External vs internal hard drives and speed
January 18, 2008 12:21 PM   Subscribe

Please explain to me how a 7200rpm SATA external hard drive (USB 2 or Firewire 800) can be written to / read from as expediently as a 7200 RPM SATA internal hard drive? It doesn't seem to make sense. Isn't the USB/Firewire a bottleneck? Why would an external drive be as suitable for audio recording as an internal?
posted by xmutex to Technology (16 answers total) 4 users marked this as a favorite
 
What's the bandwidth of Firewire-800 (800Mbps) or USB2 (320Mbps+)? What's the bitrate you're recording at?

There's your answer.
posted by LordSludge at 12:28 PM on January 18, 2008


To answer the last question, because audio doesn't need much in the way of transfer speed. Consider that uncompressed CD-quality PCM audio only needs a 1x CD drive (153.6 kB/s). Even multi-track audio at a higher sampling rate would only multiply that by a factor of 10, which is still nothing.
posted by smackfu at 12:29 PM on January 18, 2008


An internal SATA is faster than USB2/Firewire but like LordSludge said the data rate from your audio won't max out any of them.
posted by zeoslap at 12:31 PM on January 18, 2008


It can't be written to as expediently, there are two different definitions of speed and they're both reduced with an external drive. The first is the bandwidth of the interface which is how much data can be transferred across it per second. Firewire and USB are slower than the native SATA interface so this represents a bottleneck. The second definition of speed is latency, or how long it takes before the data makes the trip to the platter. Data to/from the SATA drive has to essentially go through a FIFO to match up speeds and feeds.

For audio neither is super critical with modern drives and machines which is why an external is just as suitable as an internal.
posted by substrate at 12:37 PM on January 18, 2008


Slightly off topic but apropos: My experience watching the CPU load while transferring big files (on several different machines) over the two external options you presented (USB and FireWire) is: go with the firewire if at all possible. I can't tell you why, but shifting files via firewire barely affects the CPU, while USB(two) really eats up the processor.
posted by jackbrown at 1:02 PM on January 18, 2008


USB was designed with Intel's hand in it partially to use as much CPU as possible, in order to derive benefit from faster consumer CPUs. Firewire was designed (intelligently) for portable and thin devices, hence little to no CPU usage.
posted by kcm at 1:07 PM on January 18, 2008 [1 favorite]


kcm, I'm intrigued, do you have a citation for that?
posted by PercussivePaul at 1:14 PM on January 18, 2008


While USB 2 and Firewire 400 are indeed slower than the 3.0Gbps that Serial ATA supports, a typical 7200 RPM SATA hard drives top out at about 50-75 megabytes/s (400-600Mbps) for sequential writes. This means that the fastest the hard drive can go when writing a stream of data isn't much faster than what you can stuff over USB or Firewire.

And like smackfu says, audio recording doesn't use up a whole lot of bandwidth. However, real time issues may be a problem, say if a CPU spike slows down your USB performance. Thus Firewire would be a better option.
posted by zsazsa at 1:17 PM on January 18, 2008


USB uses a bunch of bandwidth from your CPU - thats the main bottleneck (not the USB standard). They sell PCI/XC SATA cards to plug a SATA drive in directly; new laptops are coming with SATA ports too sometimes. In the meantime, I'd stick with FW.
posted by lrodman at 1:24 PM on January 18, 2008


One wonders if kcm's allegations re: USB 2 and intel are true, is that what explains the slow disappearance of Firewire from Apple machines/peripherals?

Thanks for the responses.
posted by xmutex at 1:31 PM on January 18, 2008


What kcm and Irodman said about USB. Your USB hub doesn't use much of an on-board processor chip to transfer data but rather relies on the CPU. It's the same idea of xATA v. SCSI- ATA mostly uses the CPU for processing power whereas SCSI uses an on-board processor for the bulk of its data transfer processes.
posted by jmd82 at 1:34 PM on January 18, 2008


The slow disappearance of Firewire is due to lack of 3rd party support. Not many drives / peripherals use it, so consumers want more USB ports. Plus, USB2 is more than good enough for most purposes and defines exactly how to make a new peripheral and have it "just work."

USB / Firewire disks are going to be just as fast because the real bottleneck for most applications with a 7200 RPM drive is the "7200". It's a physical disk. Spinning. How quaint. Anyway, most of the time you're just waiting for the disk to spin around or the heads to physically slew over to the right track. For sustained large-scale transfers the speed difference in connecting wires (USB &c.) matters. For anything else it doesn't.
posted by TeatimeGrommit at 1:39 PM on January 18, 2008


Answer: No, the connection is not the bottleneck.

Elaboration: The hard disk spinning is your bottleneck. A 7200RPM drive is the same speed internally or via USB2.

Granted, there are some minute differences for the amount of cache (8MB, 16MB, 32MB), or different seek times and such, but you won't notice the difference. You will see a difference if you upgrade your internal to a 10000RPM, 15000RPM, or a SSD-HD (flash drive).
posted by mhuckaba at 2:26 PM on January 18, 2008


Many digital PC surround sound speaker systems are driven directly by USB operating merely at the Full-speed mode of 12 Mbs. So there is certainly no bottleneck for audio when using the High-speed mode of 480 Mbs. There just isn't very much bandwidth required for audio.

USB was designed with Intel's hand in it partially to use as much CPU as possible, in order to derive benefit from faster consumer CPUs. Firewire was designed (intelligently) for portable and thin devices, hence little to no CPU usage.

The PC has a USB host controller embedded in it that does all of the low level transmission and reception of data. The CPU is only used to fill and empty the buffers that are used by the USB controller so it has no more CPU overhead than Firewire.

The main difference between USB and Firewire is that Firewire was designed to be peer to peer and therefore requires a fairly sophisticated controller on both ends. USB was designed as host/slave so that the slave peripheral devices would not require expensive controllers and could be dirt cheap. That's why you don't see Firewire mice and keyboards. They would be too expensive to be competitive.
posted by JackFlash at 3:22 PM on January 18, 2008


I believe there are also higher licensing fees for firewire.
posted by gjc at 8:27 PM on January 18, 2008


Whoa there are a lot of wrong answers in this thread. Modern drives easily saturate the effecti bandwidth of a usb 2.0 connection. A lot of people dont realize that drive speed is rated in megaBYTES per second and network/datalinks at megaBITS per second. Depending on where the data is on the drive, you are easily blowing through the usable 300 or so mbps or a USB 2.0 connection. Now toss in whatever latencies and delays the protocol adds to this. This is why we dont just put all our drives on usb 2.0.

Yes, USB is not as smart as firewire, thus USB being everywhere and firewire being no where. THis is not a conspiracy by intel, its grade sc.hool economics. Yes, firewire 400 outperforms usb 2.0 on disk tasks for a variety of boring technical reasons, but if your pc doesnt have a fireware card then its all academic. If you want performance on an external drive you can always shell out for a esata card/cable.

Audio work on a single workstation writing to a single usb 2.0 disk should be fine. Youre not moving that much data. Remember to do backups
posted by damn dirty ape at 9:10 PM on January 18, 2008


« Older Date protocol: when the check comes   |   Coaxingly close Newer »
This thread is closed to new comments.