Review It's finally arrived: the first retail Nvidia nForce4 SLi Intel Edition mobo. Considering this is Nvidia's first attempt at creating a chipset for Intel CPUs it's no mean feat to make it an SLi solution at the same time. Of course, Intel doesn't have a consumer-level SLi-capable chipset at the moment, though there are some rumours about it launching its own SLi-compatible solution shortly. But if this first board is anything to go by, Intel will be facing some stiff competition, writes Lars-Goran Nilsson.
One of the most interesting features of the nForce 4 SLi Intel Edition is its native 667MHz DDR2 SDRAM support. This is not yet a ratified standard, but Corsair is already offering memory which will run at this speed and faster. The new Twin2X CM2X512A-5400UL modules we were sent to test the GA-8N-SLi Royal with worked flawlessly at 667MHz on a 1066MHz bus with a 3.73GHz Pentium 4 Extreme Edition CPU.
Nvidia might be one step ahead of Intel when it comes to memory speeds, but I doubt it will take to long before 667MHz DDR2 becomes the norm. Nvidia seems to have anticipated this and the chipset's dual-channel memory controller has an independent bus for each memory module. This should make for better performance than competing chipsets which put each pair of installed modules on the same bus.
Add to this Nvidia's QuickSync and DASP 3.0 technology and you've got most of the features from the nForce 2 chipset integrated into the new nForce 4 SLi core logic. QuickSync provides performance-boosting asynchronous memory, whereas DASP 3.0 is a high-performance memory pre-fetch core-logic. This should give the CPU quicker access to data it needs by getting what the CPU will need next, ahead of time.
The nForce 4 also supports dual-channel asymmetric DIMM population, which gives you dual-channel memory performance even if you use three modules. Right now, this is a unique feature to the nForce 4 SLi Intel Edition chipset.
To my knowledge this is also the first Intel CPU solution to use HyperTransport as the chipset interconnect. However, this is a slower 800MHz 8-bit implementation rather than the 1GHz 16-bit version found on current AMD platforms. It should provide plenty of bandwidth for PCI Express devices, which is important when you start sharing the PCI Express bus between devices like Gigabit Ethernet adapters and HDTV-tuner cards, both of which require plenty of bandwidth.