Testing RAID cards is a long and arduous process. The most difficult part was deciding on a standard cluster size for all cards. While the Promise cards can have from 8k to 2048k, Adaptec’s limit is 128k, and 3ware’s is 64k. All cards default to 64k cluster size, so who am I to argue? Just for comparison sake, I ran the Promise FastTrak100 with three cluster sizes to see the difference.

I’ve indicated best scores in a range with Bold face. I can’t account for the poor write scores with HDTach, although they were quoted in the December issue of Maximum PC as indicating their benchmark is not designed for RAID systems. All tests were run using the Test Platform.

On the two disk system 64k appears to be the clear winner. With four disks, 512k kills with RAID 0 and does well again with RAID 0+1.

Cache plays an integral part in RAID types that use parity or error checking and correcting (ECC). How much cache to put on a RAID card to get the best performance is a tricky bit of business: the natural reaction is to max it out. Below is a comparison of benchmarks using three sizes: 64, 128 and 256 MB PC133 SDRAM on a Promise FastTrak SX4000 with IBM DeskStar 75GXP 20 GB drives.

Well, after a lot of benchmarks it looks like maxing the cache paid off on four disk RAID arrays. For two disk arrays 64 MB is fine, and three disk arrays can go with any size.

The BC4000 RAID HBA family, comprised of 64-bit, PCI-X low-profile compliant controller cards, are the only 2U compatible SATA RAID controller cards available today. The product family provides both four-drive (BCM4452-H) or eight-drive (BCM4852-H) support and are based on Broadcom’s new BCM5770 Serial ATA controller chip. The new Broadcom BCM5770 SATA chip and the BC4000 controller card significantly boost reliability with efficient power consumption and thermal load balancing, and feature a MTBF (mean time between failures) of 3.5 million hours, which is nearly three times greater than that of competitive SATA RAID controller cards.

Enterprise-class RAID functionality for the BC4000 RAID HBAs is provided by Broadcom’s XelCore RAID software, based on proprietary Fulcrum architecture (a highly integrated storage architecture). These enterprise-class RAID features include:

• Online RAID level migration – the ability to change the RAID level of an array without taking a system offline
• Online capacity expansion – the ability to add storage to an array without taking a system offline
• Online array creation/deletion – the ability to add, modify or delete arrays without business interruption
• Controller spanning – the ability to create a very large array that spans up to 32 disks and achieves extremely high read/write performance using a large number of spindles in a single array
• Drive roaming – the ability to move drives between systems and controllers without keeping track of the controller and channel to which a drive was attached
• N-Way disk mirroring and hiding – the ability to create a secure backup of a data array while hiding it from the operating system
• Distributed sparing – the ability to use all disks in creating a spare as a means to achieve higher performance by using all disks in an array

The above encapsulates the features of the RAIDCore controller, as defined by Broadcom. RAIDCore is now part of the Broadcom family, but a new addition. I couldn’t find anything on the Broadcom site by means of navigation and search, so I went to the old www.raidcore.com URL and was redirected to the proper page. Read the above and check out the links to get a good understanding of this product group.

In an interesting turn of fate I was contacted by the RAIDCore PR rep months ago regarding the site, and I was quick to request a review unit. Unfortunately they were in the process of merging into Broadcom and switching over to the BCM5770 controller chips. Some time passed and I received my shiny new BC4852 SATA RAID controller, my first for a PCI-X 133 slot. I had to test in a 64 bit 66 MHz slot in my test configuration since that’s what I have and I can’t seem to get a deal on a new Opteron board and the new Xeon boards aren’t available yet.

Luckily I did wait on the card, since there were some issues with the RC4000 series controllers. Broadcom is offering a card replacement, so if you have an older RAIDCore card take advantage of this offer. Broadcom is also working to remove the “RAIDCore” name from the product, or so it seems.

Let’s jump to the chase and talk about performance, then take a step back and look at the card and it’s features and interface. Open our benchmarks and look for yourself: the BC4852 stands shoulder to shoulder with the rest of the pack. It leaps ahead in WinBench, and in Atto and HDTach writes. The only issue to show itself is a high CPU utilization, 11-25% depending on which benchmark you believe. This is because there is no onboard cache or XOR engine. They’re depending on your system having a reasonably fast CPU.

Looking at the BC4852 you’ll see a clean design. Without cache or a dedicated processor there’s room to spare. In the center is the BCM5770 controller chip, to the right six SATA connectors with two on the other side. On the left are eight pin headers for individual HDD LED connections, if you want to connect your drive cages that way. As well on the back of the card are eight small lights that flash as the array is utilized. It’s a low profile PCI adapter that comes with a PCI MD2 low profile bracket.

The key to the BC4852 is it’s rich feature set; just read the laundry list on the previous page. Let’s touch the highlights, most of which center around the ability to make changes to the array without bringing it down. The two biggies are RAID expansion and RAID migration; while the array is in use you can expand the capacity or change the RAID level. It’s not a quick process, but it works. Move from a RAID 10 array to a RAID 5; the only catch is that you can’t go to a smaller array, so no four disk RAID 5 to RAID 10 migrations. Add disks to your array and then expand it. Once you’re done expand the formatted partition with Microsoft’s diskpart utility.

Lots of other goodies as well. Up to four BC4852 cards can be used in one system, with arrays stretched across. Multiple arrays can all use the same hot spare drive, minimizing disk waste. The controller marks the drives, so they can be unplugged and reconnected in a different order without breaking the array. Staggered spin up allows each disk connected to spin up in order, so there’s not one power draw on boot up. Options abound, and all are well documented in the included PDF and HTML manual. This covers the entire spectrum of using and working with the BC4852 and should be an example for all manufacturers. It sounds geeky but it was a very enjoyable read, and provided all the information most would need regarding RAID. The drawback is a lack of printed material in the box.

Installation was smooth and straight forward. Creation and management can occur through the BIOS or via the RAIDConsole application. I used both without incident. The RAIDConsole is very easy: there’s one screen to work with, and right clicking on objects gives context options. It can also be run from the command line, but I didn’t try it. Inside the box was the controller, eight half meter SATA cables and a CD.

Once an array is created a full disk integrity is run, referred to as initialization. The array can be used immediately, but a performance hit obviously takes place with the initialization is occurring. Every option available can be performed from the RAIDConsole interface.

There has been a lot of talk about RAIDCore and availability since it’s initial launch last year. A where to buy page is available from Broadcom, and I checked PC Mall for pricing and availability: $385 US. That price puts it in the mid range of SATA RAID controllers.

Bottom line on the Broadcom RAIDCore BC4852: it’s a feature rich SATA RAID Controller that provides good performance. If your server is running a 1.5 GHz processor or higher then you won’t be hit too hard by the BC4852 using the host for XOR calculations. Each of use and feature set make it an excellent choice.

Editors note: this review was migrated from the old eBabble.net site and the photos updated. Originally published August 10th 2004.

Promise controllers and I are old friends, all the way back to their first ATA RAID controller. Their latest is the Promise FastTrak S150 SX4: a four port Serial ATA RAID controller. Looking at the image, you’re probably thinking you’ve seen it somewhere before: the S150 SX4 shares the same PCB layout and size as the SX4000. The only differences are the SATA connectors and the Marvell PATA to SATA converter chips. For a complete rambling on the Marvell chip check out our 3ware Escalade 8506-8 review.

This time around I didn’t get a retail package for review, but a review’s package. What should come in the box is the controller, SATA cables, installation manual, and drivers and management software on 3.5” floppy disk and CD-ROM. In case you’re wondering the review’s package came with the controller, photocopied manual, burned CD, reviewer’s guide and 128 MB DIMM. You’re not missing out on anything, and it’s a good thing I already had SATA cables.

Looking at the S150 SX4 board layout, it’s a match to the SX4000. The same single Promise PDC20621 ASIC to handle XOR ( RAID 5 parity calculations ) and the four ATA channels. What’s been added to the S150 SX4 are four Marvell 88i8030-TBC parallel to serial ATA converter chips. As well and I2C connector is in the top left, for those motherboards that support it for system monitoring. The only mention of it in the user manual is in the board layout. A very large board that in it’s next version will be about half that size. It’s interesting that the four status lights from the SX4000 have been dropped: since they weren’t documented it’s left a mystery.

Installation was very simple. I configured an array via the S150 SX4’s BIOS, called FastBuild. Entering Windows 2000 I was prompted for drivers, then a reboot and installation of the PAM utility.

Feature wise Promise has everything there you’ll need. It’s a 32 bit 66 MHz PCI controller capable of bursting 266 MB per second. It handles four SATA drives in configurations of RAID 0, 1, 10, 5 and JBOD ( just a bunch of disks ). The manual and all configuration screens I encountered showed the S150 SX4 handling RAID 0+1 and not RAID 10, but the Promise website and it’s datasheet both list RAID 10. Check our RAID Types to see the difference. Promise calls it’s ability to recover from an error by drive hot swapping or hot replacing PerfectRAID, although all controllers support this. With Promise they would like you to use their Superswap 1100 enclosures. Cache is required for the controller to operate, and a user installed SDRAM DIMM of 64 – 256 MB must be installed. I found it funny that page 30 of the user manual shows an example of the BIOS configuration with a 512MB DIMM installed.

Management of the S150 SX4 is handled via the Promise Array Management, or PAM. This utility has been around for some time, but is updated for every generation of Promise cards. It’s very clean the menus follow a nice flow. IT has the usual features: array build, monitoring and SMART support, email and pager notification. It’s standout feature in my opinion is array expansion and conversion, covered on pages 67 -73 of the PAM user manual. This allows an array to be expanded by adding more disks, or converted to another RAID level. Some caveats of course, but it works and works well. I added a fourth disk to a RAID 5 array. It was time consuming but worthwhile. Once completed it’s best to use a partition management package to resize the partition and get the benefit of the extra space. This functionality screams out for a controller that can handle more than four drives: when will Promise match 3ware’s eight and twelve drive controllers?

Performance has been increased substantially to match the 3ware Escalade 8506, judging by our extensive benchmarks. While the previous Promise SX4000 trailed the 3ware Escalade 7500 across the board, now we see a neck and neck race. Promise pulls ahead of the competition in price: the FastTrak S150 SX4 can be found for almost half of the 3ware Escalade 8506-4, it’s direct competitor. Where it falls short is drive capacity and PCB size: let’s see some 12 port SATA controllers in a low profile PCI-X packaging. If four drives are all you need, the Promise FastTrak S150 SX4 is the SATA controller for you.

Editors note: this review was migrated from the old eBabble.net site and the photos updated. Originally published November 28th 2003.

I’ve been a big Promise fan since the original FastTrak (later renamed FastTrak33 to fit with the product line). At the time I had a bunch of 2 GB Western Digital hard drives that I didn’t know what to do with. The idea of using a RAID controller to make one big 8 GB ( wow, that’s big ) drive was fantastic. Of course there were some growing pains, but the card worked great. What’s even better is that I just requested a new BIOS chip for the card and it was sent for free; unfortunately Promise didn’t switch to a flash BIOS until the FastTrak66. But as usual I digress.

In the early days of ATA RAID controllers RAID 0, 1 and 0+1 was the limit of availability, thus relegating it to workstations and enthusiast machines. Of course SCSI RAID controllers allowed for so much more, but at an exorbitant prices which flew in the face of the entire ATA RAID concept. Adaptec was the first to release an ATA RAID 5 controller, the AAA-UDMA, which was a SCSI RAID card with a “translation” chip for IDE drives. It’s performance was adequate, but more importantly paved the way for Promise SuperTrak100, and the rest is history.

My focus has shifted to RAID 5 controllers, since they are perfect for entry level servers used in small to medium enterprises ( SME or SMB ). Today’s motherboards bundle RAID 0, 1 controllers onboard for a minimal charge, and Intel and VIA are now building them into their chipsets with the advent of serial ATA. Unlike these controllers, RAID 5 requires some hardware horsepower to calculate the parity bits and spread the data across the array. Onboard RAM is also a necessity, along with the CPU and ATA controllers bringing the complexity and cost of the product up. Imagine my surprise when I heard of the Promise FastTrak SX4000 RAID 5 controller selling for less than half of it’s competitors.

The SX4000 is a new step for Promise in a lot of ways. Their two previous RAID 5 cards, the SuperTrak100 and the SuperTrak SX6000, used Intel i960 processors along with three Promise ATA ASICs (Application Specific Integrated Circuit ). They were full length PCI cards with six IDE connections, one RAM slot and battery backup.

Let’s look at the SX4000. Cracking open the box we get the card, a quick start guide, a product manual, drivers and software on 3.5” floppy and CD-ROM, four 80 wire 40 ping ATA cables and two thick molex power “y” splitters. On to the card: it’s a mid length ( not small but not full length ) PCI card with four ATA connections. Bottom right is the 168 pin DIMM slot for SDRAM from 64 MB to 256 MB. A small speaker sits above. The only processor or ASIC onboard is the Promise PDC20621, which is described in the literature as an XOR engine. This single chip solution saves cost from the previous SuperTrak series, described above. On the faceplate are four red status lights; it’s interesting to note they’re not mentioned in any of the literature as to what they do. From the card layout you can also see what isn’t there: a spot for a battery backup in the bottom left and one can only guess what that space in the top right is for.

Installation was straightforward: plug the card into an available PCI slot, either 32 or 64 bit, 33 or 66 MHz. The SX4000 is a 66 MHz 32 bit PCI controller capable of 266 MB/s maximum. I installed an available 256 MB ECC DIMM in it’s single memory slot: you can use 64, 128 or a 256 MB DIMM. The case I’m using, and Antec SLK3700AMB, has three drive lights, so I connected one of them to the header on the SX4000. Hooked up the four drives and powered the system. The card is longer than most PCI adapters, but it isn’t full length either. I had to move one of my hard drives to accommodate it, but that was the only issue. For some reason all AMD 768 MP motherboards put the 64 bit 66MHz PCI slots in the middle of the board, instead of the bottom where most cases leave room.

Once the SX4000 initialized it prompted me to enter the BIOS and define an array. Here you can view, define, delete or repair an array. RAID levels 0, 1, 0+1, 5 and JBOD (just a bunch of disks) are supported. As well all these functions and more can be performed in the operating system via the Promise Array Management, or PAM in Windows and the Array Utility in Linux. That’s right, the SX4000 is supported in various Windows and Linux flavors. For this review Windows 2000 Server is used. The BIOS interface is dead easy, and arrays can be deleted and created with no delay or wait time.

Booting into Windows 2000 server I installed the drivers and rebooted. The SX4000 shows as a SCSI device in Device Manager, as do all ATA RAID controllers. The array shows under disk drives as a Promise Array. Installing the Promise Array Management utility was another non event, but a username and password has to be created for array management. Once installed PAM can be managed from any PC over the network by installing PAM wherever and selecting remote management. PAM is straightforward to use and allows for the gamut of options, including enabling write back cache. I didn’t find this aided performance, and the default is having it disabled. Changes to an array require a reboot, but it nice to make changes remotely and schedule a reboot without interacting with the machine.

Performance was excellent, in relation to the SX4000’s abilities and cost. Check the benchmarks and testbed for more information. Promise designed the card as a one chip solution, covering the ATA controllers, RAID calculation and the like. Check the cache size comparison to see what fits your needs; I anticipated better performance from more cache, but that’s not the case.

I’ve had this card for the better part of six months, trying to arrange competitors. It performed flawlessly across three platforms as I ran various benchmarks and changed my platform and software. I that time there were BIOS and driver revisions, which were implemented on this final benchmarking and platform.

As I went to format a RAID 5 array for the final PCMark2002 benchmark, I received an error that the format failed. I played around and ran diagnostics on all the drives individually, plus used them with other ATA RAID controllers without incident. I submitted an incident report via the Promise website, or tried to. For two days the form created SQL errors. On the third day it went through, but I didn’t receive a reply. I waited more than a week and didn’t hear back again. I’m sending an email with the same information this week and will post what happens. This is my first bad experience with Promise’s tech support, and I’m hoping it only a bug with their online reporting form.

I heartily recommend this card for users looking to utilize four or less drives in a RAID 5 array. It’s half the cost of the Adapter 2400A or 3ware 7500 series, even with the additional purchase of the required DIMM.

Update / Revision

May 30th 2003

After my email to tech support I received an answer three days later suggesting the problem lay with the DIMM I was using. Replacing it with an available 128 MB ECC PC133 DIMM seemed to do the trick.

August 9th 2003

Promise has released new drivers, BIOS and PAM for the SX4000. Included is a utility to test memory installed on the card. As well Windows Server 2003 drivers are included.

Editors note: this review was migrated from the old eBabble.net site and the photos updated. Originally published May 22nd 2003.

Last week Microsoft released the Windows Home Server Community Technology Preview (CTP).  This put a lot more fit and finish to the product.  I used the upgrade option to upgrade my existing WHS beta 2 installation.  Things went well, but I was still getting a failing service and couldn’t install the client software.  I had hoped the upgrade would resolve these two issues.

Around the same time I realized my home office needed to be cleaned up.  I had three servers in various states of assembly plus four PCs scattered around the room.  Everything needed to be organized and prepped.

Tiger Direct had a sale on AMD Athlon x2 3800+ socket 939 processors so I picked one up.  I have an eVGA Nforce4 SLI motherboard in an Antec SLK3700BQE case and an Antec Neo HE 550 power supply.  It just needed to be assembled to get things cooking.  When the CPU arrived I popped it in with an Arctic Cooling Freezer 64Pro since it was an OEM model.  Added two 1 GB PC3200 DIMMs and an LG GSA-H22L DVD re-writer to the mix and it was almost ready.  Around this time I decided this machine would be the new WHS box, so I could just transfer the four 500 GB Maxtor drives from the current dual Xeon WHS machine.  The Nforce4 SLI motherboard had PCI-E slots for the graphics and I didn’t have anything to spare so I threw in an ATI 8 MB PCI card and it was working beautifully; thank goodness I keep boxes of old hardware for no particular reason.

I moved over the hard drives in the same order and booted from the WHS CTP DVD in the hopes of running the upgrade option again, but no dice.  Reordering the hard drives on the motherboard didn’t help and I started to feel a little panicky.  I pulled one of the four and hooked it to my main system: all my files were there.  WHS and it’s folder duplication function had just saved my bacon and vividly illustrated why not to use a RAID array for WHS.  I did a fresh installation of WHS on the three connected drives and copied my data over from the pulled drive.

Everything was going so well…until WHS client installation.  Failure again loomed over me, as it had the last three installs.  I hit the WHS forums, again, and looked around until I found someone talking about adding the WHS machine to the clients HOSTS file.  Brilliant!  Doing so got everything tickety boo.  I did change the IP of the box from a DHCP address (the default) to a static address so I could feel secure with port redirection on my router.

I skipped one little item to keep continuity: the Antec Neo HE 550 wouldn’t power up the new WHS machine.  I hit the power button, and the fans wouldn’t even start: the motherboard flashed “FF” at me.  Solid FF meant booting, but flashing wasn’t really mentioned.  I plugged the power supply into my Antec power supply tester and it showed as all green, thumbs up.  I unplugged everything except CPU, RAM and video card but still nothing.  Swapped the CPU with a Sempron 3000+ socket 939 CPU I had from another system and it was no luck.  Finally desperation sank in and I pulled an Antec TP2-550EPS12V from my dual Opteron server and hooked it up: success!  Swapped power supplies and put the Neo HE aside: this was my first failed Antec power supply.  With it’s five year warranty I’ll try and RMA.

Cleaned everything up and loaded the dual Xeon system with 4 GB PC2100 ECC RAM, four Maxtor 250 GB hard drives and a 3ware Escalade 9550SX.  I downloaded Windows Server 2003 x64 Standard edition from TechNet and am planning to install next week, along with Microsoft Virtual Server.  This will be my secondary test/backup platform as I like a belts and braces approach to my data, and until WHS is a final product I’ll need a secondary file depot.

But that’s alright, as I had a spare server already set up.  A Tyan 2882-D motherboard, two Opteron 246 CPUs, four 1 GB PC2100 DDR ECC RAM, 3ware Escalade 9550SX RAID controller, two 80 GB WD SATA hard drives, four 250 GB Maxtor SATA hard drives, all wrapped in an Antec Titan server case.  Very nice indeed.  The motherboard was a bit old but had what I needed: 133 MHz PCI-X slots for the RAID controller.  Everything else was onboard, but it only has USB 1.1 connections.  The RAM was slow for the CPUs. but I had PC2100 around and didn’t have PC3200 ECC.

The big thing was getting my data off the existing server and migrating it to the new server.  Since the new would be using the old’s IP address network connection was out of the question, plus it would be too slow.  I didn’t have a PATA hard drive big enough to hold my 300 GB of data, so I couldn’t use the native ports on the old server, which was a dual Athlon MP system and pre-SATA.  I had four 500 GB Maxtor SATA hard drives that were going into my backup machine running Windows Home Server, so I pulled one of those and connected it to the old server’s 3ware Escalade 8506 SATA RAID controller.  I’m a big fan of 3ware and they’ve always been generous enough to supply me with controller cards for my ATA RAID testing.

While moving all the files off the degraded RAID array and onto the single 500 GB Maxtor, I installed Small Business Server 2003 R2 onto the new Opteron server.  It was a long process, but it was still done before the file copying.  Thankfully the RAID array didn’t degrade any further.  Once it was completed I hooked to 500 GB Maxtor to a spare port on the 3ware 9550SX and booted the system.  The drive was recognized but I had to designate it as a single drive before the OS would recognize it, something I didn’t have to do with the 8506.  Once recognized on the new system the drive appeared as unpartitioned!  Bells ringing in my head I tried a few things but saw this wasn’t going to work.

Luckily the Tyan 2882-D has an onboard Silicon Image SATA controller, the sort of in-between step vendors took when SATA was introduced.  I enabled it in the BIOS and saw the hard drive was recognized, but now the machine wanted to boot from it instead of the 9550SX.  I popped into the BIOS and checked the boot order: three slots were available, but with the Silicon Image controller enabled the 9550SX was no longer an option.

I quickly disabled the Silicon Image controller and brought the Maxtor 500 GB back to the old server.  Hooking it back up to the 8506 showed a bare unpartitioned drive, so connecting it these RAID controllers was causing some problem.  I pulled the Promise SATA300TX4 from the Windows Home Server and installed it into the old Athlon MP server; it wouldn’t be needed until I could put the 500 GB drive back into the Windows Home Server anyway.  Fired the older server back up and put the drivers in for the SX300TX4.  Started the copying process again, still hoping the degraded array would be fine for just a few more hours.

Data copied I moved the Promise SX300TX4 into the new system but found the same BIOS issue.  Pulling my hair and cursing the system I pulled the motherboard and put in a Tyan 2892 I had on the shelf that I had foolishly purchased on eBay without a need and so left it in the box for a year.  This board had the Nvidia Nforce Professional chipset with native SATA and lots of BIOS boot options.  Copied the data over without issue, finally.

I posted a review of the Promise SATA300 TX4 but needed a high quality pic.  There’s a tiny picture on the Promise website but when you click “Enlarge Image” you end up with a picture of a different model.  Contacted Promise tech support for a photo and to let them know of the problem but so far I’ve been ignored.  To be fair I’ve been emailing old Promise contacts I had from the RAID review days; who really knows if they’re still around?

Rethinking my whole home software setup. involving a hardware shuffle and software reload.  Currently my SBS 2003 install runs on a home built Athlon MP server.  With my TechNet subscription I wanted to update to SBS 2003 R2 anyways, so my big home built dual Opteron server will become the main home server.  Lots of work there.

Each day I use Windows Home Server beta 2 the more I appreciate it, and hope it’s main features will make it into a future SBS version.  Automated nightly backup with no file duplication is a fantastic feature, especially as it allows bare metal restore.  The beta team ask that RAID isn’t used, but with it’s folder duplication I feel too much space is being wasted.  I’ll rebuild this server as well with 500 MB hard drives and a RAID 5 adapter.

Finished Titan Quest Immortal Throne.  This is an expansion back, and is a Diablo-esque game with monster killing and weapon collecting.  The expansion added some much needed features such as auto sorting your loot storage and providing additional storage space in towns.  Look for a full review shortly.

When one thinks of ATA RAID, the first name that comes to mind is Promise. The first company to mass market a controller for IDE drives continues the FastTrak line from the original to the 66 and 100 models, each accommodating a new generation of ATA specs.

Looking at the features offered, very little has changed between the three models other than the ATA spec supported. All three FastTrak’s support four disks in the following configurations: RAID 0, RAID 1 and RAID 0+1. The FastTrak100 supports JBOD ( just a bunch of disks ) which takes all disks connected and uses them contiguously. The 66 and 100 models support flashable BIOS, while the 33 model requires a new BIOS chip from Promise. To compare the three cards, FastTrak ( FastTrak33 for this review ), FastTrak66 and FastTrak100, I installed them in the RAID Projects test machine and benchmark criteria.

Everything from packaging to board layout has remained consistent, aside from a chip shuffle. Opening the box reveals the FastTrak series adapter, four IDE cables, a manual and a bunch of floppy disks. Looking at the image above shows the board layout, with the Promise controller chip and BIOS, along with two ATA connectors and a four pin LED connector. The original 33 model had two controller chips, a Promise PDC20247 and a PDC20246. The 66 and 100 models use one controller chip, the PDC 20262 and PDC20267 respectively. An interesting side note: the SuperTrak100 uses the PDC20265 while the FastTrak100 uses the PDC20267, yet they both are used to control ATA100 drives.

Installation and setup was straightforward with all models. Insert the card, hook up the drives and run the BIOS setup. Again basically the same screens across the line; either let the card configure itself through a simple menu selection or configure it yourself by choosing drives and RAID levels. Unlike the RAID 5 adapters previously reviewed, RAID setup cannot be configured in software and must be done through the cards setup at boot.

I should point out that I flashed the 66 and 100 models to BIOS version 1.30 build 12. Promise uses one unified driver and utility set for all cards, and maintains BIOS versions across the flashable models. For the 33 model I ordered the 1.06 BIOS two years ago, but to date it’s at version 1.08. I ordered it but didn’t receive in time for this review. The drivers uses were 1.30 build 42.

Driver and utility installation went smoothly in my Windows 2000 Professional testbed. The OS recognized the 33 and 66 models and installed default drivers. I upgraded them to 1.30 without problems. One interesting anomaly; after installing the 33 model I found the OS was seeing the hard drives as individual 20 GB drives, not as the RAID 0 four drive 80 GB array I set up.

The FastCheck utility allows for simple maintenance of your RAID arrays. Status screens shows the drive specs hooked to the adapter and what configuration used. An options page lets you set criteria such as event logging, PCI bus utilization, synch and rebuild settings for RAID 0 and 0+1 arrays. A fair number of options present, and the ability to schedule an array synchronization for late hours gives peace of mind.

The manual is well laid out with a thorough examination of all hardware and software options and settings. While not necessary, a good read gives one a complete knowledge of what Promise has delivered. I especially liked the RAID definitions and frequently asked questions provided. Very informative.

Benchmarks paint a promising picture ( ouch, bad pun ). Unfortunately it was at this point that my FastTrak66 gave up the ghost and wouldn’t work. I’ve tried everything and resorted to returning it for repair. Looking at the scores from the 33 and 100 models show very similar results, with the variance easily credited to the improved interface. While the ATA specs’ high end only come into play during some burst operations, there is a divide between 33 MB/s and 100 MB/s.

I’m a big fan of Promise’s line of products and heartily recommend the FastTrak series. Keep in mind a RAID 0 array won’t make loading Windows or playing games that much faster. RAID should be used for redundancy, backup or when there is a need for large disks.

Editors note: this review was migrated from the old eBabble.net site and the photos updated. Originally published January 29th 2001.

]]> http://tech.ebabble.net/promise-fasttrak/feed 1 http://tech.ebabble.net/promise-supertrak100 http://tech.ebabble.net/promise-supertrak100#comments Fri, 26 Jan 2007 19:24:29 +0000 Scott VanderPloeg http://www.ebabble.net/reviews/promise-supertrak100/

The second card in our ATA RAID project is the Promise SuperTrak100.  Promise has had a long standing relationship with ATA RAID in the form of their FastTrak line, which we’ll look at next week.  With the SuperTrak line Promise adds RAID 3 and 5 along with RAID 0, 1 and 0+1.  Two models are available; the SuperTrak66 which is an ATA66 card with a four drive capacity, and the SuperTrak100 which boasts ATA100 connectivity with a six drive capacity. Let’s examine the card from box to benchmarks.

Upon opening the Promise SuperTrak100 box you see it’s a full size PCI adapter, the same length as the Adaptec AAA-UDMA we looked at last week.  Also included are six 80 wire 40 pin single drive cables, two manuals and three 3.5” disks.  The adapter has six IDE channel connections, three Promise PDC20265 controller chips, an Intel i960RD CPU, a battery, an alarm, a 72 PIN SIMM slot with a 16 MB EDO SIMM, an LED connector and four LED’s along the backplane.

Promise has followed Adaptec’s lead again in the drive setup, with one drive per channel connection. This is to eliminate any waiting on each channel; with a traditional two drive channel setup only one drive communicates over the bus at a time.  The two items needed for a RAID 3 or 5 adapter are cache and a processor to handle the parity distribution.  The SuperTrak100 handles anywhere from 8 MB to 128 MB, but the 16 MB included is more than adequate.  The Intel i960 is an I2O ( intelligent I/O ) compliant processor that handles the parity information and allows for network management of the adapter. Some complex concepts that deserve a closer look, so make sure to check out those links.

Installation was straightforward and easy enough, but perusing the “Quick Installation Guide” walks through everything you need to do.  I connected four drives and entered the BIOS setup to see what was what; Promise refers to their BIOS setup as SuperBuild and their Windows setup as SuperCheck.  The standard Promise configuration allowed me to quickly setup my array as I’ve used FastTrak adapters. Once configured I rebooted and entered Windows 2000 and installed the drivers and SuperCheck suite.

SuperCheck consists of three parts: message server, message agent and monitoring utility.  Through these the SuperTrak100 can be configured and monitored locally or across the network, even over the internet.  All the important information that’s available with SCSI RAID controllers is present here, showing Promise’s commitment to giving a quality adapter that rivals any SCSI setup.  The adapter, drives, users and configurations are represented as icons for easy navigation and I had no trouble moving around. SuperCheck’s three components are installed as services: while the SuperTrak100 requires Windows 2000 or NT for installation, it can be monitored from any Windows 9x machine.  The user’s manual covers all aspects of setup and monitoring thoroughly, and is a good read.

Also part of the monitoring available is for Promise’s SuperSwap 5.25” removable drive housings.  These only work with Promise adapters and provide a solid removable solution for hot swapping which is supported by the SuperTrak100.  In fact it’s the only ATA RAID adapter to do so, but these housings don’t come cheap.  SuperCheck monitors cable, temperature and fan status allowing for a complete picture for a remote operator.  I’m trying a few out and will have an in depth review shortly.

The benchmarks show the SuperTrak100 to be a good performer all around and the fastest ATA RAID 5 adapter available.  It’s important to remember that RAID 5 is not used for performance but reliability and fault tolerance in a reasonably fast manner.  I am disappointed that RAID 0 scores are below those of the FastTrak100; it seems the cache and i960 processor are more a hindrance in this setup.

If you’re looking for an ATA RAID solution that handles levels 3 and 5, or need hot swap capability, or require remote monitoring, then the SuperTrak series is for you.  It’s got a lot going for it, but in a server environment.  If your concern is speed or redundancy via RAID 0 or 0+1, go with another adapter and save a few dollars.

Editors note: this review was migrated from the old eBabble.net site and the photos updated.  Originally published January 27th 2001.

After spending far too many nights benchmarking, let’s take a look at the first of three ATA RAID adapters. The Adaptec AAA-UDMA was the first RAID 5 adapter designed for ATA hard drives. RAID had been the domain of SCSI hard drives until 1995 when Promise released the FastTrak, a RAID 0, 1 and 0+1 ATA33 adapter. Since then quite a few companies, Promise, Highpoint and AMI to name a few, have released ATA RAID adapters in the same vein. It took a leader in SCSI to release the first RAID 5 adapter, and so Adaptec made the AAA-UDMA.

This leads us to the first problem: Adaptec is a SCSI company, and that’s what it knows. Therefore the AAA-UDMA is based on their line of SCSI RAID adapters, right down to the BIOS. In fact Adaptec has taken one of their RAID cards, added an ATA controller chip, and created the AAA-UDMA. The card translates SCSI commands to the ATA chip and in one stroke limits it’s ability to be a contender. Let’s take a step back and look at the whole picture.

Opening the box reveals a full length PCI adapter, two manuals, four drive cables, one CD and a slew of diskettes. If you’ve never worked with full length PCI cards, be prepared for a surprise; they run the entire case and take advantage of those slot brackets by the case fan you’ve never used. The AAA-UDMA is 12.28 inches long. Examining the card shows four IDE connectors and an IDE LED connector. There is a 168 pin DIMM slot with a 2 MB ECC EDO RAM DIMM installed; why they chose EDO over SDRAM I’ll never know, but I would have liked to try a 64MB SDRAM DIMM in there to see what’s what. I put a 64MB PC100 SDRAM DIMM on the card just to see, but it gave me an error of unrecognized DIMM. Of note as well are the controller chips: standard Adaptec AIC-7890AB and an AIC-7815G. These are used on Adaptec’s SCSI RAID controllers and a sure tip off that this is a SCSI design cobbled with IDE components. The UDMA translator is an Altera Flex, which is my first experience with this company and perusing the web site didn’t give me much insight. This card is over a year old and uses ATA66 connections, giving a maximum transfer rate per channel of 66 MB/s burst speed.

Installation was quick and easy: plug the card in and reboot. If you want to boot from the array and install an operating system, a configuration boot disk is provided for most network OS’s utilizing Adaptec’s ArrayConfig utility. I was using a single disk as the boot disk and the RAID array as separate storage, so once into Windows 2000 the hardware was detected and I used the diskette to load the drivers. Up and running.

I shut down, hooked up the drives and rebooted. Unlike other ATA RAID controllers, going into the card’s BIOS doesn’t let you configure anything. It’s Adaptec’s standard SCSI BIOS and let’s you low level format the drives, but that’s about it for ATA drives.

All drive, adapter and array configuration and management is done through software, namely Adaptec CI/O Management. This package is installed from the single CD and has a thick manual dedicated to it. CI/O is used manage all Adaptec arrays, ATA or SCSI, and is easy to work with. The software serves two main functions: to set up arrays and to monitor all related equipment, locally or remotely via a client PC on the network.

Once I had everything installed I selected set up my first array, a four disk RAID 0 configuration. CI/O walked me through the process and led me to the last step: initialization. This took over four hours for most configurations using four 20 GB Maxtor drives; better safe than sorry.

Quite a few options are presented for RAID configurations: RAID 0, 1, 0+1, 5 with block sizes in kilobytes of 8, 16, 32, 64 and 128. 64 is the default setting, as it is with all ATA RAID adapters I’ve looked at. As well you can have just disks connected to the card, but it would be a lot cheaper to buy a Promise Ultra adapter if you were just looking to add two more IDE channels.

Adaptec CI/O Management is designed for just that. All aspects of the array, drives and controller are monitored and reported through CI/O or NT’s Event Viewer. I found my way around without consulting the manual and was able to accomplish most tasks easily. Everything is well laid out and logical; a minor problem was that each task opened a new window within the software. Remote management is done by installing the CI/O software and pointing it to the Adaptec array; performance over the network was good and I managed everything the same as on the machine with the card installed.

Consulting the manual showed the full depth of features developed by Adaptec for their SCSI RAID controllers, pointing out time and time again that advanced features weren’t available for the AAA-UDMA. I recommend reading through everything before deploying in a production machine, but that’s safe advice for any hardware upgrade.
Consulting the benchmarks show what a few other reviewers have discovered: this card isn’t built for speed. Adaptec bills the AAA-UDMA on their RAID site as a “sub-entry server” model. Apparently that’s reliable but slow.

To wrap it all up, I would have to recommend that most users pass on this model. If you need ATA RAID 5, go for the Promise SuperTrak line. If your environment is using Adaptec SCSI RAID controllers and CI/O Management, then the AAA-UDMA would fit right in.

Editors note: this review was migrated from the old eBabble.net site and the photos updated. Originally published January 10th 2001.

Escalade® 8506 Series

• Supports up to 12 drives with a single PCI card enabling up to 3terabytes of storage (dependent on drive capacity , 2TB per array maximum)
• Supports ATA / 133 / 100 drives with a Parallel-to-Serial Drive Converter
• StorSwitch point-to-point non-blocking architecture
• PCI 2.2 compliant 64-bit / 66MHz bus master
• RAID 0, 1, 10, 5 and JBOD support (2-port cotrollers support RAID 0,1 and JBOD)
• On-board processor to provide true hardware-based RAID
• Bootable array support for greater fault tolerance
• BIOS set up utility and 3ware Disk Manager (3DM) web-based management software
• Hot-swap and hot-spare capability
• Windows® and Linux® operating systems support

The Escalade 8506 series is 3ware’s second generation Serial ATA ( SATA ) RAID controller. Building from it’s previous generation, this series offers a 64 bit 66MHz PCI interface. On paper that’s about the only improvement over the 8500 series, but benchmarks and deep digging reveal otherwise.

The Escalade 7506 and 8506 series are 64 bit 66 MHz upgrades to their 7500 and 8500 series 64 bit 33 MHz predecessors. Check out our previous coverage of the Escalade 7500-8, and our extensive definitions of ATA RAID.

Serial ATA has really taken off the last year, mostly due to Intel adding native support in it’s 865 and 875 chipsets. Instead of worrying about master and slave settings, each SATA drive has it’s own connection via a slim cable. No jumper worries, small connectors and a slim long cable meant lots of new opportunities for system integrators. Manufacturers scrambled to create SATA products, including ATA RAID controller manufacturers 3ware, Promise, LSI Logic and Highpoint. A simple and easily employed solution was available from Marvell: the 88i8030 Serial ATA Bridge Chip. This translates Parallel ATA to Serial ATA and vice versa. Maxtor and Western Digital use them on their SATA hard drives, and the ATA RAID manufacturers mentioned above use them on their controllers. Take an existing product, add the Marvell 88i8030 and you’re in the SATA business.

For ATA RAID manufacturers this makes a lot of sense, since they can design a solution for Parallel ATA, then use the Marvell 88i8030 and have a Serial ATA solution.

Which brings us to the 3ware Escalade 8506-8. The XOR engine is the 3ware 200-0069-00 in the centre. The ATA controller chips are the same as the previous generation, the 200-0033-00, which can control four drives each. This time around two ISSI memory modules are onboard, maintaining 3ware’s pattern of 1.8 MB hardwired cache. The board layout is clean and relatively compact. The spots for the 12 port model can be seen on the PCB, saving 3ware some manufacturing costs.

3ware has two technologies for ATA RAID: StorSwitch and TwinStor. Here’s a quote from their FAQ, which is pretty self explanatory.

3ware’s StorSwitch replaces shared-bus architectures found in SCSI systems with a dedicated port for each drive, thus maximizing throughput and minimizing latency of the drive subsystem. 3ware’s TwinStor is a RAID level 1 technology that profiles disk drives and optimizes the algorithms for greater performance during reading and writing in a two-disk mirror. TwinStor is designed for applications that are read intensive, such as web serving which utilizes a mixture of small and large data sets with multimedia content.

Opening the box showed the controller, installation guide, software on 3.5” floppy disk and CD-ROM and eight SATA cables. Everything you need, except possibly SATA power adapters. They will run you a few dollars each, but may not be needed if your SATA hard drives also have molex power connections such as the Maxtor drives used for our testing.

I installed the controller, connected the drives and powered up the system. The BIOS setup hasn’t changed from the previous 7500 series, and for that matter doesn’t stray far from any ATA RAID controller. Select your drives and define your array. For RAID 0, 1 and 10 arrays the build time is instant. For RAID 5 arrays you now have a choice of having the array initialized, meaning the entire array is written to for verification, or you can go the instant route. Once into Windows 2000 the driver prompts were followed and the system was up and running after a reboot.

The 3ware Disk Management ( 3DM ) was installed: it’s HTML based and can be accessed from anywhere on your network. A very clean and easy to follow interface, it’s quite easy to use. We were fully installed and away to the races. Lots of nice management features, including email and pager notification. SMART status is displayed as well. When something goes wrong a message flashes on screen to alert the user, as I experienced when one of my drives went into power saving mode all on it’s own.

Let’s talk features. The Escalade 8506 is a 64 bit 66 MHz PCI controller capable of sending 512 MB per second across the PCI bus. It handles RAID levels 0, 1, 10, 5 and JBOD ( just a bunch of disks ). Who really spends $500 for JBOD? It supports hot swap and hot spare. Hot swap allows you to remove a failed drive and replace it while the system is running, provided you’re using a drive enclosure. I love drive enclosures but could never spend the money on them: crazy stuff. Hot spare allows you to have an extra drive connected to the controller, not part of the array, that can be added if a drive fails. Both scenarios have the controller rebuilding the array after a drive replacement in real time, so you lose performance but not uptime. All in all every ATA RAID controller should have these features, and most do.

Benchmarking was performed and compared against the Promise FastTrak S150 SX4. While the previous PATA benchmarks showed 3ware with a sizable lead, this time around things are evenly matched. The 8506-8 took most of the benchmarks, but by a very slim margin. 3ware trumps in two areas: 8 and 12 port cards, and a smaller form factor. Otherwise it’s a dead heat.

The 3ware Escalade 8506-8 is an excellent SATA RAID controller, capable of handling eight drives. If your needs lean to eight or twelve drives, 3ware is the way to go.

Editors note: this review was migrated from the old eBabble.net site and the photos updated.  Originally published November 28th 2003.

]]> http://tech.ebabble.net/3ware-escalade-8506-8/feed 1 http://tech.ebabble.net/3ware-escalade-7500-8 http://tech.ebabble.net/3ware-escalade-7500-8#comments Fri, 26 Jan 2007 18:56:55 +0000 Scott VanderPloeg http://www.ebabble.net/reviews/3ware-escalade-7500-8/

The world of ATA RAID is a small one. When I first became interested there were two players: Promise and Highpoint. Slowly I began to hear of 3ware, a company that produced great products, but at a high price. At that time RAID 0 and 1 were about it, and then Promise released the SuperTrak100 and Adaptec released the AAA-UDMA, both touting RAID 5. At that time 3ware released a BIOS and driver update for their Escalade 6000 series that added RAID 5. Of course there was no onboard RAM cache, so performance was lackluster. Then with the second wave of ATA RAID 5 3ware released the Escalade 7000 series and things looked up. We’re looking at the Escalade 7500-8, an ATA RAID controller capable of handling 8 drives in RAID 0,1,5,10 and JBOD ( just a bunch of disks ) in a Windows 2000 environment.

Escalade® 7500 Series

• Supports up to 12 drives with a single PCI card enabling up to 3terabytes of storage
• Ultra ATA/100 interface
• StorSwitch point-to-point non-blocking architecture
• PCI 2.2 compliant 64-bit/33MHz bus master¹
• RAID 0, 1, 10, 5 and JBOD support²
• Highest performance ATA RAID Controller
• On-board processor to provide true hardware-based RAID
• Bootable array support for greater fault tolerance
• BIOS set up utility and 3ware Disk Manager (3DM®) web-based management software
• Hot-swap and hot-spare capability
• Windows® and Linux® operating systems support

Opening the box revealed: the card, eight 18” ATA 80 wire 40 pin cables, a manual, a driver revision printout, a software envelope containing driver disks and a CD, and four molex “Y” power splitters. I read through the manual to get a feel for the product: it’s a thin photocopy that covers the basics. The manual covers the Escalade 6000 and 7000 series, so the photos in the manual are of 6000 series cards. What I found odd was that the 7.4 driver revision notes were thicker than the manual. Even so, the manual was straight forward and easy going for anyone with RAID experience.

Looking at the actual Escalade 7500-8 adapter, it’s layout is simple and effective. A central 3ware 200-0017-00 chip for RAID calculations and their own StorSwitch technology, two 3ware 200-0033-00 chips to each handle four ATA133 ports. Unfortunately there’s no information regarding individual ASICs on the 3ware website. Two 150 MHz 0.9 MB IDT memory chips for a total of 1.8 MB onboard cache, although I’ve read it online as 2 MB cache available. A few other chips on the board I was unable to identify, plus a four pin LED connector. The board is compact, reasonably short and well laid out.

Installation was straight forward: connected the drives, inserted the card and booted. the Escalade series are 64 bit 33 MHz PCI adapters, allowing for a maximum of 266 MB/sec transfers. Installed the drivers, rebooted and installed 3DM, 3ware Disk Management Utility for Windows. Creating, deleting and repairing RAID arrays is done via the card’s BIOS at boot up, and employs a simple and easy to understand interface. Creating RAID 0, 1 and 10 arrays was immediate: creating RAID 5 arrays took some time as X’s are written to test. This is good feature but takes some time. While it was inconvenient for testing and benchmarking, users should see this as a nice safety feature. As with all RAID products, the Escalade 7500-8 showed as a SCSI controller in Device Manager, and arrays show under disk drives.

Monitoring and reporting is handled via 3DM. This runs as a web server, so access is through your browser and accessed locally or over your network. Again, easy to use and straight forward. Linux users have the CLI, or 3ware Command Line Interface.

In the box I received driver version 7.4, which I installed and began to benchmark with. After a few weeks the board stopped working, so I took advantage of 3ware’s technical support. The system hung at the Escalade 7500-8’s BIOS and wouldn’t go any further. First you register your product online and create a profile, then submit a problem report. Within a day I received an RMA for the board, and since I selected cross shipment by providing my credit card, I received a replacement board within a week. Very nice and smooth technical support.

This time the box contained the 7.53 drivers. I began benchmarking with these, the latest drivers. Performance was ho hum and nothing special. Following Storage Review’s forums on 3ware products revealed users resorting to registry hacks and changing disks from basic to dynamic to get better performance. Then a posting revealed a new driver revision, 7.6, that gave a real performance boost without the “voodoo” tweaking. Redid all benchmarks with the 7.6 driver, firmware and 3DM releases. Doing this revealed the product forces you to update all three items, another nice feature to keep things as 3ware wants it.

Benchmarks don’t lie: the 3ware Escalade 7500-8 with the 7.6 drivers dominated all tests save WinBench99 RAID 5. Sometimes it’s a close call with the FastTrak SX4000, other times a huge margin.

Recommendations are easy: 3ware products give the best ATA RAID performance. If you need to use 8 or 12 drives, they’re the only player in town. The Escalade 7500-4LP is a four drive low profile adapter and fits just about anywhere, another great plus. The only concern is cost: expect to pay 50-100% more for 3ware cards compared to Promise’s FastTrak SX4000. In this case you get what you pay for: the best performance for the highest price.

Editors note: this review was migrated from the old eBabble.net site and the photos updated. Originally published May 30th 2003.

]]> http://tech.ebabble.net/3ware-escalade-7500-8/feed 1 http://tech.ebabble.net/maxtor-maxline-pro-7h500f0 http://tech.ebabble.net/maxtor-maxline-pro-7h500f0#comments Mon, 22 Jan 2007 14:16:43 +0000 Scott VanderPloeg http://www.ebabble.net/maxtor-maxline-pro-7h500f0/ This is not so much a review but a cautionary tale. Let me begin by saying I’ve been a longtime supporter (and purchaser) of Maxtor drives. They’ve had a few bumps in the road but I stick with them.

Just after Christmas the Maxtor Maxline Pro 500 GB drives were going on sale online. It makes sense that Seagate would be clearing them out and they’re the only drive really left available in Maxtor’s high end line that wasn’t SCSI. I’m building a new server to test out the beta of Microsoft Data Protection Manager v2 and needed some drives for a RAID array; these were a real bargain so I ordered four.

I set up the server with an AMCC 3ware 9550SX RAID controller and the four Maxlines connected in RAID 5. 3ware cards like you to initialize an array before use; basically it does a full scan and diagnostic to make sure everything is working optimally. I upgraded the card’s firmware to the latest level and installed it’s drivers in my clean install of Windows Server 2003 R2.

After setting up the RAID 5 array I watched the initialization for about twenty minutes and went out; to fully initialize 1.5 TB will take many hours. I returned a short time later to see the drive on port three had failed and the RAID array had degraded. I shut everything down and checked the cabling and connections but no problems. I deleted and recreated the array, but after a while of initialization I had the same issue on port three. I swapped the drive to port five and repeated the process. This time it failed but told me the drive had a SMART (ECC) FAIL. I pulled the drive and set it aside.

Figuring something was up I created the array with three drives instead of four. Initialization failed shortly thereafter with a DEVICE ERROR on port four. Rinse and repeat, same thing.

Somewhere between exhaustion and blind rage I pulled all four drives out and hooked them up individually to scan with Maxtor’s PowerMax diagnostics utility. Started with the SMART error drive; the HP I had it hooked to let me know drive failure was imminent but I ran the diagnostics anyway and found it was dying. Next drive tested was the DEVICE ERROR; PowerMax came to the same conclusion. The last two tested fine.

Final score: two of four brand new drives purchased are in need of replacement. My four drives were all manufactured 12SEP2006 so avoid that batch if you’re buying. Maxtor’s RMA is now part of Seagate but they have an exchange program so I’ll get my replacements in a few days.