It is good to see Aopen catering for the overclocker, but having to use a jumper in order to choose a FSB range is very annoying, although the AX3S in my opinion has the better range to choose from. Voltage selection on both boards is excellent with a maximum of 2.05volts on the AX34 and 3.5 on the AX3S!! In the right hands being able to push your chip to 2.1volts+ on a super cooled machine is useful, but to a normal person this could be DEADLY. The AX34 has a major drawback which may be able to be fixed via a BIOS update, but at the moment - when you change FSB ranges the voltage resets to default. Now imagine this scenaria. You have a Celeron 566 and in order to POST at 100mhz (850) it needs 1.6volts, and because it resets it just won't do it. You may have a lovely 566@850 using 1.7volts on one board but you won't be able to run it on this unless you can POST at the default voltage in order to set it. To an overclocker, this is a major drawback. The AX3S on the other hand has no problems at all with this. I was very lucky that my 666 celeron could infact POST at 100mhz (1000) at defualt voltage so I could run my benchmarks.

Both these boards got my Celeron 666 running at 1000mhz using 1.8volts. This is good for benchmarking as we have a 66mhz and 100mhz chip. The benchmarks where taken using Windows ME, an ATI Radeon 32mb DDR, 128mb RAM - all with the latest drivers. Now the one thing Via chipsets have always been slated for is memory performance. With exception of the newer AMD boards which have decent performance, there is no exception here. I uses Sisoft Sandra 2001 to get the following benchamarks. The ram speed was indentical to the FSB (so 66mhz for 666 and 100mhz for 1000mhz) and using CAS2.

This clearly says to me that the Via Apollo ProA chipset is optimised for 100mhz+ ram speed as it is pretty much on par with the AX3S. At 66mhz the AX34 is lagging behind the AX3S considerably. Lets see how this affects real life performance. For the following tests I used Quake 3 in fastest setting, and the default benchmark on Madonion's 3dMark.

These results show exactly what is expected, at the 66mhz mark the memory performance was lacking and this let the AX3S outperform. But at 100mhz they where pretty much on par. Now, at this point you might be thinking if I'm using my Celeron and I'm not going to overclock, the Via board is not for me. But this board has a function that the AX3S does not - the ability to run your RAM at higher/lower speeds that the FSB. For instance you could run FSB at 66mhz and your RAM at 100mhz. Or your FSB at 100mhz and run your ram at 66mhz or 133mhz. Because of this, I reran the benchmarks for the AX34, this time at the faster ram speeds.

Well, now we see the AX34 beating the AX3S on both occasions. Now seeing as RAM is so cheap now, and getting hold of PC100 or PC66 is actually more expensive than PC133 most of the time, a lot of people will be running at +33mhz at least on their ram. In my opinion that is the performance boost we are after. As far as raw performance goes, the benchmarks where almost identical so I didn't see any point including them - the chipset made no difference.

<< Back | 1 | | 2 | | 3 | Next >>