your home like microwaves, vacuum cleaners, etc. as well as other people in your area using wifi). Many of these things are intermittent, not something that you would necessarily see in an isolated test, and may get worse at certain times of day. To make matters worse, when interference causes a packet drop on bnet, it will take at least a roundtrip between you and your opponent before any re-sent packet can be dealt with (this is something ShieldBattery improves upon). You might think everyone telling you to play on ethernet is stupid and it's all fine, but please, for the sake of everyone, just believe that we are correct and play on ethernet unless you absolutely cannot. On April 19 2024 23:21 WGT-Baal wrote: possibly a stupid question but what exactly do you mean my moving? do you archive them and remove them entirely from SC (and thus you cant open them at all from the game) or do you just take them out of the main /maps folder (and if so, where and how are you able to link them to the game so you can watch them) Just to note, ShieldBattery can be set as a program to open replay files in Explorer, and you can also configure it to immediately launch the game with the replay (by default it will show some info about the replay in SB and give you the option of viewing it). | ||
Peeano
Netherlands4767 Posts
On April 24 2024 09:58 tec27 wrote: Just to note, ShieldBattery can be set as a program to open replay files in Explorer, and you can also configure it to immediately launch the game with the replay (by default it will show some info about the replay in SB and give you the option of viewing it). Nice! | ||
|
iopq
United States876 Posts
On April 24 2024 09:58 tec27 wrote: To make matters worse, when interference causes a packet drop on bnet Let me stop you right there, dropped packets on WiFi are SO rare that it's a less than once a day occurrence. I will run ping for the whole day and play starcraft and we'll see how often packets are actually dropped in ping. Note that WiFi itself will retransmit a dropped packet before it goes out to the internet - that's the 15ms delayed packet that was retransmitted it would have to fail to transmit it from my router to do a full round trip to b.net in practice, I have no noticeable delay on battle.net because I'm using 5g wifi, not 2.4g, so microwaves are not an issue because they are on a different frequency the speed at which ping sends packets is not an issue because I'm pinging while playing, so you can consider it a sampling of all the packets that's sent, I don't need to DOUBLE the amount sent just to see if the latency is high, in fact you want to increase the total number of packets by a very small amount, otherwise you're just testing whether the router can keep up with that greater amount of packets, not the amount generated by the game --- 192.168.2.1 ping statistics --- 12057 packets transmitted, 12057 received, 0% packet loss, time 12280590ms rtt min/avg/max/mdev = 0.552/0.928/49.109/0.738 ms so you see, there's no packet loss because the wifi retransmits the packets so battle.net doesn't have to do a round trip | ||
|
Bonyth
Poland537 Posts
On April 24 2024 20:06 iopq wrote: --- 192.168.2.1 ping statistics --- 12057 packets transmitted, 12057 received, 0% packet loss, time 12280590ms rtt min/avg/max/mdev = 0.552/0.928/49.109/0.738 ms so you see, there's no packet loss because the wifi retransmits the packets so battle.net doesn't have to do a round trip The spike (at least 1) of 49 ms will cause the turnrate to drop. Also pinging once per second seems not accurate. Would be better to ping 100 times per second for more accurate results. Or make more shorter tests in order to see how often such a lag spike occurs. PS. Note that i'm not saying it's Wi-Fi that was responsible for the lag spike. | ||
|
tankgirl
351 Posts
On April 24 2024 20:06 iopq wrote: Let me stop you right there, dropped packets on WiFi are SO rare that it's a less than once a day occurrence. I will run ping for the whole day and play starcraft and we'll see how often packets are actually dropped in ping. Note that WiFi itself will retransmit a dropped packet before it goes out to the internet - that's the 15ms delayed packet that was retransmitted it would have to fail to transmit it from my router to do a full round trip to b.net in practice, I have no noticeable delay on battle.net because I'm using 5g wifi, not 2.4g, so microwaves are not an issue because they are on a different frequency the speed at which ping sends packets is not an issue because I'm pinging while playing, so you can consider it a sampling of all the packets that's sent, I don't need to DOUBLE the amount sent just to see if the latency is high, in fact you want to increase the total number of packets by a very small amount, otherwise you're just testing whether the router can keep up with that greater amount of packets, not the amount generated by the game --- 192.168.2.1 ping statistics --- 12057 packets transmitted, 12057 received, 0% packet loss, time 12280590ms rtt min/avg/max/mdev = 0.552/0.928/49.109/0.738 ms so you see, there's no packet loss because the wifi retransmits the packets so battle.net doesn't have to do a round trip tec27 i know u wrote openBW blah blah but i have a certificate in IT technologics from freecourseweb and after spending a few hours with the command prompt ive conclusively demonstrated that Wi-Fi is 100% more superior than CAT-7 ethernet for PC gaming, in all contexts throughout time in perpetuity forever. trust me ive been to IEM and DreamHack and everyone there definitely always just uses wifi. but dont worry ill share some of my turing prize moneys with you. | ||
|
WGT-Baal
France3341 Posts
random lag spike, error 3: 9, failed to download settings, bugged (more than before) friend list etc... | ||
|
iopq
United States876 Posts
On April 24 2024 21:33 Bonyth wrote: The spike (at least 1) of 49 ms will cause the turnrate to drop. Also pinging once per second seems not accurate. Would be better to ping 100 times per second for more accurate results. Or make more shorter tests in order to see how often such a lag spike occurs. PS. Note that i'm not saying it's Wi-Fi that was responsible for the lag spike. again, if I'm pinging 100 times a second I might actually slow down my game since when I'm doing the test I'm actually playing the game this is actually SAMPLING every once a second and yes, if over that time I had one spike of 49ms, I didn't notice it. That's the point, I see no difference in real use | ||
|
Jjb
1 Post
| ||
|
Branch.AUT
Austria853 Posts
On June 06 2024 03:15 Jjb wrote: OK, this is the most ridiculous conversation I’ve heard. It does not matter the situation if you have the availability of fiber optic cable as opposed to Starlink, The fiber optic cable will wipe the floor with Starlink. Whether it’s latency, download, or upload speed, fiber optic cable is by far the best. The same goes for regular cable although the difference is not as overwhelming as fiber optic. Starlink advantage is connection ability almost anywhere. This is because SpaceX is covering the lower orbit with literally thousands (like 12,000 eventually) of satellites. That’s the accessibility advantage over past satellite services like Hughes Net that flys only a couple. Starlink will have much better latency and speed than those other satellite companies simply because Starlink satellites are at a much lower orbit, closer to earth using microwave signal to communicate. Consider fiber optic moves data at close to the speed of light, and that’s about all you need to know. So, it doesn’t at all “depend”. Tell me again what is the speed of microwaves in air/vacuum? The limiting factor in all multi network internet connection is always going to be switch throughput. That is entirely independent of physical transmission medium. Because outside the switch the information travels at the speed of light anyway. "It depends" is the correct answer, because what causes the latency is the routing, the number of networks/routers involved, and the bandwitdth that router needs to handle. If one router delays your packet by 100ms, everything else is immediately irrelevant. Doesn't matter if the packet arrived via eleczromagnetic wave in fiber optics or electromagnetic wave in air/vacuum. | ||
|
rtyrt7
44 Posts
On April 24 2024 09:58 tec27 wrote: Ping is not really a realistic test of this as BW will be sending packets much more quickly (and also larger packets). Wifi is worse because of potential interference (which can happen from many things in your home like microwaves, vacuum cleaners, etc. as well as other people in your area using wifi). Many of these things are intermittent, not something that you would necessarily see in an isolated test, and may get worse at certain times of day. To make matters worse, when interference causes a packet drop on bnet, it will take at least a roundtrip between you and your opponent before any re-sent packet can be dealt with (this is something ShieldBattery improves upon). You might think everyone telling you to play on ethernet is stupid and it's all fine, but please, for the sake of everyone, just believe that we are correct and play on ethernet unless you absolutely cannot. Just to note, ShieldBattery can be set as a program to open replay files in Explorer, and you can also configure it to immediately launch the game with the replay (by default it will show some info about the replay in SB and give you the option of viewing it). Thanks! Knowing that you're the creator of ShieldBattery, which I thank you for, I'm gonna reference your post, as I grew tired of explaining why ethernet is better than wifi. | ||
| ||
