ABit BP6 Motherboard: a review
by Drew Dunn
by Drew Dunn
The NOSPIN Group
Multiple processor motherboards are nothing particularly new. When mainframes ruled the world, gangs of processors made crunching data much faster because, as computer engineers discovered, if you process data in parallel, that is, many pieces of data at the same time, you can speed up computer performance immensely.
That fact has carried over into the PC arena nicely. Modern servers of two to 16 processors are common because the last thing a server administrator wants is for a server to be bottlenecked by its CPU. Workstations benefit from multiple processors for the very same reason. Graphics processing is very processor-intensive, so the more processors (within reason) in a computer, the faster the job gets done.
But, of course, all this comes at a price. Pentium II and III processors are pricey, so it's difficult to justify putting them into an every-day system.
Dual Celeron-based systems have been around since hackers discovered that a bit of judicious rewiring would allow two Slot-1 Celerons to work together. It's even possible to use a sort of kludge to install two Socket 370 Celeron processors in a dual Slot-1 motherboard.
ABIT has come up with something different. As of this writing, the BP-6 is the only dual Socket 370 motherboard on the market. And besides putting the power of dual CPU computing in your computer, it also adds a few more features.
The board comes equipped with the requisite features: ATX form factor, serial, parallel, PS/2 mouse and keyboard, USB, wake on LAN, fan and infrared headers. The board also includes 5 PCI, 2 ISA (one of each shared) and an AGP slot. Three temperature sensors monitor the two CPUs and the ambient air temperature. And you can connect up to eight IDE devices to the motherboard.
What? Eight IDE devices on one motherboard? It's true, and what's more, four of them can be ATA/66 compliant. Abit includes, along with the standard ultra 33 controller, Highpoint's ATA/66 controller. The four IDE sockets support two devices each.
I installed the motherboard in a mini tower case, added 96MB of PC100 RAM (although PC66 will work just fine with non-overclocked Celerons) and connected the rest of the peripherals. In this case, the system was outfitted as follows:
(2) Celeron 333 CPUs
Intel Etherexpress Pro NIC
Maxtor 5.5GB SCSI hard drive
Seagate 6.5GB SCSI hard drive
Buslogic Flashpoint UltraSCSI controller
Seagate 9GB SCSI hard drive
Western Digital 4.3GB SCSI hard drive
Iomega JAZ SCSI drive
HP 4020i SCSI CD writer
Plextor 20Plex SCSI CD-ROM
After closing up the box and powering up the system, the motherboard detected the two processors. I entered the BIOS setup program and configured the usual range of drive settings, time, date, and that sort of stuff. The setup program includes settings for the processor multiplier and clock speed. It's possible to overclock the Celeron processors by altering the front side bus, up to 133MHz. You can also adjust the core voltage to compensate for the faster bus speed.
ABIT claims that it is possible to operate the motherboard with different speed CPUs, although I did not test that claim. And while a prominent warning states that two CPUs should only be used experimentally, the motherboard was stable over the course of two months of continuous operation.
Of course, the bottom line is performance. I replaced a Pentium II 350MHz processor with two 333MHz Celerons. Was the difference noticeable? In a word, yes. Even applications that did not directly support multiple processors (a technique called multithreading) ran faster since the operating system (RedHat Linux 6.0) supported dual processors and could select which CPU ran which programs. Multithreaded applications, such as The GIMP and POVRay, normally very CPU intensive, performed their graphical duties up to 75% faster.
The motherboard lends itself well to overclocking. With PC100 DIMMs installed, I was able to successfully run the Celeron 333's at a whopping 500MHz. And after adjusting the core voltage slightly (and making sure that there was plenty of cool air available), the system remained very stable, a testament to both the motherboard and the apparent bulletproof-ness of the Celerons.
The integrated peripherals worked just as well. An Iomega Zip drive connected to the USB port operated just as expected (with the latest Linux USB patches). The IDE controllers did just what they were supposed to do. I did not test the ATA/66 functionality of the Highpoint controller, but as a U/33 controller, it worked well, with one exception. The controller has problems with CD writers. Abit is aware of the problem and recommends that IDE CD writers be attached to the U/33 controller. Also, they report that Maxtor U8 and Seagate Barracuda hard drives may have problems, but IBM Deskstar drives worked well.
Another issue isn't a problem with ABIT or the motherboard, but just an inherent issue with operating systems. To use this motherboard with two processors requires an operating system that supports them. Windows 95 and 98 will not. Windows NT and Linux do. If you're considering a motherboard like this, make sure that your operating system will support it.
The motherboard comes with a utility CD-ROM, floppy cable, U/33 IDE cables, ATA/66 IDE cables and a well written instruction manual. Support is available via
ABIT's web site, email and newsgroups. BIOS updates are available at the web site (the latest, November 17, adds support for 37GB drives and 600MHz processors).
The bottom line? The motherboard works. It works well. It's fast, it's inexpensive and it's reliable. You can buy one with a couple of Celeron 366's on the street for less than US$300. After two months of non-stop work in a production environment, it gets my recommendation. If you're on a budget and want the most bang for your buck, get this motherboard!
ABIT Computer (USA) Corporation
46808 Lakeview Blvd.
Fremont, California 94538
ABIT BP6 Owner's Manual
Acrobat PDF file (1.4mg)
January 5, 2000
After using the BP6 as a Linux server for several months, I decided that a moreextensive test of the motherboard's capabilities was in order. While I purchased the original BP6, Abit sent another board to us for review. I replaced a Pentium II 450MHz board with a dual Celeron 333 setup. To provide as torturous a test as possible, the system was configured as so:
- Video: STB Velocity 4400 AGP, 16MB
and Creative Video Blaster PCI, 4MB
- 3D Graphics: STB Black Magic Voodoo 2, two in SLI mode
SCSI: Symbios 8100S SCSI 2 PCI
Network: Netgear FX310 100Mb PCI
Scanners: HP Scanjet 4P, HP Photosmart, both SCSI
CD-ROM: Acer 42x, HP 8100i CD-RW, both IDE
Hard Drive: Maxtor 11GB (two) on standard IDE controller and
IBM 8GB (two) on HighPoint HPT366 controller
USB: Entrega Hub4U, Iomega Zip100 USB, Microsoft Digital Sound System, Microsoft Explorer Mouse, Microsoft Natural Keyboard
Memory: Samsung 128MB PC100 (2x64)
Installing the peripherals was no trick, since the motherboard has 5 PCI and one AGP slots. But installing the peripherals so that they would work was another story. As advanced as computers are today, we are still saddled by the same 16 IRQ lines that were introduced on the IBM PC AT.
Even though the HighPoint HPT366 controller is integrated on the motherboard, it still requires resources controlled by the PCI bus. So, the controller is hard-wired to use the same interrupt as PCI slot 3. And slots 4 and 5 use the same DMA signal. So to use PCI slot 3 along with the HPT366, the card must be capable of sharing the IRQ. And two bus mastering cards cannot be installed in slots 4 and 5 at the same time.
A little forethought saved the day in this case. The two Voodoo 2 cards were not bus mastering devices and the PCI video card did not require an interrupt to function under Windows 98. Thus they were installed in slots 4, 5 and 3 respectively. An important note, of course, is that Windows 98 will only use one of the two processors on the motherboard, and Windows NT will not allow for IRQ sharing.
After disabling the VGA IRQ in the BIOS setup, I started the computer and installed the operating system. Windows 98 installed with no problem, detecting all of the installed hardware, including the HighPoint controller, which appears as a SCSI controller in the device manager.
The board overclocks fairly well using Windows 98SE. I could run the system with a 92MHz front side bus, or 450MHz, but faster speeds resulted in system lockups or data corruption. Nonetheless, at 450MHz, the Celeron-powered BP6 was a solid, well-behaved performer.
I still like the BP6. Even with the potential headaches that come with arranging a full stock of peripheral cards to match the motherboard's IRQ and bus mastering kinks, this is still a heck of a bargain and a hot performer to boot.
Long Term Torture
Last year, I looked at Abit's BP6, a Socket 370 board that had some ground-breaking features like support for dual CPUs, ATA66 and SoftMenu II. I liked the board…a lot. And while Abit was good enough to send us one gratis, I actually went out and spent my hard-earned cash on one of my own.
Dual processor boards aren't any great news, of course…they've been around for a long time. But a dual processor Celeron board brought a lot of excitement, especially since Intel frowned on using their low-end processor in a high-end situation. After all, Pentium II's were the processor that they would prefer to see in a dual processor board. Fortunately, Abit was more responsive to their customers than to Intel.
Both of the boards are in the NOSPIN Network Operations Center. One of them, with dual Celeron 366's, hosts the stable of web sites that we maintain. The other, fitted with Celeron 333's, does general duties as a CD burner, game box and all around punching bag for whatever Linux software needs testing.
Both systems run 24 hours a day. In fact, the only time that they're ever turned off is for hardware upgrades or power failures (the UPS shuts them down). Neither board has even given the hint of a problem. I've gone through a couple of hard drives and a network card, but through it all, the motherboards have persevered.
The web server has 128MB of memory and two 10GB IDE drives using an ArcoIDE RAID controller. An IDE CD-ROM is connected to the DMA 33 socket and an old Buslogic Flashpoint SCSI controller supports an even older Conner 4/8GB DAT drive. The machine is connected to the Internet with an Intel EtherExpress 100Mb network card.
The other machine is a hodgepodge with three SCSI controllers: two Adaptecs, a 2940UW and a 3940UW, as well as a Buslogic Flashpoint. A plethora of devices connected include five SCSI-3 drives with software RAID striping, three Plextor CD-ROM drives, including a 12x burner, four SCSI-2 drives, a Tecmar 20GB tape drive and a Hewlett-Packard Scanjet 4p. An STB Velocity 4400 TNT AGP video card drives the monitor and an Intel EtherExpress network card talks to the network. It has 384MB of memory installed.
As I sort of alluded to earlier, both of these boards are very stable, something that's very important in the kind of environment that they are in. The last thing that I want is for a web server to crash in the middle of the night or when I'm out of the office. That's never happened with the BP6. I'm also impressed that these boards work well even with the very old hardware that we've thrown at them.
The only quirky part of working with these boards is dealing with the IRQ assignments. The board has five PCI slots, but the IRQs and DMA channels are shared between them and the onboard IDE controllers. The web server has a pretty minimal set of peripherals, but the other system is stuffed. The upshot is that getting everything to work correctly in that system required a balancing act that would have made the Flying Wallendas proud. After trying something like 10 different combinations of cards in various slots, however, everything did end up working. I suspect that if I actually had any IDE devices in the system, though, that I would be out of luck, since the four PCI and one AGP card pretty much exhausted the available resources. But I don't think that I'd throw the blame at Abit's feet on this one…after all, they're not the ones who limited us to 16 interrupt lines per system.
Both of the systems use Thermaltake's Golden Orb heatsink/fans. They aren't the best fans, but for the money, they're hard to beat and they look pretty darn cool, too. On the BP6, though, they are a little tough to install. Abit's engineers put the filter capacitors pretty close to the sockets, so I had to file and grind on the bottom of the Orbs to get them to fit.
In an era of 815 chipsets, the 440BX chipset on this board is showing its age, but since Abit managed to pack in a ton of adjustability with Softmenu II, the BX isn't dead yet. Even though Softmenu II makes this board a great candidate for overclocking, I never took advantage of it since stability is my game. But you can see plenty of stories of successful OC'd BP6s on the 'net.
Would I buy one now? That's a hard question. The VP6 is on the market for about the same cost as a BP6, so I'd be torn. But if I could find one in the bargain bin, or maybe a reliable used board on the cheap, I would snap it up in a second. I suspect that I'll always look at these boards with fond memories and admire Abit for looking Intel in the eye and delivering what their customers wanted.