Noob question on Multithread test

[CaliScotsman]

Greetings.

 

Time Warner tells me I'm getting 25 Mbps.

To prove this they point to Speedtest.net. (Which I now believe is inaccurate.)

 

Using Testmy.net it looks like my actual speed is @ 5 - 7 Mbps DL. And this is borne out by the time it takes for page loads.

For example when I visit www.telegraph.com it takes between 30 and 45 seconds for the home page to DL completely.

 

So, regarding a speed test.

 

If I do a multithreaded test it shows around 25 Mbps, single thread @ 6 Mbps. (Similar to Speedtest.net.)

 

Is the multithread test a special kind of test that doesn't apply to the real world?

Is the multithread test therefore not a valid test?

 

I'd like to understand.

 

Thank You

[mudmanc4]

Greetings.

 

Time Warner tells me I'm getting 25 Mbps.

To prove this they point to Speedtest.net. (Which I now believe is inaccurate.)

 

Using Testmy.net it looks like my actual speed is @ 5 - 7 Mbps DL. And this is borne out by the time it takes for page loads.

For example when I visit www.telegraph.com it takes between 30 and 45 seconds for the home page to DL completely.

 

So, regarding a speed test.

 

If I do a multithreaded test it shows around 25 Mbps, single thread @ 6 Mbps. (Similar to Speedtest.net.)

 

Is the multithread test a special kind of test that doesn't apply to the real world?

Is the multithread test therefore not a valid test?

 

I'd like to understand.

 

Thank You

I have two things. 

First, the page load of the site you mentioned is less likely to be on your end then something in between, or more likely the server itself. I just loaded it myself, and it crawls. 

 

Secondly, from my understanding, the multithreaded test, is more reality than not. Considering most sites today link multiple resources, even on the pageload itself. Can be from any one or dozens of servers worldwide. 

 

Therefore the multutest, pulls from resources located in various geographical location to simulate 'real world' downloading. 

[CaliScotsman]

Ok, now I'm confused.

When I've been doing some research it often suggests that multithread tests are not as accurate as single thread.

 

I do the multithread test here and it says 25 Mbps

I do the single thread test and it says 6Mbps

 

I get slow loading and timeouts so it seems to me that I do not get 25 Mbps.

[CA3LE]

Hi CaliScotsman, welcome.  No such thing as a n00b question here.  We don't use that word around here.

 

Actually, the multithread speed test still provides important information to understand your overall connection.  If it didn't provide value, it wouldn't be an option here.

 

Having said that, it's only part of the picture.  The complete picture is brought to light with my linear download test. The classic test that started this site.  I see it often. The multithread test will far out perform linear (single thread) test.  This means that the resources are there... they just can't all be utilized in a single stream.  In my opinion if it can't be utilized in a single thread it's not truly the speed they say it is.

 

Sometimes there's nothing you can do, other times there is room for improvement.  Better equipment, better settings.

 

I see your results before you signed up at https://testmy.net/compID/955319156813

 

You're testing to China, pick one of the test servers in the US.  Testing out to CN will definitely lower your results.  This doesn't mean that you should pick the closest server.  To be realistic pick one at a distance of over 1000 miles.  Closer servers may provide better results but they aren't going to represent what you would expect to see across the Internet.  The further you go, the harder the test becomes.  In theory you could get similar speeds to any locale if circumstances were ideal.  They never are unless maybe you were on an private, isolated network.  

 

Let us know more about what modem you have and how your network is setup.  Try testing directly to the modem and see if that improves the situation.  Sometimes the modem itself is the issue.

 

Back to the multithread test.  When you run that test, think of it as a HUGE google images page loading.  Pulling from a bunch of resources at once, much like a webpage with document elements scattered about.  It's a very good number to know.  Some people perform slower on multithread actually.  This is often due to slow processing power but can be also brought on by malware, spyware, viruses.  So if you're performing at or above your package speed that's a good thing.  The question is the ratio of your linear results.  From the on test you did in Dallas (the only one I'm considering) and the two multithread results you look like...

 

1-(9.49/((22.83+26.44)/2)) = ~ 0.61

 

So your linear result was 61% slower than your multithread. (3:5 ratio)  Mine, for comparison, is pretty much 1:1 ratio all of the time.  That's what you should shoot for. 

[CA3LE]

Ok, now I'm confused.

When I've been doing some research it often suggests that multithread tests are not as accurate as single thread.

 

I do the multithread test here and it says 25 Mbps

I do the single thread test and it says 6Mbps

 

I get slow loading and timeouts so it seems to me that I do not get 25 Mbps.

 

I didn't notice your reply while I was typing.

 

The linear test is the important test.  If you're not seeing your speed there... it's probably not there.  You can download a bunch of things at once at your full speed but that doesn't help everything.  Looking at only the multithread speed is lying to yourself.  When multithread is much faster than single thread, the speed is there but at the same time it really isn't.  Does you no good if it works how yours is working for you right now.  But if you only looked at the multithread number you'd be scratching your head.

 

In short, trust the classic test (following that link will default you to linear and test against Dallas).  It's a true representation, improvement on it's numbers always equals improved connection quality and speed.

[CaliScotsman]

Thanks for the reply. I'm not an expert in this though so I retain the right to be confused.

 

My network access is a little strange.

 

I'm in a rural area.

There's a Time Warner pipe going to a local radio tower owned by my provider, SuperNet360.

The connection between that tower and my home is via a radio signal transmitted to a receiver on my roof and then down to my Netgear Router.

 

To eliminate my home network and router I am able to connect my pc or mac, directly via ethernet cable, to the radio receiver.

I saw no difference when I eliminate my home network and router.

 

I carried out the test to the San Jose server. (I'm in Los Angeles.)

Linear Test was higher @ 8 Mbps

Multithread was slower @ 11 Mbps

 

I don't know how to interpret this.

[CA3LE]

https://testmy.net/db/yHc9TrI I see your selected on 

 

Multithread test utilizing the following domains: testmy.net, east.testmy.net, west.testmy.net and uk.testmy.net

 

Go to the multithread test and select just the San Jose server.  The uk server may have brought down your result.

 

When I look at your linear result...

 

 

It doesn't look awful to me but depending on where you're testing to it may be a different story.  It's pretty bad but you could probably enjoy some 480p youtube and cruise around pretty good.  Much better than some that I see and much better than my mobile service I often see in my area.

 

Testing Verizon Wireless in my house...

 

https://testmy.net/db/htVZOAai

 

Even testing within the same city...

https://testmy.net/db/ny8HK3kx

 

See how my speed is fluctuating like yours.  Up and down, no consistency.

 

You want to see that graph look more like...

 

https://testmy.net/db/aNjYkJnq

 

Even better... a flat liner.

https://testmy.net/db/WUAGtDMB (testing within my own city though.)

 

According to your response time (this will be integrated soon) the San Jose server might not be your best option.  Your RT dropped significantly when you were selected on multithread, which tested your Response Time against the Dallas server.  Select a few different locations and run tests to each.  Make sure you wait a little bit between tests (at least a 15 second wait between page loads) so that the system can accurately gauge your response time.  When this function is turned on you'll understand it better.  I'll help you interpret the results when you're done.

[CaliScotsman]

I did a bunch of single thread tests and some multi ones.

Then it proceeded to time out.

 

I really appreciate your input although I am still unclear what my genuine effective download speed is from my provider.

Close to the 8 or 9 figure? Or closer to 25 as they claim?

[CA3LE]

I did a bunch of single thread tests and some multi ones.

Then it proceeded to time out.

 

I really appreciate your input although I am still unclear what my genuine effective download speed is from my provider.

Close to the 8 or 9 figure? Or closer to 25 as they claim?

 

Are you testing through a proxy or something?  You're being detected under Google now, where before it said Time Warner.

 

 

I would say that your speed is closer to the 8-9 Mbps figure.  For sure.  But if you're connecting through a proxy service or a VPN, it will slow you down. 

 

On top of that your IP address keeps changing... like at a crazy frequency.

[CaliScotsman]

When I do a WhatisMy IP search it does give me two addresses..must be the way its set up I guess.

I also talked to my provider about erratic speed, timeouts etc. He thinks I may have a cabling issue.

We're going to check that and then take stock of the situation.

 

I will report back so as not to leave my thread hanging.

[CA3LE]

Hope this resolves your situation. Let us know how it goes, you may help someone in a similar situation.

[CaliScotsman]

I will certainly do that. Thank You.

 

(Can you elaborate on my IP address changing frequently? How frequently and is there a tool where I can monitor over time?)

[CA3LE]

if you look at your results you can see the computer ID changing. When that changes it means your detected IP changed. Those tests were back to back and it changed each time, that makes me think it was changing frequently.

tracemy.net monitors that information and will be re-released soon. It remebers all of your past IP addresses.

[CaliScotsman]

I think the google IP was because I've been using the google compression proxy.

 

Anyway, we checked cabling and it seems to be ok.

 

With Testmy and Speedof it indicates my real world speeds are @ 6-10 Mbps DL....even when flipping speedtest indicated 30Mbps.

[Sean]

I gave the Chrome Google Data Saver extension a quick try as I was curious to see how well it performs on my DSL connection with my fixed wireless connection being practically unusable for the past few weeks.

 

One thing I can definitely confirm is that it will heavily skew the TestMy testing:

 

 

That multi-threaded test was performed on my DSL connection which is physically synced at 5120kbps:

 

 

So the Data Saver extension must be turned off before running the speed test. 

 

On the other hand, it does make for an interesting test to see roughly how well the Chrome Data Saver extension is performing.