The enthusiast market is composed of the most discerning customers. They demand unparalleled performance. They want systems capable of cutting through old, single-threaded legacy applications and the latest parallelized programs. Most important, power users are intimately familiar with hardware. They do their homework and they know what it means to build balanced PCs. What enthusiasts often lack is the time to track down the lowest prices for each component, build everything up, and then troubleshoot, should a problem crop up. That's great news for system builders. High-end parts give you lots of room to add value, whether you get there through comforting warranty protection, a guaranteed overclock, or relevant software bundles.
Up until just recently, however, the upper echelon of boutique hardware seemed to be stagnating a bit. The fastest platform you could buy was X58 Express—a more than three-year-old chipset that launched with Intel's very first Nehalem-based Core i7 900-series processors. Although it remains remarkably modern, the CPUs it accommodates drop into an LGA 1366 interface and are consequently limited to the quad-core Bloomfield and hexa-core Gulftown designs.
An Advanced Interface. The latest Sandy Bridge-E-based CPUs require motherboards with LGA 2011 interfaces. The new socket is a necessary component of enabling the architecture’s feature set. Those first-generation Core processors are still very fast, particularly in well-threaded workloads like video rendering and transcoding. However, Intel really outdid itself early in 2011 with its second-generation Core architecture, also known as Sandy Bridge. The application of a mature 32 nm manufacturing process to a greatly evolved layout paved the way for an even faster portfolio of new products armed with as many as four cores. Intel took the extra step of launching two special enthusiast-oriented K-series SKUs (the Core i7-2600K and Core i5-2500K) that featured unlocked multipliers, allowing overclockers to easily ratchet clock rates up above 4 GHz without much difficulty.
With prices on those new parts significantly lower than the LGA 1366-based flagships, Intel's premium CPUs faced competition from Sandy Bridge. Even still, the enthusiasts able to really push X58-based machines to their limit held off in anticipation of a true workstation-class successor. That platform is upon us now, and it's called X79 Express. Built to host the new LGA 2011 interface and the second-gen Core i7-3000-series chips that drop into it, today's flagship combines the very best elements of Intel's mainstream Sandy Bridge architecture and performance-oriented server philosophies.
Laying The Foundation
A Smart Value Play. Although it’s not the family’s flagship, Core i7-3930K’s six cores, 12 MB of shared L3 cache, and sub-$600 price make it an attractive offering for performance-seekers. The introduction of Sandy Bridge took a lot of folks by surprise. While the architecture borrowed heavily from Intel's first-gen Core design, a shift to 32 nm lithography allowed Intel's engineers to pursue integration much more aggressively. All of the sudden, you saw execution cores, lots of cache, a memory controller, graphics processing, and PCI Express all built onto one piece of silicon.
The LGA 2011-based Core i7-3000-series processors (code-named Sandy Bridge-E) are similar in many ways, though they trade away some of Sandy Bridge's mainstream components for more performance-oriented specifications. Nevertheless, the new CPUs still leverage a common core design infused with great single- and multi-threaded performance potential.
Despite Sandy Bridge's apparent complexity, the very highest-end products derived from it consist of 995 million transistors, populate a dainty 216 square millimeter die, and dissipate up to 95 W of heat. In other words, Sandy Bridge is both efficient and cost-conscious, even as it proffers unprecedented speed.
The key to Intel's approach is balance. All of the products that center on Sandy Bridge come with either two or four cores. They're fed by 32 KB of L1 instruction and 32 KB of L1 data cache, plus a 256 KB L2 cache dedicated to each core. The next step up the memory hierarchy is an L3 cache shared between all of the execution resources (plus the graphics engine). In its most complete implementation, that cache is up to 8 MB in size. From there, two 64-bit channels of DDR3 memory pump information into the chip. Officially-supported data rates as high as 1,333 MT/s theoretically push more than 21 GB/s of data, which, at some point during the design phase, Intel's architects must have deemed ample.
Beyond its cores, cache, and memory control, Sandy Bridge also incorporates 16 lanes of on-die PCI Express connectivity. That's another carefully-chosen number. Sixteen lanes built into the processor are enough for one discrete graphics card to operate at full speed: 8 GB/s of bidirectional bandwidth. Really, that's all most folks need. However, the controller is flexible enough to divide its 16 lanes into a pair of x8 links, enabling dual-card configurations capable of serious gaming performance. Although each link's bandwidth is halved to 4 GB/s, we've run enough benchmarks to know the loss of throughput is barely, if at all, quantifiable.
Intel’s Scalable Architecture. Core i7-3000-series CPUs center on a die that sports more than two billion transistors. The same piece of silicon will soon enable Intel’s Xeon E5 processors.
Click to Enlarge.Moreover, the P67, H67, and Z68 Express platform controller hubs that complement most of the consumer-oriented Sandy Bridge-driven PCs include an additional eight lanes of second-gen PCI Express, used to add extra USB 3.0, storage, and Ethernet controllers. Between the second-gen Core processors and their accompanying chipsets, nearly all of your customers get access to more expansion than they'll ever need.
The processor's built-in PCI Express lanes don't necessarily have to be used for graphics, of course. A great number of Sandy Bridge's transistors are used to enable the HD Graphics 3000 and 2000 components. In environments where big 3D processing muscle (and the power consumption it imparts) isn't necessary, Intel's own solution is capable enough to handle mainstream gaming. In addition, Quick Sync support makes it possible to accelerate a number of video-oriented tasks in hardware. Blu-ray DVD playback is one example; converting a movie to work on your smartphone is another. That leaves the CPU's PCI Express connectivity available for storage or Ethernet controllers.
Evolving An Architecture
Built For Enthusiasts. A top-down view of Intel’s LGA 2011 platform puts its big data pipelines front and center, from four memory channels to lots of PCIe and plenty of integrated features.
Click to Enlarge.Some of the decisions Intel made in bringing second-gen Core processors based on Sandy Bridge to market don't necessarily map over to the enthusiast and workstation segments, though. Integrated graphics is a good example. Extremely useful in a mainstream context (especially with the benefit of Quick Sync behind it), the HD Graphics engine eats up a lot of die space that could otherwise be used to further beef up the CPU's computing muscle. Thus, Intel's Sandy Bridge-E-based Core i7-3000-series chips shed that circuitry altogether.
In its place, the company enables as many as six processing cores, all of which are comparable to the cores found in existing Sandy Bridge-based processors. That means the same 32 KB chunks of L1 data and instruction cache per core, along with a dedicated 256 KB L2 cache repository for each core. All of the older four- and six-core Core i7-900-series processors feature similarly-sized caches, so those specs aren't particularly newsworthy. However, Sandy Bridge introduces a number of very significant enhancements that affect the number of instructions successfully executed each clock cycle.
At any given frequency, then, Intel's more modern architecture offers better performance than the company's previous-generation enthusiast parts. Per core, the improvements put Sandy Bridge-E at the same level as the company's mainstream Sandy Bridge CPUs. But because the Core i7-3000-series chips boast as many as six cores, high-end desktop and workstation-class applications optimized to exploit those resources get 50% more execution hardware pushing them along. Video editors, engineers, architects, and capable multitaskers running several programs at a time all stand to reclaim serious productivity from a processor with the headroom to do more and not get bogged down. Those same customers are also more likely to run software optimized for the latest instruction sets like Advanced Vector Extension (AVX), which Sandy Bridge-E supports, but Gulftown did not. Floating-point-intensive multimedia, scientific, and financial calculations most readily benefit from AVX, making its inclusion particularly exciting.
Sandy Bridge-E's L3 cache configuration is also clearly intended for heavy lifting. Whereas Sandy Bridge included up to 8 MB, Sandy Bridge-E wields a maximum of 15 MB. Even last-generation's Gulftown-based Core i7-900-series SKUs topped out at 12 MB. Extra cache gives the processor more room to store frequently-used data, and any time it can pull information from cache rather than go to system memory or magnetic storage, performance increases. Though the first- and second-level repositories are dedicated on a per-core basis, L3 cache is shared. Thus, if only one or two cores are active, they're able to dynamically access the full capacity.
The second-gen Core i7-3000-series also boasts a more complex memory controller. Instead of Sandy Bridge's two-channel solution, Sandy Bridge-E accommodates four 64-bit pathways. And while the mainstream Core processors officially run at up to 1,333 MT/s, Intel pushes its enthusiast-oriented processors up to 1,600 MT/s. The result is theoretical memory bandwidth in excess of 51 GB/s. Workstations based on the X58 Express chipset can't even compete. Their triple-channel controllers are rated for DDR3-1066, capping bandwidth at a still-impressive 25.5 GB/s.
Lots Of Connectivity. X79 Express includes two SATA 6 Gb/s ports, four 3 Gb/s connectors, up to 14 USB 2.0 ports, an integrated gigabit Ethernet MAC, HD Audio support, and eight PCIe 2.0 lanes.Paving the way for more than two times the potential throughput wasn't an arbitrary decision. Rather, more cores able to operate on information faster than before necessitate a wider data pipe to keep execution resources fed. There's also the expectation that Sandy Bridge-E will encounter more demanding workloads than Intel's mainstream chips. A quad-channel design helps prevent bottlenecks in today's software and in the apps that'll spring up to tax this platform in the years to come.
Following the same theme of preempting possible performance limitations, the Core i7-3000-series CPUs incorporate a ton of PCI Express connectivity. Instead of 16 lanes, which easily satisfy most desktop customers, Sandy Bridge-E is armed with no fewer than 40 lanes. Intel also makes it a point to specify compliance with the third-generation standard, doubling peak throughput to 1 GB/s per lane and per direction. That makes these new processors the first to enable PCI Express 3.0, ahead, even, of any compatible peripherals. They won't be far behind, though. We've already talked to graphics, networking, and storage vendors planning to launch add-in adapters shortly after torch-bearer Intel enables the standard.
What makes third-gen PCI Express necessary? Again, as a result of running lots of comparative tests, we know that an eight-lane PCI Express 2.0 link allows today's fastest graphics cards to just about achieve their potential. It's worth noting, though, that we're on the cusp of seeing a new generation of GPUs capable of better performance and more likely to utilize additional PCI Express bandwidth.
Perhaps more important, though, is the fact that discrete graphics cards aren't just being used for gaming any more. General-purpose compute workloads are more commonly being offloaded to highly parallelized hardware. Today, gaming-oriented 3D accelerators greatly enhance facial recognition, security, and video processing workloads. However, Intel recently showed off its Knights Corner co-processor, featuring more than 50 cores manufactured using the company's new 22 nm tri-gate transistors. Said to be capable of more than 1 TFLOPs of double-precision floating-point math, Intel expects Knights Corner to realize quick adoption, since it works with today's x86 programming model. Already, the hardware has been shown to benefit weather modeling, tomography, protein folding, and advanced materials simulation. Those data-intensive workloads depend on a very fast connection between Knights Corner and the CPU feeding it with information. As a result, Sandy Bridge-E's connectivity-laden PCI Express controller stands to really bolster professional applications more sensitive to I/O than consumer-oriented software. And when you consider that Intel's X58 platform stood for three years as the desktop flagship, delivering up to 36 lanes of second-gen PCI Express, you can be sure that the Core i7-3000 series/X79 combination is built to persist across successive iterations of mainstream- and professional-class peripherals.
Between its armada of processing cores, the massive collection of cache, a potent four-channel memory controller, and incredibly robust PCI Express support, Sandy Bridge-E unquestionably becomes the most complex desktop processor Intel has ever introduced. All told, it's composed of 2.27 billion transistors that occupy 434 square millimeters of die space. Both numbers more than double the biggest Sandy Bridge-based chips.
It's more appropriate to size Sandy Bridge-E up against its predecessor, Gulftown, which gave the same group of discerning high-end customers access to six cores on Intel's same 32 nm lithography node. Made up of 1.1 billion transistors on a 248 square millimeter die, Gulftown-based CPUs like the Core i7-990X are significantly smaller in comparison. Remember, though, that processors designed to drop into the LGA 1366 interface didn't include on-die PCI Express. They were also armed with triple-channel (rather than quad-channel) memory controllers. And they topped out at 12 MB of shared L3 cache.
Giving The Models Names
Liquid Cooling Rocks. Combining the best features of water cooling without the maintenance requirements, Intel’s RTS2011LC is a great match to overclocked Core i7-based platforms.Intel is launching its Sandy Bridge-E family as a trio of distinct SKUs. They all center on the same physical die and the three processors support a common set of features. Certain knobs and dials are tweaked, however, to create differentiation in an otherwise beastly family of high-end CPUs.
Up top, the Core i7-3960X represents Intel's most exclusive model. It's armed with six cores and the full 15 MB of shared L3 cache. The -3960X benefits from Hyper-Threading, which allows compatible operating systems like Windows 7 to schedule two software threads to each physical core.
Most obviously, the feature manifests itself as 12 logical processors in Microsoft's Task Manager. But Hyper-Threading's true benefit is felt when applications aren't able to fully utilize execution resources. Scheduling a second thread to a partially-loaded core lets software extract more performance from the processor. When they're not needed, the Hyper-Threaded logical cores can be "parked" by Windows 7 to help save power.
The Core i7-3960X also enjoys Intel's second-generation Turbo Boost technology. As you no doubt already know, Turbo Boost serves to increase clock rate in lightly-threaded applications, improving performance when the thermal headroom allows for it. There's actually a power control unit built into the die able to measure a number of different variables. Using that data, Turbo Boost permits the processor to exceed its power ceiling until it reaches a defined thermal limit, at which point it backs off the accelerated frequency. Simply, Intel's second generation of Turbo Boost plays the same role as the version popularized by its first-generation Core processors. But it enables those higher clocks for longer durations, resulting in better overall responsiveness. In our tests, we've seen Turbo Boost improve performance up to 15% in single-threaded applications like iTunes and up to 10% in more thoroughly parallelized titles like Photoshop.
A base clock speed of 3.3 GHz for the Core i7-3960X is largely academic. So long as you use ample cooling, the processor is actually capable of driving all six of its cores at 3.6 GHz with Turbo Boost turned on. The same frequency holds with five cores loaded down. Run an application that only needs three or four cores, though, and Turbo Boost cranks the chip up as high as 3.8 GHz. With only one or two cores active, it speeds up to 3.9 GHz.
Of course, as with all of Intel's flagship Extreme Edition processors, the Core i7-3960X also boasts an unlocked multiplier ratio. That's a critical feature in the enthusiast world, as it lets overclockers experiment with even higher clock rates without affecting memory speeds, the PCI Express bus, or SATA. Again, the utility of this addition is going to depend wholly on the cooling solution you bundle. However, the thought of this processor cranking away at 4.4 or 4.5 GHz with all of its cores active is an exciting proposition for power users (even when exceeding Intel's factory settings voids their warranty).
A 130 W thermal design power is common across all three Core i7-3000-series chips, as is compatibility with Intel's LGA 2011 interface. As its name would seem to suggest, 2,011 pins inside the socket make contact with pads on the bottom of the CPU. That's a significantly higher number than LGA 1366, the interface that previously covered Intel's highest-end desktop chips. Credit the more complex memory controller and on-die PCI Express logic for the shift.
At nearly one thousand dollars, Core i7-3960X is one of those trophy parts that everyone wants, but only a fortunate few will own. The next model down in Intel's stack, Core i7-3930K, is much more accessible. It's very similar to the flagship in many ways. Armed with six cores, the -3930K is equally adept at cutting through the heavily-threaded applications most power users find themselves running. And very similar clock rates translate to very little compromise in performance. A base frequency of 3.2 GHz is just 100 MHz lower than Intel's Core i7-3960X. Of course, Turbo Boost technology means more performance when the thermal headroom is available. With five or six of the processor's cores loaded down, a 300 MHz speed-up to 3.5 GHz kicks in. Four active cores yield 3.6 GHz, while three cores run at up to 3.7 GHz. With just one or two cores busy, expect to see 3.8 GHz clocks.
For as much performance as Sandy Bridge-E facilitates at those frequencies, overclockers will want to push the architecture further still. That's why Intel equips the Core i7-3930K with an unlocked multiplier ratio, too. A mature 32 nm manufacturing process makes it relatively easy to achieve more than 4 GHz with the right cooling setup.
The Key To Overclocking. All three of Intel’s Core i7-3000 CPUs center on the same complex die, making effective cooling absolutely imperative. The RTS2011LC steps up to this challenge.The only other notable differentiating feature separating Core i7-3930K and the more expensive flagship is its 12 MB shared L3 cache. Interestingly, the piece of silicon Intel uses to create its Sandy Bridge-E-based processors actually plays host to eight cores and 20 MB of shared L3 cache. Granular control over each piece of the design lets Intel turn components of the die on and off to hit certain price, power, and performance targets, though. Disabling 3 MB of shared memory is a great way to set two of the company's highest-end CPUs apart without impacting performance in a significant way. And as a result of its much lower price tag, the Core i7-3930K quickly becomes a favorite for savvy power users looking to maximize bang for their buck.
Intel's third model, Core i7-3820, was announced alongside the other two processors. However, it won't be available until the first quarter of 2012. There's no doubt that its arrival will be anticipated, though, as a low estimated price is expected to dip in under the fastest mainstream Sandy Bridge-based Core i7 CPUs. Like those 95 W quad-core parts, the Core i7-3820 also features four cores, a result of Intel fusing off more of the Sandy Bridge-E die. It also enjoys less shared L3 cache. Though, at 10 MB, each core has access to 2.5 MB—same as the flagship Core i7-3960X, and plenty to ensure great performance.
Because it features fewer cores, the Core i7-3820 is able to run at higher frequencies without exceeding its 130 W thermal ceiling. A 3.6 GHz base clock rate is enhanced by Turbo Boost technology, which bumps up to 3.7 GHz when three or four cores are active and 3.8 GHz with one or two working.
Intel does not equip its entry-level LGA 2011-based processor with an unlocked multiplier ratio, making that feature a differentiator unique to the two higher-end parts. However, the Core i7-3820 is still partially unlocked, meaning enthusiasts seeking additional value can push the multiplier up to six 100 MHz "bins" beyond the top Turbo Boost setting of 3.8 GHz. That's a maximum of 4.4 GHz.
Overclocking Sandy Bridge-E
Six Cores To Rule Them All. Core i7-3960X runs at 3.3 GHz by default and up to 3.9 GHz with Turbo Boost. But its unlocked multiplier ratio allows enthusiasts to push the chip even faster.Although Core i7-3820 is technically limited by its available ratio settings, Intel builds in a bit of additional flexibility that effectively uncorks the platform's potential.
You see, in its effort to lower the cost and simplify motherboard layout of Sandy Bridge-based systems, Intel integrated the clock generator responsible for setting the DMI and PCI Express buses. The result was very limited overclocking scalability, aside from the multiplier-unlocked K-series processors.
Sandy Bridge-E circumvents that by employing an external clock and buffer chip. Fixed PCI Express and DMI buses prevent interference with the BCLK; they run at 8 GT/s and 5 GT/s, respectively. But the buffer chip facilitates a number of specific ratios able to increase the BCLK without affecting those other sensitive interfaces.
The short story is this: in addition to the 100 MHz reference clock that's normally multiplied by the ratio, you can also pick 125, 166, and 255 MHz frequencies without disturbing the platform, so long as the CPU itself can take whatever ratio you also throw at it. Using the Core i7-3820 as an example, jumping from 100 MHz to 125 MHz and using the same 36x default multiplier (which would have yielded 3.6 GHz previously) results in a 4.5 GHz base frequency. So, again, although that entry-level Core i7 3000-series part isn't fully unlocked, resellers who know Sandy Bridge-E's ins and outs will have little trouble giving customers exactly what they want, be that a completely stock or heavily modified machine.
A High-End Motherboard. Intel’s motherboard team continues to excel. The DX79SI is its most advanced creation yet, with eight memory slots and support for three-way graphics configs. Naturally, the same strap-based overclocking technique works just as well on the six-core parts (perhaps better, in fact, since they both sport lower, more easily attainable ratio settings). Because they're both rocking unlocked multipliers, though, simply tweaking the ratio up and down is the most convenient way to achieve elevated performance.
But there's another trick to overclocking unlocked chips for even more effect. Rather than setting all six of the -3960 or -3930's cores to the same frequency, which is common practice, use Intel's Turbo Boost technology to dial in more precision. Let's say that a certain Core i7-3960X is stable at 4.4 GHz during a stress test. Though it's a good baseline, you already know that the processor won't use as much power or get as hot when only one or two of its cores are under the same load. The configurable Turbo Boost settings make it possible to try for 4.5 GHz when five cores are active, 4.6 GHz when three or four are loaded, and 4.7 GHz with one or two cores busy. It does take a bit of hands-on tweaking to figure out which frequencies are attainable. Fortunately, enthusiasts are particularly appreciative of subtle adjustments that fire the afterburners on high-end hardware.
X79: Latest And Greatest
An Emphasis On I/O. The DX79SI’s back panel exposes a number of important connectors, including six USB 2.0 ports, two USB 3.0 ports, two gigabit jacks, and audio connectors.Intel's previous-generation flagship chipset, X58 Express, was revolutionary in the amount of I/O it enabled. The I/O Hub component delivered as many as 36 lanes of PCI Express 2.0 connectivity, while the ICH10 added six additional lanes. It also included support for 12 USB 2.0 ports, HD Audio, a gigabit Ethernet MAC, and six 3 Gb/s SATA ports.
In the three years since its introduction, a lot has changed, though. To begin, we've seen Intel shift from three-chip platforms to two-chip designs. The process of moving memory and PCI Express control from its chipsets to its processors essentially rendered traditional northbridges obsolete. Removing those components frees up room on enthusiast-oriented motherboards for third-party storage, USB, and Ethernet controllers. As an added bonus, pulling the I/O hub eliminates a piece of hardware rated for nearly 25 W thermal design power. The resulting two-chip configurations, such as P67 and Z68 Express, facilitate greater efficiency. And because they're more modern, Intel is able to arm then with updated feature sets, too.
X79 Express is the successor to X58, and a quick peek at its block diagram makes it quite clear that the crusade toward integration marches forward in a big way. This time around, memory control and most of the PCI Express connectivity originate at the processor level. Then, there's a 20 Gb/s Direct Media Interface between the CPU and X79 Platform Controller Hub. That bi-directional link is fast enough to support whatever devices you connect. The same gigabit-class Ethernet MAC and HD Audio controllers are both exposed. USB 2.0 support expands to 14 ports, though, and two of the six SATA connections now accommodate 6 Gb/s data rates. That's particularly notable in light of the latest solid-state drives that push sequential read speeds beyond 500 MB/s. Intel's X79 also features eight additional lanes of PCI Express 2.0, which motherboard manufacturers can use to add FireWire, USB 3.0, more SATA, or a second networking interface to the chipset's already-robust feature set.
Even though X79 is positioned as an enthusiast- and workstation-oriented piece of core logic, Intel supports the chipset with its Rapid Storage Technology enterprise 3.0 software. As with previous versions of RST, this one enables a number of software-based RAID levels able to help optimize for either performance or data protection. RAID 0, RAID 5, and RAID 10 all help improve speed versus individual drives operating on their own, while RAID 1, RAID 5, and RAID 10 are the more safety-oriented choices. Notice the overlap. When you put at least three drives together in RAID 5 or at least four in RAID 0, customers realize more speed and true redundancy. RST enterprise 3.0 takes protection one step further by pre-configuring certain Windows-based options to minimize the chance information could be lost in the event of a power outage.
Making X79 Motherboards
More Than Good Looks. Sure, its black and blue scheme has power user appeal. But the DX79SI’s strengths are unquestionably its beefy power circuitry and copious expansion potential.Using the X79 Express Platform Controller Hub, Intel designed two motherboards of its own. Both target the enthusiast and workstation markets for which LGA 2011 is intended. However, they hit different price points that make it easier to build PCs as diverse as the three Core i7-3000-series models themselves.
The DX79SI serves as Intel's flagship product, equipped to support any LGA 2011-based CPU, but most naturally matched up to the Core i7-3960X or Core i7-3930K. Eight memory slots correspond to the Sandy Bridge-E's quad-channel controller, accommodating up to 64 GB of RAM. Despite the fact that Intel officially rates its processors for DDR3-1600 data rates, its DX79SI is robust enough to push memory at DDR3-2400 and higher. The impact of a speed-up like that is profound. At 1600 MT/s, four 64-bit channels top out at just over 51 GB/s. Operating at DDR3-2400, the same memory configuration can potentially move in excess of 76 GB/s.
Of course, achieving ambitious memory overclocks necessitates high-quality modules, giving resellers a great new opportunity to add value through go-fast gear. Understand, though, that more memory bandwidth will help improve performance only in specific applications. Demonstrate your technical skillset by helping your customer determine if faster DDR3 memory will improve his experience first.
The X79 Express chipset naturally assumes a prominent position on the DX79SI. Intel exposes its complete feature set through a number of internal, back-panel, and header-based connectors. For instance, twin gigabit Ethernet ports are enabled by the integrated MAC and an additional controller. All 14 of the supported USB 2.0 ports are accessible via six rear-panel connectors and enough on-board headers to expose eight more. Ten-channel HD Audio aims to match the performance of an add-in sound card by delivering exceptional clarity and diverse connectivity without incurring much processing overhead. Eight of the sound system's channels correspond to four 1.8" rear-panel analog mini-jacks, while the other two map over to the headphone port found on the front of many cases. There's also an optical S/PDIF output on the back panel for HDMI support. Full access to X79's complete storage package translates to two 6 Gb/s SATA ports and a quartet of 3 Gb/s connectors, which are cumulatively managed by Intel's Rapid Storage driver and support RAID 0, 1, 5, and 10 arrays.
Intel augments its own chipset by adding two USB 3.0 controllers and FireWire connectivity. The PCI Express-based USB 3.0 chips hook in to a couple of X79's eight available lanes and the FireWire hardware exposes two ports capable of 400 Mb/s transfers. Most competing motherboards leave you to integrate your own wireless networking hardware, adding cost. However, Intel takes the extra step of bundling a module that contributes Wi-Fi and Bluetooth connectivity. The module plugs into an on-board header, leaving the platform's expansion slots vacant for other upgrades.
Given the Core i7-3000-series' copious 40 lanes of third-gen PCI Express, expansion is a really big part of this story, too. The DX79SI features three x16 slots, two of which run at full 16-lane data rates, while one is wired for eight-lane operation. In an enthusiast-oriented build, those slots are most often going to be populated by graphics cards. As such, it's a good thing that Intel licenses CrossFire and SLI support for two- and three-way multi-GPU arrangements. The company even bundles SLI bridge connectors with the DX79SI.
All three 16-lane connectors have enough space between them for graphics cards that consume two slots—a result of high-end cooling needs. In between, Intel exposes two PCI Express x1 slots and one PCI slot. So, if your enthusiast customers don't fully load the DX79SI down with GPUs, they still have room for other add-ins.
Beyond the list of specifications most likely to attract enthusiasts, Intel retrofits its most feature-complete desktop motherboard with a handful of less prominent (but equally important) features to help prepare it for life in a power user's PC. Built-up power circuitry, for instance, protects whichever processor you install from too much voltage, even as it delivers the stable current needed to hit the highest possible frequencies. On-board power and reset switches facilitate easy operation on an open-air test bench. And Intel's own Overclocking Assistant feature makes it easier for less technical users to dial in a profile with a high probability of success, without much risk to the hardware. A software package consisting of Intel's Windows-based Extreme Tuning Utility and ESET's Smart Security 5 rounds out the extras, along with UV-reactive SATA cables and a free mouse pad.
At first glance, the DX79TO looks incredibly similar to the DX79SI. And truth be told, the less expensive model's list of capabilities follows in Intel's flagship's footsteps. That's great news for value seekers lusting after potent hardware, yet still tied to a budget.
Eight DIMM slots continue offering support for up to 64 GB of memory running at overclocked speeds of DDR3-2400 and beyond, for the most talented tweakers. Any of Intel's three Core i7-3000-series chipsets is capable of enabling this board with up to 40 lanes of third-generation PCI Express connectivity. However, not all of them are used here. The DX79TO comes equipped with two x16 slots and a x4 slot, chewing up as many as 36 lanes. Even still, the majority of enthusiasts willing to drop multiple graphics cards into their machines don't need more than a pair of connectors. Intel's DX79TO has those folks covered. Use the x4 slot for an add-in RAID controller or PCI Express-based SSD.
Top Of The Stack. Although the DX79SI is Intel’s flagship X79 Express motherboard, there are actually two platforms in the company’s line-up, targeting a broader range of price points. Of course, in the context of a cost-conscious performance build, perhaps it's more realistic to find a blend of solid-state and magnetic storage attached to this motherboard. Two 6 Gb/s SATA ports and four 3 Gb/s connectors deftly handle a mixed-drive subsystem with room to expand. And the bundled Intel Rapid Storage Technology software supports the same handful of RAID modes as the more expensive platform. Install performance-sensitive applications to the SSD and save user data to a couple of hard drives slung together in RAID 1. Given a relatively modest investment, that's the best mix of performance and reliability available on the desktop.
The DX79TO's other connectivity features represent a carefully calculated step back from Intel's flagship, without giving up any critical I/O. You still get access to 14 USB 2.0 ports, but a single add-on USB 3.0 controller enables two high-speed ports (instead of the DX79SI's four). The one gigabit Ethernet controller found on DX79TO is all most enthusiasts will ever use. However, the DX79SI comes with two, accommodating more complex network setups. Although most of us don't even have 7.1-channel speaker systems attached to our big-screen TVs, Intel's flagship enables eight channels out of its rear panel and another two up front. The company's DX79TO more practically serves up six-channel HD Audio with an S/PDIF optical output.
A number of value-adds found on the DX79SI persist. You still get the easily accessible on-board power and reset switches, diagnostic LEDs, the back-to-BIOS switch for fast recoveries from overly aggressive settings, and a sophisticated voltage regulator. SLI bridge connectors, Intel's Extreme Tuning Utility, and ESET's Smart Security 5 anti-virus app make the bundle accompanying Intel's DX79TO motherboard even more comprehensive.
Performance Questions
Portable Performance. Falcon Northwest is building its iconic FragBoxes using Core i7-3000 processors and X79-based platforms, cramming tons of speed into a form factor you can carry around.Complete platform upgrades are the most dramatic.Though they often require new processors, motherboards, and memory (along with storage and graphics when the upgrade is truly significant), the resulting performance jump can be incredibly satisfying. Knowing full well that Intel's Core i7-3000-series CPUs are high-end gear, it's important to know where six-core processor are most effective and when a less expensive quad-core chip might work just as well. The easiest way to pinpoint the right applications is to identify customers who'd be interested in premium gear and look at what they run. The best source we could think of for insight on enthusiasts was a system builder who deals with them on a daily basis. Kelt Reeves, president of Falcon Northwest, shared with us that some of his most demanding customers are gamers, video editors, artists who spend a lot of time in apps like Photoshop, and professionals involved in modeling and rendering. Perhaps not surprisingly, the categories aren't mutually exclusive, either. Power users by day often turn into gamers at night.
A few years ago, customers would have had to choose between more cores running at lower frequencies, or less-complex processors sporting higher clocks. Although the idea that heat and power have to be balanced still applies today, technologies like Turbo Boost make the compromise automatic, allowing one piece of silicon to address single-threaded and more thoroughly parallelized programs. Now, the decision to go with a four- or six-core chip is driven more by pricing.
When, then, does it make sense to pay a premium for a hexa-core processor? The choice is perhaps most easy for workstation buyers running software on their machines that make money. From Adobe's Creative Suite to Autodesk's many offerings, the most demanding professional apps are often tuned to exploit all available hardware resources. Speeding up those workloads improves productivity. And it's only a matter of time before a higher-end CPU pays for itself.
Gamers can use the extra horsepower as well. An increasing number of titles are using processor cores to handle physics, artificial intelligence, and extra on-screen content that less-capable systems never see. Moreover, very few power users run a game all by itself. Software running in the background is also handled more adeptly by a CPU with resources to spare.
Resellers Rule The High-End
The channel is always going to face competition from multiple directions. On one hand, you have the top-tier vendors who do tons of volume and can consequently source the custom-looking enclosures and lighting. On the other, do-it-yourselfers are often eager to hit online shopping sites under the guise of saving tons of money.
An Imposing Monolith. Talk about an imposing presence. You can almost sense that Falcon Northwest’s ICON2 centers on a high-end Core i7-3000-series CPU that is, in many cases, overclocked. Integrators, meanwhile, search for the best ways to add value in a meaningful way, giving customers some feature or capability they wouldn't get from a home brewed machine, while creating a more boutique experience than anything a larger builder could achieve.
Intel's Core i7-3000-series processors and their accompanying platform favor SIs and resellers more than the other two categories. Clearly, the new hardware caters to a particular segment with money to spend, discerning performance expectations, and a need for reliability. That's not the same group you'd expect to see piecing their own boxes together. Power users often lack the time to order, build, and configure their systems. They can't afford to spend two or three days stress testing overclocked settings. At the same time, uptime is a must. Troubleshooting random crashes isn't an option; it's much better to have a phone number to call for answers. There goes the do-it-yourself mentality. But, at the same time, big-name vendors trusted to deliver a reliable experience certainly aren't known for their enthusiast credibility. In the interest of minimizing support calls or dealing with the demanding upper echelon of buyers, they don't push performance as hard as smaller, more specialized system builders.
Now, with Sandy Bridge-E-based processors, resellers who understand what power users and performance-sensitive professionals really need have a chance to shine. In fact, more so than any other time, modifying X79-based platforms is very much skill-based, according to Falcon Northwest's Reeves. "At their stock settings, the new Core i7-3000 CPUs have 130 W TDPs," he says. "But when you overclock them, that more than two billion transistor chip quickly dissipates a lot more power. At that point, managing thermals, which is one of our responsibilities as a system designer, gets tricky."
Here's where it really pays to know this hardware intimately. All three Sandy Bridge-E-based chips are less tolerant to high temperatures than the mainstream Sandy Bridge parts. Intel's throttle point for those 95 W processors with up to four cores is 98 degrees Celsius. The more complex 130 W CPUs are kept under 91 degrees Celsius, which Reeves says is pretty easy to hit when you start overclocking.
As a result, it's very important to handle cooling skillfully. Intel isn't bundling heat sink and fan combinations with these CPUs. Instead, it's letting integrators pick their favorite LGA 2011-compatible coolers. The company does have its own branded closed-loop liquid cooler, which it says is built for overclocking enthusiasts. Manufactured by Asetek, the propylene glycol-filled unit quickly shuttles heat away from the processor through two tubes attached to an internally-mounted radiator. A 120 mm fan blows 74 cubic feet per minute of air through the radiator. Several third-party cooling partners offer their own takes on closed-loop liquid cooling, some of which are even more aggressive than the Intel-branded unit.
Intel also has a value-oriented air cooler available, intended for government/business machines and entry-level workstations. Though it's able to effectively dissipate the heat from a Core i7-3000-series chip running at stock settings (Intel even guarantees compatibility with the next generation of LGA 2011 processors), look to a more enthusiast-oriented design if you want to go the air cooling route.
The reason we like closed-loop liquid designs for Sandy Bridge-E is that their radiators can be mounted against chassis vents. Heat is then exhausted straight out, rather than getting blown around internally. Maintaining a cool ambient temperature is particularly important when you're dealing with quad-channel memory kits and multiple double-slot graphics cards. According to Reeves, his team is noticing that the latest low-voltage DDR3 modules are giving off more heat, compounding the thermal issues with which you're forced to contend.
Of course, at the end of the day, your task is to not only pick components that let you hit attractive performance levels, but also impart unflagging reliability. At stock speeds, that shouldn't be a problem. But taking advantage of an unlocked multiplier for easy overclocking, as an example, isn't covered by Intel's warranty protection. Pushing performance, then, means embracing the support side of a tweaked gaming or workstation box. Systems built on Intel's Core i7-3000-series CPUs running at their default frequencies are already the fastest single-processor systems in the world, so don't feel obligated to push them. In fact, Falcon Northwest's Kelt Reeves makes it a point to mention that very few customers are even interested in buying their hardware overclocked.
The recipe for success, then, becomes a matter of picking the right Sandy Bridge-E-based processor, choosing between Intel's two motherboards built on X79 Express, finding a quad-channel memory kit, adding one or more powerful graphics cards, and mixing in a combination of solid-state and magnetic storage.
