ASUS M3N-HT Deluxe/Mempipe will be one of the first motherboards to be launched on the market based on the forthcoming nForce 780a chipset – which, by the way, is delaying a lot to be released. But don’t think that nForce 780a is simple the AMD version of the nForce 780i – it isn’t. One of the main differences between the two is the support for HybridSLI, which features HybridPower, a feature that disables the video cards installed in the system when you are not playing games and generates 2D images – i.e., when you are using Windows, a word processor or a spreadsheet – through the chipset to save a lot of power, and GeForce Boost, which uses the on-board video to increase 3D performance. The problem is that both features require compatible video cards and so far only two video cards support each feature. Other features from this motherboard include three PCI Express 2.0 x16 slots supporting 3-way SLI and Quad SLI, a passive cooling solution for the memories, an eSATA port, coaxial and optical SPDIF outputs and more.
click to enlarge
Figure 1: ASUS M3N-HT Deluxe/Mempipe motherboard.
This motherboard is a socket AM2+ board. There are two main differences between socket AM2 and socket AM2+: the support for HyperTransport 3.0 and Dual Dynamic Power Management (DDPM) technology, present on AMD CPUs based on K10 core (e.g., Phenom CPUs). You can still install Phenom on socket AM2 motherboards but you won’t have these two technologies available and the CPU will talk to the chipset only up to 2,000 MB/s (HyperTransport 2.0) instead of up to 10,400 MB/s (HyperTransport 3.0) – it is important to notice that current Phenom processors work under the lowest HyperTransport 3.0 transfer rate, 7,200 MB/s. For more information about those technologies we recommend you to read two other articles, Inside AMD K10 Architecture and The HyperTransport Bus Used by AMD Processors.
NForce 780a is, of course, compatible with HyperTransport 3.0, allowing the communications between the chipset and the CPU to be performed at a higher speed – if an HT3-based CPU is installed on the system, of course; otherwise, i.e., if you install a plain socket AM2 Athlon 64, the maximum transfer will be 2,000 MB/s. HyperTransport 3.0 maximum transfer rate is of 10,400 MB/s (this rate is also called 5,200 MT/s) but Phenom CPUs released so far have their HyperTransport bus working at 7,200 MB/s (3,600 MT/s) It is important to remember that since on AMD CPUs the memory controller is embedded on the CPU and not on the chipset this transfer rate is available for I/O transfers only.
The second biggest feature of this new chipset is its support for the PCI Express 2.0 bus, which doubles the bandwidth available for video cards from 2.5 GB/s to 5 GB/s, if they are also based on PCI Express 2.0 (you can see which video cards are based on PCI Express 2.0 on our NVIDIA Chips Comparison Table and on our AMD ATI Chips Comparison Table). This motherboard has three x16 slots and according the motherboard manual the blue slots work at x16 when one or two video cards are installed, but they will go down to x8 if you use three video cards under 3-way SLI.
This motherboard also has one PCI Express x1 slot and two standard PCI slots.
click to enlarge
Figure 2: Motherboard slots.
As you can see in Figure 2 this motherboard has two standard PCI slots and one PCI Express x1 slot.
Keep in mind that if you use an AMD CPU that isn’t based on HT3 you will limit the graphics performance of your system. Let’s do some math to explain this better. In our calculations we are assuming you have two video cards installed.
The maximum bandwidth between nForce 780a chipset and the video cards is of 10 GB/s (5 GB/s, which is the x16 transfer rate, times two – if you use three video cards, the required bandwidth will be less, 7.5 GB/s, as each slot will work at x8, not x16), if PCI Express 2.0 video cards are used. If PCI Express 1.0 cards are used, the maximum bandwidth available is of 5 GB/s (2.5 GB/s, which is the x16 transfer rate, times two).
In theory HyperTransport 3.0 offers a maximum transfer rate of 10 GB/s, what would match perfectly the required bandwidth if you use two PCI Express 2.0 video cards. However Phenom processors released so far use a lower 7 GB/s transfer rate, generating a bottleneck. This bottleneck does not exist if you use PCI Express 1.0 video cards, as the maximum required bandwidth will be of 5 GB/s.
But if you use an Athlon 64 CPU, for example, you will limit the graphics performance. Since this CPU can talk to the chipset only up to 2 GB/s you won’t be able to achieve the maximum performance video cards are capable of delivering. In fact, even on regular socket AM2 motherboards with PCI Express 1.0 this occurs, as the maximum I/O transfer rate of the CPU is of 2 GB/s and the maximum graphics transfer rate if of 2.5 GB/s if one video card is used or 5 GB/s if two cards are installed.
So HyperTransport 3.0 unlocks the maximum graphics performance of your system, even though for a perfect system we still need to wait for the release of Phenom CPUs based on the 10 GB/s HyperTransport bus.
However the numbers provided above are theoretical and in fact no performance impact may be seen, as video cards transfer data at a rate far below their maximum theoretical transfer rates.