Xbox in the test: Microsoft dares to revolt against Sony
First a central processing unit had to be found, and which processor manufacturer is the best fit for the monopoly Microsoft? Clearly, the dominant giant from Santa Clara, Intel. MS didn't even need the absolutely fastest processor available from Intel. Rather, it had to offer sufficient performance at an affordable price. On the one hand, the price was lowered by opting for an FC-BGA Pentium III soldered directly to the circuit board; on the other hand, the circuit board itself also comes from Intel, which has lowered the processor price not only due to possibly higher efficiency in production should - monopolists among themselves.
What As far as the clock is concerned, there was initially talk of a PIII with 600 MHz, later it was increased to 733MHz - probably less because of the processor clock than the system clock. In the Xbox, as in any other console, the CPU clock is by no means as decisive as it is in a PC, since in such a closed system, in particular, its share in the 3D calculation can be reduced to a minimum through optimized programming. As a PC user, you shouldn't be fooled by the comparatively low MHz number. The PS2, for example, works at only 294MHz. Even that is enough for a lot more than what a PC with a 300 or 400MHz processor can do, because the CPU's tasks are mainly limited to data management and the calculation of artificial intelligence. This is why the level 2 cache could be halved to 128 Kb compared to the normal PIII. The level 1 cache was left at 2 x 16Kb. The PIII definitely offers enough computing power, the 600 would probably have done it too, but then another weak point would have arisen. Because, as is well known, the maximum bandwidth between the CPU and the memory controller in the northbridge integrated in the graphics chip depends directly on the clock rate of the front side bus. With 100MHz FSB it would have been 800MB per second. The 133MHz * 2 * 32bit/8bit/byte that is now available results in 1064MByte, i.e. about 1GB per second. That is at least 20% more bandwidth with which the processor can access the main memory. Nevertheless, at around 26 bytes per polygon, that would be enough for 'only' 40 million polygons per second (the maximum polygon rate of the PS2 is 77 million/second).
When playing games that actually reach the limit of their capabilitiesthe Xbox, but most polygon-specific data shouldn't flow between the CPU and memory anyway. Rather, in complex scenes of such games, for example, so-called 'static meshes' should be used. Certain recurring elements (networks of vertices) can then be calculated and stored in the memory and then accessed directly from there, namely from the real heart of the Xbox, namely the graphics chip; because it also has direct access to the RAM and has to share it with the other components in the Unified Memory Architecture.
On the next page: Memory architecture