There's more to the speed the CPU accesses the memory than the number of lines it has - I don't know very much about the GVP board but it could be that the memory chips GVP use simply run slower than EDO SIMMs. I'm going on hazy memory myself here, but ISTR the Apollo cards are faster because they're synchronous, whereas the Blizzard cards are asynchronous, with extra buffer logic used to interface between buses.