Page 3

Dual 16x PCI-Express lanes and other great features

Another big feature for SLI gamers is the dual 8GB/s bandwidth on two graphics ports. I mean two times x16 PCIe bandwidth here between the SPP, MCP and the GPU's undoubtedly fantastic. Next to that we already mentioned it... there's another PCI-express 8x slot to be found in-between them... which could be used for graphics cards.

Another great feature is the fact that you'll get two 1GBit Ethernet ports, which you can actually bind together as one 2 GBit/s connection. Agreed a little excessive and only 0.00001% of you will use it, but it might be "neat" function to have on a LAN party as a server. Some other good features obviously is the support for 10 USB ports and the HDA multi-channel audio solution from Realtek.

With that being said, let's move onwards to the terminology you just read.

SLI Memory

You already have been able to learn about DDR2 EPP (SLI memory) recently as it was covered in our AMD NFORCE 590 SLI review. It's very clever technology for the overclocker who does not dare or know how to configure memory at it's best in an overclocked environment.

DDR2 memory (actually pretty much any memory) uses an SPD (Serial Presence Detect) chip to store its JEDEC certified timings and setup specifications on. Now a massive chunk of that SPD chip is being written as ZERO; meaning it's largely unused space. A couple of brainy guys at NVIDIA figured... hmm, let's talk to some memory allies and see if we can do something with that extra space and utilize it for our new products to offer our buyers more options.

Since most memory is configured at JEDEC specifications the standard timings are not exactly enthusiast presets, most of us know, we can do a heck lot more with memory timings and frequency wise then the rather safe JEDEC SPD settings tell the BIOS to fire off on your memory.

Once you overclock the system bus, your memory clock will run faster in MHz also. A higher FSB usually means that we'll need slower timings and/or different voltages, that kind of information could be stored into the SPD in several profiles.

So basically additional non-JEDEC specified timings related towards a certain clock frequency or related to multipliers/dividers/voltages/drive strengths will be stored on that unused SPD space. This way that memory can be setup in a different way with the click of a button, or even better detected by your nForce 680i mainboard and set automatically.

So in short: you can tweak memory really easy yet this is intended for the folks that have a hard time overclocking themselves. It's a reasonably fool proof method of getting the best out of your DDR2 memory with an NFORCE series 590 or 680 mainboard. Obviously you can manage memory timings and everything related to it manually also. This is just an extra feature.

Look out in the stores for memory with a "SLI Ready" logo and I honestly think this is a great feature !

NVIDIA LinkBoost

Introduced with the last gen nForce 590 boards and now also in the 680i series is LinkBoost. I'm actually moderately excited about this feature as right now it doesn't do much in terms of measurable performance, with the right components you'll receive automated tweaks straight out of the box when it comes to inter component bandwidth. Example: imagine that you, for whatever reason, believe that PCIe x16 bandwidth is not enough for that so GeForce 8800 GTX combo and you'd need like PCIex20 bandwidth... well hey good news.

So basically you instantly will get 25% more performance on the internal clocks (interconnect bandwidth and PCIe bus). Meaning your PCIe x16 bandwidth just jumped from 8 GB/s towards 10 GB/s.

The chipset will detect your graphics card and once validated with a certified GPU this feature will be automatically enabled (can be disabled as well). And even with that standard 25% overclock on all interconnects NVIDIA ensures that there still will be a little more headroom left to overclock.