• Content count

  • Joined

  • Last visited

  • Speed Test

    My Results

Everything posted by CA3LE

  1. Can you post examples? I looked through your results and didn't find any that looked like that. Maybe you deleted them. You were talking about the TiP result, within the individual test results, correct? Keep in mind that your overall score is calculated completely separate from the TiP result. The TiP minimum, maximum and average are taken from calculations that are separated from the main event. Some of your tests are running more smoothly... ...and some look like a struggle to get to the end. Obviously, it's best if it runs like the first scenario most of the time. But if you see an extreme spike either way and then it's followed by normal plots, simply disregard it. If you have time, send me a private message with the TID so I can look at it and use the information to improve the algorithm. But keep in mind when that happens, your overall result is not affected. It's not as if the TiP result aggregates into the final result... TiP is a totally separate entity. I may put a 'report inaccurate results' link in an upcoming version to make it easier for users to flag results for a closer look. In this case, I personally never see spikes like you're seeing. It's hard to program a fix for something when you can never see it yourself. But I do learn from your results and apply it. I often spend hours going through live results to look for anomalies. TiP is already much more accurate than the original concept but I know there's always room for improvement. I think the issue on some machines is that something in your software configuration is causing intermittent delays in browser transactions. When these delays happen during the test it causes a misreporting of the TiP numbers. Probably because the times are taken so closely together. Sometimes the fact that a computer is not able to run TMN like the majority of the population is a clue. Example, my friend _____'s laptop can hardly run TMN right now. It used to run it perfectly (another clue that something new is up) and all of the other computers and devices on his network run TMN to the best of the networks ability. The issue is in loading. The test loads and runs but seems to get hung up on the forwarding around that TMN does during the test. Disabled his virus and internet security protection, because sometimes those programs have been known to cause issues... nope. Still does it. Now if this were anyone else I might dig deeper but he admits that he does things online that he probably shouldn't, opens and clicks things that he shouldn't, etc. Per my recommendation he's backing up his files and I'm going to format it and reinstall windows for him. I guarantee, when it boots up... it will no longer have that issue. Sometimes you don't even have to get to the results page to get a result at TMN. Hell, I often start a test and cancel it part way through because I got the point of how it's running just by how it loaded. Please forward me those results in the future so I can make sure it's not something I can improve on.
  2. If the other people on the network are currently downloading or uploading it will affect the result for sure. Most of the time however it's not necessary to disconnect anyone. Chances are, they're pretty much idle in the grand scheme of your connection, even if they're hitting webpages at the time. Now, if you suspect that one of the users is eating up bandwidth it may help to test first to get a baseline with everything connected normally, then remove the other connections one in order to determine which machine is at fault. It can be a quick, easy way to troubleshoot. It's always accurate. It's just not going to give you favorable results if other people on your network are using a significant portion of the connection. That lends to the accuracy. If you add the speed of what they're doing and your result... then run an identical test without the variable of the other users, unless there are other variables at work, you'll always find that it matches. No, you shouldn't. I never do. But if my result is way lower than expected... I'll start looking at the other connected machines... and then the internal network before I ever assume it's my providers fault. 9 times out of 10, it's my end. Something needs a routine reboot, I forget about a file transfer on another machine.. etc. But the variation in my normal result shows me that there's something not right and helps me track down the culprit every time. Someone told me about how it helped them figure out that their neighbor was leeching their connection, torrenting. Although like mudmanc4 said, there's a definite benefit to isolating variables. When you're at the end of your rope. First, get your baseline hooked up normally. Then if you're on wifi, test directly connected to the router. No improvement? Then test directly wired to the modem (if your modem is separate from the router). If you still have the same issue, connecting a different computer to the same direct connection will prove that it's not isolated to the one machine. Remember, the modem will need to be power cycled between connecting and disconnecting the modem from anything. It must be assigned to the new mac address of the device it's plugged into.
  3. Add yourself to our new frappr map [LINK REMOVED ... Site no longer exists] just a fun way to see where our members live, will also give you an idea how close you are to the test server. - Damon
  4. Read this ... long story short, you most likely have the 7-11 Mbps TMN is telling you. But you're in South Africa and the closest server you've tested on is in the UK. I also see that you did multithread tests but selected all of the server, which can bring your score down. Try multithread again, this time only select UK or DE servers... you could even try both at the same time. Just don't select the wide array you did before. These results will probably be your highest. Both results are true, they're just tested differently. The default single thread test will show you how fast single files are able to download and streaming ability. When you multithread, imagine that it's not downloading a single source, it's grabbing multiple files at the same time. You may be able to perform better in aggregate but not in a single thread. You also may be able to perform better on that single thread if the server is closer to you. By the way, TCP optimizer helps older windows based machines perform better at distance by adjusting MTU and RWIN values.
  5. Welcome to!
  6. Probably more likely that it's because Comcast has a wider sample, with more variety. Doesn't mean that they're worse or less reliable. They have options for lower packages, the people with the lower scores may be happy. Without knowing the package speed and the scenario the tests were taken under it would be hard to fully make an assessment like that. I have improved charting that's on the way that will give you much more useful information when doing those comparisons.
  7. Log in next time you test at your office so I can find your results on that connection.
  8. I had temporarily disabled the larger upload tests because I noticed an issue but it appears to be working correctly so you can now test upload up to 100 MB again. So you're talking about a different connection from the one you're posting from right now, right? Because your test sizes are more than adequate at those speeds. Even at 100 Mbps (12.5 MB/s) a 100 MB test gives you 8 seconds of testing which is more than enough to get accurate numbers. Having said that, larger tests are coming. Both upload and download.
  9. The most common question that people ask us is why their results here differ from I want to explain something that may help you understand why this is. According to the Ookla Wiki the following is true about your tests taken at The fastest 10% and slowest 30% of your results are DISCARDED The remaining data is averaged together to determine the final result Throwing away the fastest 10% and slowest 30% of the results in my opinion does not make for an accurate test. Isn't THAT the data that you're here to see? Also note this isn't only true for Ookla is huge, most of the speed tests online run their software. Source: Multithreading makes your connection look better Ookla speed tests are also multithreaded. Meaning that they open more than one connection to the host and combine the speeds. This often can mask congestion issues. For instance, imagine that you have a pipe along your route that's limiting you to 10 Mbps. If you open more than one connection through that pipe you'll be able to achieve a faster speed... but testing that way will not clue you in that there is actually a problem along the route. Remember, you shouldn't have to multithread your connection to pull your full speed. Multithreaded speed tests, like the majority out there, are designed to benefit the Internet provider not the consumer. They often display your maximum throughput not your throughput over the course of an entire upload or download. Omitting the worst portion of your test resulting in inflated scores that may make you feel warm and fuzzy but aren't going to help you see and resolve connection issues. Multithreading can show that you're able to max out your connection by combining the speed of concurrent connections but a great connection can max out without having to multithread. So, if you have to multithread your connection to get your full speed, you should be asking why that is. With it's instantly apparent if there is a congested route. Your speed here reflects the actual loading time of data within your browser... not the combined efforts of multiple threads being altered (dropping the top 10% and bottom 30% of the results) and added together, all through a plugin that isn't suited for the task. :: EDIT :: If you'd like to multithread with TMN, I offer that as well. Try the multithread speed test. TMN's multithread test gives you the option to select from an array of servers. Allowing you to test across multiple routes, to more than one server at a time. is the only Internet speed test with this ability. So THAT'S why your results differ. -- is a harder test to ace and a lot less forgiving than other speed tests. But isn't that what a benchmark should be? :: EDIT :: Another thing to consider is if you're running windows, particularly if it's older than Windows 7 you may need to tune your TCP stack to see your full speeds. Windows doesn't always come out of the box optimized for fast connections. Flash based speed tests fail to detect this problem. To make this change for free I recommend TCP Optimizer. This makes changing those settings very easy and nearly fool proof. Just open it, slide the bar over to your speed that you're supposed to have, check "modify all adapters" ...apply the settings and reboot. You should have faster speeds after your re-test if that was the problem. Here's an outside source talking about exactly this issue, I'll hunt down more examples. >> Download Speed Test - Something fishy? A few related topics :: Accuracy.. Slow upload, drastically different from other speed tests Resuts Vs. Actual Download speeds? shows different results than other speed tests TestMy v Ookla Difference Between Speed Test Sites? Satellite Service Testing Locking at around 14Mbps Questioning resolved with Outside Sources :: Internet providers caught inflating speed test results [] Use or else!! NO MORE posts will be allowed!!! by the Admin on [] Accurate Speeds Or Speed Test Error? on []
  10. Here's a couple more multithread tests on different servers in Sao Paulo. If you perform better on these then it may be that the peering to that server I just threw online for you isn't very good. Like I said, they were untested. These are two examples taken from's public list and entered into Mercury Speed Test. By taking any of the URL's listed you can reference the files used by to test... but use for the actual testing. <server url="" lat="-23.5000" lon="-46.6167" name="Sao Paulo" country="Brazil" cc="BR" sponsor="Vivo" id="6696" url2="" host=""/> So you find the base URL Files used random350x350.jpg 240K random500x500.jpg 494K random750x750.jpg 1.1M random1000x1000.jpg 1.9M random1500x1500.jpg 4.3M random2000x2000.jpg 7.5M random2500x2500.jpg 12M random3000x3000.jpg 17M random3500x3500.jpg 23M random4000x4000.jpg 30M So if you paste... ...into the Mercury Test, it will create a multithread download speed test using that 240 kB test file. Faster connections benefit from larger test files, as these are repeated to meet the final test size. I do see other people testing at high speeds on that server but it may be that those specific people's ISPs peer better with the providers that host uses. I need to see more data before I look for a different solution.
  11. This post cannot be displayed because it is in a password protected forum. Enter Password
  12. I have A LOT of upcoming releases, many of which are going to be built around a new concept. Because there is so much core programming being changed I'm going to do something I've never done. I normally work on the site live and things are released as I write them... if your in the right place at the right time and you pay enough attention you can see me building sometimes. Well, I really feel that the version I have out right now is INCREDIBLY stable... I don't want to disturb any testing or use of the site. So I'm going to offer a beta program for those who wish to join. To join, vote Yes on the poll above. I'll add you to a special member group and you'll be emailed with updates. You'll also have an option in your menu that will allow you to switch between beta and release. Smaller updates will be globally available but the major ones will be held back for the beta testers to play with first. This will help find bugs before public release. When I'm satisfied with the code I'll offer an option for a while to the general public to turn on the beta... after general public beta testing I'll then release it to the final. We'll then start the whole process again for another round. This is going to be really hard for me to do, I really hate to hold new stuff back. But maybe doing it this way will help me spend less time on bug checking and more time of actually building. I'm one person. Your help with this is appreciated. Everyone who has ever submitted a comment... you've already helped build this site. I started with a vison but you guys expanded it. If I'm going to keep bringing those ideas to paper I really need your help. I have at least a couple thousand hours of programming planned just for the next 3-6 months. I need to use my time effectively, spending hundreds of hours searching for bugs is not using my time wisely. Register (If you haven't already), vote yes above and I'll email you later when I have more information. Then, you can tell your friends that you helped build this site. Thanks -D
  13. Especially since you mention issues with streaming I would compare the multithread speed test to your previous results (tested with a single thread). (ookla) does a number of things that make things appear faster and make it less likely to help you in your case. They multithread, which makes it difficult to see issues affecting single thread performance. They also adjust the result, removing the top 10% and bottom 30% of the result. In my opinion, negating the entire test. So if you want to get something fixed, is a far better choice. Comparing single thread to multithread with TMN can identify these issues, a healthy connection is normally able to perform equally well on both tests. Default linear test (bottom) and multithread result. Both taken against Dallas on my home connection (in Colorado Springs > 700 miles away). Enable multithread, select and test again so we can see.
  14. - Sao Paulo BR is online This can be found on -- should be immediately available to most but may take 24 hours to propagate DNS for some people. Let us know how it works for you.
  15. Still waiting on activation. They're taking longer than most to get something online.
  16. Hey @That_Eriksson I'll have a server online in Sao Paulo Brazil for you in just a few. I'll update this topic when it's online. It's with a hosting company I haven't tested so we'll have to see if it provides good comparison for our South American friends.
  17. Your post may have been in moderation or something, because you're new. is a independent third party, no ISP has influence over the content here. They're able to post and read the same as you are and can't delete or moderate anything.
  18. I have renovated a bit.. not only looks wise but I also made one major change. When / if you score over what I think is acceptable for the test size you have chosen you will automatically be redirected to the next size up. This increases accuracy a whole lot ::: Download Stats ::: Connection is: 3911 kbps about 4 Mbps (tested with 1544 KB) Download Speed is: 477 Kb/sec Auth Code: 2058666 (validate a Bottom Line: 70 times faster than 56K you can download 1MB in 2 second(s) Validation Link :: So you don't get a burst score... you get a nice steady download score.
  19. I added a couple more options for our members. Going up to 500. Keep in mind, you must be signed in the see and use these options. Hope this helps!
  20. Verizon's speed test also multithreads. If you want to multithread for comparison toggle the option on the multithread speed test, you can then use the site normally and the download speed test and automatic speed test will be multithread also. I'm curious to see what your results are using the multithread option. Let me know after you've tested.
  21. ... super delayed. Happy New Year!
  22. Welcome not_a_happy_camper! First, the averages are always current. They are just brought down by people with slower results. By default it shows the last 50 download and 50 upload test results. change the drop down Looking at Verizon FIOS speed test log I find some people with faster results but when I look closer at the 'connectID' results, some display a seesaw pattern. But then I find a result for someone in Belcamp, MD and they're consistent. Although they haven't tested as often. Here's one in NJ that's incredibly stable Are you using wifi? Have you tried testing directly connected (ethernet) to the modem or router? A 50 Mbps is a high overall average for an ISP on Look at the top Internet performers.
  23. It should do that automatically for you. If you're below 1000 Kbps it will show in Kbps, above shows in Mbps. Next to that you'll see the conversion to kB/s and MB/s. Again, if you're lower than 1 MB/s it will display in kB/s. Where it may become a little confusing is when you look at your graph. Which currently only displays in Mbps, rounded to the 2nd decimal. Your 658 Kbps result converts to Mbps by dividing 658 by 1000.
  24. Note: you can also do this for states. Examples: Arizona download speed city rank with a minimum of 5000 recent results. California download speed city rank with a minimum of 5000 recent results. Colorado download speed city rank with a minimum of 5000 recent results.