What is Athlon 64?
First of all, it must be emphasized that in the AMD64 architecture, the main attention was paid to the issue of compatibility with 32-bit applications that are widespread now. The new processors are fully compatible with 32-bit (generally speaking, even with 16-bit) applications. Compatibility is achieved due to the fact that the AMD64 architecture is an extension of the x86 architecture. The registers existing in the x86 architecture are expanded to 64 bits, while the lower 32 bits of each register can be used by 32 bit applications in the same way as in the x86 architecture:
Thanks to this solution, the Athlon 64 and Opteron processors to a certain extent now occupy the same position that the i386 processor once occupied. On a machine with an Athlon 64 processor, you can run all the applications that the user is currently working with. As 64 bit applications appear on the same machine, it will be possible to run both old and new applications. Thanks to this, the transition to 64-bit computing promises to be painless.
The structure of the processor core has undergone quite serious changes. Some changes concern only 64 bit applications. This, in particular, is the possibility of direct addressing in excess of 4 GB, which is why, in fact, everything is started.
Certain measures have been taken to improve performance on 64-bit applications. So, in the AMD 64 architecture, you can use twice as many registers, which, of course, can improve the speed of application execution. Naturally, this feature cannot be used in 32-bit applications, since these applications “do not know” about the presence of additional registers.
Many other processor nodes have also been improved. So, in particular, x86 instruction decoding blocks convert the input sequence into a noticeably smaller number of microcommands.
But about one architectural feature that helps improve performance in both 64 and 32 bit mode, you need to stop. This is a “memory controller integrated in the processor.”
In fact, this phrase is not entirely correct. Recall how a computer works. A specific chipset is sealed on the motherboard – a set of chips that ensure the interaction of all elements of the computer. Any chipset is traditionally based on two microcircuits, one of which is the “north bridge”, the other is the “south bridge”. The processor, memory, video system (AGP) and the south bridge are connected to the north bridge. Southbridge provides interfaces to all standard peripherals.
As a rule, a computer is judged primarily by the processor installed in it. Therefore, it is believed that the north bridge provides the processor with other elements of the computer. However, in reality, the north bridge provides primarily the exchange of information between all elements (including the processor) with memory. That memory is the bottleneck of the computer. Indeed, almost all elements of a computer work only through memory. Therefore, it is more correct to consider that the north bridge is an element that provides memory connection with all elements of the computer (including the processor). Let’s look at the block diagram of the Opteron processor.
The processor microcircuit includes Crossbar – a switch that provides memory interaction with the processor and peripherals (via HyperTransport buses) It is Crossbar, not the processor, that interacts with the memory controller. Thus, it will be more correct to assume that 64-bit AMD processors are actually the north bridge with a processor built into it. As a matter of fact, this is exactly what the developers intended. Take a look, for example, in Figure 4, taken from the official AMD OpteronT Processor. Data Sheet document.