NVIDIA's nForce 780a
NVIDIA has been one of the biggest players for AMD back when Socket A was in its infancy. NVIDIA started its nForce series chipsets off by making platform solutions for AMD, and it's good to see NVIDIA making more headway here.
nForce 780a is based around a two chip design, but in all honesty it could have been done in a single setup. However, for some reason beyond our comprehension, NVIDIA decided to do things the hard way, and quite honestly, the wrong way. The two components to the 780a is a single chip 780a GMCP and an nForce 200 PCI Express 2.0 controller hub. The 780a GMCP is designed with Hyper Transport 3.0 interface to the CPU, allowing Phenom customers to get the full benefits associated with HT3.
One of the most interesting things about this is NVIDIA has put a GPU on the 780a, that's right, it has an integrated GPU for if you're not going to use a dedicated graphics card. But really, if you're going to get this chipset, chances are you're going an SLI setup, leaving the onboard GPU useless. If you are only going to use one graphics card, Hybrid SLI is also possible with this setup with certain graphics cards. While Hybrid SLI has its place in the more value positioned PCs, it really has no place in the enthusiast level PC environment.
Hybrid SLI allows the integrated GPU on the chipset to work in conjunction with a discrete graphics card to boost the performance; however this does come at a price. Since the integrated GPU requires a frame buffer, it will steal up to 256MB of your system memory for this task, a big no-no among hardcore gamers. If you're going for a budget system, this will be fine, but for a high-end setup you are likely to see the integrated GPU going to waste.
In order to gain PCI Express 2.0 compliancy, the nForce 200 PCI Express 2.0 root hub is resident, the same chip that gives 780i its PCI Express 2.0 ports. Depending on the board, you can have two PCI Express x16 slots or one PCI Express x16 and two PCI Express x8 slots all running 2.0 specs. The hub however connects to the 780a MCP using a single PCI Express x16 lane; this means that if the nForce 200 chip has to send any commands from CPU to GPUs, it's limited to a 4GB/s bottleneck. However, NVIDIA did manage to set the chip up so that any GPU to GPU communications do not need to go to the CPU, so this is where some bandwidth savings go on.
The peripheral setup on the 780a is a better setup than that of the AMD 790FX. First off, it has six SATA ports for mass storage devices, a single Gigabit Ethernet port and HDA audio; in all, a well rounded setup. However, this will soon be lost to SB700 from AMD.