yashitkumarsingh

why??
and also if parallel is faster than what about USB ....how can this be fast??

Nigel Goodwin

If you're talking about a PC?, then parallel is faster than serial, and USB is faster than parallel (and certainly USB2 is much faster).

But really a LOT depends what you're doing, and why you're asking?.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

yashitkumarsingh

yes, about PC...

but usb is used for transferring data serially , isnt it??
though i dont have much idea??

plz explain the cases??

Nigel Goodwin

Data is written to the disks in hard drives serially, and they are faster still, serial simply depends on how fast you send it!.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

dknguyen

Ideally...parallel is faster since you are sending a bunch of bits at the same time instead of one after another...

However, in reality when parallel gets really fast, the bits that are sent at the same time don't arrive at the same time, and there is crosstalk because there are so many wires involved. What can happen then is you use differentially signalled serial communications (and other considerations) and then hike up the speed a LOT to surpass the speed of parallel which doesn't have the problem of getting all the bits to arrive at the same time relative to each other and doesn't have the problem of crosstalk because there are less wires.

THat's what's happening between the older harddrives and newer harddrives now- with old technology then parallel is faster, but with newer technology you can get faster speeds with serial. So in short, being parallel or serial doesn't determine speed- how you implement it does.

tboydva

dknguyen,

Thanks for your answer to this question!!! I have posed this in hushed tones to all the tech people I know and noone was ever able to give me a plausible answer. I took computer science in 1979 and we learned (quite convincingly) that parallel was the faster transfer protocol (one byte transferred per clock cycle vs 8 clock cycles for serial). This was 1979 mind you, so a byte was a respectable amount of data to transfer. I've been curious why everything has become serial (from PATA to SATA)... I greatly appreciate your explanation. I'm surprised someone hasn't come up with some form of shielding for parallel transfers (but I guess 64 wires would get out of hand!). Thanks again.

Tom

3v0

dknguyen's answer was very good.

They have but you will not see it in consumer products because of the high cost.

dknguyen

Cost is an issue too to switching to serial. It's just cheaper (and looks prettier). The cost of a serial (with it's slightly more complex signalling) and the better materials involved probably costs less than making a parallel interface that runs just as fast (even if you didn't have to spend money to overcome the crosstalk and timing problems) since the parallel interface changes all the circuitry at a fundamental level taking up more space and more circuitry (to handle each bit in the parallel interface).

Of course there might be a parallel interface somewhere really high end that uses the same high speed signalling used in the high-speed serial interfaces. But it probably costs a fortune.

It's also easier to bend 4-wire cables around the inside of your PC than a 32-pin ribbon cable.

Leftyretro

Well certainly when asked as a academic questions, and assuming identical bit clock rates, one would have to say, for example, an 8 bit wide parallel interface will be at least 8 time faster then a serial interface.

For serial data one should also add the time needed to create the serial stream (the parallel to serial conversion done in UART or software) and to convert back to parallel data. The hardware costs should be somewhat less for the serial link but probably not 8 times cheaper if we are talking standard wiring costs.

So you take your trade-offs and make your choice. It's really a application + specifications and + budget driven decision, no?

I suspect the fastest buses in use (supercomputer?) most likely utilize parallel fiber optics channels.

Lefty

__________________
Measurement changes behavior

mneary

Documents on the BlueGene/L (fastest supercomputer in the world) suggest that the communication among the 65,536 dual-processor modules is a combination of "Tree link" and gigabit ethernet. It isn't stated, but both would appear to be serial on copper. When it needs more than a gigabit/s it appears they use multiple serial links.

Nigel Goodwin

I would have thought it's likely - optical isn't much advantage because you've got to convert it either end.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

donduvall

Isn't multiple serial lines called parallel.

dknguyen

No. THe data in a parallel interface is split up into chunks and sent down the parallel lines simultaenously. In a serial message, the chunks of a single message are sent down the same line sequentially.

In multiple serial lines, the data chunks sent down each line is not part of the same message (and therefore does not have to be time synced among the lines). All chunks that are part of the same message are sent down it's own dedicated line sequentially.

So suppose one line gets corrupted somehow, in parallel, the message is destroyed. In multiple serial lines, only one message is destroyed and the messages on the other lines are unaffected.

Lesh

PCI Is a pretty darn fast parallel interface. 32 bits x 66 mhz is 2.1 Terrabits per second.

mneary

Uh, almost. It's 2.1 gigabits.

And it's spelled terabits. Terrabits would be Earth bits.