Dual processor boards are nothing new

WITH THE announcement of AMD's enthusiast-oriented 4x4 platform, featuring two Athlon 64s on a consumer board, it brought back some fond memories of the first foray into dual processor motherboards not aimed at workstations or servers.

Way, way back in June 1999, Abit released a real landmark motherboard - something completely different to anything available at the time, and a very clever piece of engineering - the BP6 motherboard.

The low cost BP6 took two off the shelf Socket 370 Celeron processors and ran them in Dual processor mode - something reserved at the time for only high end motherboards and slot-based Pentium IIs.

Combined with the overclockability and high performance of the low speed Celerons of the time, this was a dream for enthusiasts, as you could buy two dirt cheap processors and a motherboard for less than the cost of a Pentium II, run them at a higher clock speed, and get the advantages of the dual core processing - six years before the launch of the current generation of dual core chips!

AMD's 4x4 platform will mark the first dual processor motherboards targeted at consumers since the BP6, so it has some big boots to fill.

The history of the BP6
The first dual Celeron systems were not actually socket based; a Japanese chap by the name of Tomohiro Kawada had spent a few bored afternoons in late 1998 looking at Intel datasheets, and worked out that the only difference between the Celerons and the Pentium II cores at the time was down to one pin on the package: the one that told the motherboard to use 100MHz bus instead of 66.

However since both Celeron and PII at the time used a slot package, Intel had decided to keep the chip the same, and not connect it on the PCB that fitted into the slot. By a process of drilling packages and soldering on extra thin wires, Kawada san managed to bully a dual processor workstation board into accepting the chips.

When the Celeron moved to a PGA Socket 370 platform, he found that it was even easier using and modifying the readily available "Slotkets" to put them into the board.

Five months later and Abit had pushed out the BP6 board, based on the high-performing and reliable 440BX chipset, combining all the modifications needed to get the PGA Celerons needed, and disposing of the now unnecessary slots in favour of two sockets.

All the users needed to do was to find two matching Celerons and drop them in… and overclock them. No drilling or soldering!

The Celeron 300/366A
At this time, the Pentium 2 had reached 450MHz, and the Pentium 3 Katmai had just about stretched to 500. These featured 512KB of external cache, running at half the clock speed, compared to 128KB at full clock speed on die for the Celerons. But they also ran at the 100MHz bus instead of the anaemic 66MHz used by default by the cheaper chips.

Both Celeron and Pentium had locked multipliers, introduced to stop dodgy resellers re-labelling the parts. This meant that, in order to get a Celeron running on the same bus speed as the Pentiums, where they would often perform even better at a given clock speed (courtesy of the faster cache), you would have to overclock them by a whopping 50%!

You could, of course, run them at lower bus speeds and still get a healthy overclock, but the BX chipset ran the PCI bus at a multiple of the processor FSB, and would overclock this, leaving you with some nasty side effects, such as unreliable hard disk controllers.

But, of course, the Celerons were sold at artificially low clock speeds, so this was not a problem with the low speed parts: first the 300A which would happily run at 450MHz, and then the 366A, which could get up to a huge 550MHz!

These chips were so predictably good at overclocking that companies such as OCUK were selling them with guarantees of hitting the higher speed.

The Experience
I remember the exciting moment where I took my two black organic looking processors with their metal heatsinks out of their box, specified at 333MHz, opened up the sockets and put them into the motherboard, before installing the over-indulgent 256MB of RAM.

On paper, this should have been a big step down from my 450MHz PII system - a 366MHz cheap Celeron? A risky move for sure!

But no, first time it booted up, I moved into the BIOS and hit the bus speed up to 100MHz…. and prayed. Up it came running at 550MHz, default voltage - no problems…. Windows NT 4 installed perfectly, and the computer absolutely flew!

In almost every single benchmark, the Dual 550MHz Celeron would thrash the top of the range 500MHz PIII, and even then some games were becoming SMP aware (Quake III, for example).

But that wasn't the main thing - the change from one to two processors is the most incredibly feeling when you experience it for the first time. No longer does one process have the ability to lock up your life. No longer do you need to wait 20 seconds for task manager to appear, so you can kill that dead Outlook window!

Way above and beyond any minor benchmark improvement, the added day to day usability of the system was astonishing, even if the scream of two golden orb heatsinks and six case fans trying to keep the chips cool would keep the entire house awake at night.

Moving back to single core.
To say it had been a revelation was an understatement, and my consequent move back to a single processor system, even a 1.5GHz Athlon XP 1800 was a very painful one in many ways, and was done for reasons of noise and stability rather than anything else.

Yes - a few years of running at over 50% clock speed (606MHz in fact) had slightly ruined the ability of the system to stay crash free, even on a fresh install of Windows, and even when throttled back to 366MHz.

Whether this was electromigration on the chips, or more likely degradation of the capacitors on the CPU core voltage rail, I will never know, but for those few years, it was the only computer I have ever owned that would make people say "wow".

I know that I was not the only one to experience this dual CPU goodness - two of my friends built similar BP6 systems, my boss at the time ran on an identical dual 550MHz BP6 doing advanced CAD work without any failures right up until 2005!

4x4 and the future
So is there any way that 4x4 can achieve the same level of excitement? Well in a word, no, I suspect. One problem is that moving from four cores to two does not deliver anywhere near the same "feeling" that one to two does, as you've already removed the problem of the locked system. The other is the price and poor overclockability of the Athlon FX parts.

That's not to say there won't be a market for it. It may well turn out to be the best way to get the fastest system around.

My hope is that they will remember the enthusiast crowd, and allow people to pick up a real bargain by dropping two 3800 X2s into a 4x4 board and experience the benefits of massive parallelisation, but for not much money. Preferably with some headroom for good overclocking!

Let's be honest: having double the normal number of anything in a computer is inherently cool - just ask the SLI crowd! µ

L'INQS
Kawada san's dual Celeron website