ESATA vs. USB 2.0

[ruffhouseutah]

I'm trying to find the fastest way to back up hundreds of gigs of pics and video from my internal SATA drive to an external drive.

My board is an EVGA E758 X58. Internal drive is WD Caviar Black w/ 7200 RPM, 32 MB Cache. My external drive is also 7200 RPM, with 16 MB cache, and has both USB 2.0 and eSATA connections.

Using USB 2.0 the transfer goes about 10 MB/sec before I cancel it.

Using eSATA the transfer starts at over 100 MB/sec, then over the a few minutes it slows to around 7 MB/sec, then slowly climbs back to just over 40 MB/sec.

I'm guessing that if I hadn't cancelled the USB transfer if it also would have climbed to over 40 MB/sec, but I didn't have the patience to wait around and find out...

I had just purchased the eSATA cable hoping that eSATA would be much faster than USB.

Any ideas on how to make this go much faster? Is 40 MB/sec all I'm going to get, even though SATA II is, in the specs, supposed to be capable of up to 300 MB/sec?

 

[sminlal]

USB 2.0 has a theoretical bandwidth of about 60MB/sec - but because of protocol overhead in practice you're usually limited to about 30-35MB/sec. I personally have never seen a USB drive with a sustained transfer rate more than about 37MB/sec.

There is no question that the eSATA connection will give you much better throughput than the USB connection.

You'll get much better performance if you're copying very large files. When you copy a lot of small files the drive ends up doing a lot of seeking in order to find the files and write the directory entries - that could be a reason why you saw the transfer rate drop.

Another reason would be if the source or the target disk is doing some other, non-copy work at the same time. Any drive that's accessing multiple files at once will be much slower due to the extra seeking required.

 

[ruffhouseutah]

USB 2.0 has a theoretical bandwidth of about 60MB/sec - but because of protocol overhead in practice you're usually limited to about 30-35MB/sec. I personally have never seen a USB drive with a sustained transfer rate more than about 37MB/sec.

There is no question that the eSATA connection will give you much better throughput than the USB connection.

I'm not doing any other tasks, and the files are large, so it's a pretty fair comparison.

You've convinced me that eSATA is quicker than USB 2.0, but shouldn't eSATA be quicker than about 45 MB/sec?

Is it possible that my HD enclosure's eSATA connection/controller just isn't very good? Or is that about as good as it's gonna get with any transfer?

I've heard of others getting 90 MB/sec--double mine--but why can't I get closer to the 300 MB/sec spec of SATA II?

 

[sminlal]

SATA is the pipe between the disk and the memory. Just because it's rated at 300MByte/sec doesn't mean that a disk can actually deliver data at that rate. It's just like a posted city speed limit of 30MPH doesn't mean that all traffic can actually go that fast - bicycles will go considerably slower, for example.

Most hard drives these day can get up to around 90-100MByte/sec - but that's only under optimal conditions. If the files you're transferring are on the slower, inner tracks rather than the faster outer ones, 45MB/sec is probably about right. Remember that there are two drives involved in your copy, and the slower one will be a bottleneck for the faster one. So everything has to be just right on BOTH drives in order to get the fastest possible transfers.

 

[Paperdoc]

sminlal is dead right. The speeds you are seeing are what I've seen as "typical" for those systems. Actual average data transfer rates over a long set of file transfers are often in the range 30 to 40 (max) MB/s for USB2, and 40 to 60 MB/s for true eSATA. Anyone getting close to 100 MB/s sustained on eSATA should be really happy!

 

[juxtawill]

sminlal wrote:

If the files you're transferring are on the slower, inner tracks rather than the faster outer ones...

Constant Angular Velocity is used on magnetic disks, this means that the r/w speeds are the same on inner and outer tracks. The bit spacing is more dense on inner tracks and becomes less dense the farther out the head goes. If you've wondered why sectors are shaped like a piece of pie this is it. This allows the r/w speed to be constant while the disks revolution is constant (5400, 7200, etc...) and concentric circles, that make up track formation, circumference to increase as the r/w head moves outward from the center.

Pie shaped sectors containing bits: (ignore the = signs)

Inner track bit spacing =========|-1-0-1-0-1-0-1-0-1-|
Middle track bit spacing ====== |--1--0--1--0--1--0--1--0--|
Outter track bic spacing = |----1----0----1----0----1----0----1----0----|

In contrast optical discs use Constant Linear Velocity (CLV), have one spiraling track with evenly spaced bits, and to achieve a constant r/w speed the drive adjust the spinning velocity.

 

[WyomingKnott]

It's just like a posted city speed limit of 30MPH doesn't mean that all traffic can actually go that fast - bicycles will go considerably slower, for example.

A gross calumny against many fine bicyclists. <i>Most</i> bicycles will go considerably slower. In my youth, I got above 50 MPH (of course, that was coming down off a Rocky Mountain pass, not on the flat).

 

[Paperdoc]

Looks like WyomingKnott has 15,000 rpm legs! VelociKnott, he is.

 

[fionah0]

Looks like WyomingKnott has 15,000 rpm legs! VelociKnott, he is.

Yoda, Paperdoc is!