Review Price free/subscription
nForce 5 and AM2 Revealed - SLI Memory / EPP
Possibly one of the most talked about features nVidia has introduced, is SLI Memory and EPP. Unlike LinkBoost, which is a mandatory feature for 590 SLI motherboards, SLI Memory is not necessary for a motherboard to be certified SLI ready.
All memory modules have a 128 byte SPD chip which contains information such as the manufacturer name, rated speeds and most importantly the timings/latencies. However, there are other timings that are not included as part of the SPD configuration that if not specified mean that the memory won’t run at its optimal settings.
nVidia and Corsair teamed up to create what is known as EPP – Enhanced Performance Profiles. Only 96 bytes of the SPD chip is actually used by SPD, so Corsair use the rest of this space to store extra timings and any other useful information such as DIMM voltage, to ensure that users will get the best performance out of their memory when they plug it in.
Implementation of this is purely at the BIOS level so it’s not strictly a feature of nForce 5 and could be included into on any platform. It is an open standard that has yet to be approved by JEDEC (the organisation that runs the SPD specifics) and it’s quite possibly that it won’t be. It’s also quite possible that a limited number of motherboard manufacturers and memory manufacturers will use this feature. At the moment only OCZ has decided to follow Corsair’s footsteps.
When we plugged in our 6400C4 modules, just using SPD it booted up with 5-5-5-12 timings, which is incorrect. This is a little strange as these are basic timings that SPD should cope with. However, upon turning on SLI Memory it automatically set it to the 4-4-4-12 that it should have been.
However, SLI Memory is a little more than just extra timings. In the Foxconn board we were supplied with, it had options for “CPU OC 0 %” all the way to 16 per cent. At 0 per cent with our 6400C4 modules it kept the frequency at its 800MHz. When we introduced the 8400C5 modules, the memory was obviously not running at 1,066MHz as the processor does not yet support it. By selecting the “CPU OC 0%” function, it lowered the multiplier of our CPU to 12x and increased the internal clock from 200 to 234MHz. This enabled the memory to operate at 940MHz while not overclocking the CPU at all. This is really clever and allows people who buy high-end Corsair memory to automatically get that little bit of extra performance without having to understand what’s going on.
Selecting any of the non 0 per cent options meant overclocking the CPU, which is also a quick and easy way of getting a mild overclock on your system.