ECS' UEFI gave us the most competitive overclocking experience we’ve ever had with one of the company's products, though a few teething problems are still present in this version. We hope to see a fully-patched version very soon.
The first problem we found was that some of the settings were mislabeled. DIMM voltage controls the CPU core, and we believe that CPU core voltage controls the DIMMs. This mix-up is also reflected in a DIMM voltage reading that doesn’t correspond to anything realistic. Overclocking adjustments were easy after we got to the root of this problem.
Our next problem was again one of naming, but only because it forces you to convert numbers in your head. Rather than write its base clock in MHz with a decimal point, ECS chooses a 10 kHz scale.
Most P67 motherboard firmware has a hold-up time for turbo multipliers, but ECS’ appears to go up to nearly nine hours. Perhaps this is a millisecond scale?
Timing controls are extremely limited, but do work. Setting the main M.I.B. X menu to an XMP profile overrides these, and there are no individual or global controls to revert to “automatic” mode.
I'm just missing benchmarks like SATA/USB speeds etc. Please Tom's get those numbers for us!
1. SLI "support". Do not understand why end-user has to pay for mythical SLI "sertification" (all latest Intel chips support SLI by definition) and a SLI bridge coming with the board (at least 75% of end users would never need one). The bridge should come with NVIDIA cards (same as with AMD ones). Also, in x8/x8 PCIe configuration nearly all NVIDIA cards (exept for low-end ones) will loose at least 12% productivity - with top cards that is about $100 spent for nothing (AMD cards would not see that difference). So, If those cards are coming as SLI-"sertified" they have to be, in the worst case, equipped by NVIDIA NF200 chip (though, I would not recommend to by cards with this PCIe v.1.1 bridge). As even NVIDIA GF110 cards really need less than 1GB/s bandwidth (all other NVIDIA and AMD - less than 0.8GB/s)and secondary cards in SLI/CrossFire use no more than 1/4 of that, a normal PCIe v.2.0 switch (costing less than thrown away with x8/x8 SLI money) will nicely support three "Graphics only" x16 slots, fully-functional x8 slot and will provide bandwidth enough to support one PCIe v.2.0 x4 (or 4 x x1) slot(s)/device(s).
2. Do not understand the author euphoria of mass use of Marvell "SATA 6G" chips. The PCIe x1 chip might not be "SATA 6G" by definision, as it woud newer be able to provide more than 470GB/s (which is far from the standard 600GB/s) - so, I'd recommend to denote tham as 3G+ or 6G-. As it is shown in the upper section, there is enough bandwidth for real 6G solution (PCIe x8 LSISAS 2008 or x4 LSISAS 2004). Yes, will be a bit more expensive, but do not see the reason to have a palliative solutions on $200+ mobos.