This was supposed to be a mini-review on a QTEC USB 2.0 Hub I bought recently, but I got distracted …

I'd always wondered if using a USB hub would slow down data transfer, so I moved 43.4 MB of data through it repeatedly (specifically mp3s of Buddy Holly's '57-'59 Rarities Collection, to/from a 64MB SDRAM card in a Samsung Digimax 250 camera), and compared the results to those found with a direct connection between the external device & the motherboard (Abit NF7 running an AMD XP 2800+ chip).

Starting the stopwatch when I pressed CTRL+V & stopping it when the "copy" dialogue box disappeared gave the following results with a direct connection:
* To: 1 minute 15 seconds
* From: 47 seconds

Performing the same test with the camera connected via the QTEC hub gave these results:
* To: 1 minute 32 seconds
* From: 1 minute 12 seconds


So the answer to the question is yes - a USB hub seems to significantly slow down data transfer. Not having another USB hub handy to make comparisons limits the value of this test - it's possible (or even likely) that all USB hubs exhibit the same performance loss. I would have also liked to repeat the tests with a larger amount of data, to determine whether the performance loss is fixed amount of time, or a fixed percentage loss: what I'm getting at here is that in using the hub, I lost 17 seconds on average on downloads ... or 22.7%.

On a small amount of data, the nature of the loss isn't important - but if I was shifting 10gigs, it would make an enormous difference. Extrapolating from the results I got, transferring those hypothetical 10gb (assuming this hypothetical external device was as slow as an SD card) would take 4 hours, 54 minutes and 56 seconds. If the hub causes a 17 second loss, that time would be 4 hours, 55 mins & 13s - a trivial difference.

If however, the loss is 22.7 percent, the transfer time would be 6 hours, 1 minute and 53 seconds. Eek.


As I haven’t come to any conclusions, I’d be interested to hear from anyone else with a USB hub and a mass storage device (faster or otherwise) – do me a favour & repeat the tests with your kit & post the results?

I didn’t notice any performance hit with the other two devices I have connected to my hub (an HP printer and a Saitek joystick), but the chances are it’s there … & any extra lag from an input device is going to hurt when gaming

Nice information Jonny! Never even thought about the possible slow down. Wonder what other variables would change if they come into play with more testing.

My understanding has always been that USB, as a standard, is basically parasitic[sp?]

In other words, USB always uses the CPU to do it's dirty work.

For example, you may get "lag" with a USB microphone that you don't get with a sound card microphone simply because the CPU is doing the work with that USB connection.

So on a hub - even if nothing is connected - the CPU may have some polling/overhead duties that must be performed. How else is it gonna "know" when a new USB device is plugged in?

Another interesting benchmark might be to max out USB [64 ports?] and see how much overhead you get.

-MF

Think of a hub like a 4-1 header on a 4cyl. engine, now think of a direct connection like a 4-2-1.

Besides, with all the ports onboard these days, who needs a hub...

That's weird...every time I back up my audio library to my external HD (Seagate 200G) through a USB 2.0 hub, it only takes about 68 minutes on average for 45G of music.

Is it possible the type of external storage would make a difference in this case?

Rob

Yes Rob, a SD card will run significantly slower than a harddrive.