Desktops, servers, and workstations. Those are the three staple segments that every VAR needs to master in order to address the needs of small- and medium-sized businesses. At least, they’re are the areas where you should be focusing, given Intel’s campaign to enable its Nehalem micro-architecture across the company’s broad product portfolio. But covering the bases doesn’t necessarily mean you’re hitting every opportunity out there for margin, upsells, and service. There are a couple more markets to which we think you should cater: mobility and small form factor computing.
The buzz surronding Intel's Nehalem micro-architecture has escalated over the past year and a half, beginning with an excited murmur and crescendoing into a more frenzied roar after the recent debut of its 32nm mainstream CPUs and Xeon 5600-series refresh in Q1. It’s understandable, then, that resellers have their eyes on that prize, focusing on building the best desktops, servers, and workstations for their SMB customers.
However, don’t let Intel’s other channel-friendly gems slip through the cracks. Advancements in manufacturing technology are paving the way for VARs to compete more aggressively in a couple of markets traditionally dominated by larger builders with the resources to build with an eye for aesthetics: ultra-thin laptops and phone book-sized SFF PCs.
The Channel Goes Mobile Beyond the company's presence in more conventional business markets, remember that Intel is also the driving force behind mobile processors, chipsets, and wireless networking technology. Its Centrino brand is one of the most widely recognized. And although it’s no longer a platform marketing initiative encompassing the pieces that go into a notebook, you’ll still find Centrino in Intel’s Wi-Fi and WiMAX adapters.
Channel-Optimized Mobility. Intel’s SP13 whitebook employs 10W Core 2-based processors, serving up copious battery life and ample dual-core performance, all in a 24.5mm Z-height.
Over the years, we’ve covered several attempts to make notebook sales more accessible to the channel. One of the first was the Verified by Intel program, which helped expand standards-based mobile computing in much the same way motherboard form factors and common slots make it possible to piece together desktops today. From there, we went full circle, settling on the Authorized Integrator program. Mobile AIs sell fully-configured notebooks based on their own unique value adds. So, a reseller might partner with one, two, or more AIs to get the solutions their customers want to buy, custom-branded by the AI for a more professional presentation.
VARs are getting a new piece of mobile firepower called Spring Peak for their efforts in 2010, though. Spring Peak takes us back to the concept of having ODMs manufacturing notebook shells that resellers buy from their distributors. This time, however, you’ll have the option of buying built-up systems or procuring barebones configurations. Either way, Spring Peak supports a range of different CPUs, graphics options, and upsell opportunities.
According to Bill Davidson, marketing manager at Intel, Spring Peak is the culmination of everything the company has learned about mobility in the channel. “We’re listening to our customers,” Bill says. “They’ve been telling us they need a family of notebook products in order to compete, and by family they mean models that have a similar look and feel.” Indeed, Spring Peak includes a number of 13” variants, 15” derivatives, and a docking station that supports both chassis.
Starting Off Small The 13" SP13 is considered an ultra-thin design, sporting a sleek 13.3” LED LCD display with a gorgeous 1366x768 resolution. Its 24.5mm Z-height compares favorably to competing tier-one designs, many of which actually feature smaller screens with lower resolutions. The platform centers on a Mobile Intel GS45 chipset, which is optimized for embedded applications, while still delivering a surprisingly modern feature set (up to a 1,066 MHz front side bus, a dual-channel DDR2/DDR3 memory controller, dual display outputs, PCI Express connectivity, USB 2.0, and more). Intel’s GS45 also contributes the platform’s GMA 4500MHD integrated graphics core. Full DirectX 10 support is complemented by hardware acceleration for MPEG-2, VC-1, and AVC video playback, giving this platform the horsepower to smoothly play back Blu-ray content.
Phoning Home. SMBs using their Spring Peak notebooks as office workstations will love the complementary dock, which offers an optical drive, DVI output, gigabit Ethernet, plenty of USB, and of course, power.
Although Intel recently launched its first 32nm Nehalem micro-architecture-based processor back in January, the 13” Spring Peak platform supports a handful of Core 2 Duo CPUs. Now, your customers might wonder why these notebooks employ Core 2 Duos when faster Mobile Core i7, Core i5, and Core i3 models are available. The answer is simple. The four dual-core chips you can use in the SP13, which include the Core 2 Duo SU9600, SU9400, SU7300, and Pentium SU4100, are all 10W models soldered down to the motherboard. The lowest-power Core i5 is an 18W processor—almost twice the peak thermal ceiling of the low-voltage Core 2 Duos—and it drops into an LGA 1156 interface. Thus, achieving an ultra-thin form factor, managing heat, and extending battery life means using the still-potent Core 2 family.
Of course, the SP13 doesn’t sacrifice connectivity in its mission to cram compelling mobile performance into a diminutive chassis. It boasts a gigabit Ethernet port, headphone output, microphone input, a 3-in-1 card reader, two USB 2.0 ports, a USB/eSATA connector, VGA output, HDMI output, and a Kensington lock. Expansion is made possible via 34mm ExpressCard, and you get 802.11 Wi-Fi as a standard feature.
Making The Connection Business-class notebooks are increasingly being used not just on the road, but also as a computing platform in the office. But when your customers are at their desks, they don’t necessarily want to type on a compact laptop keyboard or stare at a 13” display all day long. That’s where docking stations come into play, augmenting the mobile platform’s capabilities. Intel lists the Spring Peak docking station as a fantastic differentiator, and we’d agree. It supports a SATA optical drive, offers gigabit connectivity, USB 2.0, VGA, and DVI outputs. You can order the dock with whatever drive your customer wants, be it the DVD-ROM, a DVD multi-recorder, or a Blu-ray/DVD-RW. Best of all, the dock is compatible not only with the SP13 13” notebook, but also the 15” version, which will see initial availability at the end of Q1.
More Space, More Options Stepping up to the 15" Spring Peak platform naturally translates to a larger LED LCD, a roomier enclosure, and more space for powerful hardware. Here’s where Intel’s latest-generation processors and chipsets have the room to stretch their desktop-derived legs.
The SP15 will be available in a couple of different configurations: one that employs integrated graphics and another (which will follow at the end of April, according to Intel) with discrete graphics. Both center on a common chassis with an extended keyboard and full numeric keypad. They also employ Intel’s Mobile HM55 Express chipset, which extends support for Mobile Core i3, Core i5, and Core i7. So, on top of the distinct graphics options, you have the flexibility to configure the CPU, memory, and storage subsystems, too.
Exploring The Latest Tech. The SP15’s larger chassis offers enough room to accommodate the latest Intel mobile technology, complemented by integrated or discrete graphics.
Of course, all three processor families are based on a two-die package. One of the chips is a 32nm dual-core CPU, and the other is a 45nm graphics, memory, and PCI Express controller. This advanced level of integration is what allows Intel to offer faster clock speeds, Turbo Boost technology, Hyper-Threading support, and better graphics performance than any of its previous-generation mobile platforms and still hit aggressive 35W, 25W, and 18W TDPs.
The first SP15s to hit distributors will be based on Intel HD Graphics built into the company’s processors, utilizing the chipset’s Flexible Display Interface to enable a secondary VGA output. Sporting ample 3D performance and the engine's full suite of video decoding and audio output features, the SP15 does double duty as an entertainment center during off-hours. They’ll include four USB 2.0 ports, standard 802.11n Wi-Fi support, and a 54mm ExpressCard for expansion. Unified support for the Spring Peak docking station means SMBs with a blend of 13” and 15” notebooks can standardize on the same dock without worrying about compatibility.
When the discrete version emerges, it’ll employ Nvidia’s GeForce GT 330M graphics processor. Best suited to gamers on the go, the GeForce GT 330M sports 48 shader cores, giving it up to 182 gigaflops of compute horsepower. Support for DirectX 10.1, OpenGL 2.1, and Nvidia’s PhysX acceleration technology are all hallmarks of a product designed with 3D duties in mind. But the discrete adapter is also well-equipped for video playback, accelerating all three codecs used in Blu-ray movies, and general purpose GPU computing. Plug the discrete SP15 into Intel’s Spring Peak docking station for access to HDMI and VGA outputs.
Going To Market With More It's all well and good to sell your customers the latest and greatest hardware. After all, Intel is in the hardware business and the company has seemingly mastered differentiating through features, speeds, and value-adds. The problem is that you and your competitors are all trying to sell the same differentiated hardware. So, you need your own special sauce in order to do battle against tier-ones. This is especially true in relation to mobility, a segment as of yet unconquered by the channel.
A True Business Notebook. Unlike some of the consumer notebooks out there with piano black finished shells, Intel’s SP13 and SP15 are more functional with their classy textured keyboard and polished body.
Intel’s Spring Peak platform includes two significant enablers of value beyond the basic hardware story. One is called Always Aware. Enabled by a triple-axis accelerometer and custom software, Always Aware serves three unique functions: hard drive protection, physical security, and display orientation control. The hard drive protection feature, dubbed SmartDisk, senses if the notebook is dropped and reactively parks its heads to prevent damage on impact. SmartAlarm is a theft deterrent. Let’s say you order a coffee at Starbucks and you get up from your seat to grab your drink. With the feature enabled, anyone who tries to make a grab for the notebook on your table will set off an audible alert. Finally, SmartScreen automatically rotates the LED display based on the notebook’s orientation. Tilt the display on its side to view in portrait mode, and the screen responds accordingly.
Anti-Theft technology is one of the most impressive additions to Spring Peak, addressing a problematic result of keeping critical business information on a notebook without robust data encryption. Should that system get misplaced or stolen, your customer is immediately looking at a serious security issue. The implementation of Anti-Theft, a combined effort between hardware, software, and a service provider, is perhaps more accurately described as a data protection technology.
Ready To Rock. Your customers can use the SP15 attached to Intel’s docking station, thanks to a comfortable, tilted interface. Or, attach a conventional keyboard, mouse, and monitor to the station.
The hardware component is built-in to Intel’s processors and chipsets. The software and service comes from a provider like Absolute Software (though Intel says others will be coming online soon with Anti-Theft support). In a small business application, you’d be upselling Absolute’s LoJack. Larger organizations should be stepped up to the company’s Computrace offering. In either case, a theft leaves you with a few different options for either recovering the hardware or simply preventing access to the data.
Option one is referred to as a track and trace, whereby Absolute is able to locate the machine when it connects to the Internet. Though this sounds like something from a science fiction movie, Intel’s Bill Davidson says that, actually, in three of four cases, local police are able to issue a warrant and recover the hardware. Option two is to simply deny the thief access to your information by locking it down. As ominous as that sounds, it’s intended to keep your information safe. Anti-Theft can be configured to perform a lock-out if the wrong password is entered a certain number of times. Or, if the hardware hasn’t rendezvoused with Absolute’s service within a preset timeframe, it can be locked as well. Your customer can even call Absolute, which will send a “poison pill.” The next time the stolen machine goes online, it’ll download the code and, again, lock the thief out. As a final resort, the third option is to have Absolute wipe the drive remotely, removing all doubt that any sensitive data on it is inaccessible. Should your customer be forced to go to this extreme, note that there’s a persistence module in the machine’s BIOS that still makes it possible to track and trace the hardware, even if the thief reinstalls Windows or replaces the drive entirely.
Integration Enables Low-Power. The Atom D510 and D410 sport loads of integration, which helps cut overall processor power consumption and enable simpler motherboard designs.
There’s clearly a lot to like about Spring Peak, regardless of whether you’re looking at hardware or software. The fact that Intel now offers a mobile product family with a common aesthetic across the 13” and 15” form factors is a good start. Factor in support for everything from low-voltage Mobile Core 2 Duos to high-performance Mobile Core i7s and you’re looking at a broad range of potential customers. Then there’s the docking station, which accommodates all Spring Peak notebooks, giving businesses the freedom to mix and match the models they order without worrying about compatibility. Intel even paid attention to the finish of its enclosures. “Spring Peak is designed for the VARs selling to SMBs,” says Bill Davidson. “We wanted the chassis to look good, regardless of whether it’s new or being used in the field. It doesn’t pick up fingerprints like many of the notebooks resellers will be competing against.” It sounds like Intel has a well thought-out channel-friendly notebook family, inside and out.
Smaller Than A Breadbox… Mobility isn't the only way to upsell good things in tiny packages. In addition to its Spring Peak initiative, Intel also has new small form factor desktop hardware ready to enable silent computing. Internally dubbed Pine Trail, this is the company’s next-generation Atom-based platform, emphasizing power consumption so low that it doesn’t even require active cooling.
Atom isn’t as heavily marketed as some of Intel’s other brands. But it is a very well-established in the netbook and nettop marketplace, two segments that have ramped up at an unprecedented rate over the past couple of years (Intel says it has already shipped tens of millions of units). Don’t worry if those two words still sound a bit foreign. Because nettops and netbooks are inherently small, low-power devices, they’ve been most quickly adopted by manufacturers able to put them into customized enclosures. Increasingly, though, this is becoming territory where the channel can compete.
Of course, Pine Trail isn’t Intel’s first Atom-based project. Its freshman platform, launched back in 2008, employed a conventional three-chip organization made up of processor, northbridge, and southbridge. Though this arrangement worked well enough, and the Atom CPU proved itself an incredibly energy-efficient chip, its accompanying chipset components were derived from desktop-class logic and therefore not as purpose-built for the job. This time around, though, Intel is coming to market with an intelligently-designed two chip solution that includes a processor and its attached I/O, similar to what we’ve seen on the desktop with the LGA 1156 interface.
Next-Gen Nettops. The dual-core Atom D510 features an on-die memory controller and graphics engine, helping cut power consumption even as integration helps bolster the CPU’s performance.
The new Atom processors themselves actually use more power than their predecessors—13W for the dual-core Atom D510 and 10W for the Atom D410 versus 8W for the older dual-core Atom 330. There’s a good reason for this, though. The memory controller and integrated graphics core that comprised the northbridge of previous-generation platforms are now part of the CPU, eliminating the need for a 6W+ graphics and memory controller hub.
Functionally, little changes. You still get a single channel of DDR2 memory support and a DirectX 9-class graphics engine driving two display outputs. The execution cores themselves carry over mostly-unchanged as well. The Atom D510 and D410 both run at 1.66 GHz with 64-bit support. They also include Hyper-Threading, allowing each physical core to appear as a pair of logical cores in Windows. Really, the only thing separating these two CPUs is that one sports two cores, while the other is a single-core chip.
Both models communicate with Intel’s new NM10 Express chipset over a DMI link, similar to the company’s other two-component mainstream desktop platforms. But the NM10 is a little leaner than a P55 Express, for example, shedding a bit of functionality in favor of compact dimensions and a lower thermal ceiling. When you take into consideration where the NM10 is meant to operate, those compromises are just what the doctor ordered, though. You still get access to a pair of SATA 3 Gb/s ports, up to eight USB 2.0 ports, four PCI Express lanes, and HD Audio support—plenty of I/O for any small form factor build.
Clearly, most of Intel’s changes to Pine Trail center on integration, translating to bill of material savings, lower power use, and better performance. The benefits of the latter two points to your customers are obvious—they get a more responsive system that pulls less power from the wall. But the simpler two-chip platform is more of a win for resellers. Condensing the system architecture means four-layer motherboards, the thermal headroom to go fanless, and physically smaller enclosures.
Building On Pine Trail Intel offers two motherboards in its Essential Series that center on the updated Atom processors: D510MO and D410PT. Both conform to the mini-ITX standard, measuring 6.7” x 6.7, and look very similar at first glance.
A Passively-Cooled Platform. Intel’s D510MO motherboard features an embedded dual-core Atom D510 CPU, passively cooled to enable compact mini-ITX desktops that go light on power and noise.
Of course, the D510MO is armed with an Atom D510 processor, the dual-core model running at 1.66 GHz, soldered right onto the board. Its single-channel DDR2 memory controller enables two slots that take up to 4GB of memory running at 800 MT/s. An Intel HD Audio subsystem exposes six channels of output, and a gigabit Ethernet controller helps facilitate screaming network performance. Despite the fact that this board is intended to drop into a mini-ITX chassis, expansion is still possible via one PCI slot and one PCI Express mini-card. Intel offers a number of wireless networking solutions that'd serve as great drop-in upgrades here, enabling 802.11n connectivity. Or, upsell a PCIe-based SSD, adding as much as 64GB of storage able to move more than 100 MB/s.
As mentioned, the D410PT is aesthetically similar. It is set apart by a few notable feature changes, though. Most obvious is the Atom D410 processor that runs at the same clock rate as the D510, but sports one physical core instead of two (complemented by Hyper-Threading, of course). It also takes up to 4GB of DDR2-800 across a pair of memory slots, but drops audio output to four channels and steps network connectivity down to 10/100 Mb/s. The good news here is that, for a home or office nettop, Intel’s D410PT fits the usage model perfectly. Get the board into a compact mini-ITX chassis, add a slim DVD-ROM drive, and finish it off with 2.5” storage. You’re looking at a very capable e-mail, word processing, and Web browsing machine at a value-oriented price.
Tomorrow’s Storage Today Spring Peak and Pine Trail both represent great opportunities for the channel. They’re solid entry points into segments that most resellers haven’t yet conquered. Intel enables enough differentiation on the hardware side to build right-sized notebooks and small form factor desktops. The company also addresses the importance of software through its Always Aware and Anti-Theft mobile suites and its contributions to the Moblin project for small form factor systems. Moblin is a build of Linux specifically optimized for a rich Internet/media experience on netbooks and nettops powered by Atom processors. The implications of Moblin are significant, given the low-cost nature of the two Atom-based motherboards. While we’re big fans of the improvements made to Windows 7 (and have confirmed that Microsoft’s latest OS runs well on the Pine Trail platforms), the software costs almost as much as the hardware. Deploying Moblin as an alternative gets you up and running without any additional expense.
Another benefit is that Moblin is much leaner than Windows; you don’t need to sell huge hard drives with your nettop builds. What to use instead? Consider solid state storage. An SSD delivers massive performance gains in the metrics that matter to customers using low-power hardware. Operating systems boot faster, application load times fall, and shut downs are near-instantaneous. We could wax poetic about read performance, write speeds, and sub-millisecond response times compared to conventional magnetic storage, but it really takes getting in front of an SSD-equipped machine to realize the experiential impact those specifications have.
Cost is naturally a concern when you start talking about premium storage on value-oriented terminals most commonly used in productivity tasks; for many folks, the addition of an SSD won’t be warranted. This is one of those upsells that deserves its own demonstration, though. And SSDs don’t have to break the bank, either. A 40GB Intel X25-V, which employs the slim 2.5” form factor you’d want in a mini-ITX machine, costs roughly what you’d pay for a 500GB notebook drive—only it’s significantly faster, completely silent, and more energy efficient. Regardless of whether you’re building Pine Trail-based nettops or Spring Peak notebooks, adding solid state storage is going to result in a usability boost your customers will remember.
Opportunities Abound With the launch of its Clarkdale-based Core i5, Core i3, and Pentium processors, Intel now has solutions centering on its Nehalem architecture to address all of your desktops, workstations, and servers. The company’s SIPP, EPSD, and ESAA programs really take the work out of picking components and validating configurations—a boon to the channel. Resellers also have a chance to get in early on channel-friendly notebooks and nettops, both segments that have eluded VARs up until now.
Intel’s efforts on Spring Peak are the culmination of past attempts to not only make mobility more accessible to system builders, but also to give those resellers a real shot at competing against tier-ones. Between the 13” and 15” Spring Peak designs (not to mention the docking station that supports both), you have plenty of scalability, a real performance story to tell, and the I/O SMBs need in the field. Choose the 13” model for ultra-thin applications. The embedded processors built into Intel’s SP13 are plenty potent, yet they sip battery power. Any customer considering a 12” or 13” netbook should compare the SP13’s performance and think hard about an upsell—especially if they’re looking for a system to last several years.
Stepping customers up to the SP15 affords room for higher-performing processors, plus the integrated HD Graphics engine, capable of reasonable 3D and accelerated video playback. It’s certainly nice that the two shells match, enabling a coherent lineup for VARs to customize.
Pine Trail is more of an upsell opportunity in itself, serving as that Web-enabled appliance in the lobby of a business or kitchen of a home. Low-cost, low-power, silent, and yet plenty responsive, both of Intel’s Atom-equipped mini-ITX motherboards are great values to customers with limited IT budgets.
Almost exactly one year ago, Intel launched its Xeon
5500-series processors—some of the fastest, most efficient dual-socket-capable CPUs ever seen. At the same time, the company also unveiled a rich ecosystem of chipsets, motherboards, enclosures, and expansion cards, giving resellers modern tools on which to build the next generation of servers and workstations. Today, Intel is refreshing a number of its highest-end models thanks to a shift to 32nm manufacturing. The good news? These CPUs are compatible with existing chipsets and motherboards, simplifying adoption immensely.
Regardless of whether you're talking about cars, cameras, or computer parts, we always expect this generation’s offerings to be better than what came before. But every so often, there’s an inflection point that stands out for a distinct reason. Intel’s Xeon 5500-series launch one year ago was such an introduction.The company unveiled an architecture that accelerated multi-threaded applications via four execution cores bolstered by Hyper-Threading technology. It improved performance in single-threaded titles through Turbo Boost technology. Gated transistors shut down processing cores when they weren’t in use, allowing Xeon 5500-series CPUs to draw less power, dissipate less heat, and require less cooling than any other server and workstation processor family seen before.
All of those points have been proven with numbers—concrete benchmark results showing exactly how much better the Xeon 5500s are than hardware that preceded them. But perhaps the most telling indicator of the processor family’s significance came from Pat Gelsinger, former senior vice president and general manager of Intel’s Digital Enterprise Group, who called the Xeon 5500-series the most important processor since its Pentium Pro, which was almost 10 years old at the time. He also said he expected he wouldn’t make a similar announcement for another 10 years. That’s quite an endorsement.
And yet, here we are, 12 months later, ready to discuss the successors to the Xeon 5500-series processors. Was all of the excitement over the 5500s just passing euphoria over having a strong server platform to sell into SMB environments? Decidedly not. After all, the CPUs being discussed today interface with the same chipsets and drop into the same motherboards introduced a year ago. The Xeon 5500-series debut was the real thing, and for customers who made the jump and adopted the optimized platform architecture already, there’s a wide-open upgrade path available in today’s announcement. For those that haven’t, Intel’s latest release is a gentle tap on the shoulder asking, “why not?”
Manufacturing Leads The Way We've discussed Intel's "tick-tock" manufacturing cadence many times before; it continues to work in such a way that the company makes regular improvements to its process technology, followed by innovative architectural changes. With its 32nm die shrink, Intel carries over the wins it achieved with Nehalem, decreasing oxide thickness and reducing gate length to help improve transistor performance, plus helping minimize leakage current.
Six Cores Of Fury. Manufactured at 32nm, the Westmere-EP die that constitutes Intel’s Xeon 5600 family features as many as six cores, 12MB of shared L3 cache, and three memory channels.
At mainstream price points, we see these manufacturing improvements manifest as the new Clarkdale-based processors, which combine a 32nm dual-core, Hyper-Threading-equipped core paired up to a 45nm graphics, memory, and PCI Express controller on a single package. The integration-heavy combination of technologies goes a long way to drive down manufacturing cost and platform complexity, all the while improving the performance of subsystems like graphics, which were previously built into northbridges where they’d be starved for data. But when it comes to the upper echelon of performance, smaller transistors aren’t used to facilitate integration so much as they’re used to open the door to additional compute performance.
From its inception, Intel’s Nehalem micro-architecture was designed to be scalable. The first Core i7-, Xeon 5500-, and Xeon 3500-series derivatives were all quad-core designs, sure. However, the mainstream Core i5, Core i3, and Pentium processors demonstrate how easily Intel can maintain the differentiating features that make Nehalem-based platforms so well-balanced. Now the company is doing something similar in the other direction, exploiting its move to 32nm manufacturing to add cores and cache.
A Processor Refresh Is Born The new Xeon 5600-series, formerly refferedto as Westmere-EP, is the product of Intel’s “tick.” Enabled entirely by the company’s 32nm manufacturing process, the Westmere-EP die is made up of as many as 1.17 billion transistors occupying 248 square millimeters. Compare that to Nehalem-EP, which we know as the Xeon 5500 family, made up of 731 million transistors on a 263 square millimeter die.
Of course, the transistor count increase can be attributed to an additional two cores and 4MB of shared L3 cache (totaling six cores and 12MB), maintaining the same core-to-cache ratio seen on the two- and four-core variants. Impressively, though, all of this extra logic fits onto a piece of silicon that’s actually smaller than the generation before. Impressive? You bet. But that’s just the beginning. Intel says that these new processors should be drop-in replacements for the folks who’ve already made an investment in the Xeon 5500-series ecosystem. Will SMBs who’ve jumped on-board with last year’s launch be looking for new CPUs already? Probably not. It is good to know the upgrade path is available, though. More likely is the scenario where an SMB was forced to hold off on new IT equipment last year in light of the business climate, and as the economy has warmed up, is now in the position to start replacing servers and workstations. The good news for the channel is not needing to validate another line of chipsets, motherboards, drivers, etc.
New Options For SMBs. The dual-socket-capable Xeon 5600-series now includes hexa- and quad-core options at 130, 95, 80, 60, and 40W thermal ceilings, appealing to every budget and performance target.
You’re going to be familiar with most of the features being offered with the Xeon 5600-series CPUs, but there is a handful of value-adds debuting here. Chief among them is AES-NI, or Advanced Encryption Standard New Instructions. The AES encryption standard is a publically-accessible and open cipher used by the US government and approved by the NSA to protect top secret information. It’s implemented in software by a number of popular applications, including archive and compression tools, full disk encryption utilities, and LAN security standards. But because performing encryption and decryption in software is potentially taxing in terms of performance, it often goes unused. AES-NI aims to change that by accelerating AES operations in hardware. The feature simultaneously improves security by circumventing possible implementation-based vulnerabilities, referred to as side-channel attacks. Though Intel also makes AES-NI available on mainstream Core i5 processors, it’s an even more exciting feature to discuss with SMBs buying servers. Deploying full disk encryption right from the get-go is a great way to prevent unauthorized access to data storage on a machine housing sensitive information.
Another capability that has been the subject of much discussion and is now accessible though the Xeon 5600-series processors is Trusted Execution Technology (TXT), a family of security-oriented features that is already available on the desktop via vPro. TXT is responsible for protecting five different points: the processor, chipset, user I/O, the display interface, and the Trusted Platform Module. Generally, TXT helps detect and prevent software-based attacks by creating and controlling access to separated execution environments. For instance, it protects keyboard/mouse states, so a user interacting with a trusted platform application doesn’t risk falling victim to a keylogger that might be running on the system. TXT also protects against more sophisticated security breaches, like rootkit hypervisors on a virtualized machine or BIOS update attacks.
More Power In Less Space. Despite the fact that the Xeon 5600 processors include more than a billion transistors, they populate less physical space thanks to Intel’s new 32nm manufacturing process.
TXT works in conjunction with Virtualization Technology for Directed I/O (VT-d), one of the headlining features we talked about when the Xeon 5500-series processors debuted. As a quick refresher, VT-d allows virtualized operating systems direct access to I/O devices, instead of routing requests to hardware through the hypervisor. Ideally, you’ll realize near-native performance, allowing a graphics card, for example, to deliver 3D acceleration to a specific virtual machine running CAD software. But VT-d gets a speed-up with Westmere-EP. The increase in cores and cache is responsible for much of this, as are the lower-latency components used in Intel’s 32nm manufacturing process. Regardless of the reason, Intel says virtualization performance increases by as much as 35% moving from Xeon 5500-series to 5600.
Two of the Xeon 5500’s most key features—Turbo Boost and Hyper-Threading—naturally carry over into the Xeon 5600-series as well. Neither is changed, but they both continue balancing performance in single- and multi-threaded applications. Hyper-Threading, the simultaneous multi-threading technology that makes two logical processors available to the operating system for every physical core present, is even more effective here. With six physical cores per processor, the hexa-core implementation is able to address as many as 12 threads at a time. In a dual-socket server, that’s an astounding 24 threads in flight. Turbo Boost similarly operates as before, turning available thermal headroom into extra clock frequency in applications less-optimized for parallelism. Granted, a six-core Xeon 5600-series chip is going to employ a different binning structure than a four-core Xeon 5500-series CPU. Intel works around this by dialing in its fastest models to use the same single-bin boost that a quad-core model would enjoy.
Climbing The Stack Choice is a beautiful thing. At the same time, it’s important to maintain a product stack that resellers can navigate. Intel knows this, and has organized the Xeon 5600-series into a handful of classifications based on thermal design power. For instance, the Advanced lineup consists of six processors—two 130W models and four 95W variants. The flagship, Intel’s Xeon X5680, is a six-core part running at 3.33 GHz. Its contemporary, the Xeon X5677, is a quad-core derivative running at 3.46 GHz. Both dissipate up to 130W, employ a 6.4 GT/s QPI link, include 12MB shared L3 cache, and support three channels of DDR3-1333 memory (with as many as two slots per channel populated, unlike the Xeon 5500s, which only supported a single slot per channel at that speed).The two processors are already running at aggressive clocks, and consequently utilize conservative 1/1/1/1/2/2 and 1/1/2/2 Turbo Boost configurations. But you’re still looking at an extra 266 MHz on both chips with one or two cores active.
All-Around Server Platform.The S5520UR motherboard represents all that is good about Xeon 5600 in the enterprise. High performance, density, expansion, and capacity all set this board apart.
The other four Advanced processors—three hexa-core models and one quad-core chip—are similarly equipped with the fastest QPI rating, DDR3 ceiling, and large shared L3. Xeon X5670 (2.93 GHz), Xeon 5660 (2.8 GHz), and Xeon X5650 (2.66 GHz) are all six-core versions, while the 3.06 GHz Xeon X5667 includes four cores. Because they’re running at lower standard clock rates, Intel has more room for higher Turbo Boost settings: 2/2/2/2/3/3 and 2/2/3/3 (up to 400 MHz), depending on the 95W model your customer needs.
All three of the new Standard-class Xeons are quad-core models with 5.86 GT/s QPI links, 12MB of shared L3 cache, and support for DDR3-1066. The fastest offering, the Xeon E5640, runs at 2.66 GHz. The Xeon E5630 and E5620 run at 2.53 GHz and 2.4 GHz, respectively. The trio is configured with an 80W TDP. So again, Turbo Boost is limited to a 266 MHz speed-up with one or two cores active. However, it really takes a comparison to the Xeon 5500-series to see how much value is being enabled here. The Xeon E5540, for instance, runs at 2.53 GHz, and includes 8MB of L3 cache. The Xeon E5640 runs one speed bin higher by default, boasts an extra 4MB of cache, and sells for the same price. While the Xeon 5500-series CPU is still a tremendous performer, why wouldn’t you enable your customer with a faster processor for the same investment?
Cutting Edge On A Budget. The S5500BC serves up 24 lanes of PCI Express 2.0 connectivity and it delivers six channels of DDR3 memory to a pair of 5600-series Xeons.
The Basic lineup is also three models strong. Intel’s quad-core Xeon E5507 runs at 2.26 GHz, the quad-core E5506 operates at 2.13 GHz, and its entry-level dual-core E5503 cruises along at 2 GHz. As you can see, the trio carries over from the Xeon 5500-series family. Thus, it gets a 4.8 GT/s QPI interface, 4MB shared L3 cache, and support for DDR3-800 memory.
Its 32nm manufacturing process lets Intel do some interesting things with the low-power lineup, too. Previously comprised of three 60W quad-core CPUs, customers now have access to one six-core 60W processor (the L5640, running at 2.26 GHz with up to four bins of Turbo Boost) and two 40W quad-core options (the L5630 and L5609, running at 2.13 GHz and 1.86 GHz, respectively). These are particularly attractive SKUs not only because they consume a fraction of the power used by the standard-voltage parts, but also because they serve up commanding performance in the process. If you have customers buying into HPC, these processors work well in high-density applications.
Intel’s message here is simple: there’s a lot of differentiating going on, so the Xeon 5600-series isn’t just about finding the SKU your customer can afford. Rather, it’s about selling the features that are most important to the work those customers are doing. Do they need four cores and more clock speed, or are their workloads heavily threaded, favoring the six-core models? Those are the questions that’ll allow you to optimize performance.
Building With 5600-Series CPUs It's no coincidence that the Xeon 5600-series processors drop into the same LGA 1366 interface as the models that came before. It’s also no surprise that the thermal ceilings for these CPUs—despite the new 32nm process—fall in line with the same general peaks. You have high-end 130W processors, Advanced 95W offerings, 80W Standard/Basic variants, and a few special low-voltage options. Intel designed the Xeon 5600-series chips to work on existing 5500- and 5520-based motherboards. Customers running Xeon 5500-series servers today can, with the existing processors installed, update their board’s BIOS to enable support for the Xeon 5600 family. The only thing missing will be the TPM required to enable Trusted Execution Technology. No to worry, though. TXT is still very young as far as the server world goes. So, while it might be one of the Xeon 5600’s differentiators, the software able to employ it is still forthcoming.
Something For Everyone. With thermal ceilings ranging from 40 to 130W, there’s a Xeon 5600-series CPU for every application. Intel’s 1U SR1625 is great for deploying lower-TDP processors.
Of course, we’re making the assumption that SMBs who just bought these fantastically speedy servers and workstations are going to want to upgrade them a year later. You’re more likely to see businesses with Xeon 5300- or 5400-series machines from 2006 or 2007 taking a jump straight to the 5600-series after weathering last year’s economic tumult. For those customers, look to the same Intel motherboards you met back in 2009. Intel says it will be changing some of the product order codes to differentiate the boards with new BIOSes already installed, but the product names won’t be changing. That leaves you with six possible choices for building new servers on the latest Xeons—three boards based on the 5500 chipset and three that employ the 5520.
Picking the right platform starts with understanding the differences between Intel’s two chipset options. The 5500 gives you 24 lanes of PCI Express connectivity with which to work, while the 5520 includes 36 lanes total. They both attach (via a 2 GB/s Direct Media Interface) to the ICH10R I/O hub, which provides USB 2.0, gigabit Ethernet, HD Audio, and SATA storage. Thus, the core logic you choose boils down to how many add-in cards and controllers you plan to install. Despite the simpler chipset lineup, Intel manufactures a board for every environment, from rack-mounted models optimized for energy efficiency to workstation-enabled platforms designed to address graphics-intensive applications.
Connectivity Galore. A pair of second-generation PCI Express x16 slots makes Intel’s S5520SC an ideal motherboard for graphics workstations, as does support for up to 96GB of ultra-fast DDR3 memory.
That energy-efficient model is a particularly interesting better-together story when you match it up to Intel’s new low-voltage Xeons. The engineers who designed the S5500WB spread its voltage regulation circuitry out to dissipate heat across more of the board’s PCB. They also spaced out the LGA 1366 interfaces leaving more room for effective cooling. Less heat means less cooling and, pleasantly, less fan noise. Because the S5500WB runs exclusively on 12V power (converting 12V to 5V and 3.3V using its own on-board logic, rather than relying on a power supply), total board power is cut by as much as 30W.
If you’re looking for a more expandable server platform, the S5520HC is your board. The 5520 chipset gives this board copious PCI Express connectivity, exposed through four x8 slots (perfect for high-end storage and networking cards) and a x4 I/O module meant to accommodate Intel’s new four-port internal SAS controller.
The S5520SC centers on the same chipset. However, it’s not a server board at all. Rather, the S5520SC is all decked out for workstation duties, wielding a pair of second-gen PCI Express x16 slots supporting professional graphics cards. This is the platform you’d want to use for enabling the 130W Xeon 5600-series SKUs in a pedestal enclosure. Twelve DDR3 slots support up to 96GB of memory, and with the Xeon 5600’s ability to run two slots per channel at 1,333 MT/s signaling rates, you’re looking at improved memory performance at higher capacities.
Intel does plan to launch two new board SKUs, both with integrated TPM chips to enable TXT support with Xeon 5600-series CPUs installed. Dubbed S5520URT and S5520HCT, you’ll notice similar model names as existing platforms with a T-suffix tacked on to indicate the TPM. They’re otherwise identical, sporting the exact same PCB layout and feature list.
Differentiating Solid State As Intel's processors get more powerful, increasing total potential platform performance, we hammer home the importance of balance harder than ever. Specifically, you’ll want to keep an eye on the storage subsystems used to feed these powerful configurations with data. Conventional magnetic storage is great for enabling large capacities, but it’s no secret that mechanical hard drives can bottleneck the performance of even a single-processor Xeon machine; a dual-processor setup able to crank through 24 threads at a time is easily starved for information. When it comes to maximizing throughput, SSDs deliver the best results.
Intrepid Performer. Solid state drives are not all created equal. OCZ’s Intrepid lineup is available in SLC and MLC flash varieties, offering read performance up to 260 MB/s and writes as fast as 210 MB/s.
If you want a solution customized for the workloads your customers run, check out OCZ’s Intrepid lineup. Not available at e-tail, the Intrepid drives are validated for the enterprise, which often requires specific optimizations on a per-application basis.
Thus, it's completely apropos that the Intrepid lineup already features plenty of variety.. The 2.5” drives are available in multi-level and single-level cell flavors, enabling several different performance and capacity points. A drive based on MLC technology is able to store multiple bit values per memory cell. It’s less expensive and higher-density. It follows, then, that the MLC Intrepid drives ship in 30, 60, 120, and 250GB capacities, enabling burst reads of up to 250 MB/s and burst write performance as high as 180 MB/s. The drives are able to sustains write performance as high as 100 MB/s. The SLC-based Intrepids hold a single bit value per cell, and thus ship in lower capacities (60GB and 120GB). They’re able to move significantly more data, though, with burst reads of up to 260 MB/s and writes as fast as 210 MB/s. As with the MLC drives, the SLC versions sustains writes around the 100 MB/s mark. Where should you use one flash technology instead of the other? Typically, SLC storage is the way to go in business-class applications. It’s faster, less susceptible to cell leakage, and able to withstand an order of magnitude more write cycles than MLC flash. When it comes to designing 2P servers built for heavy lifting, adding SLC-based SSDs to a SAS storage environment is the fastest, more reliable way to go.
Both of OCZ’s lineups share a number of features in common, such as 64MB of on-board cache, 1.5 million hour MTBF ratings, three-year warranties, and extremely high shock resistance, which is especially desirable in rack-mounted machines. They also employ SATA connectivity—not a problem, even if you’re rolling out SAS controller cards. Because SAS technology includes the SATA Tunneling Protocol, it’ll support a mixed environment of SATA SSDs and SAS hard drives. By mixing and matching familiar magnetic storage with the latest in solid state technology, you’re able to maintain capacity levels, deliver on uptime requirements, and drastically improve the performance of servers built on Intel’s new Xeon 5600-series CPUs.
Putting It All Together Intel's Xeon 5500 platform was a revolutionary launch. It involved a new processor, interface, chipset, motherboards, and add-in peripherals to take advantage of the second-gen PCI Express connectivity. Fortunately, the company’s ESAA program helped make it easy to build validated servers and workstations.
Smart planning and a well-executed transition to 32nm manufacturing make the move to Xeon 5600 significantly less disruptive. Customers who jumped into 5500-series CPUs can update the BIOS on their current boards and drop Xeon 5600s right in. For those who held off, sticking with older Xeon 5300- and 5400-series platforms, now is the time to upgrade. Intel is maintaining its already-established price points and enabling more cache and more clock speed for each step up the stack.
Almost exactly one year ago, Intel launched its Xeon
5500-series processors—some of the fastest, most efficient dual-socket-capable CPUs ever seen. At the same time, the company also unveiled a rich ecosystem of chipsets, motherboards, enclosures, and expansion cards, giving resellers modern tools on which to build the next generation of servers and workstations. Today, Intel is refreshing a number of its highest-end models thanks to a shift to 32nm manufacturing. The good news? These CPUs are compatible with existing chipsets and motherboards, simplifying adoption immensely.
Regardless of whether you're talking about cars, cameras, or computer parts, we always expect this generation’s offerings to be better than what came before. But every so often, there’s an inflection point that stands out for a distinct reason. Intel’s Xeon 5500-series launch one year ago was such an introduction.The company unveiled an architecture that accelerated multi-threaded applications via four execution cores bolstered by Hyper-Threading technology. It improved performance in single-threaded titles through Turbo Boost technology. Gated transistors shut down processing cores when they weren’t in use, allowing Xeon 5500-series CPUs to draw less power, dissipate less heat, and require less cooling than any other server and workstation processor family seen before.
All of those points have been proven with numbers—concrete benchmark results showing exactly how much better the Xeon 5500s are than hardware that preceded them. But perhaps the most telling indicator of the processor family’s significance came from Pat Gelsinger, former senior vice president and general manager of Intel’s Digital Enterprise Group, who called the Xeon 5500-series the most important processor since its Pentium Pro, which was almost 10 years old at the time. He also said he expected he wouldn’t make a similar announcement for another 10 years. That’s quite an endorsement.
And yet, here we are, 12 months later, ready to discuss the successors to the Xeon 5500-series processors. Was all of the excitement over the 5500s just passing euphoria over having a strong server platform to sell into SMB environments? Decidedly not. After all, the CPUs being discussed today interface with the same chipsets and drop into the same motherboards introduced a year ago. The Xeon 5500-series debut was the real thing, and for customers who made the jump and adopted the optimized platform architecture already, there’s a wide-open upgrade path available in today’s announcement. For those that haven’t, Intel’s latest release is a gentle tap on the shoulder asking, “why not?”
Manufacturing Leads The Way
We've discussed Intel's "tick-tock" manufacturing cadence many times before; it continues to work in such a way that the company makes regular improvements to its process technology, followed by innovative architectural changes. With its 32nm die shrink, Intel carries over the wins it achieved with Nehalem, decreasing oxide thickness and reducing gate length to help improve transistor performance, plus helping minimize leakage current.
Six Cores Of Fury. Manufactured at 32nm, the Westmere-EP die that constitutes Intel’s Xeon 5600 family features as many as six cores, 12MB of shared L3 cache, and three memory channels.
At mainstream price points, we see these manufacturing improvements manifest as the new Clarkdale-based processors, which combine a 32nm dual-core, Hyper-Threading-equipped core paired up to a 45nm graphics, memory, and PCI Express controller on a single package. The integration-heavy combination of technologies goes a long way to drive down manufacturing cost and platform complexity, all the while improving the performance of subsystems like graphics, which were previously built into northbridges where they’d be starved for data. But when it comes to the upper echelon of performance, smaller transistors aren’t used to facilitate integration so much as they’re used to open the door to additional compute performance.
From its inception, Intel’s Nehalem micro-architecture was designed to be scalable. The first Core i7-, Xeon 5500-, and Xeon 3500-series derivatives were all quad-core designs, sure. However, the mainstream Core i5, Core i3, and Pentium processors demonstrate how easily Intel can maintain the differentiating features that make Nehalem-based platforms so well-balanced. Now the company is doing something similar in the other direction, exploiting its move to 32nm manufacturing to add cores and cache.
A Processor Refresh Is Born
The new Xeon 5600-series, formerly refferedto as Westmere-EP, is the product of Intel’s “tick.” Enabled entirely by the company’s 32nm manufacturing process, the Westmere-EP die is made up of as many as 1.17 billion transistors occupying 248 square millimeters. Compare that to Nehalem-EP, which we know as the Xeon 5500 family, made up of 731 million transistors on a 263 square millimeter die.
Of course, the transistor count increase can be attributed to an additional two cores and 4MB of shared L3 cache (totaling six cores and 12MB), maintaining the same core-to-cache ratio seen on the two- and four-core variants. Impressively, though, all of this extra logic fits onto a piece of silicon that’s actually smaller than the generation before. Impressive? You bet. But that’s just the beginning. Intel says that these new processors should be drop-in replacements for the folks who’ve already made an investment in the Xeon 5500-series ecosystem. Will SMBs who’ve jumped on-board with last year’s launch be looking for new CPUs already? Probably not. It is good to know the upgrade path is available, though. More likely is the scenario where an SMB was forced to hold off on new IT equipment last year in light of the business climate, and as the economy has warmed up, is now in the position to start replacing servers and workstations. The good news for the channel is not needing to validate another line of chipsets, motherboards, drivers, etc.
New Options For SMBs. The dual-socket-capable Xeon 5600-series now includes hexa- and quad-core options at 130, 95, 80, 60, and 40W thermal ceilings, appealing to every budget and performance target.
You’re going to be familiar with most of the features being offered with the Xeon 5600-series CPUs, but there is a handful of value-adds debuting here. Chief among them is AES-NI, or Advanced Encryption Standard New Instructions. The AES encryption standard is a publically-accessible and open cipher used by the US government and approved by the NSA to protect top secret information. It’s implemented in software by a number of popular applications, including archive and compression tools, full disk encryption utilities, and LAN security standards. But because performing encryption and decryption in software is potentially taxing in terms of performance, it often goes unused. AES-NI aims to change that by accelerating AES operations in hardware. The feature simultaneously improves security by circumventing possible implementation-based vulnerabilities, referred to as side-channel attacks. Though Intel also makes AES-NI available on mainstream Core i5 processors, it’s an even more exciting feature to discuss with SMBs buying servers. Deploying full disk encryption right from the get-go is a great way to prevent unauthorized access to data storage on a machine housing sensitive information.
Another capability that has been the subject of much discussion and is now accessible though the Xeon 5600-series processors is Trusted Execution Technology (TXT), a family of security-oriented features that is already available on the desktop via vPro. TXT is responsible for protecting five different points: the processor, chipset, user I/O, the display interface, and the Trusted Platform Module. Generally, TXT helps detect and prevent software-based attacks by creating and controlling access to separated execution environments. For instance, it protects keyboard/mouse states, so a user interacting with a trusted platform application doesn’t risk falling victim to a keylogger that might be running on the system. TXT also protects against more sophisticated security breaches, like rootkit hypervisors on a virtualized machine or BIOS update attacks.
More Power In Less Space. Despite the fact that the Xeon 5600 processors include more than a billion transistors, they populate less physical space thanks to Intel’s new 32nm manufacturing process.
TXT works in conjunction with Virtualization Technology for Directed I/O (VT-d), one of the headlining features we talked about when the Xeon 5500-series processors debuted. As a quick refresher, VT-d allows virtualized operating systems direct access to I/O devices, instead of routing requests to hardware through the hypervisor. Ideally, you’ll realize near-native performance, allowing a graphics card, for example, to deliver 3D acceleration to a specific virtual machine running CAD software. But VT-d gets a speed-up with Westmere-EP. The increase in cores and cache is responsible for much of this, as are the lower-latency components used in Intel’s 32nm manufacturing process. Regardless of the reason, Intel says virtualization performance increases by as much as 35% moving from Xeon 5500-series to 5600.
Two of the Xeon 5500’s most key features—Turbo Boost and Hyper-Threading—naturally carry over into the Xeon 5600-series as well. Neither is changed, but they both continue balancing performance in single- and multi-threaded applications. Hyper-Threading, the simultaneous multi-threading technology that makes two logical processors available to the operating system for every physical core present, is even more effective here. With six physical cores per processor, the hexa-core implementation is able to address as many as 12 threads at a time. In a dual-socket server, that’s an astounding 24 threads in flight. Turbo Boost similarly operates as before, turning available thermal headroom into extra clock frequency in applications less-optimized for parallelism. Granted, a six-core Xeon 5600-series chip is going to employ a different binning structure than a four-core Xeon 5500-series CPU. Intel works around this by dialing in its fastest models to use the same single-bin boost that a quad-core model would enjoy.
Climbing The Stack
Choice is a beautiful thing. At the same time, it’s important to maintain a product stack that resellers can navigate. Intel knows this, and has organized the Xeon 5600-series into a handful of classifications based on thermal design power. For instance, the Advanced lineup consists of six processors—two 130W models and four 95W variants. The flagship, Intel’s Xeon X5680, is a six-core part running at 3.33 GHz. Its contemporary, the Xeon X5677, is a quad-core derivative running at 3.46 GHz. Both dissipate up to 130W, employ a 6.4 GT/s QPI link, include 12MB shared L3 cache, and support three channels of DDR3-1333 memory (with as many as two slots per channel populated, unlike the Xeon 5500s, which only supported a single slot per channel at that speed).The two processors are already running at aggressive clocks, and consequently utilize conservative 1/1/1/1/2/2 and 1/1/2/2 Turbo Boost configurations. But you’re still looking at an extra 266 MHz on both chips with one or two cores active.
ALL-AROUND Server PLATFORM. The S5520UR motherboard represents all that is good about Xeon 5600 in the enterprise. High performance, density, expansion, and capacity all set this board apart.
The other four Advanced processors—three hexa-core models and one quad-core chip—are similarly equipped with the fastest QPI rating, DDR3 ceiling, and large shared L3. Xeon X5670 (2.93 GHz), Xeon 5660 (2.8 GHz), and Xeon X5650 (2.66 GHz) are all six-core versions, while the 3.06 GHz Xeon X5667 includes four cores. Because they’re running at lower standard clock rates, Intel has more room for higher Turbo Boost settings: 2/2/2/2/3/3 and 2/2/3/3 (up to 400 MHz), depending on the 95W model your customer needs.
All three of the new Standard-class Xeons are quad-core models with 5.86 GT/s QPI links, 12MB of shared L3 cache, and support for DDR3-1066. The fastest offering, the Xeon E5640, runs at 2.66 GHz. The Xeon E5630 and E5620 run at 2.53 GHz and 2.4 GHz, respectively. The trio is configured with an 80W TDP. So again, Turbo Boost is limited to a 266 MHz speed-up with one or two cores active. However, it really takes a comparison to the Xeon 5500-series to see how much value is being enabled here. The Xeon E5540, for instance, runs at 2.53 GHz, and includes 8MB of L3 cache. The Xeon E5640 runs one speed bin higher by default, boasts an extra 4MB of cache, and sells for the same price. While the Xeon 5500-series CPU is still a tremendous performer, why wouldn’t you enable your customer with a faster processor for the same investment?
The Basic lineup is also three models strong. Intel’s quad-core Xeon E5507 runs at 2.26 GHz, the quad-core E5506 operates at 2.13 GHz, and its entry-level dual-core E5503 cruises along at 2 GHz. As you can see, the trio carries over from the Xeon 5500-series family. Thus, it gets a 4.8 GT/s QPI interface, 4MB shared L3 cache, and support for DDR3-800 memory.
Cutting Edge On A Budget. The S5500BC serves up 24 lanes of PCI Express 2.0 connectivity and it delivers six channels of DDR3 memory to a pair of 5600-series Xeons.
Its 32nm manufacturing process lets Intel do some interesting things with the low-power lineup, too. Previously comprised of three 60W quad-core CPUs, customers now have access to one six-core 60W processor (the L5640, running at 2.26 GHz with up to four bins of Turbo Boost) and two 40W quad-core options (the L5630 and L5609, running at 2.13 GHz and 1.86 GHz, respectively). These are particularly attractive SKUs not only because they consume a fraction of the power used by the standard-voltage parts, but also because they serve up commanding performance in the process. If you have customers buying into HPC, these processors work well in high-density applications.
Intel’s message here is simple: there’s a lot of differentiating going on, so the Xeon 5600-series isn’t just about finding the SKU your customer can afford. Rather, it’s about selling the features that are most important to the work those customers are doing. Do they need four cores and more clock speed, or are their workloads heavily threaded, favoring the six-core models? Those are the questions that’ll allow you to optimize performance.
Building With 5600-Series CPUs
It's no coincidence that the Xeon 5600-series processors drop into the same LGA 1366 interface as the models that came before. It’s also no surprise that the thermal ceilings for these CPUs—despite the new 32nm process—fall in line with the same general peaks. You have high-end 130W processors, Advanced 95W offerings, 80W Standard/Basic variants, and a few special low-voltage options. Intel designed the Xeon 5600-series chips to work on existing 5500- and 5520-based motherboards. Customers running Xeon 5500-series servers today can, with the existing processors installed, update their board’s BIOS to enable support for the Xeon 5600 family. The only thing missing will be the TPM required to enable Trusted Execution Technology. No to worry, though. TXT is still very young as far as the server world goes. So, while it might be one of the Xeon 5600’s differentiators, the software able to employ it is still forthcoming.
Something For Everyone. With thermal ceilings ranging from 40 to 130W, there’s a Xeon 5600-series CPU for every application. Intel’s 1U SR1625 is great for deploying lower-TDP processors.
Of course, we’re making the assumption that SMBs who just bought these fantastically speedy servers and workstations are going to want to upgrade them a year later. You’re more likely to see businesses with Xeon 5300- or 5400-series machines from 2006 or 2007 taking a jump straight to the 5600-series after weathering last year’s economic tumult. For those customers, look to the same Intel motherboards you met back in 2009. Intel says it will be changing some of the product order codes to differentiate the boards with new BIOSes already installed, but the product names won’t be changing. That leaves you with six possible choices for building new servers on the latest Xeons—three boards based on the 5500 chipset and three that employ the 5520.
Picking the right platform starts with understanding the differences between Intel’s two chipset options. The 5500 gives you 24 lanes of PCI Express connectivity with which to work, while the 5520 includes 36 lanes total. They both attach (via a 2 GB/s Direct Media Interface) to the ICH10R I/O hub, which provides USB 2.0, gigabit Ethernet, HD Audio, and SATA storage. Thus, the core logic you choose boils down to how many add-in cards and controllers you plan to install. Despite the simpler chipset lineup, Intel manufactures a board for every environment, from rack-mounted models optimized for energy efficiency to workstation-enabled platforms designed to address graphics-intensive applications.
That energy-efficient model is a particularly interesting better-together story when you match it up to Intel’s new low-voltage Xeons. The engineers who designed the S5500WB spread its voltage regulation circuitry out to dissipate heat across more of the board’s PCB. They also spaced out the LGA 1366 interfaces leaving more room for effective cooling. Less heat means less cooling and, pleasantly, less fan noise. Because the S5500WB runs exclusively on 12V power (converting 12V to 5V and 3.3V using its own on-board logic, rather than relying on a power supply), total board power is cut by as much as 30W.
If you’re looking for a more expandable server platform, the S5520HC is your board. The 5520 chipset gives this board copious PCI Express connectivity, exposed through four x8 slots (perfect for high-end storage and networking cards) and a x4 I/O module meant to accommodate Intel’s new four-port internal SAS controller.
Connectivity Galore. A pair of second-generation PCI Express x16 slots makes Intel’s S5520SC an ideal motherboard for graphics workstations, as does support for up to 96GB of ultra-fast DDR3 memory.
The S5520SC centers on the same chipset. However, it’s not a server board at all. Rather, the S5520SC is all decked out for workstation duties, wielding a pair of second-gen PCI Express x16 slots supporting professional graphics cards. This is the platform you’d want to use for enabling the 130W Xeon 5600-series SKUs in a pedestal enclosure. Twelve DDR3 slots support up to 96GB of memory, and with the Xeon 5600’s ability to run two slots per channel at 1,333 MT/s signaling rates, you’re looking at improved memory performance at higher capacities.
Intel does plan to launch two new board SKUs, both with integrated TPM chips to enable TXT support with Xeon 5600-series CPUs installed. Dubbed S5520URT and S5520HCT, you’ll notice similar model names as existing platforms with a T-suffix tacked on to indicate the TPM. They’re otherwise identical, sporting the exact same PCB layout and feature list.
Differentiating Solid State
As Intel's processors get more powerful, increasing total potential platform performance, we hammer home the importance of balance harder than ever. Specifically, you’ll want to keep an eye on the storage subsystems used to feed these powerful configurations with data. Conventional magnetic storage is great for enabling large capacities, but it’s no secret that mechanical hard drives can bottleneck the performance of even a single-processor Xeon machine; a dual-processor setup able to crank through 24 threads at a time is easily starved for information. When it comes to maximizing throughput, SSDs deliver the best results.
If you want a solution customized for the workloads your customers run, check out OCZ’s Intrepid lineup. Not available at e-tail, the Intrepid drives are validated for the enterprise, which often requires specific optimizations on a per-application basis.
Thus, it's completely apropos that the Intrepid lineup already features plenty of variety.. The 2.5” drives are available in multi-level and single-level cell flavors, enabling several different performance and capacity points. A drive based on MLC technology is able to store multiple bit values per memory cell. It’s less expensive and higher-density. It follows, then, that the MLC Intrepid drives ship in 30, 60, 120, and 250GB capacities, enabling burst reads of up to 250 MB/s and burst write performance as high as 180 MB/s. The drives are able to sustains write performance as high as 100 MB/s. The SLC-based Intrepids hold a single bit value per cell, and thus ship in lower capacities (60GB and 120GB). They’re able to move significantly more data, though, with burst reads of up to 260 MB/s and writes as fast as 210 MB/s. As with the MLC drives, the SLC versions sustains writes around the 100 MB/s mark. Where should you use one flash technology instead of the other? Typically, SLC storage is the way to go in business-class applications. It’s faster, less susceptible to cell leakage, and able to withstand an order of magnitude more write cycles than MLC flash. When it comes to designing 2P servers built for heavy lifting, adding SLC-based SSDs to a SAS storage environment is the fastest, more reliable way to go.
Intrepid Performer. Solid state drives are not all created equal. OCZ’s Intrepid lineup is available in SLC and MLC flash varieties, offering read performance up to 260 MB/s and writes as fast as 210 MB/s.
Both of OCZ’s lineups share a number of features in common, such as 64MB of on-board cache, 1.5 million hour MTBF ratings, three-year warranties, and extremely high shock resistance, which is especially desirable in rack-mounted machines. They also employ SATA connectivity—not a problem, even if you’re rolling out SAS controller cards. Because SAS technology includes the SATA Tunneling Protocol, it’ll support a mixed environment of SATA SSDs and SAS hard drives. By mixing and matching familiar magnetic storage with the latest in solid state technology, you’re able to maintain capacity levels, deliver on uptime requirements, and drastically improve the performance of servers built on Intel’s new Xeon 5600-series CPUs.
Putting It All Together
Intel's Xeon 5500 platform was a revolutionary launch. It involved a new processor, interface, chipset, motherboards, and add-in peripherals to take advantage of the second-gen PCI Express connectivity. Fortunately, the company’s ESAA program helped make it easy to build validated servers and workstations.
Smart planning and a well-executed transition to 32nm manufacturing make the move to Xeon 5600 significantly less disruptive. Customers who jumped into 5500-series CPUs can update the BIOS on their current boards and drop Xeon 5600s right in. For those who held off, sticking with older Xeon 5300- and 5400-series platforms, now is the time to upgrade. Intel is maintaining its already-established price points and enabling more cache and more clock speed for each step up the stack.
------------------------------------------
Six Cores Of Fury. Manufactured at 32nm, the Westmere-EP die that constitutes Intel’s Xeon 5600 family features as many as six cores, 12MB of shared L3 cache, and three memory channels.
New Options For SMBs.The dual-socket-capable Xeon 5600-series now includes hexa- and quad-core options at 130, 95, 80, 60, and 40W thermal ceilings, appealing to every budget and performance target.
More Power In Less Space.Despite the fact that the Xeon 5600 processors include more than a billion transistors, they populate less physical space thanks to Intel’s new 32nm manufacturing process.
ALL-AROUND Server PLATFORM.The S5520UR motherboard represents all that is good about Xeon 5600 in the enterprise. High performance, density, expansion, and capacity all set this board apart.
Cutting Edge On A Budget.The S5500BC serves up 24 lanes of PCI Express 2.0 connectivity and it delivers six channels of DDR3 memory to a pair of 5600-series Xeons.
Something For Everyone.With thermal ceilings ranging from 40 to 130W, there’s a Xeon 5600-series CPU for every application. Intel’s 1U SR1625 is great for deploying lower-TDP processors.
Connectivity Galore.A pair of second-generation PCI Express x16 slots makes Intel’s S5520SC an ideal motherboard for graphics workstations, as does support for up to 96GB of ultra-fast DDR3 memory.
An Eye On Power.Intel’s S5500WB motherboard was designed with energy-efficiency in mind. From its layout-optimized voltage regulation circuitry to exclusively 12V input, it’ll help cut power costs.
Intrepid Performer.Solid state drives are not all created equal. OCZ’s Intrepid lineup is available in SLC and MLC flash varieties, offering read performance up to 260 MB/s and writes as fast as 210 MB/s.
The DX58SO motherboard includes all of the slots, ports, and connectors needed to build potent gaming platforms and workstations.
In the year and a half since Intel launched its X58 Express chipset and Core i7-900-series processors, the company has gone uncontested in the high-end desktop market. The PCI Express 2.0-rich platform continues to be top-choice for enthusiasts eyeing multi-GPU rendering technologies like CrossFire and SLI. Moreover, Intel’s Nehalem micro-architecture (with its four cores, Hyper-Threading support, and Turbo Boost technology) still serves up the best single- and multi-threaded application performance.
But the wheels of progress keep turning. Intel recently started shipping mainstream processors based on its 32nm lithography process, and is now ready to apply the next generation of manufacturing technology to its fastest desktop chips. By further miniaturizing transistors and improving their power handling characteristics, Intel is able to create a six-core monster with 12MB of shared L3 cache (totaling 1.17 billion transistors) that fits onto a smaller die than its predecessor!
There currently exists a single model enabling Intel’s latest architectural evolution: the Core i7-980X Extreme Edition. Operating at the same 3.33 GHz as the previous flagship (Core i7-975), performance naturally increases in threaded software. But because the 12MB shared L3 cache can be allocated dynamically, some single-threaded apps will even see a speed-up.
Now here’s the really cool part. Despite its many improvements, Core i7-980X is designed to drop into the same LGA 1366 interface as existing Core i7-900-series processors. And because Intel used the 130W thermal design power ceiling already in place when it conceptualized this CPU, the same X58 Express-based motherboards you’ve already sold support Core i7-980X as well. Simply roll out a BIOS update to arm those boards with the microcode needed to properly recognize a brand new processor.
So, you don’t need to stock another platform in order to sell six-core powerhouses. Just pick a motherboard with the feature set that works for enthusiasts and workstation customers alike, such as Intel’s DX58SO. The board’s LGA 1366 interface accommodates not only the entire line of Core i7-900-series CPUs, but also a number of the Xeon 5500-series processors. A quartet of memory slots takes up to 16GB of DDR3-1600 memory for blazing-fast transfer rates across three channels.
Intel’s Core i7-980X wields six cores (with Hyper-Threading, of course) and 12MB of shared L3 cache to become the latest flagship in Intel’s desktop processor family.
A pair of second-gen PCI Express x16 slots is complemented by support for the CrossFire and SLI technologies. An open-ended PCI Express x4 slot addresses the I/O-constrained video editing crowd with room for a high-end storage controller and the productivity-oriented customer needing additional display outputs. Standard on-board I/O includes 12 USB 2.0 ports, six SATA 3 Gb/s connectors, a pair of eSATA ports for attaching external storage, two FireWire 400 interfaces, and an IR receiver. Intel’s HD Audio implementation consists of 7.1-channel analog output and two digital audio outputs (coaxial and optical) via Realtek’s popular ALC889 codec.
Paired together, Intel’s new Core i7-980X and well-established DX58SO are a powerful combination, comprising the fastest desktop hardware currently available. And of course, the fact that you’re selling an Intel processor on an Intel motherboard employing an Intel chipset is good news for gamers and professionals alike, all of whom value stability just as much as unadulterated speed.
Current Issue
