AOpen AX6BC Pro II Millennium Edition
Last modified 03-Dec-2011.
First, there was the AX6BC. Which begat the AX6BC Pro, which had jumperless voltage adjustment as well as jumperless CPU speed adjustment.
The Pro begat the AX6BC Pro II, which is more stable at high FSB speeds, and ditched the two ISA slots in favour of one more PCI slot. Then came the Pro II Millennium Edition, which is essentially the same thing as the plain Pro II, only prettier.
There are things in this world that look better and better the drunker you get. Dragsters. Narn starships. Model tanks. And AOpen's latest, limited edition, motherboard, the AX6BC Pro II Millennium Edition. The name's not much, and neither's the packaging - one of AOpen's ordinary, anonymous, "Computer Motherboard" boxes, without even a special badge for your PC - but the board itself is... well, it's distinctive.
Shiny black (AOpen call it a "Special Military Grade BLACK PCB"). Platinum plated heatsink on the main motherboard chip (AOpen describe it as "25 karat platinum", which doesn't make sense, but I can forgive them). A little "Millennium Edition" plaque, also platinum. Designer's names.
Truly, this is the turbo-supercharged six-wheel-drive nitrous-enhanced super-ultra FAFMO (For A Few Men Only) edition of the AX6BC.
It's sort of a shame that you're going to put it inside a PC case, really.
But there's good reason to do so. All of that show-off stuff is just the icing on a quite impressive cake. Like earlier boards in the AX6BC line, this is a serious piece of gear.
Hey, sometimes those cars covered with racing team stickers and goofy badges and big look-at-me wings actually are fast, you know.
The Millennium Edition board sells for $310 (Australian dollars).
What you get
As a motherboard for the cognoscenti, the Millennium Edition doesn't bother with those awfully plebeian ISA slots. It's PC 99 ATX form factor, with six PCI slots, one AGP, no ISA. The PC 99 spec also explains its gaily coloured rear connectors.
Colour-coded rear panel connectors, according
to PC 99 spec. Check out the hefty capacitors
beside the CPU slot!
These big caps aren't entirely an advantage,
for extreme overclockers. They're part of the
reason why the AX6BC Pro II can run stably at
FSB speeds above 133MHz, but they're tall enough
to foul some of the outrageous CPU coolers that
the nuttier overclockers like to use.
You get three DIMM sockets for RAM (up to 768Mb memory supported), and all the usual ports and connectors, including USB and IrDA. The motherboard supports Wake On LAN if you've got a network card with the Wake On LAN connector. Similarly, both the plain serial port and zero-voltage power-up Wake On Modem modes are supported. Plain Wake On Modem requires a powered-up external modem, activity from which will power up the computer. Zero-voltage Wake On Modem works with internal modems, via another special connector. There's also an SB-LINK connector, in case you want to use a PCI Sound Blaster card under DOS.
There are the usual two ATA drive connectors, each of which supports the usual two drives, at up to Ultra DMA/33 speed. Some speed freaks insist on UDMA/66, which has a bandwidth of 66 megabytes per second, but since no ATA drive configuration can actually saturate that much bandwidth for more than the tiniest fraction of a second before the data in the drive caches is exhausted and the far slower real drive transfer rate takes over, there's no real advantage to the faster standard.
The Millennium Edition comes with a cheerful, comprehensive Easy Installation Guide, reminiscent of the setup sheets that accompany Micronics' boards (I review the C400, an older Micronics model, here). This double-sided A2-sized Guide is of necessity rather more technical than the what-to-plug-where guides that come with less capable motherboards, and its English is less than impeccable, but it's still not too terrifying for newcomers. It is specific to the Millennium Edition - you don't just get a re-treaded Pro II guide.
The full-blown manual for the motherboard is on the AOpen Bonus Pack CD-ROM. The manual is one of those obviously-intended-to-be-printed jobs with crop marks and page numbers that alternate from side to side, but that doesn't matter.
The Bonus Pack disc also contains Norton AntiVirus and Crash Guard, a Hardware Monitor Utility for various models of AOpen board, motherboard chipset and IDE driver installers for people running older versions of Windows than Win98, a system error detector program, and software for a huge variety of other AOpen motherboards.
Setting up
Installing the Millennium Edition board is no more difficult than setting up any other motherboard. The screw holes are all in the right places and the sockets all fit the plugs, but that's normal; the tiny pin-headers for the case switches and lights are, as always, slightly cryptic, but the Easy Installation Guide covers them. Connecting LEDs backwards just means they won't work; it won't hurt them.
System configuration is, also, no harder than for any other motherboard; in fact, aside from making sure your CPU speed is set correctly, it's pretty much a plug-and-go proposition, since ATA auto-detection is turned on by default so you don't even have to set up your hard drives.
The whole BIOS configuration process is made a bit less worrying by the fact that the Pro II-series boards also let you manually back up and restore all of the BIOS settings to an on-board EEPROM (Electrically Erasable Programmable Read Only Memory), so if you're doing advanced fiddling with RAM modes and have your hard drive settings just the way you like them you don't have to imperil your changes by trying for higher CPU speeds.
Thanks to the EEPROM, AOpen says the Pro II boards have "Battery-less Design". Because of the BIOS backup, you don't strictly need a button cell on the motherboard to maintain the CMOS settings. The CMOS memory is retained whenever the ATX power supply is plugged in and in standby mode, and if you lose power completely you can restore the BIOS config from the EEPROM. You'll still have to reset the system clock manually, though.
Despite this, the boards still come with a lithium button cell. The manual says it just comes in the box with the motherboard, but it was already installed on the one I checked out.
The most interesting feature of the Pro II boards for many users, though, is their overclocking ability.
Winding it up
At a given FSB speed, no Intel BX chipset board is inherently any faster than any other. But some boards can run higher FSB speeds than others.
If you just want to run your CPU at stock speed, any current motherboard will do. The Pro IIs are perfectly happy running processors at the sticker speed, but they can do rather more if you ask them. As well as the standard 66 and 100MHz FSB speeds, you can also pick 68.5, 75, 83.3, 103, 112, 117, 124, 129, 133.3, 138, 143, 148 and a frankly alarming 153MHz speed. You can also crank up the CPU supply voltage, which is how you get borderline processors to run stably at higher speeds - at the cost of rather higher heat output.
The Pro series AOpen boards are favourites of many overclockers because of their imposing feature set, and excellent stability at very high FSB speeds. The Intel BX chipset isn't meant to operate at FSB speeds above 100MHz, but the Pro series boards are perfectly happy at 133MHz or even faster.
The thing that really attracts the overclockers is the AX6BC's software CPU setup. This is a feature, pioneered by Abit, which allows you to set the FSB speed, CPU power supply voltage and multiplier (well, probably not the multiplier - more on this later) from the BIOS setup utility. Just press Delete when prompted during the boot process and you can play with your CPU settings all you like; no jumpers, no DIP switches. If you overdo it and end up with a computer that doesn't want to boot, holding the Home key when you power the machine up should reset the settings to the default. If that doesn't work, there's a jumper on the motherboard that resets the BIOS settings manually.
The manual sternly warns you that higher-than-100MHz Front Side Bus speeds "may cause system damage". Elsewhere, though, it cheerfully dedicates a section to overclocking, and says "when proper setting and qualified components were presented, we feel quite comfortable overclocking to 133.3MHz", following this statement up with a list of various components that were happy at 133MHz FSB in the AOpen lab.
The 153MHz top FSB speed is, in truth, of little interest. This is because all Slot 1 processors for rather a while now have had what's known as a "locked multiplier"; the speed of the processor is a set multiple of the FSB speed. Early Pentium II processors had unlocked multipliers, but all of the current Intel models are locked. The only unlocked processors Intel makes now are the rare sample processors used for testing purposes.
With a locked multiplier and an intended FSB speed of 66 or 100MHz, very high FSB speed inescapably means very high processor core speed. Take a 300MHz Celeron, for instance; it's meant to run at 66MHz (66 and two thirds, actually) with a 4.5X multiplier. Give it a 153MHz FSB speed and that same multiplier means the processor is going to try to run at almost 690MHz, which is definitely not going to work. The 300MHz Celeron cores are very often good for 450MHz - the standard overclocking strategy for some time was to just drop a C-300 into a board that could run it at 100MHz FSB and enjoy your dirt cheap rock solid 450MHz PC - but no P-II class processor, Celeron or "full" P-II, is likely to be useable at more than 550MHz without huge cooling systems of rococo complexity.
A 400MHz P-II, meant to run at 100MHz FSB with a 4X multiplier, will only be clocked to 612MHz if you give it 153MHz FSB. Once again, though, this is well over the probable maximum tolerance of the processor core. Pentium IIs are further handicapped by their Level 2 cache memory; they've got 512 kilobytes of it, versus the mere 128 kilobytes on the Celeron, but it's separate chips running at half CPU speed, and it's not good at running faster. The Celeron's smaller cache is integrated into the main CPU chip and runs at full processor speed. This gives Celerons exactly the same performance as P-IIs at a given clock speed for all domestic applications, including games, and also makes them more overclockable. The first two Celerons, the original 266 and the first 300, had no Level 2 cache memory at all.
Even if you can only easily wring about 10% more speed out of a processor - say, by running a 500MHz CPU at 560MHz with the 112MHz FSB setting - you'll still end up with a substantial saving over getting a "real" processor that runs at the faster speed. As I write this, 500MHz P-IIIs sell for about $AU500, and 550MHz ones cost more than $AU700. Personally, I wouldn't pay the extra even if overclocking were impossible (10% more speed is barely noticeable, but 40% more dollars is a lot more conspicuous...), but when you can easily and stably get the 500MHz processor running a little faster than the standard speed of the 550MHz one, why not? The more-than-$200 saving certainly covers the price of even a show-off motherboard like the Millennium Edition.
(The BIOS version (1.08) on the Millennium Edition board I reviewed permits outrageous CPU voltage cranking - you can set the voltage to anything from 1.3 to 3.5V if you like! Given that winding a 2V CPU up to 2.3V is, generally, about as far as anybody should ever go, this is something of a recipe for disaster. Abit made a similar mistake with their early software-setup boards, and a few people apparently managed to fry their CPUs.)
Because of the multiplier lock problem, there's a peculiar phenomenon at the bottom of the Celeron market. Slower CPUs - C-300s and 333s - cost more than faster ones (see tech-review's CPU Price Chart if you don't believe me). This is because the slower ones will, often, work at 100MHz - 450 and 500MHz core speed, respectively. But a C-366 is fairly unlikely to be able to manage 550MHz - well, not without voltage goosing and hefty extra cooling, anyway - and speeds above that are only very occasionally attainable by any means. Since the core silicon in the different Celerons is very much the same, the "slower" models are actually just as good as the "faster" ones for overclockers' purposes, or better - running at 100MHz FSB is preferable to using the non-standard 83MHz, because the PCI and AGP bus speeds end up normal.
Intel has long since stopped production of the slower Celerons, though, so the only ones you're likely to find at your local computer store are far too fast to endure 100MHz FSB. C-400s are about the slowest commonly available Celerons at the moment, and even those are being phased out rapidly, making C-433s the new low end. Watch this space; Intel wants the last of the PII-based Celerons out of the market as soon as possible, to make way for the new PIII-based one.
Fortunately, overclocking the faster Celerons is far from impossible - you just have to use 75MHz or 83MHz FSB speeds, instead. Many C-400s at 83MHz steam along cheerfully at 500MHz core speed. I'm using one right now. But you're only getting a 25% overclock, versus the 50% you can usually wring out of an old C-300 by winding it to 450MHz. If it's possible to get a C-333 and run it at 500MHz, and you can pay less for it than you'd pay for a C-400, go for it.
Three victims
I tried three processors on the Millennium Edition board - a 500MHz SL3CD Pentium III, a 433MHz PPGA Celeron, and a good old Celeron 300A. The first two of these are representative of the CPUs you can actually find on the shelves of computer stores at this moment. The 300A is one of those old-model low-multiplier chips which you may or may not be able to buy from one or another online vendor. This one I bought when 300As were new and shiny, and it's been steaming along happily at 450MHz at its stock voltage, and with its stock Intel CPU cooler (a "cooler" is a heatsink-and-fan assembly), ever since.
I'd previously noted that I could run the 300A at 504MHz (112MHz FSB), but not for very long; the AOpen board didn't magically fix this. At 2.2V (against its stock 2.0V), 504MHz worked just fine for maybe ten minutes. This means that with a big fat after-market cooler, I could indeed wring another 12% out of this CPU, but whether that's enough of a speed-up to justify the cost of a hefty cooler is another question. Remember, this CPU is already running 50% faster than it's supposed to; whaddaya want from an entry-level processor?!
The P-III ran stably at 112MHz FSB, for a 560MHz speed. No voltage goosing was needed, and the CPU didn't run much hotter than normal. A 12% overclock, for free. 117MHz wasn't stable, no matter what voltage I used, although it did run long enough to start Windows 2000.
(If a CPU runs at a given speed, but only for a while, it's a thermal problem. It's overheating. As a general rule of thumb, a processor that doesn't need its supply voltage pumped up to run at a particular speed will be happy with a stock CPU cooler, provided your case has decent ventilation. Once you start increasing the supply voltage, though, the processor's heat output increases markedly.)
Windows 95 or 98 or NT or 2000, by the way, are all an excellent way to tell if your overclocking efforts have been at all successful. Getting past the POST (Power On Self Test) stage is easy enough for a borderline processor; chewing through the system-flogging startup of one of Microsoft's fine dancing-bear operating systems is quite another.
It looks as if 550-odd megahertz is all this particular P-III's core and/or cache RAM want to do, and asking for 585 is pushing it.
The 433MHz Celeron was more pleasing. To make the PPGA (Socket 370) CPU fit in the Slot 1 AOpen board, I popped it into an ASUS S370 adaptor board. This gizmo is often included in the $AU200-or-so price of the Celeron itself; if it isn't, it's $AU30. The ASUS S370 board has on-board jumpers to adjust the core voltage it requests from the motherboard. This lets you tweak the CPU voltage even if your motherboard has no manual voltage adjustment; it's superfluous on the AX6BC Pros.
Since Socket 370 Celerons all come with a separate chip cooler (boxed Slot 1 processors come with the cooler pre-installed, and often quite hard to remove), I got off my lazy rear and improved it slightly. So can you.
Scrape off the conductive pad that's stuck to the Intel cooler by default, and smear a thin layer of silicone heatsink grease over the shiny metal top of the processor to replace it. A few little dabs of grease, spread over the top of the processor with a screwdriver, will do; the aim of the exercise is to fill in all of the tiny air gaps, not put a huge squishy layer of silicone marmalade between the CPU and its heatsink. Heatsink grease can be had for a few dollars for a 10 gram tube, and 10 grams is lots. Unless you're going into production, don't bother buying more. Also don't go nuts trying to find the very finest grease you can; the difference between the worst and best of the silicone greases is only a few per cent.
The freshly greased and assembled PPGA Celeron package behaved itself perfectly at 2V for both 75MHz and 83MHz FSB settings, for final speeds of 488 and 541MHz. Since no voltage boost was needed, it probably would have been fine with the standard conductive pad still in place, too.
A better-than-25% overclock out of a budget processor is quite respectable, but it's hardly stretching the capabilities of the AX6BC. Any old board capable of 75 and 83MHz, preferably with some voltage-tweaking ability, will do for cheap Celeron overclocking. For two-thirds of the AOpen board's price, for instance, you can pick up an ASUS P2-99, which I review here.
Just on the off-chance that I had the fabled Lost Celeron of G'Quon, I gave it a fling at 100MHz FSB as well. Since this would have meant a 650MHz core speed, it's hardly surprising that it never looked even vaguely like working.
For Celeron overclocking, the best option is still an Abit board, like the dual-processor Socket 370 BP6. Celerons aren't even meant to be capable of dual-processor operation, but Abit made a board that did it, provided of course that you're running an operating system that supports dual CPUs - Windows NT, Windows 2000, or Linux, for instance. With only one CPU the BP6 is still a great board, and it supports no fewer than 30 FSB speed increments between 66 and 133MHz, with no fewer than 23 speeds below 100MHz! This means you're likely to be able to wring about another 10% out of a Celeron on one of the Abit boards, versus a board like the AOpen with no speeds between 83 and 100MHz.
Still, 540MHz out of a cheapo processor meant to do 433 isn't a bad result. 433MHz Celerons are selling for less than $AU200 as I write this, and at this level of overclock will, overall, outperform a 500MHz P-III that costs twice as much, run at its sticker speed. The P-III has fancy multimedia extensions the Celeron lacks and is thus capable of doing things rather faster, but in the real world the difference is less than exciting.
The 75 and 83MHz FSB speeds also have the advantage that any old PC100 SDRAM modules, rated for 100MHz FSB, will work fine. Heck, better PC66 modules will often work at 75MHz, and sometimes 83 as well. PC133 RAM is not a great deal more expensive than PC100, but why pay the extra if you don't need to?
The new "Coppermine" Pentium IIIs (still practically unobtainable, as I write this, despite being announced almost two months ago, on the 25th of October) have their Level 2 cache RAM built into the CPU, like the current Celerons, instead of using separate chips. Older P-IIIs and P-IIs, though, have individual cache RAM chips which aren't touched by the stock Intel CPU coolers. These present a significant overclocking obstacle; without an after-market cooler that cools the cache RAM too, abnormally high FSB speeds are unlikely to work.
I haven't yet had the chance to play with a Coppermine P-III, but the Millennium Edition board supports them. The stock 133MHz FSB speed of half of the Coppermines (most of the ones that run at some multiple of 50MHz are 100MHz FSB chips) is on the high side for a BX chipset board, and you might have to pick your other system components a bit carefully to successfully overclock them, but the upgrade option is there. The new 100MHz FSB P-III based Celerons will be an excellent match for the AX6BC boards.
Above 100MHz the AX6BC Pro II series offers a fine selection of speeds - 103, 112, 117, 124, 129, 133.3, 138, 143, 148 and 153MHz. According to the reports of those who've been able to test the board with unlocked processors (like overclockers.com.au, for instance, whose review of the board is here), the 153MHz top speed is actually stable. This is pretty darn startling; it might not be any use for anybody stuck with a locked-multiplier processor, but it bodes well for the stability of the lower speeds. 133MHz FSB is not out of the question for processor meant to run at 100MHz FSB - the faster Pentium IIs, and all of the Pentium IIIs - and a one-third overclock is a thoroughly worthwhile improvement.
Another reason why 153MHz is less than totally useful is that the AX6BC Pro II, like all other BX chipset boards, has only two possible multipliers for the AGP bus speed. The AGP bus is meant to run, according to spec, at 66MHz. You can run it at "1/1" of the FSB speed - full FSB speed, in other words, the standard setting when the FSB is 66MHz. Or you can run it at 2/3rds of FSB speed, which is the standard setting for 100MHz FSB. For really high FSB speeds, though, it's nice to have a lower AGP multiplier. This feature is supported by some other chipsets, which allow a 1/2 multiplier, but without it you're overclocking your AGP bus by 53% at 153MHz FSB. This is fine, if your video card can take it, but it probably can't.
The Pro II boards may have software CPU speed selection, but they don't have software AGP multiplier setting - that's done with a nice old fashioned jumper on the motherboard. By default, the jumper's in its "Auto" position, where it uses the 1/1 multiplier for 66MHz FSB and 2/3 for 100MHz and 133MHz. Behaviour at other speeds is... undefined. There are two other settings for the jumper, which lock the multiplier to one or the other setting, regardless of the FSB speed; if you're playing with trans-100MHz FSB, make sure to set the multiplier manually to 2/3.
Overall
This is a very nice bit of gear, but it and the less ludicrous green-board AX6BC Pro II are not for everyone.
They don't have the extraordinary flexibility of Abit's recent boards, which means they're not much better for Celeron overclocking than any current board with manual FSB setting. If you want to run a Celeron fast, save your money and get an ASUS P2-99, or something. You'll presently get change out of $AU400 for a P2-99 and C-433 together, and the price drops by the week.
But the AX6BC's ordinary Celeron capabilities are, I think, the only thing not to love about this board. Despite its boy-racer looks, it's not ridiculously expensive. It's very stable indeed at all speeds. It's about as easy to install and set up as any "power" motherboard could be.
I like my Millennium Edition a lot, and you'd better believe it's going to stay in my test machine for a while yet. This board's a class act; if you buy one (or even if you miss out on the black version and have to buy the plain Pro II), you won't be sorry.
Evaluation board kindly provided by Servex Australia.
