Written By:
Date Posted: August 6, 2001

Most of you are probably familiar with , but for those of you that are not, Abit has been around for a long time, making some of the most popular retail motherboards. Their claim to fame is their tweakability, solid design, and adding some advanced features to their motherboards such as RAID support and their Soft Menu BIOS. Abit has always been near or at the top of performance and stability. The KT7A-RAID is no exception.

The KT7A-RAID is an AMD Socket A motherboard. It's licensed the Northbridge chipset to power this board. Basically, in a nutshell, the officially supports the 133FSB for the CPU and ram. This is good for a few reasons. People can now use the newer AMD Athlons with the 266Mhz bus at it's full potential, and they can possibly overclock the older 200Mhz bus Athlons to higher, more stable speeds. The previous KT133 chipset was unable to go faster than 115, on average, so the KT133A certainly opens the door. As you will see later on, the small jump from 100FSB (older Athlons and the original KT133) to 133FSB yields impressive performance gains in most applications. The other usual suspects are supported as well, such as ATA100 support, AGP 4x, ACPI and USB.

Specifications

CPU: Supports AMD Athlon/Duron 600MHz-1.2GHz 200/266MHz FSB Socket A Processors
Chipset: VIA KT133A /VIA 686B
Ultra RAID DMA 100, RAID 0/1/0+1Technology
High Point HTP370 IDE Controller
Memory: Three 168-pin DIMM sockets support up to 1.5 GB PC100/PC133 SDRAM module
Multi I/O: Four channels of Bus Master IDE Ports supporting up to 8 Ultra DMA 33/66/100 devices
Miscellaneous: 1 AGP slot, 6 PCI slots and 1 ISA slots
BIOS: SoftMenu"III Technology to set CPU parameters

Abit did a good job with the motherbooard design. The last motherboard I had made almost no sense with it's layout. The PCB is very clean here, and the board should have no problems fitting in the majority of ATX cases. The only major issue I had was with the DIMM slots. If I plug in a video card into the AGP slot (which you need to do eventually), removing or inserting DIMMs in the first slot may be difficult. Oh yes, I mention DIMMs. This is a SDRAM only board. It's fully compatible with PC100/133 ram, so anyone who bought a few truckloads will be happy here. Thankfully, Abit does not include any onboard sound, since it sucks usually, which cuts down a bit on the total cost.

One big selling point is the Soft Menu III BIOS. Many motherboards in the past required you to flip jumpers and switches to properly configure. Newer motherboards don't, and are pretty much plug and play. Abit has gone the extra mile in adding in all kinds of extra features in there Soft Menu III, such as voltage tweaks and memory timing adjustments. All this is important for enthusiasts who like to squeeze every drop of performance out of their rigs.

Another feature is the integrated RAID support. The brains behind the RAID is the . According to documentation, it supports RAID 0, RAID 1, and RAID 0+1. To explain, RAID stands for Redundant Array of Inexpensive Disks. Back in the past, RAID was popular only with servers and workstations that supported SCSI disks. Hardly inexpensive, these disks were the only ones that were capable of working in a RAID array. Times have changed, and RAID has become very popular in the retail market.

RAID 0, also known as "Striping" configures several drives, a minimum of 2, so that data is written at the same time on all the drives in the array. Since data, ones and zeros, is written between the drives, where ones go to one drive, zeros to the other, your performance improves. When one disk fails, the entire array is lost. This may sound risky, but the same would happen if you relied on one drive. Just keep your backups current, and you'll be fine. For optimal performance, it's generally recommended you use identical drives. The reason is because all drives will perform only as fast as it's slowest member. Another reason for identical drives, is that RAID 0 will use the full capacity of it's smallest drive. So, if you combine a 40GB drive with a 20GB drive, your array total will only be 40GB, rather than 60GB.

RAID 1, known as "Mirroring", is exactly that. It'll take 2 drives, and "mirror" them, writing the same data on both drives. This is the slowest method of RAID, but it provides full redundancy in case of failure.

RAID 0+1, or "Striping + Mirroring" borrows techniques from both striping and mirroring. It performs well by reading and writing at the same time off two disks, while mirroring on another.

Click here to read a small article I wrote on the Soft Menu III BIOS and HPT370 RAID controller.

The IDE slots are placed next to the RAID slots, out of the way of interfering with your other components. Everything is properly labeled on the PCB and in the manual, so you won't accidently plug your hard drive into your floppy slot ;) The expansion slots are neatly arranged, and with the exception of possible AGP cards getting in the way of the DIMM slots, you shouldn't have any problems adding any other expansion cards. You got a total of 6 PCI slots, so there is plenty of room to grow here.

Abit has included a thermal sensor withen the socket slot. This is good, because it will allow you to properly monitor the approxiamate temperature of the CPU. Since AMD does not include a thermal diode with their procs. To cool the KT133A northbridge, AMD has added active cooling in the form of a heatsink and fan to avoid lockups. I question the usefullness of it since I don't think it really needs the fan, but it does look nice.

Speaking about lockups, ...I hate to spoil it for some multimedia design, workstation type people, but the RAID controller here is problematic with the Matrox RT2000 DV card. It's , which is sort of true. You can disable the controller, and the card will work, but you lose the RAID. I'm sure there is other hardware with similar issues, so check with the manufacturers beforehand.

Performance

The test bed consists of an unlocked, AVIA core Thunderbird, at stock, underclocked and overclocked speeds:

Tbird 1.2GHz (1.75v, 9x133), 1.2GHz (1.75v, 12x100), 1.3GHz (1.85v, 10x133), 1.4GHz (1.85v, 10x140)
Abit KT7-RAID (duh!)
512MB PC133 Ram CAS2
RAID 0 Array w/2 x Quantum Fireballs

As I stated on the first page of the review, the KT133A should provide a performance boost using the 133FSB. Since I didn't have an older 1.2GHz CPU handy, I simply changed the multiplier to 12x100 on my CPU. The first series of benchmarks will be the 1.2GHz w/different bus speeds, 100FSB and 133FSB.

Sisoft Sandra - CPU

(left) 1.2GHz (1.75v, 9x133) and (right) 1.2GHz (1.75v, 12x100)

As you can see from the results on the previous page, the results are fairly close at either 12x100 or 9x133. This doesn't surprise me so much, since in either case, the CPU is still running at 1.2GHz.

Sisoft Sandra - MMX

(left) 1.2GHz (1.75v, 9x133) and (right) 1.2GHz (1.75v, 12x100)

It's a little more obvious with the MMX benchmarks that the 133FSB provides a small boost. Hardly worth mentioning it though since the difference is so small, but the improvement is there.

Sisoft Sandra - Memory

(left) 1.2GHz (1.75v, 9x133) and (right) 1.2GHz (1.75v, 12x100)

Now we're talking baby. For the memory bandwidth tests, the extra 33MHz jump yields impressive gains. Now, before we get too excited, keep in mind that these are synthetic tests, but you can see it's a huge improvement. For the record, the same CAS2 ram was used in all testing, and no changes in the BIOS were made other than user defining the CPU. If you have PC133 ram, and an unlocked Thunderbird or Duron CPU, getting a 133FSB capable motherboard will add new life to your computer.

Overclocking

Well, other than features like RAID (which I'll get to later on), Abit is renowned for it's overclocking potential. When sifting through the BIOS, you got adjustments for the multiplier (so long as your CPU is unlocked), the front side bus, memory timings, vcore voltage tweaks and more. Using my 1.2Ghz AVIA core Thunderbird, I managed to get a 1.33 (vcore 1.85) overclock, as well as a 1.4 (vcore 1.85) overclock. I was successfully able to boot into Windows, as well as running a number of applications without crashing. You'll notice in the screenshots for 1.4GHz (note the hand hovering over funky coloured text), that I went with a 10 multiplier and a 140FSB. I tried 10.5x133, but Windows would not load. It would post properly, but as soon as it gets to the login screen, it'd hard lock. I also did get some failures in Prime95 at both 1.33GHz and 1.4GHz however. I would always have the test abort after starting up the application. Cooling was adequate, since the load temperatures never passed 45 degrees Celcious, and I got a newer Enermax 350W PS, so I figure it's the CPU or I didn't supply enough juice to the proc.

Sisoft Sandra - CPU

(left) 1.33GHz (1.75v, 10x133) and (right) 1.4GHz (1.75v, 10x140)

Nice! At 1.2GHz, the ALU mark was 3312 and FPU was 1612. The 1.33GHz overclock yields a nice increase, but the 1.4GHz really kicks ass. Uh, just for you to know, I'm positive a lot of the improvement at 1.4GHz is due to the fact that it's on a 140FSB. Still, since we're talking about overclocking, it doesn't matter how you get there, so long as you get further than where you were before.

Sisoft Sandra - MMX

(left) 1.33GHz (1.75v, 10x133) and (right) 1.4GHz (1.75v, 10x140)

Well, the MMX tests seem to be following the same pattern as CPU tests. Again, no surprise, since we're talking about CPUs 133MHz to 200MHz above stock speed.

Sisoft Sandra - Memory

(left) 1.33GHz (1.75v, 10x133) and (right) 1.4GHz (1.75v, 10x140)

This kind of perplexed me. Due to the increased FSB, the 1.4GHz easily stomps the 1.33GHz, but barely outperforms the 1.2GHz. I did these tests a few times, and the results were withen 2% of each other, so I'm at a bit of a loss to explain this.

Gaming Performance

Yay! Now we're talking. Synthetic benchmarks are fine and dandy, but let's get to the important stuff, video games! I've decided to benchmark Quake 3 (like every other website in the world) and 3D Mark 2001. It happens that I recently aquired a MSI Starforce 822 GeForce 3, so I'll be able to run all the tests in 3D Mark, which is what we'll start with first.

The picture above is the settings I chose to use for the benchmarks. For those who don't already know, with todays powerful video cards, and the GF3 in particular, benchmarks at low resolution, low detail will stress the CPU more than at higher resolution, which is more video card dependent.

Fot testing, the only factors worth noting are the MSI GeForce 3, running nVidia drivers v12.41, Norton Antivirus running in the background, and the OS, which is Windows 2000, SP2. vSync was disabled for all the tests, and video driver Display Properties are installation defaults.

(left)1.2GHz (1.75v, 12x100) and (right)1.2GHz (1.75v, 9x133)

(left)1.3GHz (1.85v, 10x133) and (right)1.4GHz (1.85v, 10x140)

As expected, 3D Mark scores improve as we boost the CPU speed. Nothing surprising, but check out the 1.2GHz scores. At 12x100, we get a score of 5457, but at 9x133 it's 6006. An improvement of 549 points! Like I said before, the small 33MHz FSB boost kicks ass.

Next up is Quake 3, same test bed as before. Quake 3 is version 1.29h, using demo four.dm_66. Settings are Fastest, at 640x480, with sound on.

To do the Quake 3 timedemos as I have, simply bring down the console. "~", without the quotes, in the main menu, then type:

/timedemo 1
/demo four

The demo will play very fast, and then exit back to the main menu. Bring the console back down, and the frames per second is displayed.

Here are the results:

1.2GHz (1.75v, 12x100): 126 Frames Per Second
1.2GHz (1.75v, 9x133): 143 Frames Per Second
1.3GHz (1.85v, 10x133): 150 Frames Per Second
1.4GHz (1.85v, 10x140): 149 Frames Per Second

Like the 3D Mark scores, there is a large jump from the 100FSB to the 133FSB. Things seem to top out at 1.33GHz though, which is odd, since I've seen online scores much higher when surpassing that speed. I redid the test a few more times, and the results were the same.

RAID Performance

Sisoft Sandra - Hard Disk

1.2GHz (1.75v, 9x133)

You're probably wondering why the RAID 0 scores seem low, compared to the comparison RAID 0 test bed. I was a bit curious myself, until a friend pointed out that it's likely because I split the RAID array up into 7 partitions. That'd make some sense since the drive heads have to move around more. Other than the scores, coming from a non-RAID motherboard before, I can say that applications installed off the hard drive is much faster. When I load up Quake 3, I am always one of the first ones to connect to a new map on a server. I'm pretty sure the RAID 0 array I set up is helping me load maps up faster.

Final Words

At the time of writing, Abit is releasing the DDR powered KG7 motherboard, so getting the KT7A-RAID may be of some question. I have no doubt that the KT7A-RAID is one of the best SDRAM based motherboards out there. If you happen to have a 700MHz+ Thunderbird CPU, that is unlocked, you can take advantage of the KT133A chipset, and get a nice performance boost. I had a few lockups when I tried overclocking, but overclocking involves some luck, something I think I lacked with some of my overclocking attempts.

The RAID version only adds about 30$ to the cost of a non-RAID board (the Abit KT7A). Considering a decent IDE RAID controller runs about 75$ or more, this is a bargain. Unfortunently, some of you may not be able to use the RAID features, depending on your hardware, so you may want to consider the KT7A and save some money. For the rest of you, if you're still hesitant about the DDR plunge, but want the RAID performance, and you have perfectly good PC100/133 ram, I strongly urge you give the KT7A-RAID a hard look.

Pros: Great performance, Soft Menu III, loads of tweaking, jumperless, RAID support, great manual.

Cons: RAID can cause incompatibilities with some hardware.

Home>>