CardBus S-ATA Controller
Notebook goes S-ATA

First Tests

Introduction

It all started when I was doing benchmark tests with external 3.5" harddisks. Current 3.5" harddisks have a sustained data rate that exceeds by far the performance of FireWire 400 (1394a) and USB 2.0. It is very frustrating to see that the transfer rate is clipped across 2/3 of the available capacity. Besides that, you never know whether the interface controller of the Notebook or PC is the limiting factor, or if the bridge controller inside the external enclosure is the culprit.
I would have never thought that S-ATA and Notebooks would have anything to do with each other. But when I was browsing a mail order for external FireWire enclosures, I came across a CardBus S-ATA Controller. But how should I know whether both, the CardBus slot of my Notebook and that S-ATA controller, can provide the appropriate throughput? Buying an S-ATA harddisk was a little to much risk, since my PC is an old Ultra-160 SCSI system running under Windows NT 4.0. But then I came across an ATA to S-ATA bridge at a different mail order. That looked like the right equipment to start with, and I was lucky enough to receive the two orders at the very same day!

CardBus S-ATA Controller

The S-ATA CardBus Controller is from an undisclosed manufacturer and seems to be distributed by Tragant Computer here in Germany: "Serial ATA CardBus PC Card, 61605 ATA PCMCIA CardBus zu S-ATA, EAN 4 043619 61605 3". The card itself has front and back labels: "32-bit PC Card, Serial ATA CardBus Host, 5307-0737-1" / "5307-0445-1 V1.0, S/N 200300031831 -".
When inserted in the CardBus slot, it identifies as "Silicon Image SiI 3112 SATALink Controller" (Vendor ID 1095, Device 3112). The supplied drivers installed without problems.
For connecting the disk, a special cable is provided to connect the 15-pin connector of the CardBus Controller with the 7-pin S-ATA connector. The cable has a length of only 0.24m (0.31m tip to tip) and is a little stiff.

ATA to S-ATA Bridge

The ATA to S-ATA bridge is distributed by DeLOCK: "DeLOCK IDE zu S-ATA Converter, 61301, EAN 4 043619 61301 4". The small box is labelled "SERIAL ATA TO ULTRA ATA SD-SA2IDE-A1, S/N NO. TS 03120007791". The size is 61.5mm×29.8mm with a thickness of 13.7mm excluding and about 21mm including the connectors. The weight is only 17g, so you won't notice the extra weight once it is connected to a 3.5" harddisk.
Thanks to the clear plastic case it is easy to identify the controller: Silicon Image SATALink, SiI3611CT80 Q34123.1 0343 1.4. The box has three connectors: One 3.5" type power connector, one 2-pin HDD Busy LED connector and the 7-pin S-ATA connector. A Y-cable for splitting up one 5.25" type power connector to one 5.25" and one 3.5" type connector is provided.
The Bridge can be snapped onto the 40-pin ATA connector of the harddisk and adds about 14mm to the length of the disk, not including the space needed for connecting the power and S-ATA cable.

Test Results

The fastest 3.5" ATA harddisk I have available for testing is the IBM IC35L180AVV207-1, a 180GB 7200RPM disk. I used an external 3.5" enclosure as power supply, but could not mount the disk due to the additional length. But for testing purposes, this was perfectly OK. For running the benchmark I use c't h2benchw, a special WIN32 console application designed for harddisk benchmarks. The first output was very satisfying: 54.7MB/s sustained data rate, highest interface transfer rate achieved: 82.4MB/s.
The test showed a perfect transfer diagram with all 25 media zones reproduced in detail. The 25 media zones have an average sustained data rate from 54.66MB/s down to 26.42MB/s with an average sustained data rate of 44.34MB/s across the whole disk surface. The diagram looked much cleaner than that of the same harddisk inside my external Ultra-160 SCSI enclosure using the ACARD AEC-7726H Ultra-ATA to Ultra-160 SCSI bridge. However, the difference in data rate between the two diagrams is very small.

Summary of First Tests

Both items, the S-ATA CardBus Controller and the ATA to S-ATA bridge, work perfectly. The interface test showed that the whole chain can support at least 82.4MB/s transfer rate for reading. This is even a little more than any current ATA or S-ATA harddisk can provide. Unfortunately, I could not perform the write test, because the disk already contains data I still might need.
If you want the maximum performance from your external harddisk, this is the way to go. My next task will be to find a nice external S-ATA enclosure.
If you are interested in related stuff, check out my hdinfo page that I am currently building up.

Further Tests

CardBus Dual Channel S-ATA Controller

Further investigations showed that there are dual channel S-ATA controllers available. The card even uses the same controller from Silicon Image. The only difference is that this card has the usual bulge holding the connectors. The connectors are the common internal S-ATA connectors and there was one S-ATA cable supplied.

S-ATA Harddisk

Because of the positive test results, I decided to get a native S-ATA harddisk. If you are looking for the highest storage capacity available, there's only one choice at the moment (October 2004): The Hitachi HDS724040KLSA80 with 400GB (372GB in binary arithmetics). First tests were very disappointing, because I only achieved 43.7MB/s in the fastest media zone. But after several tests I found out that the power supplies of the enclosures I used had insufficient power. The detailed specifications for the disk reveal that the peak current during read/write operation is 5V, 1600mA and 12V, 2100mA. That is even more than during startup. As soon as the disk is properly powered, it shows a maximum sustained data rate of 58.6MB/s. The interface speed between disk and controller is reported as 86.3 MB/s by h2benchw. The true limit might even be a little higher.

Single Channel Tests

The tests with the properly powered Hitachi HDS724040KLSA80 showed a perfect transfer diagram with 19 media zones having a sustained data rate from 58.59MB/s down to 28.41MB/s. It would be interesting to see where the actual limit is. The fastest S-ATA disk available right now would be the WD Raptor WD740GD with about 70MB/s, but judging from the result of the interface test, I think that the limit is well above that. Proper write tests confirmed my first observations that the write transfer is limited to 17.3MB/s. That's not very thrilling.
Tests with a different Notebook showed a little better write performance of 19.5MB/s but a read performance of only 41.3MB/s. This means that the read performance depends a lot on the Notebook. The write performce will be different, too, but not that much.

Dual Channel Tests

To get a better idea about the total throughput, I hooked up two harddisks to the adapter. Although I used two completely different harddisks, the throughput was equally balanced. The result is a total combined throughput of 66.2MB/s. Parallel write tests showed an equally balanced total combined throughput of 13.36MB/s.
If one can assume that dual channel operation causes some constant overhead, then the dual channel write performance is 77% of the single channel write performance. If the overhead would be the same for write and read, this would lead to a read performance of 85.7MB/s. This is about the same as the result from the interface test, so one could safely assume that the read performce will be around 86MB/s.

External S-ATA Enclosures

There are a few external S-ATA enclosures available. Some with external S-ATA interface, some with combined S-ATA and USB 2.0 interface. I am not sure, which connectors will be used. Some sources state e.SATA (most likely for "External S-ATA"), but the connector shown on product photos looks a lot like 6-pin FireWire 400.
A better solution would be combined S-ATA and FireWire interface, but I did not see any products advertized, yet.

Overall Conclusion