Today we continue reviewing IDE RAID controllers. And this time we have only one sample - Promise FastTrak TX4. Why? Because it is only one of its kind in our lab; all other 4-channel controllers have a RAID5 support.

Promise FastTrak100 TX4

Let's open the box and see what's inside.

• The card.
• User Manual (it is identical for TX4 and TX2).
• 4 standard 80-conductor cables.
• 2 diskettes - with drivers and FastCheck utility.
• Booklet advertising Promise products.

Of course, the card is the most interesting for us.

You can see here two chips Promise PDC20270 - the same chip is installed onto the TX2, it just has a different sticker

and the chip is from Intel.

Unfortunately, Promise provides no information on the design of its controllers, but it quickly identified it according to the marking. It is a transparent PCI-to-PCI bridge. Here you can get detailed information on this chip. And in brief, this chip can be used in expansion cards the following way.

(Courtesy of Intel)

In our case network chips are replaced with two RAID controllers.

I think that it won't beneficial to use this controller for connection of one or two discs, but an access time can get higher. In case of a RAID array of the 0 level on three or four discs much depends on the BIOS - it is necessary to adjust interaction of two controllers, and it must be beneficial. At last, in case of the RAID0+1 the benefit must be considerable as an array of two discs is created on each controller, mirroring is implemented via its own bus, and if the BIOS distributes reading/recording operations correctly between two controllers, it will be simply a dream.

Let's see whether I'm right.

Installation and operation

The installation of the controller under the Windows 2000 Professional SP2 was flawless, its operation doesn't differ from that of the TX2 - everything is simple. If you are interested in it see the review of two-channel controllers.

Tests

Since the controller supports 66 MHz of the bus, the tests were carried out on the following stand.

• Mainboard - Tyan S2567 (ServerWorks HE-SL chipset)
• Processor - Intel Pentium III 800EB
• Memory - 512MB Registered ECC PC-133 SDRAM
• Video card - Wicked3D Vengeance (3Dfx Banshee 16 MBytes)
• Hard drive - Seagate Barracuda ATA III ST340824A
• Discs for an array - 4XFujitsu MPG3204AH-E

The tests are based on the same technique as for the two-channel controllers. For comparison we used the Promise FastTrak TX2 controller (because it is based on the same chip and it was the best in the testing of the two-channel controllers). Of course, usage of another chipset may affect the results, but I think the situation will be clear. In the tests we used BIOS and drivers coming with the controller - ver. 2.00 build 11.

The results are gathered according to the RAID levels. And here we give only brief summary.

RAID0

On the whole, the Promise FastTrak TX4 doesn't have a great advantage over the TX2, and it yields to the latter when working on the PCI 33 MHz bus. The only configuration where the TX4 performs much better is a 3-disc one.

RAID1

It is not beneficial to use the TX4 in the RAID1 mode if both discs are connected to one controller.

RAID0+1

The Promise FastTrak100 TX4 thrives in the RAID0+1 mode as compared with the junior brother. This advantage and hot replacement support (which is lacking in two-channel controllers) let me recommend this controller for such arrays.

Conclusion

Some results of the Promise FastTrak100 TX4 controller can't be explained, but figures will be figures. Maybe the new version of the BIOS will eliminate those shortcomings. But experts consider that it is peculiar to almost all RAID controllers, including highly-intellectual SCSI RAID controllers just after their release. Sometimes such problems live up to the end of such controllers.

The only way-out is to know the problems and to choose controllers which suit the tasks required. Well, this is what we aim in our reviews. In this case I can recommend the Promise FastTrak100 TX4 controller in the RAID0+1 mode on a 66-MHz PCI bus.

But if you have already bought a controller and you face such problems you should contact the company's service, and if it's possible, directly a certain engineer you trust to (remember that in some cases it is necessary to change a board).