Conclusions
If we summarize the results we get to the following conclusions:
- 440BX at 133 MHz FSB is not a fully valid solution because BX and the AGP are running beyond specifications. At the same time 'BX133' is scoring best in consumer and office applications. Therefore I recommend conservative people to stay away from this combination, but whoever is not afraid of overclocking will find that BX at 133 is not only one of the cheapest, but certainly the fastest solution for Intel's Coppermine-processors.
- VIA's Apollo Pro 133A is currently the most sensible chipset for the average Coppermine-user. Systems with this chipset come at a very attractive price point, since motherboards and PC133 SDRAM are cheap. People who are using professional OpenGL-software should stay away from it though.
- Intel's 820 chipset is in my eyes simply obsolete. Why would anyone buy a system with it? The performance is rather bad and the price is high. By the time when Intel releases i815 or 'Solano', i820 will simply disappear. Until then I would advise everybody to stay away from it, especially in combination with PC700 RDRAM.
- The Intel 840 chipset has its place in the workstation area, where it performs very well. I am not that impressed with it anymore though, since I had to see how badly it got beaten by the old 440BX-chipset in a large number of benchmarks.
I know that the inclusion of an overclocked 440BX-platform will upset quite a few of you, but it will delight a lot of others at the same time. Even though this old chipset was running beyond spec, its exceptionally good results should make us wonder what is going on in the platform market right now. Can we really accept that the latest Intel-platforms with the super-expensive RDRAM are not even able to beat the 'tuned' version of its two-year-old predecessor? Shall we sit idly, nod and go buy those platforms without even getting any real benefit? For most computer users the new chipsets i820 and i840 present the biggest fraud in the IT-industry that I have personally ever encountered. RDRAM seems nothing but a very bad tasting joke, unless you really use workstation type OpenGL-software. Maybe we should wonder why AMD never jumped on the RDRAM-bandwagon. Maybe AMD was wiser to stay away from it and count on the upcoming DDR-SDRAM memory solution. Could it be that Intel is only trying to get its return of investment out of Rambus? I really wonder what is going on here, but I know for a fact that I will stick to the BX-platform in my personal systems or that I will switch to an AMD Athlon-platform. I am simply disappointed and can only shake my head.
VIA was wise to take advantage of Intel's i820/i840 RDRAM disaster. The Apollo Pro 133A is looking very good against i820, and it even beats it as soon as an i820-system is equipped with the common PC700 RDRAM. However, there's huge room for improvement. Intel's old 440BX-chipset at 133 MHz FSB is scoring so much better than the Apollo Pro 133A that you wonder if VIA will ever be able to design a fast performing chipset. It also shows how much room for improvement is left for Athlon chipsets. If the KX133 had an SDRAM-performance close to BX, Athlon could run a whole lot faster.
Intel will release i815 or 'Solano' sometime in June 2000. If Solano isn't suffering an artificial slow-down, it should perform even better than the overclocked BX-chipset. In this case all the RDRAM-chipsets would look really bad and nobody will find a reason to buy i820-systems anymore. By that time VIA will also have trouble selling Apollo Pro 133A chips.
