Intel Sandy Bridge platform and Core i7 2600K first impressions

AFTER A LONG WAIT, the first members of Intel's Sandy Bridge processor generation entered the market earlier this month. The initial mainstream Intel Core i5 and Core i7 launch, topped by the 3.4GHz Core i7 2600K CPU, was an incremental performance upgrade over the Nehalem-based Core i5 and Core i7 700 and 800 series. However, the socket change from LGA1156 to LGA1155 precludes mainboard upgradeability, partly to accomodate the faster south bridge connection.

The initial reviews were quite euphoric, in line with the "revolutionary" impact of the new processors lauded in the Intel CEO's speech at CES. Comparing the 2500K non-hyperthreaded 6MB cache CPU and 2600K hyperthreaded processors with 8MB cache, some of these reviews occasionally showed odd results by leaving the CPUs in their default states. Of course, the Sandra 11 CPU benchmark results will shoot up when hyperthreading is enabled, while the Cinebench 10 results will go down in the same situation. So, when comparing these, it's important that the CPU settings are the same, too.

In most cases, the difference compared to the previous generation can't be called exactly revolutionary. The 10 per cent to 30 per cent average benchmark differential between the Core i7 2600K and the previous generation Lynnfield Core i7 875K is good, but does not yet warrant a system replacement from the older chip. Those with the quad-core Core 2 Extreme 9770 from the famed record-setting Core generation of three years ago, which itself wasn't exactly slower than the Nehalem Core i7 4-core Lynnfields, might not be tempted to upgrade yet either. However, for the dual-core Core 2 or Pentium generation or the AMD initial Athlon/Phenom user base, this is a perfect upgrade, bringing at the very least three times the performance in real life applications under pretty much any scenario.

On the other hand, platform level innovation is important here. The south bridge chipset connection improvements, with doubled bandwidth between the CPU and I/O channels interface, now allow for plenty of extra headroom even for SATA3 SSD RAID arrays, with SATA3 itself now embedded in the chipset, as well as USB3 controller bandwidth allowance through the spare PCIe v2 lanes from the south bridge. Therefore, the headaches in fixing the USB3 and SATA3 in the highest gear in the PCIe lane scarce P55 chipset are now a thing of the past.

In this initial look, I put the top bin, the 2600K, through its paces on two boards: the Asus Maximus IV Extreme and the Gigabyte P67-UD5, both based on the P67 chipset. The two boards, both high end, have very different specifications and design goals. The oversize Maximus IV Extreme, a Republic Of Gamers (ROG) product, is one of the most feature-rich P67 boards, with Tri-SLI or Trifire capability - yes, using the infamous Nforce bridge - top notch components and even voltage probes onboard. The UD5, on the other hand, is a standard ATX sized board with the usual dual-graphics card capability, however with a tremendously beefed up power portion and a good set of features to boot.

Here are both mainboards side by side. The red trimmed Asus ROG motherboard is on the left.

The Asus board is the firm's first to have a graphics rich UEFI enabled BIOS, which in the ROG edition looks like a next generation game control panel more than a typical BIOS screen, as you can see here:

And here, the Gigabyte BIOS is very plain and classic looking:

But the Gigabyte mainboard reaches top notch, stable productivity overclocking easily, even in Auto mode for this first run.

Here are the Sandra benchmarks at the defaults as well as the 4.3GHz overclocked settings, which were reached with just 1.3 volts on the Asus Maximus and 1.29 volts on the Gigabyte board:

And the memory bandwidth and latency benchmarks as well:

The other shot:

As you can see, the overclocking latency is better than the default frequency as well.

Then, for better measure some Cinebench 11:

Now, the rest of the configuration was the GEIL EVO TWO high end DDR3-2133 gaming memory running at CL9, as well as an Asus HD6870 DirectCU customised factory overclocked version graphics card with a better PCB and improved cooling. The cooling was provided alternately by Intel's supplied heat pipe cooler, as well as the higher capacity Xigmatek AEGIR heat sink fan, delivering up to three degrees less heat at full load while keeping the system quieter if not silent.

Generally, while the default performance is good, but nothing spectacular, the overclocking experience is fantastic. The degree of overclock by default, without going to the extreme voltages, is even higher than on the default Nehalem or Westmere processors, with 30 per cent or above easily achievable in stable operation without any major voltage adjustments or heat increase, only 4C higher in peak use with the standard Intel fan, and nearly the same when using the higher performance Xigmatek AEGIR 4-heatpipe direct touch heat sink fan.

In summary, before we carry out further tests this coming month, for a mainstream user looking for high end performance at a competitive cost, Intel's latest generation Sandy Bridge platform delivers. While it is slower than Intel's own X58 high end six-core Westmere processors, it relegates those systems firmly to the very high end. It also pushes the current AMD Phenom processors further down the price slide, a situation that AMD will only be able to reverse once its quad-core Llano fusion processors, as well as Bulldozer high end 8-core entries get into the market mid-year.

The two boards we used here, the Asus Maximus 4 Extreme and the Gigabyte P67 UD5, are top notch, and either can be recommended to the high end user wanting to overclock the maximum out of Intel's latest chips while keeping the operation stable and power and heat down. The Asus board is more flashy, from the packaging to the BIOS user interface, however the Gigabyte board delivers just as good overclocking performance in a more classic package.

Howewer, Intel's latest Sandy Bridge platform does have some limitations. The number of PCIe lanes is not sufficient for any dual or more parallel graphics. Asus and Gigabyte tried to put the old fashioned Nforce 200 bridge into their top end P67 mainboards to address this, however this platform, in my mind, should be only for single card graphics - anything more, and you should stick with the X58 chipset, which we will be looking at again here soon, or the coming X68.

Talking about the X68 and its associated 8-core socket 2011 processors, that will likely be the ultimate platform for the year. It's a pity we have to wait at least half a year more for it. But for the true high end afficionados, the wait might just be worth it, though. µ