Android Geniuses: Do speedtests provide bogus numbers, or...?

[nightrocket]

This is particularly glaring on my Asus MemoPad 8 tablet. (But I've seen something similar on my laptop and phone, though perhaps not so slow). I run speedtest and get download speeds of 2.5 Mbs. However, if I am downloading a file or streaming something, my speed goes down anywhere from 5 kilobytes to 250 kilobytes on average. To give some context, in case it matters, I am using a Wi-Fi connection and use speedtest consistently so that their results are not sharply varying from day to day.

By the way, (and this does NOT have to do with the Asus I'm talking about here) when something is listed in kilobytes as 0.5 - say on a bitmeter widget - is that essentially saying .05% of a Mb? And would that be 5 kilobytes?

Here's what I need to know from you. Are speedtests just a hypothetical, best case scenario speed that should not be used to assess whether my laptops wi-fi is faulty? (And, if that is the case, what exactly is the point of them?) On the other hand, given the significant difference I am seeing, is it time to send it in under warranty to have the wi-fi hardware switched out so that I'm downloading/streaming at that 2.5 Mbs rate?

(At the low speeds I'm getting, how is it that I am able to watch any Crackle at all? Granted, it buffers at the front end and lags, but I'm surprised that I get anything at all.)

I'd love your help. Thanks

- JX

 

[B. Diddy]

There are multiple factors involved. One of the most important factors is the server that you're trying to download or stream from. If that server is slow, then your download will be slow. On what sites are you noticing these slow speeds?

Also, if the wi-fi channel your router is using is congested (i.e., many other networks nearby using the same channel), that can impact the wi-fi data speeds. Use an app like Wifi Analyzer to see what your local channel traffic is like. If there are lots of other routers using the channel your router is using, log into your router's settings using your computer's browser and change the channel to a less congested one.

 

[nightrocket]

B

Thanks for the thoughts. Here's one of the things that they underline...If the server is slow, wouldn't that fact be reflected in the speed results of speednet? Aren't those results a reflection of the server, as well, such that there wouldn't be a difference?

Let me take a crack at this. Let me know if this is correct. You are saying that each website has its own server; including the server for speedtest. If the server for speedtest is faster than the server for crackle.com or rapid gator or logging me into Gmail, then it's results don't really matter. Some site servers will be slower and, in fact, others may be faster than the server for speednet. If I have that right, what practical purpose does speedtest provide? Not clear on this

In answer to your question, I monitor my tablet's speed consistently and I get these slow speeds even when I am doing something as simple as web browsing. 15 or 20 kilobytes per second can still get it done but it's nowhere near the seamless - if not blazing speed - of 2.5 Mbs

Again, wouldn't router congestion be reflected in speedtest? Either way, logging into the router is not an option when using public wi-fi. Correct? I would need to know the password and admin name, I believe

All that having been said, I will review your advice and see how I might implement it. Some of the fundamentals you may be alluding to are not familiar to me but I have to say that I did a fair amount of digging on the net for answers and didn't find much. Again, thank you.

 

[nightrocket]

B/

I jjust spent 20 minutes composing a response and it went poof. I'll try to condense. Shouldn't server speed and router congestion be reflected in speed test results to begin with? Both of those factors are at play when running speedtest, so not sure why it would make a difference

The speeds I reference apply to everything from loading pages while surfing to the higher end (250 kbs if lucky) on Crackle, logging into Gmail, maybe YouTube, uploaded.net, etc.

Public wi-fi doesn't provide password and admin name to log into router and choose a different channel. Correct?

Thanks. I'll review your advice and try to implement. Peace.

 

[Rukbat]

The main thing to remember is that the information highway is much like a real highway. How fast does traffic move on a highway? I see traffic going 70 on I-95, but does that mean that it's still going 70 3 states north of me?

You're measuring how long it takes to transfer a known amount of data. From that you derive how fast that connection was. Do it again (on the same Ookla server) and it might be different.

Try an experiment. Go to speedtest.net. Do a speed test on the server it chooses as the default. Now choose a server across the country and do another speed test. Now choose a server halfway around the world and do a speed test. Which one is correct?

All of them. Your speed is the speed data traveled from point A (the server) to point B (you). The default server you get with Ookla is almost always the fastest one from you. (My default server varies from the one geographically closest to me to one about a 2 hour drive away, depending on when I'm doing the test.) that has nothing to do with how fast data travels from Microsoft's download server to you, or how fast data travels from Mega's server in New Zealand to you. So you can get a 15mbps speed test and a download speed from somewhere else of 45kbps. And both numbers are "correct" - for 2 different servers.

I've gotten downloads (within the past yer - it used to be slower) running anywhere from 100kbps to just over 2mbps. My speedtest.net speeds are in the 14-15mbps range on my phone (and about 20mbps on my desktop). And all of them are correct, and all the equipment is working properly. If the widget is displaying kbps, and you get 0.5, that's 0.5kbps, not 0.5% kbps, so 500 bits per second (or about 50 bytes per second, which, other than for text or small emails, is essentially useless).

Buffering simply means that the data isn't coming down the line as fast as it's being used. It's a geeky term for "waiting for more data".



And that's the main determinant of your download speed - where the server is, and how the traffic is moving between that server and you. If there's a major news story, like WW 3 just broke out, connecting to your provider will be so slow you'll give up (and that path is normally the full speed you're paying for). If you're downloading from a decent server halfway around the world when nothing is happening on the internet, you'll get great speed. (Christmas morning is a good time to see this. Everyone got a new computer, laptop, tablet, phone, something that goes on the internet. And speeds drop like rocks. A really cold Sunday evening in the winter, when half the US is having a blizzard is another slow time. [Satellite TV reception is so bad, you give up and get on the internet to kill the evening. So does 1/4 of the population of the country.] Back around 1997, one really bad Sunday night, I measured a ping time of 30 minutes to one server. Not 30ms, or even 30 seconds - 30 minutes. Anyone who was on IRC back then probably remembers it. MCI lost the main feed from the Eastern US to the Western US. It's like everything but dirt roads were closed at the height of the rush hour in a large metropolis. The internet became the "probably not net".)

Speed isn't absolute. That's one reason you don't buy 20mbps, you buy "speed up to 20mbps". The only speed your provider can guarantee is from their equipment to you. That should be the speed you're paying for. If their entry point to the internet is close to an Ookla server's entry point, your speed test to that server should be about what you're paying for. If you're located in such a place that the nearest Ookla server is over 1,000 miles away, you're never going to see decent download speeds from that server. But if you download a file from your provider, you'll get it at the speed you're paying for almost all the time.

Should you return your ASUS? Not if you're using 3G data, or your internet connection is "up to 2.5mbps", or the speedtest server isn't within a couple of miles of your provider's entry to the internet (which may not be anywhere near where you are - I was staying in a motel just outside DC, and their IP address showed up as being in Indianapolis - that was their entry point to the internet. [Don't ask me why - I'm still fairly sane, so I have no idea why insane people do the things they do]).

BTW, "To give some context, in case it matters, I am using a Wi-Fi connection and use speedtest consistently so that their results are not sharply varying from day to day." is completely incorrect - they vary through a single speed test. The internet's speed, over any given path, varies so fast that an actual display of the speed of a given path is a blur if seen in real time. It may not vary by much, say 20mbps to 22mbps, but it's constantly varying, depending on what traffic which routers in the path are routing to where, and how much data there is on the path (compared to how much it can handle). If you're doing a speed test on a pipe that can handle 1tbps, and it's handling 1tbps, then someone sends a packet through that pipe to this server (it's going to be one packet - you can't type fast enough to send a long stream of packets), the connection slows down during the time that additional packet is moving down the pipe. Not much - one packet on a 1tbps pipe may slow it down by 5bps, but it's still a varying speed. Watch a long data transfer from drive C to drive D in Windows 8+ (say a 4GB file). You'll see the speed varying on a short wire connection. The computer itself doesn't maintain a constant speed, so it can't download at a constant speed. Neither can all those computers that are routing all those packets on the internet.

But if your other computers are all showing 20mbps and your ASUS is showing 2.5mbps - and you're using a router that's known to not have the TKIP problem (and the only one I've personally tested is the Netgear WNR2000), and you're using AES encrption (or, better yet, just for the test, no encryption), then there's probably something wrong with the tablet. But if your desktop and laptop get 3mbps? The ASUS is just not powerful enough to download and store that file at full speed, and that's probably about as good as it's going to be with that hardware, whether they give you a new one or not.

What's the silliest thing, IMO, is people paying through the nose to get these "super" 50mbps or 100mbps connections. If you're running a large internet cafe, there may be a good reason to do so. But for a home connection? Not this year, and not next year. Assume you're downloading or streaming 3 connections at once, and everyone else in your family of 4 is doing the same thing. That's 12 connections. If you're all getting really fact paths on all those connections, that's 24mbps. If you're paying for 50mbps, you're paying for 26mbps of bandwidth that you'll never use. (Bandwidth and speed are different ways of expressing the same thing, and the numbers are the same - your bandwidth is 20mbps or your speed is 20mbps - they mean exactly the same thing.) I laugh at those commercials showing a family of 5, each running a single connection - "you get such fast connections with our company that everyone can enjoy high speed". High? 5 connections is 10mbps maximum. And videos don't run even 1mbps, unless you're watching 4k video. Oh, sure, the packets may be coming down the pipe from your provider to you at 50mbps, but you're getting 20 "packet widths" of nothing for every packet you get. Burst ... wait ... wait ... wait ........... burst. Because the server is sending the video out at around 800kbps.

Sagan was so on the money. In a world in which more and more depends on science and technology, a smaller and smaller percentage of the population understands anything about either science or technology. (It's the second decade of the 21st century - and there are still people who believe that you can fall off the edge of the Earth. Or that the evolution we see isn't really happening. Or that the fact that there is now a Northwest Passage has nothing to do with the planet getting warmer. And some of them make our laws. If you're not terrified, you're not paying attention.)

 

[B. Diddy]

^^^ Um ... what he said.

 

[dpham00]

My Netflix account is the medium tier, so no uhd. I I measured data usage multiple times when watching twd, and it runs about 3-5gb per hour on my surface pro, tethered to my note 4. This equates to about 6-10mbps. Keep in mind that you would need faster than that, to get consistent playback at the highest fHD setting. I noticed that on slower connections, Netflix will reduce the quality, so you would still stream smoothly.


If you have uhd then it could run substantially more. Netflix recommends 25 mbps for the highest tier

Sent from my Verizon Samsung Galaxy Note 4

 

[UJ95x]

I tried Tidal HiFi's streaming service a few weeks ago. Every song streams at 1,411 Kbps. Even music can run you over 1Mbps (although it isn't typical, obviously). But video will easily run you into the 4-5 Mbps if you're streaming HD or higher (which most people do, meaning there can be multiple streams at a time). My home Internet is 50Mbps down, and I still get videos buffering if there are more than a few connections (usually on weekends when relatives come over). So I'd say if anything, under 10 Mbps is unacceptable, and even more if there are multiple people in your home. 50 and 100 Mbps connections are absolutely reasonable

 

[LeoRex]

I have 25/25 on my FIOS account.... and even in peak times, I'm pretty close in both directions (22/23 both ways).

Now, I have two kids... so when they are online, they are on Youtube, watching Minecraft videos on Youtube (not even in any kind of crazy HD).... I usually am only on my phone, so when I see them worshiping at the alter of Stampylonghead, I switch over to LTE if I want to do more than just browse forums because I know that sticking on my WiFi would just get on my nerves.

So 25Mbps for a family of 4 is pushing the lower limits... I would go to 50+ if I wasn't so cheap and didn't have access to a zippy LTE connection. 10Mbps for 5? I wouldn't want to have to deal with that.

 

[Golfdriver97]

I think it would depend on the price you pay for what speed you get. Let's take a couple examples: Comcast has a package that has TV and 50 mbps internet for $99.xx. Google Fiber, on the other hand, GigaBit and TV is $130. So, which is more cost effective? The 50 speed for $100, or 20 times faster, for $30 more? I honestly can't wait for Fiber to get here.

 

[Almeuit]

My 2 cents ...

Speed matters for what you're doing. If you're just browsing the net .. You don't need that much. 5 people could survive on 10 Mbps if just lightly browsing / checking email, however, if you toss in any amount of video streaming (YouTube, Netflix, etc.) one person could easily eat up that 10 Mbps to themselves.

 

[Golfdriver97]

one person could easily eat up that 10 Mbps to themselves.

Quite possibly the best thing to consider here. Also, to make an extension of this point, what does one download often? Sometimes, I download ROM files for devices. When they are usually 500MB+ I wold rather have a faster speed so they are done faster.

 

[Crashdamage]

I feel like I should chime in on this a bit...

I've had Google Fiber gigabit internet for over a year now. Actually, we have the gigabit internet + TV package. The gigabit internet alone is $70/mo. Is it worth it? Is there any point in having so much speed? Is Google lying or is it really gigabit speed?

YES and YES and YES!

I hear what Rukbat's saying. Why have 500 horsepower when the roads are covered with snow and you can only use 75? And he's right in saying, to put it in a nutshell, that it's true most of the internet is not up to gigabit speed. Most of the internet highways are covered in snow.

But with all respect, Rukbat is wrong. Because when the snow melts and you can put the pedal to the metal that 500 horses sure is fun! When you connect to sites that are go up to half gig or even quarter gig speed, you never want to go back to pokey old 100Mbps maximum cable at about the same price. IOW, if you can have a gigabit internet for about the same cost, why not?

Besides, some of the internet IS ready for gigabit speed. When you hit those spots Google Fiber is freakin' amazing. Plus, it's gigabit downloading AND uploading. We use cloud storage a lot and while none of the cloud storage services are gigabit fast, they're fast enough that Google Fiber makes a very significant difference. Uploads that took hours are done in minutes now. I can download a new Linux live CD in a blink. Gigabit IS actually useful right now.

Bill Gates famously said he didn't think networking would ever matter and left the TCPIP stack out of DOS. He also said no one would ever need more than 256kb of RAM. I bet that now he wouldn't say that there's no use for gigabit internet. He's learned his lesson about what the future might bring.

Anyway, back to the subject of this thread.

As you can imagine, with gigabit speed I can really stretch out most any site's servers including speed test sites like Ookla's Speedtest. Turns out none of the speed test sites I tried, including Ookla, are up to the challenge of Google Fiber. Because of this, Google has setup their own speedtest site. Using that site I routinely get 920-970Mbps both up and down. Not quite gigabit speed, but close enough.

The Google Fiber supplied network box (router) has state-of-the-art chips and software. Google didn't skimp on the hardware they give you. I've connected many phones, tablets, laptops, a Nest thermostat, Nest smoke detectors, etc and never had a problem connecting anything. We normally have 7 - 8 devices connected with no trouble or significant speed loss.

Now, with effectively no restrictions on speed capabilities and WiFi working flawlessly I can get true test results, right? Well, not so fast. Turns out speed test results are sometimes inaccurate to the point of being bogus.

Example 1: I take a Nexus 5 or 6 and go to Ookla and run speedtests (Ookla can keep up with the WiFi) and get what is, apparently at least, fairly accurate results on 2.4MHz and 5MHz bands. The Nexus 6 wins at around 320-250Mbps on 5MHz. But, if I turn WiFi off and connect to T-Mobile LTE, Ookla says it only does about 2.5 - 3.0Mbps! Some websurfing with the phone quickly makes it obvious that the LTE service is much quicker than that.

Example 2: I take the Nexus 5 and run speed tests on WiFi. Ookla, Speakeasy and XFINITY all give different results. If I turn off WiFi and use LTE, Speakeasy and XFINITY closely agree and give reasonable results, but Ookla refuses to connect to a server at all and gives a connection error. Again, a little websurfing proves the Nexus 5 and LTE are running pretty good.

So...don't put too much stock into results from internet speed tests. As Rukbat correctly explained in detail, they actually test the speed of the overall connection rather than the speed of YOUR connection. There can easily be be 20 - 30 'hops' through servers located thousands of miles apart between you and Ookla's server. Run traceroute on a few sites and be amazed.

To add to what B. Diddy said to nightrocket about troubleshooting his WiFi...

You need to determine the best setup for your situation and configure the router and phone accordingly. To find the best setup try using WiFi Overview 360. The free version is probably all you need but the Pro version is cheap. The Google Fiber guys use it to setup WiFi when doing installations.

https://forums.androidcentral.com/e...?id=de.android.wifioverviewpro&token=CBH-xK_F

With a dual band router, give the 5GHz and 2.4 GHz bands different SSID names. When setting up the 5GHz band usually it's best to set the router to use a channel well over 100. Over 150 if there's a clear channel. For both bands of course use channels with the least interference and best signal.

Android since v1.0. Linux user since 2001.

 

[UJ95x]

Bill Gates famously said he didn't think networking would ever matter and left the TCPIP stack out of DOS. He also said no one would ever need more than 256kb of RAM. I bet that now he wouldn't say that there's no use for gigabit internet. He's learned his lesson about what the future might bring.

He actually never said that. Just internet legend

 

[Crashdamage]

That's too bad...it would be way cooler if he actually did say it.

Figures that I wrote that long post and the only comment it gets is on the stuff Bill Gates didn't say (no offense UJ95x).

Android since v1.0. Linux user since 2001.

 

[anon8380037]

Silly question - I understand the main advantages of speeds up to 1 Gb, but is a heavy website like Forums.androidcentral.com going to refresh or open posts any quicker?

I still have to strum my fingers before I can start scrolling. I could get up to 70 mps, and tried various other systems.

(I don't get any non-AC advertising here either. )

Sites have to be 'fetched' I guess, so if you have just enough speed for any interactive content, eg - car configuration sites, browsing will be the same?

 

[UJ95x]

That's too bad...it would be way cooler if he actually did say it.

Figures that I wrote that long post and the only comment it gets is on the stuff Bill Gates didn't say (no offense UJ95x).

Android since v1.0. Linux user since 2001.

Yeah, but I'm sure you could find plenty of real quotes from major people in different fields that sound silly in hindsight.

And sorry...your post is marked as best answer though (visible on desktop)

 

[Crashdamage]

@Madd54: Rukbat covered your question pretty well in post #5. To put his point more simply, you're limited to the slowest link in the communication chain. If it's the server on AC no amount of speed on your internet connection will make AC load faster. If it's your internet connection a screaming fast new server for AC won't make AC load faster. Or maybe the slowest link is somewhere between you and AC.

But, as I tried to explain in my post, there's still plenty of good reasons why a gigabit connection is useful and generally a Good Thing.

BTW Android Central is a kinda slow site. But nowhere near as slow as it seems to be for you. 70Mbps is not blazing fast, but it's very respectable and you shouldn't be strumming your fingers waiting for stuff to load.

Android since v1.0. Linux user since 2001.

 

[B. Diddy]

Crashdamage, you're just killin' me with your constant references to Google Fiber. I WANT IT NOW!

 

[Crashdamage]

Danger! Google Fiber is highly addictive! Use with extreme caution!