Arjan meldt: dat Johan De Gelas van Ace's Hardware een artikel heeft geschreven waarin hij onderzoekt wat de grootste bottleneck is bij een modern systeem. De conclusie is dat het geheugen te langzaam is, iets wat ook erg logisch is als je bedenkt dat de multipliers elke dag hoger worden:
Performance is limited by certain bottlenecks, and that is what this article is all about. We will try to find out what exactly are the bottlenecks in a top notch system. Do we need more memory? Faster memory? Faster videocards? A faster Front Side bus (FSB)? Bigger faster L2-caches? Which one of the mentioned solutions will have the most impact?
[...] The speed difference between the CPU and the main memory has become so big, that performance is hampered by the speed of the memory in a lot of desktop applications. Even a 512 KB L2-cache and 128 KB L1-cache can not prevent that. The latency of the memory seen by the CPU is very high with such high multipliers and big caches can not hide all the memory latencies.
An Athlon with SRAMs that run 2 times slower than the CPU has no real advantage over a CPU with the L2-cache at one third of the CPU's clockspeed. Until the Thunderbird arrives, we can understand AMD's decision to use a 1/3 divider for the Athlon 900-1000. Faster SRAMS (>350 MHz) would be extremely costly and would not offer significant higher performance. Intel has anticipated this and has never shipped with CPU's with SRAMs faster than 300 MHz (PIII-600).