Why are real speeds different to specifaction speeds? 1

[KimLeece]

Hi-
It has puzzeled me for a long time as to why data transfer speeds are not the same as the quoted specs. For instance - USB 2.0 is sold as a device that transfers data at 480Mb/s - but it doesn't! Network cards of 100Mb/s also don't! If someone could fill me in on the facts and reasons for this type of thing I'd be really pleased. Probably I'm missing something simple - but I would appreciate a lesson!

Kim.

'Everybody is ignorant - only on different subjects.'
Will Rogers.

 

[KimLeece]

Sorry for the spelling error in the title - I of course meant specification!

Kim.

'Everybody is ignorant - only on different subjects.'
Will Rogers.

 

[Blujacket]

I've noticed that USB 2.0 ports are frequently listed as "up to 480Mbps". Maybe the definition of Mbps is needed, it is Megabits per second. Small b denotes bits whereas capital B is Bytes. My network does get its full 100Mbps speed.

 

[KimLeece]

Thank you Blujacket! I knew it would be simple! I was not really aware of the b's (bits) and the B's (bytes) - doooh - So I suppose that one normally has to divide by 8 for the correct answer - and like that, my hardware does work properly!

Kim.

'Everybody is ignorant - only on different subjects.'
Will Rogers.

 

[franklin97355]

You also have to take into account what is driving the device. Your source may not be able to supply the full rated data stream.

 

[dakota81]

They do specify what speed bits are transfered at, because that is really the only way they can accurately label them; you need to accept that even if it sounds far fetched right now. Network controllers do run at 100mbits/sec and usb2.0 controllers devices to run at 480mbit/sec.

If you observe that you are receiving maybe just 10mbit/sec over the network connection to toss out an example; the network card is still transfering data at 100mbit/sec, but the key is data is only being sent across the wires 10% of the time.

So the specified numbers tell you what the link provides in terms of speed. But the observed speed rather is how fast can the device provide data to the link controller to be sent.

 

[KimLeece]

Thanks for the replies. I noticed recently when I wanted to transfer a game file from a laptop to a pc that using a PCMIA network card (Xircom 10/100) to the wireless router (router also 10/100 and PC also Realtek 10/100 - 100 link lit up on both interfaces)and then to PC that is was incredibly slow - more than 2 hours for about 180MB!! When I changed the laptop PCMIA card to a wireless card it did it in about 40 mins - I really couldn't understand why there was such a difference. In another setting we use a firewire harddisk to capture DV video - no problems - it's very fast and can transfer 18GB in realtime - about an hour for an hour tape.

Kim.

'Everybody is ignorant - only on different subjects.'
Will Rogers.

 

[skiflyer]

Networks are funny beasts. You've probably noticed the 10/100 mbits issue... alot of times a protocol won't match up, or some chain in the network is running slow, and so you'll get the 10mbits. Longer the chains, more likely it is to happen.

There's also issues with hubs v. switches and whether you're splitting up the bandwidth evenly or if one hog is taking 90mbits and so on.

Then there's the discussion of resends, if it's a noisy line you can be resending data at very high percentages.

Then there's this and that and I'm no expert on network topology, just learned enough to know it's not so simple, and that alot of setups get that 10 rather than 100 out of their gear... often because one computer or old hub has outdated equipment... or they never got the right driver for the card or whatever.

Random anecdote, relating to the source point... I once had the good fortune to work at an office connected to whatever it is they call the academic portion of the internet (I'm not sure what these speeds are, they're not gigabit, I don't think, but they're definately faster than anything I've ever used). I was downloading the Debian iso's from one university to my desktop. It actually took less time to transfer the 2.4 gigs over the wire, than it did to copy from the tmp file (I'm assuming mostly in ram) to the hard drive... in large part because it was a windows machine trying to save 3 very large files at once, but still amused the heck out of me.

-Rob

 

[cdogg]

Another key point is that hard drives are rarely able to transfer faster than 45MB/s. Unless the file is completely sequential, you'd actually be lucky to get more than 35MB/s even from a fast 7200RPM 8MB cache drive.

Now of course, that's plenty fast for a 100Mb/s connection, but hardly fast enough for either a USB 2.0 or gigabit ethernet connection.

~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle

 

[1Drisnil]

In a nutshell,

it takes time to optimize for maximum capacity. As mentioned above by skiflyer, Networking devices have their own sets of idiosyncracies to get them to behave @ maximum.

In a network you need to take into account networking devices, protocols in use on the network, the servers (if you have some), types of NIC devices, drivers that drive said hardware, and then you have the "bps" -- bits per second. 100Mbps = 12.5MBps. In addition, their is the cabling is you are using wired networks -- is the wired network certified to run at a certain speed. There are settings such at Duplex that affect speed performances.
Hubs share bandwidth, switches minimize collision domains effectively providing 100% bandwidth per port between 2 systems provided that its one system per port!

In the wireless world, I am sure there are many more parameters that affect speed and bandwith -- some which are standards and some are vendor-customer driven.

Hope this adds some insights to your issues.

-Don't be afraid of the dark. Be afraid of whats in it!