Promise FastTrak100 TX4 Ultra ATA/100 RAID Controller

Read also:

In the RAID0 mode one to four discs can be connected to the controller. The tests were held in each configuration. The discs were connected according to the marking on the controller's board:

1 disc - IDE1;
2 discs - IDE1, IDE2;
and so on

As the material is quite sizeable you can directly pass to the configuration you are interested in.

1 disc
2 discs
3 discs
4 discs
Dependence of the Intel IOMeter results on the number of discs
Conclusion

1 disc

There is no any array in this configuration - we just test operation in the additional IDE controller mode.

Ziff-Davis WinBench 99

This test wasn't conducted for one disc because of absurdity to use the TX4 in this configuration; besides, the results are almost the same.

Intel IOMeter

In this test it is interesting to look at the TX4 when the load changes. I think that the IOMeter is a more revealing test for controllers than the WinBench.

On the whole, the results are close given the inaccuracy of measurements. Surprisingly, the TX4 wins in the Workstation pattern in the PCI 33 mode, and I do not have an appropriate explanation.

In the streaming read mode the 66 MHz bus has an advantage. In the recording mode the TX4 can successfully fight against the TX2 only at 66 MHz, but it still loses to the latter at the maximum load. I think that with a bridge the recording caching is not so efficient.

The TX2 sweeps the field. In the reading mode a bus frequency doesn't affect the results much. In the recording mode the TX2 copes better with a high load.

The conclusion is that it is not rational to use the Promise FastTrak100 TX4 as an additional IDE controller.

To the beginning of the page

2 discs

Theoretically, the results must be very close. But I think that in the reading mode the TX4 working at the 66 MHz will win, while in the recording mode the winner will be the TX2.

Ziff-Davis WinBench

Touching consensus :)

The TX2 leads in an access time, while the TX4 excels in the CPU utilization (and at 66 MHz the load is noticeably lower).

For the FAT32 - the more, the better. In case of the NTFS the situation is more complicated. It is only 66 MHz that saves the TX4.

Intel IOMeter

The results are identical again. The TX4 lags behind on a standard 33 MHz bus and at a large load.

The same as with 1 disc, but to a greater extent: the TX4 performs even better in the reading mode, and even worse in the recording one.

In case of the random recording the TX4 performs differently: there is almost no difference in the results when a bus frequency changes.

To the beginning of the page

3 discs

Ziff-Davis WinBench

Well, there is a gain in the linear reading speed.

The results are almost equal.

Intel IOMeter

Now the TX4 is definitely a winner in case of larger loads.

Well, the FastTrak100 TX4 has too low results in the streaming write mode on a 33-MHz PCI bus.

In the random write mode everything is just excellent!

To the beginning of the page

4 discs

Well, this is a configuration the Promise FastTrak100 TX4 was designed for.

Ziff-Davis WinBench

Well, there is some gain, but I'd like it to be higher!.

The scores are also almost equal (taking into account that the results are averaged for three tests).

Intel IOMeter

The TX2 works with 4 discs better than with 3. The TX4 falls flat at 33 MHz.

I don't understand why the TX4 doesn't get any gain when the load increases?

To the beginning of the page

Now let's see how the outcome depends on the number of discs. First of all comes the Ziff-Davis WinBench.

While the Business Disk WinMark demonstrates a logical alignment of forces, the High-End Disk WinMark shows that the TX4 falls behind on a standard (for the most of computers) PCI bus. I think it is connected with a performance lag in the streaming write mode - the High-End Winmark works with larger files than the Business does.

With the NTFS we have a real mishmash. But we should account for the difference in the results. It is minimal. This difference can be considered an experimental error.

And now some graphs demonstrating throughput in case of a different number of discs according to the Intel IOMeter results. There are two modes: Light load - 16 outstanding I/Os, and Heavy load- 256 outstanding I/Os.

These models do not benefit from 4 channels and from a 66 MHz bus.

In the streaming read mode there is no any difference in the scores, but in the recording one the TX4 performs awfully on a 33 MHz bus. I think the problem is in a bridge which is meant for a 66 MHz bus.

In a 3-disc configuration the TX4 has a real advantage. In the 4-disc one the advantage remains only for a 66 MHz bus. In the random write mode the TX4 excels in the 2-disc and 3-disc configurations, but it is crushed in the 4-disc one.

To the beginning of the page

Conclusion

Well, I wouldn't recommend to use the Promise FastTrak100 TX4, and first of all, on a 33 MHz bus. Besides, I wouldn't use it as an additional IDE controller. And whether it is worth using at all and how if yet is for you to decide. There are, however, some advantages for those who have a mainboard with the PCI 66 support.

Full test results are in the table.

Ziff-Davis WinBench 99

Intel IOMeter:

Read also:

Write a comment below. No registration needed!