Earthlink DSL → DSLR speed test versus...

I was checking my DL speed last night because I'd been getting slowdowns during peak hours... Anyway, I happened to have the Win System Monitor running, showing bytes per second received. I don't know how accurate it is, nor how well it correlates to BITS per second over TCP/IP; but relatively speaking it gives a good, quick indication of whether my DSL is getting choked.

Someone here had mentioned using a speed tester at »www.computers4sure.com. I had used it before, and the rates were consistently higher than those reported by DSLR (all three servers). I had wondered about that, but wasn't sure if it was an apples to apples comparison. Using the System Monitor, though, clearly indicated a higher speed from computers4sure, than using the DSLR servers.

I'm supposed to have a 1500kbps Covad connection. From the DSLR servers, I top out at around 1250kbps. Since this is only 83% of 1500kbps I figured that 'overhead' was higher than I could explain, or my line quality wasn't the greatest. Using the computers4sure site, I get 1350kbps. This is about 90 percent of 1500kbps, which is LESS overhead than I expected; and my line must be pretty good, too!

Again, I don't know if this is apples and oranges (TCP versus UDP or who knows what?), but the computers4sure test server can definitely get more bps into my computer than the DSLR servers; and it seems to be a better indicator of the raw speed capacity of my DSL connection. I'm now telling people that my line can support at least 1350kbps download. (I didn't try using the System Monitor AND another download session to see if I could get an even higher rate from two simultaneous connections to different servers...)

(Note that I still think that the DSLR test is superior, because of the UL rate measurement and the auto logging of test results.)

87% 1500 would be the best accurate speed that you can get.
Thats about 1305/334

[text was edited by author 2001-03-16 15:30:28]

Does ADSL perform data compression? The DSLR speed tests are done with "noncompressable" files, so they measure the actual speed of the line. If the files used by other test programs can be compressed, they'll give you a higher reading than the actual bit rate of the line.

I just tried that test, and it appears to use a very small file for testing, and appears to use page loading as part of the factor, and this type is usually less accurate in real world results. DSLR uses much larger test files (which is why it takes more time to run), and should be in line (normally) with what you get in real world downloads.

Well, this is what I meant about apples and oranges...

I agree that the computers4sure test uses a smaller download size, but it's not like an order of magnitude smaller. I also agree that DSLR is probably a more intelligently designed test and more representative of real world results (like winmark et al tests); but the fact remains that the bps into my computer is higher using the computers4sure test.

As for the 87% number, what is that considering? And are all the considerations relevant to the modem or computer port bit rate or the effective bit rate? For instance, DSL is ATM, which alone has a 10% penalty (48/53 bytes). Ethernet has its own overhead, which is packet size dependent... but that probably should be included in the bps figure.

I've found another relatively accurate speedtest site : »speedtest.umnet.umich.edu/ This seems to be fairly close to DSLR's speed rating which leads me to believe it's close enough to true speed. Most other site's such as MSN's or CNET's speed check come up short which also can be related to the number of hops to get there.

13% is the ATM overhead

Actually, the file (for higher speed connections), is a order of magnitude higher, watching the computers4sure test, it couldn't have been more than 150KB, and the DSLR test uses a 1MB test file or larger, pretty big difference. The bps isn't necessarily any higher, just their calculation of that number is higher.

The 87% number takes into account the approximate average overhead from the line sync speed to the maximum throughput to the internet.

said by des588:

---

I've found another relatively accurate speedtest site : »speedtest.umnet.umich.edu/ ...

---

That *is* DSL Report's test applet. They've simply licensed it... Some people will get higher numbers from it; some will get lower numbers. When you're using the same test, the path (network route) you take will then be a factor.

I'm about the same number of hops and get roughly the same ping times going to either.

Then you have the DSLR applet at www.stickdog.com. I'm only (8) hops away and the worst ping I get is 67ms, yet I get lower numbers from it...

Well, let me take those statements back Just tried both sites and got 1200+k from both !

mballard: According to Netscape under Linux, while running the test, it downloaded a file 1693k in size...