does your internet connection have bufferbloat?

Bigger packets take longer to deliver, dummy.

>>105688265 (OP)
Starlink + Pfsense meme

link?

>>105690042
https://www.waveform.com/tools/bufferbloat

mehhhh

Openwrt makes this so easy
Currently in a video call btw

>>105688265 (OP)
I thought BBR congestion controller + fq_codel was supposed to detect bufferbloat automatically?

>>105690317
It becomes a lot harder beyond 1gbps

You either limit your throughput rather significantly, or deal with a bit of bufferbloat when at the extreme range of your bandwidth.

Or you buy a very expensive router and make sure your client device is decently powerful as well.

https://www.waveform.com/tools/bufferbloat?test-id=c60fe5ca-8730-48c6-b009-2a8b95207f41
idk what i expected

>>105690795
Yeah there really isn't much point caring about bufferbloat on multi-gig connections since the only time it'll come up is if you're downloading or uploading at or near your bandwidth limit.

>>105690795
I'm too cheap for gb internet plus it doesn't make that big of a difference for my typical day to day and games. I only run a few local services that don't take much either

>>105690851
Phone wifi? Here's mine, router is shitty tplink, but i have no issues on the 50mb connection.

>>105690871
Yeah, just saying bufferbloat becomes a lot harder to deal with on faster connections, but on the plus side, it doesn't generally matter anywhere near as much as it does on slower connections.

On 300mbps it's realistic a single game download can take up that entire 300mbps and if you're also in a video call during that download without good bufferbloat management you'll have shitty video calls. Where as on 1gbps+ a single game download is unlikely (though technically still could) to take up your entire available bandwidth.

Why tf do you need 1 gig. 100 does it for me. Even porn sites load efficiently with 100. Maybe 300 at best if your routing

>>105690903
>Phone wifi?
no, its a really expensive plan where i live, at least according to my parents. the router is good, down is 150, up is like 20 or something, pretty sure its my shitty laptop thats throttling

>>105690903
NTA, but here is my connection with my phone wifi >>105690795

wouldn't this be skewed by how far away I am from their servers?

>>105690973
No, it's measuring your idle latency (no upload or download), then it's comparing that idle latency to your loaded latency when downloading, and then when uploading.


Basically if your idle latency is 20ms, but your loaded latency is 22-25ms, that's a good result.

But if you idle latency is 30ms and your loaded latency is 50-60ms, that's a bad result.

It's looking at the DIFFERENCE between idle and loaded latency, so server distance isn't really a factor.

Didn't bother stopping qBittorrent to run the tests but it seems fine without doing that too.

>>105690965
>1.6gbps over wifi
but why

>>105691030
gotta go fast

>>105688265 (OP)
Yes:
LATENCY
Unloaded

5 ms
Download Active

+50 ms
Upload Active

+61 ms

>>105690795
>>105688265 (OP)
Is there a fix for the latency spike during the download test?
I have a pfSense router that is more than capable for gigabit but every test there are those few measurements that fuck up the rating. It's only a few so its not like something is underperforming.

Or is it an ISP level issue?

>>105691067
it CAN be an ISP issue, but it's usually not.

https://docs.netgate.com/pfsense/en/latest/recipes/codel-limiters.html

>>105690851
are you in outer space or somewhere in the uk?

>>105688265 (OP)

I have no idea what bufferbloat is, but I'm OK

>>105691205
bufferbloat is the difference between your idle latency (unloaded latency) and your latency while uploading/downloading (loaded latency).

>>105688265 (OP)
Bufferbloat grade: B
What can I do to fix it? I never noticed any issues, but I rarely play video games these days, especially online ones.

>>105691239
buy a better router

>>105691261
That explains it, it's the cheapest tplink money can buy. Probably has backdoors from every nation known to man, even some internationally unrecognized ones.

>>105691307
Yeah, it can also be caused by shitty old devices (usually laptops).

How fucked am I?

>>105688265 (OP)
yes, this is the only option where I live

>>105691401
Wanna come live with me? I have great internet here.

>>105691394
Not at all

could be worse

Waveform fucks with the results to get you to buy one of their slop routers from the referral links at the bottom.
https://devina.io/speed-test
if you use this, which still uses Waveform's servers, you don't get the huge random 200-300ms spikes

>>105688265 (OP)
Yes and I use sqm on my owrt.

Sure does, not that it seems to make much difference.

I used to have like +100ms on upload but turns out I had really old NIC drivers with some shitty configurations
a simple update didn't work, had to uninstall the device first then reinstall new drivers so if you have issues try it

TP-Link BE9300, no QoS, no custom firmware, >1Gbps

https://www.waveform.com/tools/bufferbloat?test-id=a17fe1b9-d276-4939-9181-d6de1ac8a7c5

>>105691907
forgot image but i guess link is there

>>105688265 (OP)
Seems like a meme to get you buying their affiliate shit.

>>105691940

Then why don't we all fail?

>>105691766
I have to doubt the accuracy of their test, considering they show me doing 9gbps for a brief moment

>>105691766
Waveform directly gives me a max idle latency of 16ms with a median and mean of 6ms.
Devina.io gives me a max idle latency of 55ms with an average of 8.9ms

Waveform directly gives me a max active latency with 2 spikes, one at 249ms and another at 251ms. But the rest of my active latency pings are below 50ms, and most are below 25ms with an average of 17ms.
Devina.io gives me a max active latency of just 75m, but constantly shows my latency going from 10-20ms up to 60-75ms, then back dozens of times unlike the waveform test that had 2 spikes of ~250ms but it was FAR more consistent overall instead of going up and down and up and down.

So yeah, they're certainly not doing the exact same testing if I'm getting such widely different results between the two.

>>105692050
You're a retarded good goyim.

>>105688265 (OP)
post link in the OP next time, faggot.

also anyone who smart queue management should have an A or A+.

>>105691612
turn on SQM
https://openwrt.org/docs/guide-user/network/traffic-shaping/sqm

>>105692654
i would say nice but you're sub gigabit so your hardware isn't even being stressed

>>105692681
I don't see a reason to pay like +$25/mo more for GiB when half that is like $50/mo.

>>105692670
as far as i know openwrt isn't supported on my router

>>105691401
Almost no excuse to have this level of garbage connection with Starlink around.

>>105692854
>Starlink
hahahahahahaha

>>105692864
Be stuck with 5mbps forever just because you hate rocketman lmao

>>105692933
gladly

>>105692935
I had 15mbps back in 2005, I couldn't go back to less than that.

>>105692933
too many trees whether or not I want any kind of satellite service

>>105693069
Trees can be trimmed or removed, not to mention you can mount these things on tall polls or towers.

>>105693112
not if you rent or can't afford it

>>105693116
Sure but that's another argument.

Starlink is an option

>>105688265 (OP)
Im amazed my old archer c7 1750 v2 with openwrt still does SQM decently enough

>>105693426
it's a lot easier when your WAN speed is only 80-100mbps.

Go to 500mbps+ and it'll need a beefy CPU.

>>105690317
I'm >>105688265 (OP) and have OpenWRT but can't get below ~22ms on Downloads. I don't download at 1gbps often, but I'd like to get rid of it regardless

I didn't have this issue with cable internet

>>105694587
faster WAN speeds require more CPU power to do QoS.

Okay now what

>>105692854
Muskrats make me fucking hurl, starlink has some insane latency, you dumb fuck.

>>105694689
It's better than what he had...
Starlink latency is 20-50ms, the anon he was talking to has 83ms latency

I've had problems with this with all my ISPs in multiple houses in southern ontario with both rogers and bell and their resellers for decades but finally my latest switch back to rogers it seems to have gone away.

the fix for this for this with a shit score is to get a router that will allow you to set a bandwidth limit, cap it at like 90% of your total. ideally install custom firmware like ddwrt or merlin or tomato as they have good QoS algorithms to limit the effects of this.

for those wondering, what happens is that once you hit near the max of your download OR upload speed, your latency skyrockets. this happens because the buffer gets bloated with packets and doesnt know what to prioritize. so instead of giving you that UDP gaming packet or loading that website first, it loads another second of that netflix show which it didn't need to.

this is not always the fault of the users hardware, for me it was the fault of the ISP.

>>105694756
My router only lets me limit bandwidth in 1mbps increments between 1 and 1000. Since I have 2500mbps internet, I can only limit my speeds to 1000mbps, so not worth it.

>>105688265 (OP)
yes, i spent a little bit trying to set up sqm on my router but i don't think it's worth it. i don't game anyway

>>105691205
bufferbloat is basically just
>ping is fine when not downloading/uploading stuff
>ping gets worse when downloading/uploading stuff

>>105694756
> the fix for this for this with a shit score is to get a router that will allow you to set a bandwidth limit, cap it at like 90% of your total. ideally install custom firmware like ddwrt or merlin or tomato as they have good QoS algorithms to limit the effects of this.
wrong
you have a shitty ISP or a shitty router.
plenty of anons in this thread with gigabit+ speeds that do not have issues with this. stop coping with fucking bandwidth caps and shitty custom firmwares that can’t handle anything above 700Mbps
These shitty CPU algorithms that custom firmwares use can’t handle real internet speeds, they’re only “““““useful””””” (a cope) for poors.

>>105695036
most routers don't have the CPU horsepower for more advanced SQM methods on high-speed (500mbps+) connections.

>>105694612
>>105695052
ah damn, that makes sense. i have a Linksys LN1301 I got for like $15

>>105695106
the router's cpu cores get to 50% but they aren't super saturated...

>>105695049
both openwrt and opnsense can route gigabit if the hardware allows for it. yes they're not as fast as propiretary options, but they're still fast as fuck

>>105695052
yes, my xr500 is not really powerful enough. i could still get less loaded latency if i wanted to, it's not worth the speed reduction tho

with multiple other devices on the network, it's an ISP-provided Fritzbox

>>105695399
Fritzboxes are good from what i've seen over the years.

>>105694907
Sounds like ubiquiti

Even if you limit to 1000mbps you'll probably be lucky to break 800mbps with rate limiting since it disables hardware accelerated routing. Though you'd remove the bufferbloat entirely.

>>105691920
>>105691907
>BE9300
that thing looks fucking huge
>>105691940
what router?
>>105691874
under 10ms isnt bad
>>105692722
ln1301s end up on woot for like $15 occasionally

What are some generic SQM settings? I don't have any PCs I want to prioritise over others specifically.

I already know I need to replace that network card on the PC.
But i didn't expect it to be that bad

Custom OpenWRT router
SQM QoS with fq_codel + simplest_tbf

>>105698866
what hardware?
also it's easier to get A+ with slow speeds. see >>105694587

>>105694680
now you get faster internet

>>105699213
NanoPi R6S

That's because the higher the throughput the higher the computation needs and most routers can't keep up because they're low energy devices.

>>105698866
What speed do you get without the SQM?

>>105699309
Not much more, I have a 500/150 connection. Don't need more, don't want to pay for more.

>>105699327
Just wondering, I'm personally unwilling to sacrifice 10-20% of my bandwidth for slightly better latency under load. Thankfully not a big problem on 2gig internet

>>105688265 (OP)
At first it was C rank, and I did not like that.
Then I set up FlowQueue Codel on my OPNSense firewall.
Now it's A+.
Thanks OP.

>>105699474
How much performance did you lose?

>>105699543
Performance? I have dedicated hardware for my opnsense firewall router, so none.
The CPU can handle it easily.

>>105699222
Use case for faster internet?

>>105688265 (OP)
maybe?

>>105698866
same, i have openwrt with sqm on a thin client after i upgraded to 100/40 internet and the best regular soho router i could find couldn't keep up with that. now i've recently upgraded to 800/40 and this can no longer keep up, it tops out around 550/40, so i'll need to find something new

>>105692864
>>105689987
works fine for me

>105689987
>105690317
>105691021
>105691205
>105691394
>105691874
>105691881
>105692654
>105693426
>105694587
>105694680
>105695399
>105698866
>105699474
>105699637
> <1Gbps
result discarded

>>105691401
Damn where the fuck do you live?

>>105688265 (OP)
I could fix it but my pfsense is virtualized and I'm too lazy to pass the Nics directly to it.

>>105699950
>saar

>>105699592
So you're paying for 480mbps internet?

Come on, it's okay to admit you're paying for 500mbps and 20mbps is lost due to the SQM.

>>105691078
>https://docs.netgate.com/pfsense/en/latest/recipes/codel-limiters.html
Thanks anon

>>105701207
No, I pay for 500/500.
This is pretty normal, especially for download speed to average to that.
It still jumps to like 1300mbps.
Here you can see me getting more than what I pay for my upload.
The results just change every time I run the test.
Also I'm not as close to the test servers as the other people here.

>>105701477
Disable SQM and run the test again a few times, you'll likely average higher.

>>105701493
Nah. I tuned it a bit and I'm happy with the results.
The tiny loss doesn't matter, it's worth it.

>>105701794
If you find yourself frequently downloading heavily while on video calls or gaming, sure.


I personally don't bother since I have more than 1gbps and the odds of my download hitting my bandwidth limit are quite low as most services barely deliver 1gbps, let alone 1.5-2gbps.

I have video calls regularly and never experienced any issue.

This test is just to trick morons into buying from their affiliate links.

>>105702090
Now try downloading a game while on a video call and let us know how that goes.

Bufferbloat occurs when you're at the limit of your available bandwidth.

You shouldn't expect any issues if all you're doing is watching youtube videos while on a video call.

>>105702090
You could you know.... skim the thread
Such as
>>105690903
>>105691401
>>105698485
The grading system isn't good, but EF and beyond it looks like you have latency spikes so big it can start softlocking the entire connection

I can only get 180 down (out of my 300/150) with sqm enabled on my hAP AC2 running openwrt
I wanted to test rOS again now that it's on v7 but for whatever reason it refuses to boot into the pxe server to revert so it's stuck like this
Might upgrade to the RB5009 at some point

>>105703683
rOS v7 is apparently using different CLI commands so it can be hard to figure stuff out since most of the tutorials online for earlier rOS versions is out of date for the current rOS release.

I hate DOCSIS so much, look at this shit.

by the end of summer I'm moving to an apartment with direct Ethernet to the ISP, but turns out those fuckers skimped on proper 4-pair cabling so I'm only getting 100 Mbps

>>105704016
>direct Ethernet to the ISP, but turns out those fuckers skimped on proper 4-pair cabling so I'm only getting 100 Mbps
What the fuck kind of redneck shit is that?

>>105704016
>I hate DOCSIS so much
It's decent in my area

2100mbps down and 300mbps up.