Promise FastTrack S150 SATA TX4 and SX4

Storage/Other by BlackRabbit @ 2004-06-13

We take a look at two SATA RAID solutions from Promise, the high end SX4 PCI card which supports up to RAID5 and also the more budget minded TX4 which still packs a punch. Read on to find out how they perform.

  • prev
  • next

Test Setup & BenchMarks Used

BlackRabbit's Test Setup
CPU 2* AMD Athlon MP2000+
Mainboard Tyan Tiger MP
Memory PC3200 512Mb
System Hard Drive Maxtor Diamondmax Plus8 40GB
Raid Hard Drives 4* Maxtor Diamondmax Plus 9 80GB 8MB

Windows 2000 Professional + SP4 was installed on the system and the SX4 was equipped with 256Mb SDRAM (cache). 4 SATA hard drives were kindly provided by

Why the dual-setup? Because this motherboard has 4* 66Mhz/64bit PCI slots, allowing me to test the controllers at 66Mhz.

The following benchmarks were used:
- ATTO Disk Benchmark
- SiSoft Sandra HD Benchmark
- FCtest from X-bit Labs (create/read(2x)/copy)

I like the FCtest benchmark tool, as it gives a pretty good impression of how well the controller (& disks) perform.

The test contains 4 parts, each stressing the HD’s and Controllers differently:
1. create: files are being created, testing write-speeds.
2. read: files created are being read, testing read-speeds.
3. read_2: files are read again, cached files (SX4) should be read significantly faster)
4. copy: files are copied to another directory on the same array, testing write-read combination.

- we originally included HDTach 2.61 also in our benchmark results, but due to it not handling RAID setups as it should, we elected to dismiss it and remove it from the pages. As you can see below, the results obtained are quite weird:

Madshrimps (c)

3*striped should be faster then 2*striped, but 4*striped is not THAT much faster in real life.

Madshrimps (c)

The read speeds are extremely, abnormally low, we redid these tests several times and have unfortunately no other choice then to simply dismiss the results.

LostCircuits experienced the same weird HDTach results as we did, further proving the point that it can not be relied on.

So now onto the REAL results ->

  • prev
  • next