In response to the release of the new quad-core CPUs by Intel, AMD launched its Quad FX platform. Contrary to what many people could think, Quad FX isn’t a new AMD quad-core processor, but a platform, as we will describe in this article. Check it out.
Intel released their first quad-core processor, Core 2 Extreme QX6700, in November 2006. At the beginning of this year Intel has already announced three more quad-core CPUs: Core 2 Quad Q6600, Xeon X3220 and Xeon X3210. So far AMD didn’t release any quad-core CPU.
The solution found by AMD to counter-attack Intel’s quad-core CPUs was to release Quad FX platform, formerly known as 4×4. This platform uses two dual-core Athlon 64 FX CPUs and several other “heavy” specs, as we will see in the next pages. Thus a Quad FX PC has two dual-core physical processors working in parallel, for a total of four CPU cores on the system. These two CPUs are interconnected using a dedicated coherent HyperTransport bus. In summary, Quad FX platform is essentially a SMP (Symmetrical Multiprocessing) system with two dual-core Athlon 64 FX CPUs.
It is important to note that from the software point of view dual-core and quad-core technologies are also symmetrical multiprocessing technologies.
On Intel quad-core CPUs the cores are organized in pairs. The cores of each pair can exchange information directly between them – the same way it happens on dual-core CPUs from AMD and on Core 2 Duo CPUs from Intel – but in order to exchange information with any one of the cores located on the other pair they need to access the CPU external bus – what is exactly what happens on Quad FX platform, where the CPUs talk to each other using an external bus, the coherent HyperTransport bus.
In order to clarify the difference between Quad FX platform and the architecture used by Intel quad-core processors, see Figures 1 and 2.
Figure 1: Architecture currently used by Intel quad-core CPUs.
As you can see comparing Figures 1 and 2, Quad FX platform has some advantage on memory access. On Intel quad-core CPUs the CPU external bus (a.k.a. FSB, Front Side Bus) is used for accessing the RAM memory, other devices present on the PC and for the communication between each pair of cores. The communication between each pair of cores can be done up to 8 GB/s.
On Quad FX platform the CPUs use a dedicated communications channel (the coherent HyperTransport bus), which transfers data up to 4 GB/s in each direction. What is important here is that the HyperTransport bus provides two communications channels, one in each direction. Also, since on AMD processors the memory controller is embedded on the CPU, the memory is accessed using a dedicated bus, separated from the channel used by the CPU to access the rest of the PC.
As Quad FX platform uses symmetric multiprocessing architecture, each CPU accesses its own RAM memory. Processors used on Quad FX platform can also access the memory that is controlled by the other CPU, as we will explain in the next page.
A detailed explanation about Intel’s quad-core architecture can be found on our article Intel Quad-Core CPU Overview and Roadmap.
Let’s see the details of this new platform.