Review Index:

Mylex eXtremeRAID 2000 Review

Subject: Storage
Manufacturer: General

WinBench, HD Tach and Conclusion

This content was originally featured on and has been converted to PC Perspective's website. Some color changes and flaws may appear.

Click to Enlarge

The eXtreme 2000 kept up with the ATTO UL3D but fell a little short with the multimedia applications.

Click to Enlarge

Little differences shown here. The ATTO UL3D bests Mylex’s $1,700.00 controller by a broad margin.

Well, the results speak for themselves. For an enterprise-class high-end controller such as the Mylex eXtremeRAID 2000, it performed poorly. At no point did the eXtreme2000 distinguish itself from the likes of Adaptec’s 2100S or ATTO’s UL3D. For a $1700.00 RAID controller, this is almost impossible to neither support nor recommend. It is important to note that Mylex does not support or acknowledge Intel’s Iometer, ATTOTech’s ATTOMark, and or WinBench99. The only measurable benchmark Mylex acknowledges is TPC-C. Unfortunately, TPC-C means nothing to ninety-nine percent of the computer public. Therefore, it is almost impossible to quantify the performance of the eXtreme 2000. That is, anything short of purchasing an 81,000.000 server in which TPC-C was run on. Actually, most of the mentioned manufacturers have no benchmark results from an AMD platform. It should also be noted that the eXtreme 2000 is positioned as an enterprise-class RAID solution, not a 2-4 hard drive workstation environment, which may help to understand its undistinguished performance.

There is of coarse another possibility. After referencing many SCSI storage-related websites and forums, I noted that Intel-based PC’s perform better with almost identical RAID and storage configurations. Whether the device resides in a 32-bit or 64-bit slot, the differences between Intel-based solutions and AMD-based solutions are immense. I can only theorize that the KT133 chipset does not adequately manage the overhead or demands required by a RAID configuration. Perhaps there are architectural bus bandwidth restrictions. In addition, I have very recently discovered that the ATTO UL3D may have a compatibility problem with the KT133 chipset. The UL3D has an option within its BIOS to create or delete a RAID group. Unfortunately, this option does not function. Therefore, to format or delete a RAID group will require that the user connect the SCSI hard drives to a different SCSI controller to perform said functions. This frustrating procedure is not present on an Intel-based platform. Currently, ATTO is investigating the phenomena on the Asus A7V. The environment with which these tests were performed was ideal. That is, a fresh OS installation was performed. No PCI devices were present except the SCSI RAID controller. All non-essential features were disabled within the BIOS such as IDE, Serial, Printer, USB, etc. For utmost compatibility, the Mylex eXtreme 2000 was place in PCI slot-3. PCI slot-3 is the only “free slot” available that does not share an IRQ. Multiple block sizes were implemented from 8KB up to 128KB. Write-Back cache was also enabled. Working with Mylex, tech support also corroborated that the controller’s configuration was properly configured. My conversations with Mylex were to the extent that they were dumbfounded as to the performance problems. Again, I can only guess that these high-end components and or configurations are currently reserved to an Intel, or server-class environment; that is until AMD’s SMP offerings materialize. With this said, I expect to run these similar storage configurations and benchmarks on the new AMD 760 chipset. Perhaps I won't run into these unique performance bottlenecks.

Lastly, a very special thanks goes out to Hyper-Microsystems for providing us with Seagate’s X15 hard drives!