ABIT KT7A-RAID Socket-A KT133A ATX

Overclocking: New Chipset, New limits

It’s clear that the KT7A-RAID is designed for hardware enthusiasts, just like other ABIT motherboards, and it would definitely be useful to see how much overclocking this board can handle.  From our coverage on the KT133 motherboard roundup, we have seen the advantage of overclocking by increasing the multiplier ratio setting.  The same case still carries on to the KT133A chipset.

On the other hand, focusing on FSB overclocking is worth a closer look with the new found official support for 133MHz FSB operation.  We have all seen the poor FSB overclocking by the KT133 chipset, where  speeds of higher than 110 - 115MHz are basically unachievable, which means at most a 10 - 15% overclock.  Now with the KT133A chipset, we are sure that FSB speeds of up to 133MHz are no longer a problem.  But how high above 133MHz can we get the FSB speed to?  Is it still limited by the 10 - 15%?

In order to find out about that, we took a 1GHz Athlon processor, a Mushkin PC133 CAS2 PC133 SDRAM, and a NVIDIA GeForce2 GTS as our test bed.  We started the testing by setting the 1GHz CPU to run at 133MHz FSB, thus a multiplier ratio of 7.5.  Then we gradually increase the FSB speed and run tests to make sure the system is stable at that speed.  Given the choice, we chose a FSB speed with a PCI bus speed as close to 33MHz as possible to keep as many components within spec as possible.  If needed, we also tried to set the CAS latency to 3 and all other memory timings to as slow as possible, so that the PC133 SDRAM would not be the limiting factor.  Moreover, when the FSB gets to a certain point, we lowered the multiplier ratio setting so that the clock speed of the CPU remained close to 1GHz.

The first thing we noticed was that when running 133MHz, any clock speed of under 7.5 (for a CPU speed of 1GHz) the system would not even POST.  That means when you are running 133MHz or higher, you have to at least use a multiplier ratio of 7.5.  Our guess is that AMD officially only has 133MHz FSB CPUs starting at 1GHz, so when designing the chipset and motherboards, manufacturers did not keep the possibility of 133MHz FSB with a lower multiplier in mind.  What that means is if you have a chip that doesn’t run anything higher than 800MHz for example, you will be stuck at 100MHz x 8, rather than a faster 133MHz x 6.  That definitely poses a problem for users with lower clocked CPUs. If we’re lucky, we’ll se a BIOS update that fixes this in the future.

For our testing, initially with CAS2 and normal memory settings, the highest FSB speed we achieved was 144MHz.  At that speed we could still run SYSMark 2000, Quake III Arena, and Content Creation Winstone 2000 with no problem.  However, setting the FSB speed to 145MHz the system couldn’t even finish any of the three tests.

We then lowered the CAS setting to 3 and reduced any other memory related category in the BIOS, allowing us to get the FSB up to 147MHz and ran all the tests with no problem.  At 148MHz, the system would not even boot.

However, that does not necessarily mean that 147MHz is the FSB limit of the KT7A-RAID or the KT133A in general.  As we have shown in our PC133 memory roundup, the Mushkin memory we used for testing actually maxes out at around the same 146 - 148 MHz speed.  We will try to gather up some faster memory if possible and continue the testing to find out the true FSB limit.

Moreover, 147MHz would be a mere 10% overclock compared to the default 133MHz FSB speed.  From the fact that people get 10 - 15% when overclocking the KT133, we should be looking at up to 153MHz for the KT133A chipset.  That would mean that memory really has become a limiting factor in our testing.

Assuming in general the VIA 8363A North Bridge is really capable of running FSB speeds of 150MHz or even higher, that immediately poses a potential problem.  In our KT133 coverage, we have already discussed that the VIA 8363 North Bridge got relatively hot during testing.  But it was not hot enough to require a heat sink, so the HSF unit on the KT7-RAID was more or less overkill.  But at that time, the North Bridge was running at a max of 110MHz, whereas now the FSB speed could push 150MHz, which is about 36% faster.  That would generate considerable amounts of heat at the North Bridge, and possibly affect its operation. 

As an experiment, we decided to take away the HSF unit on the VIA 8363A North Bridge to see if we can still achieve the same kind of FSB speed. To our surprise when we tried to set the FSB speed to 147MHz with no HSF on the North Bridge, the system did not even POST.  That proves our thought that the North Bridge can’t sustain the heat generated.  Now the question is how much it is affected?  We repeated our FSB tests and we found out that 137MHz was the maximum FSB speed we could achieve without any HSF unit on the North Bridge.

Several things can be inferred from all this.  For starters, the North Bridge is really dissipating quite a bit of heat.  Thus, a heat sink on the North Bridge seems to be mandatory for the KT133A chipset.  With that in mind, the addition of a fan on the KT7A-RAID’s North Bridge doesn’t seem to be overkill anymore.  In fact, it helps ensure that the heat generated by the North Bridge would not be a limiting factor for FSB overclocking.

So there you go - a simple migration from the KT133 to KT133A chipset poses several new limits to overclocking.  Seemingly FSB speed does not seem to be a factor anymore.  On the other hand, the ability to run your chip above 1GHz becomes a key to enjoying the nice performance boost of using the 133MHz FSB.  Further, the quality of memory can be a critical factor as well.  A normal PC133 memory stick can still be a limiting factor as you continue to increase the FSB speed of the system.  Finally, if the cooling on the VIA 8363A North Bridge is not good enough, the heat generated by the chip can act as the limiting factor as well.