We get a lot of stuff to test, either by buying it or by manufacturers sending it to us and testing is a lot of fun for most of us. What is a little frustrating at times is that a lot of the equipment that we get doesn't lend itself to comparison with anything else that we have. That's not too awful, since the web is full of product reviews, but it's always nice when we can save you the work of paging back and forth between sites, trying to interpret two different reviewers' evaluations and measurements.
Fortunately, we were blessed with having two directly competing products show up on our doorstep within days of each other. DMA/100 gear has been on the market since June and the two leaders in controller technology have really hit the ground running. Both Promise and Abit have offered high performance IDE controllers before and both have really stepped up to the plate with their new products.
The Promise FastTrak 100 and the
Abit HotRod 100 Pro are DMA/100 RAID0+1 controllers that can improve your data integrity, data transfer performance or both. Each offers disk striping or mirroring via hardware, a BIOS-based setup utility and Windows drivers. Both products are also supported by Linux.
Each of the controllers has a capacity of four drives. Since they are DMA/100 controllers, they are also backward compatible with all of the IDE specifications of the past.
The idea behind RAID (Redundant Array of Inexpensive Disks) goes back years. In the past, powerful workstations and servers used arrays of SCSI disks to create either single drives of large size or multiple copies of identical disks. RAID0, or striping, "layers" data throughout the drives so that as the data is written, it is stored on more than one drive, with each drive simultaneously reading or writing the data. Theoretically, that can cut down on transfer rate times substantially, if enough drives are used. In practice, the transfer rates do show a marked improvement, but due to limitations of the IDE bus and overall computer performance, they aren't the 2 or 4 time improvements that theory predicts. Each controller tested can support up to four disks as a single or combination of striped sets.
RAID1, or mirroring, simply makes an exact copy of your data on a second disk as the data is written. If the primary disk fails, the controller will detect that and switch to the secondary, or mirrored, disk, protecting your data from hardware failures. The controllers here will support two sets of two mirrored drives, for a total of four drives.
RAID0+1 or RAID10 combines the features of RAID0 and RAID1, mirroring a striped set of disks. Both of the controllers tested support this configuration, with the ability to mirror a two disk striped set, for a total of four disks.
The Test Environment
I tested each card on the same system, with a fresh installation of Windows 98SE on each configuration (striped or mirrored). I also did the same tests on a single drive using the motherboard's built-in DMA/100 controller. The benchmarks were performed using SciSoft's Sandra Drive Benchmark utility. Remember, since these are just benchmarks, chances are good that they won't reflect what goes on in the real world, but they are valuable for comparing the relative performance of each card.
I ran three passes of each configuration and averaged the numbers to generate the statistical data.
The test platform was:
Abit SE6 Motherboard
256MB Crucial PC133 SDRAM
STB Velocity 4400 (Nvidia TNT) AGP Video
Intel PIII-600EB Processor
(2) Maxtor 30.7GB ATA/100 7400RPM Hard Drives
Digital Research 44x CD-ROM
Intel EtherExpress 10/100 NIC
Thermaltake Golden Orb Fan/Heat Sink
The system was loaded with the latest device drivers for all of the peripherals, including the ATA/100 RAID controller cards. The latest Microsoft critical updates were applied as well. Networking was limited to TCP/IP only, with a static IP address. Each manufacturer's supplied cabling was used.
As a baseline test, I connected a single Maxtor drive to the on-board ATA/100 controller of the i815 chipset. I installed Windows 98SE on it and configured the same driver set as for the Abit and Promise controllers. I installed Intel's i815 ATA/100 drivers and performed the same benchmarks as with the other two controllers. The three passes with SciSoft Sandra gave an average of 19922 which seemed pretty respectable for an on-board controller. Next, I put the two external controllers through their paces.
Promise FastTrak 100
The Promise FastTrak 100 includes two cables and a set of disks with drivers for several different operating systems. The manual is extremely well written and easily understandable by just about anybody. There are plenty of illustrations and instructions that make the process of creating the RAID arrays relatively painless.
The card installed with no problems. My initial tests were with the two drives as a RAID0 striped array. The FastTrak 100's configuration program suggested a series of default settings that I accepted to create the striped set. Then the system rebooted and I installed Windows 98SE.
After installing Windows 98SE and all of the appropriate drivers, I updated the controller's firmware to the latest revision using a DOS boot disk. The BIOS flashed with no problem. The drivers installed as described in the manual.
I didn't see any obvious performance improvement with the drives in a striped array, but that might be difficult for the eye to notice. The Intel chipset's ATA/100 driver is pretty speedy and the Maxtor drives are no slouches either. I installed SciSoft's Sandra utility and performed some benchmarks.
As each test ended, I could see that, at least on paper, there was some significant performance gains over the built-in ATA/100 controller - almost 50%! The card benchmarked at 28786.
Reliability is always at the front of my mind, so I ran the system for about 10 hours, copying a variety of files, from 600MB CD images to small text files. The system reported no errors.
I reconfigured the drives as a RAID1 mirrored array. As with RAID0, the FastTrak 100's BIOS program guided me through the simple task of creating a 30GB disk set. After designating the boot device, I mirrored it to the hidden mirror volume. This did take quite a bit of time, maybe a little longer than formatting the drive would take. Of course, it only needs to be done once. After duplicating the drive, I created a partition with fdisk, formatted and installed Windows 98SE as before.
Again, the performance didn't seem obviously faster or slower than before, but the Sandra benchmarks were telling. The FastTrak 100's RAID1 performance was significantly slower than when striped, with an average benchmark of 16339. It was even slower than the i815e onboard controller. I don't really view that as a significant drawback, however, because the aim of RAID1 is to provide data security and the penalty for that is, of course, speed.
After running the benchmarks, I tested the FastTrak 100 by unplugging the power to one of the hard drives. The FastCheck monitoring utility detected the error and indicated that the drive was in "critical" condition. Regardless of which drive I powered down, the controller maintained the system's operational status by switching to the remaining functioning drive, exactly as it should have.
In RAID0 (striping) mode, removing power to one of the drives gave an indication of a drive failure. In a case like this, however, data will be lost, since there is no redundancy as in RAID1.
The FastTrak 100 also as another mode called spanning or JBOD (for "Just a Bunch Of Drives…it's true!). This mode offers neither security nor performance, but allows you to create one large drive volume so that when one drive fills up, the data is written to the next drive.
Abit HotRod 100 Pro
Abit's ATA/100 RAID offering is based on the HighPoint HPT370 chipset, also found in several motherboards just entering the market. It features support for RAID0, RAID1 and RAID10.
The board was packaged with a manual, a set of ATA/100 cables and a CD-ROM with drivers and a utility program.
The manual, usually one of the strongest features of Abit products, wasn't up to their normal high standards. It had a number of confusing grammatical errors, obviously due to translations, but there were plenty of illustrations and diagrams to make up for it.
I initially installed the card in a Dell Poweredge 2200 server, but the card was not even detected by the system. I tried every trick I could think of to get it to at least show up in the initial power on process, but it wouldn't. The Promise card, however, did work in the server. To be fair, Abit's manual does state that the card is to be used in systems whose built-in controllers are only IDE and the Dell's controller was SCSI.
One of the things that I did to attempt to resolve the issue with the server was to flash the HotRod 100 Pro's firmware to the latest version. This proved to be a much more difficult task than it should have been. The firmware update on Abit's web page was two revisions out of date and the firmware installer was apparently designed with several flash chips in mind because it requested the model number of the chip that I was flashing.
After finding the newest firmware release on HighPoint's web page, I peeled the Abit sticker off the firmware chip and found the correct model number. I was able to upgrade the firmware to the latest version, although it did not correct the problem with the server. According to HighPoint, however, it did fix an incompatibility with Maxtor drives, an important feature given that I was testing the controller with them.
Once I upgraded the firmware, I used the HotRod 100 Pro's onboard BIOS to configure the drives as a striped array. The process was very simple and quick. Once I had Windows 98SE installed and all of the appropriate drivers configured, I tested the drives with Sandra. Apparently Abit picked the HighPoint controller for a reason because it turned in a blistering benchmark average of 36431, nearly double that of the onboard controller and about 30% faster than the Promise controller.
In RAID1 mode, the setup process was equally simple, but excruciatingly slow for the array to duplicate. In fact, I actually got so bored waiting that I left the network operations center for a little shopping. I suspect that duplicating a 30GB drive probably took a good 3 hours. But, as with the Promise card, it's a process that should only have to be done once.
Benchmarks for the mirrored set were in line with the striped performance, with a Sandra benchmark of 19969, about on par with the onboard controller and just a little faster than the Promise controller.
The HotRod 100 Pro's reliability was just fine as well. After undergoing the same file copying testing as before, the controller performed flawlessly.
I removed the power cable for each drive in turn and Abit's monitoring program noted when a drive went offline. When mirrored, the controller switched to the appropriate operational drive and when striped, the controller indicated that the drive was offline.