Review from Toms hardware - Question regarding transfer performance (USB, PATA etc)

junii

http://tomshardware.co.uk/2004/12/06/plug_and_save/page12.html

If you scroll down to the end of the page there are two graphs. On one the performance seems to increase as time goes on and on the other it decreases. What is the reason for this increase/decrease? It seems to be the same for all enclosures reviewed in this article.

Cuddlesworth

The top graph is read performance over Usb, it states that the Usb interface allowed a transfer rate sustained of around 16megs peaking to around 22 megs.

The bottom graph is Sata performance, it states that the sata interface over a Ide bridge(not a true sata enclosure) performs read speeds of around 40megs dropping after sustained use to around 25megs.

So if you plan to hook this up with a Sata cable you will see better read times/speeds or in simple terms your data will transfer much quicker over Sata rather then Ide.

junii

krazy_8s wrote:

The top graph is read performance over Usb, it states that the Usb interface allowed a transfer rate sustained of around 16megs peaking to around 22 megs.

The bottom graph is Sata performance, it states that the sata interface over a Ide bridge(not a true sata enclosure) performs read speeds of around 40megs dropping after sustained use to around 25megs.

So if you plan to hook this up with a Sata cable you will see better read times/speeds or in simple terms your data will transfer much quicker over Sata rather then Ide.

Yes but why does it peak at a certain time or decrease? Is it due to heat buildup or something?

Cuddlesworth

If it was a true sata connection there would be a high burst speed at the start, then slowing down to a steady high transfer rate.

At a guess I would say the differences in the speed over a time and the size of files transfered would be limitations in the Ide-sata bridge and usb interface. But without a full explanation of what exactly tomshardware were doing, type of files, size of files, what computer they were testing on and what it was doing at the time its hard to tell. Transfer speed drops and rises could be attributed to host pc itself.

Snowbat

The enclosure on the previous page is SATA and shows the same falloff for Sustained Read Speed.
http://tomshardware.co.uk/2004/12/06/plug_and_save/page11.html

In Zoned Bit Recording used by all modern drives, media transfer is considerably faster in the outer sectors. Unlike CD/DVD, hard drives record from the outside in so you would expect to see the highest sustained transfer rates in the beginning of a big transfer, with a gradual falloff.

Sometimes, people benchmark their disks when new, and then many months later, and are surprised to find that the disk is getting slower! In fact, the disk most likely has not changed at all, but the second benchmark may have been run on tracks closer to the middle of the disk.

I'm not sure what's going on with the USB2 speed though. The speed increase above 30GB appears to be in all USB2 graphs though with some variance. Unfortunately the tester neglected to mention the drive model(s) used for testing or how the tests were carried out.

junii

Ok, so this may explain why systems often run slower when the hard disk is fairly full. I have witnessed on various occasions machines running faster when files (large amounts of data) have been removed.

Are partitions just logical or does it physically divide the disk. (so to speak)
If there just logical, would this suggest that for optimal system performance you should be using a seperate hard disk just for the OS to try and maximize outer platter usage?

Snowbat

You can have primary partitions and you can have logical partitions within those primary partitions. These map to corresponding cylinders (set of tracks under all the heads) in the drive. The first primary partition will have the fastest transfer rate, the last primary partition will have the slowest transfer rate.

For optimal desktop system performance (assuming drives of similar performance), separate drives for OS, Applications, and data would be nice. This is a situation where the UNIX/Linux way of splitting applications into their constituent parts makes sense - binaries, config files, help files, library files, and log files are each grouped in their own directories and can be moved to separate partitions/drives easily.