mind your b’s and K’s: the arcane art of measuring download speeds

I’ve just upgraded to the 30 MB/s internet plan at Charter cable (and added HBO so we can watch Game of Thrones), so here’s the obligatory speedtest results.

It occurs to me that the units for download can be incredibly confusing. Charter advertises the download speed plan using units of Mbps. So, the question naturally arises, how long should it take to download something 18.3 GB in size? (and a related question, if I am downloading something at 300 KB/s, am I getting my max download speed?)

1 GB refers to a gigabyte (10^9 bytes) in this context, since we are talking about file sizes and network speeds. If we were talking about RAM, a GB would actually refer to a gibibyte. However, 1 Mb is a megabit (10^6 bits), not a megabyte (10^6 bytes), because of the small-case b. So 1 Mb is actually 1/8 MB (since there are 8 bits per byte).

So 18.3 GB downloading at 30 Mbps should require:

(size) / (speed) = (time)

(18.3 x 10^9 bytes) / ( (30 x 10^6 bits / sec) x (1 byte / 8 bits) = 18.3 x 10^9 * 8 / 30 x 10^6 = 4880 seconds = 81.3 minutes

Wolfram Alpha gets the answer right, too (and I like teh natural language query – very intuitive).

Now, suppose I’m rocking 300 KB/s according to a certain beta software download client. How am I really doing? The capital B means it is kilobytes, so that’s actually 300 x 10^3 x 8 = 2400 x 10^3 = 2400000 = 2.4 Mbps. Wait, what??

I’m only getting 1/10th my actual download speed for this??

This is why it’s important to do the math. Of course, the download speed may be limited by a lot of other factors, most notably how fast the server at the other end can deliver the data. I clocked almost 40 Mbps doing a speedtest with some local, low-ping server somewhere, but for downloading this big file I’m probably going a lot further and their server has a lot more to do than humor my ping requests. I guess I should be satisfied.

(But, I’m not. grrr….)

3 thoughts on “mind your b’s and K’s: the arcane art of measuring download speeds”

  1. I usually use 10:1 as the bit:byte conversion rate, with the extra 2 being for overhead of various kinds. In other words, if I have 1 megabit of download, then I’ll get about 100 kilobytes per second.

  2. This may be hazardous on high-overhead networks, such as radio (WiFi) and USB.

  3. My preference is to calculate the exact number i should be getting, then compare to measured reality. That gives me a sense of my performance – bundling in “overhead” as a rule of thumb obscures the information I am trying to divine for my purposes.

    Pete, I should have mentioned that yeah my PC on which I am doing the test is on WiFi, not directly connected to the router. Good point. Since I am using 802.11n 5GHz, I am assuming that the speed of that connection is not the bottleneck (though I havent carefully parsed the table here to verify)

    at any rate, the more important takeaway here is that your max download speed is more a measure of your simultaneous download usage capability than for any one file. There’s a plateau on how much bandwidth any one device can actually use, even if for HD-quality streaming (did someone calculate the BW requirement of an iPad HD streaming a movie somewhere?). The more BW you have the more devices you can support simultaneously, so I can game on my PC while my kids both watch Netflix on separate iPods and iPads. I think we have 5 devices in the house that support video streaming, not including the PCs, so this matters.

Comments are closed.