Infrastructure support

Unlike NVIDIA’s last SLI launch, where ASUS got a quick start out of the gate with their A8N SLI line and it took other motherboard manufacturers a month or two to catch up at retail, this time around NVIDIA has multiple launch partners that will have nForce4 SLI Intel Edition SLI-ready motherboards available. Motherboard launch partners include ABIT, ASUS, Biostar, DFI, ECS (Elitegroup Computer Systems), EPoX, Foxconn, Gigabyte, and MSI.

According to NVIDIA, the first nForce4 SLI Intel Edition motherboards should hit retail sometime around the end of this month, with retail prices hovering in the $200 range, just as you saw with AMD-based SLI motherboards when they first launched. NVIDIA expects that as additional partners release their boards, prices will slowly drop to the $160-$190 price range most AMD-based SLI motherboards sell for today.

Memory

Launching alongside the new chipset and motherboards is NVIDIA’s new validation program for DDR2 memory manufacturers who have been tested and validated on the platform for speeds of up to 667MHz. (Specifically, compliant modules must surpass commodity JEDEC memory requirements and comply with minimal targeted performance levels, including 667MHz clock speeds with memory timings of 4-4-4-12-2T.)

Since a JEDEC spec for DDR2-667 doesn’t exist yet, the validation program is the best way for enthusiasts who plan on building nForce4 SLI Intel Edition systems to ensure that their memory modules will work properly with their motherboard at high clock speeds. Modules that have been approved will carry a special “Recommended for NVIDIA nForce4 SLI Intel Edition” badge. We used these Corsair 667MHz modules for our testing. These modules should hit retail later this month alongside the nForce4 motherboards.

Power

In terms of power required, we were told that NVIDIA’s general recommendation for the nForce4 SLI Intel Edition, as well as the AMD-based SLI platform, was a 550-watt power supply. In particular, the new nForce4 SLI Intel SPP and MCP consume about twice as much power as the CK8-04 chipset used in AMD platforms, 30-watts in the Intel chipset versus 15 in CK8-04. You’ll also need to take into account the higher power consumption of Intel processors. For instance, the 3.73GHz Pentium 4 Extreme Edition consumes about 119 watts. In comparison, the Athlon 64 FX-55 dissipates 105W.

When you factor this in with the additional power requirements of the Intel chipset, you’ll need an additional 29 watts for the Intel system in the example above.

Setup

If you’re familiar with the procedure for setting up SLI on NVIDIA’s AMD platform, the Intel SLI chipset will be pretty familiar. Simply orient the SLI selector card into the motherboard for dual GPU configuration, plug in two SLI-ready NVIDIA graphics cards (GPUs supported include the GeForce 6600 GT, 6800, 6800 GT, and 6800 Ultra), and connect them with the SLI connector that shipped with your motherboard.

Once the hardware is in place, boot up your system and install the latest ForceWare drivers from NVIDIA’s website. After a quick reboot, you can then check the “Enable SLI MultiGPU” box in the driver control panel, followed by another reboot, and you’re up and running with SLI!

If you recall, NVIDIA’s SLI implementation can run in one of two modes: alternate frame rendering, or split frame rendering.

Split frame rendering mode works just as it sounds – the card splits the workload horizontally across the screen. One card takes the upper portion while the second card takes the lower segment. The frame buffer data is then combined and sent to the monitor. It’s important to note that SFR doesn’t necessarily split the screen directly down the middle; in some scenes the lower portion of the screen may be more complex than the upper portion, or vice versa. NVIDIA has developed custom load-balancing algorithms that are designed to take this into account, and split the screen appropriately – if one GPU takes longer to render, no problem, the driver just gives that GPU less work to do.

In alternate frame rendering mode, each graphics card handles the entire frame in a scene, with the cards splitting up the workload on a frame-by-frame basis. In other words, graphics card one will handle frame number one, while graphics card two tackles frame two. The only downside to AFR is the perceived “lag” that may be felt in some fast-paced games, although we haven’t seen this, nor have we seen any reports of current SLI enthusiasts who have run into this problem.