By Chris Angelini
 
 
Whatever you call them, product names can be a major source of confusion for customers. Take AMD as an example. Sure, it’s easier to call the chip powering AMD’s fastest graphics cards the R600 or Pele instead of the Radeon HD 2900 GPU. But outside of technical discussions, product code names are of limited value. Nobody can buy a bare “R600.” Most large tech manufacturers face this nomenclature nonsense and the problems it poses. Intel is no exception.


However, while Intel might prefer talking about official products, its code names are arguably better known. I was never able to hash through the Pentium 4 product families or their single-letter suffixes. But I was painfully aware that the 90nm Prescott ran hot, and I knew the 65nm Yonah was a faster, cooler architecture. I know that McKay Creek is the company’s all-star storage server platform, and Boiler Bay is the SAS controller that gives it so much storage scalability. Sometimes it’s just easier to identify technology when it isn’t buried under a confusing model sequence.

No matter how you feel about code names, get ready for a barrage of new ones from Intel. The company recently launched its first 45nm CPUs and is promising product in all segments-—server, desktop, and mobile-—by the beginning of 2008. You’ll already find 16 new processors on Intel’s price list. Take the initiative and put names to faces so you can point your customer in the direction of actual SKUs when they start asking about Intel’s latest offerings.

New Processor Family, Incoming

The code name thrown around most often seems to be Penryn. Incidentally, Penryn refers to the entire family of 45nm parts, from the mobile Core 2 Duo to the dual- and quad-core Xeons. By simply switching from 65nm to 45nm transistors, Intel says it’s able to nearly double transistor density while reducing switching power by 30 percent. Additionally, all of the processors contained within Penryn sport a handful of notable upgrades over past Core-based processors, helping them yield more performance and better energy efficiency.

Back when the Pentium 4 ruled Intel’s roost, clock speed was paramount. We all know how that tale ended, and today the tune is about getting more work done in less time, or maximizing instructions per clock cycle (IPC). The Core microarchitecture effectively sacrifices raw megahertz in favor of a shorter execution pipeline that accomplishes more work during each cycle. Penryn chips boast a faster divider, which accelerates divide operations by 1.6 to 2 times, and a Super Shuffle Engine that can reposition bits (a common task in image editing apps,Intel says) up to three times faster. Combined, the two microarchitecture enhancements further improve IPC throughput.

Because the chips are manufactured on a 45nm node (compared to 65nm), Intel can add more transistors per processor and still reduce overall die size. Let your customers know they can expect CPUs with L2 caches as large as 12MB, adding even better responsiveness to Intel’s multi-core offerings. Pay particular attention to the cache story. Unlike AMD’s processors, which leverage chip-to-chip communications pathways, Intel is still taking flak for its front side bus configuration. Adding more cache means less data traveling over the bus and big boosts to audio, video, and image editing apps working on large data sets.

Even as information moves from the hard drive to system memory and over the front side bus, Penryn is still going to deliver faster performance. The majority of Penryn-based chips already shipping employ a 1333 MHz bus speed. One still cruises along at 1066 MHz. A full five employ an official 1600 MHz bus speed, pushing throughput even higher.

TAKING IT TO THE EXTREME
The Core 2 Extreme QX9770 is the fastest desktop processor to date. Operating at 3.2 GHz on a 1600 MHz front side bus, the quad-core flagship is an enthusiast’s dream come true.

Last on the list of Penryn upgrades is the long-awaited addition of SSE4. According to Intel, SSE4 consists of 54 total instructions divided into “Vectorizing Compiler and Media Accelerators” and “Efficient Accelerated String and Text Processing.” Hardly self-explanatory, right? The first bundle of instructions helps developers generate performance-oriented code. The instructions are very much optimized for media operations, so any audio, video, encoding, image-editing, and gaming application written with SSE4 in mind will realize strong gains. The second set allows multiple compare and search operations to run simultaneously. Of course, database and data mining apps should see some benefit there, along with any other program that parses, searches, and matches patterns, such as virus scanners.

Much of Penryn’s secret sauce is available to customers right now. Drop a chip into one of Intel’s platforms, fire it up, and benchmark the new core’s advantages. The larger cache, Super Shuffle Engine, Radix-16 divider and Cache Line Split Load all speed things along today. Some of the other features need to be “turned on” by software developers before SMBs can really see them in action. Getting the full benefit of SSE4, for example, requires a recompile using the Intel compiler version 10. A revisit from some of Intel’s other software tools, such as VTune, the Integrated Performance Primitives, and Math Kernel Library couldn’t hurt either. If you haven’t done so already, this is a great chance to introduce programming customers to those development tools.


More Names In The Server Space

With a little more knowledge of the Penryn microarchitecture, we can get into the specific products peppering Intel’s desktop, mobile, and server segments. The first 45nm chips Intel launched were in its Xeon family. All of the new Xeon offerings are divided into two families: quad-core Harpertown and dual-core Wolfdale chips.

Harpertown processors have four cores each, along with 12MB of L2 cache. They all belong to the 5400 sequence and are prefixed by the letter X, E, or L, specifying performance, standard, and low-power versions, respectively. A Xeon X5460, for example, runs at 3.16 GHz and has a 120W TDP rating. The Xeon E5450 is just a little slower at 3 GHz, but it drops TDP to 80W. Although the low-voltage models haven’t been launched yet, expect TDPs in the 50W range for those. Intel is understandably focusing its efforts on quad-core, so you’ll find the most selection and best availability within the 12 Xeon 5400 series chips launched early in November.

Then again, not everyone needs a quad-core CPU in the SMB space. And because Harpertown consists of two dual-core processor die tied together, yields, power, and cooling all play a role in limiting what the chip can do. Wolfdale DP is Intel’s dual-core play, enabling faster core clocks, lower TDPs at the high end, and, of course, more conservative price tags. Intel took the wraps off of three dual-core Xeon models: the X5272 at 3.4 GHz (1600 MHz FSB), the X5260 at 3.33 GHz (1333 MHz), and the E5205 at 1.86 GHz (1066 MHz). Half as many cores mean half as much cache-—all Wolfdale DP chips top out at 6MB of L2. And because the core is less complex, it consumes less power. At 3.4 GHz, the X5272 is rated at 80W—-far less than the quad-core X5482’s 150W.

CACHING IN
This dual-core Penryn die shows Intel’s obvious emphasis on leveraging a large, shared cache. The 6MB of L2 tops the list of many performance enhancers in this latest CPU design.

Intel Waits For The Desktop Play

According to Intel, its desktop chips based on the Penryn microarchitecture are slated for release in Q1 of 2008. However, the company already took the wraps off of one model, which should give your customers some idea of what to expect when the whole family is unveiled.

The first 45nm desktop offering centers on Intel’s Yorkfield design and is officially named the Core 2 Extreme QX9650. Like the quad-core Harpertown Xeons, Yorkfield wields 12MB of L2 cache. The 3 GHz CPU runs on a 1333 MHz front side bus and is rated for a 130W TDP. Technically, the desktop chip looks a lot like a Xeon X5450. The main difference, of course, is that Intel’s Core 2 Extreme uses an LGA775 interface. As long as the motherboards you’re selling feature compatible chipsets and BIOS files, interoperability with any Yorkfield chip should be a non-issue.

When the second wave of 45nm CPUs hits early next year, the dual-core Wolfdale design will start seeing some play. Care for an easy way to keep tech specs straight on the two desktop cores? Remember that each Yorkfield processor is made up of two Wolfdales. So while the quad-core chip features 12MB of cache, the dual-core model includes 6MB. Intel hasn’t yet divulged specs, model numbers, or pricing on the Wolfdale lineup, but resellers should expect the dual-core lineup to step up as Intel’s mainstream focus.

We usually don’t spend much time talking about hardware that’s still months away. However, you and your customers want to know about Yorkfield and Wolfdale because the motherboards you sell today determine the chips SMB buyers can use down the road. Sell the enthusiast X38 chipset, the G35 with integrated graphics and official support for Intel’s 1333 MHz front side bus, or the P35, which takes discrete graphics and supports the same bus settings. All three work with Intel’s 45nm cores.

A Platform Interlude

Before you grab hold of a P35-based motherboard and Intel’s Core 2 Extreme QX9650, be aware that the company is pushing enthusiast performance a step further before the end of 2007.

To begin, it’s introducing a Core 2 Extreme QX9770 based on the Yorkfield design and running at 3.2 GHz. The big news is official support for a 1600 MHz front side bus—-a first on the desktop. Considering the new flagship features similar specifications to Intel’s $1,279 Xeon X5482, it remains to be seen how Intel prices the chip. But if past Extreme chips serve as any indication, expect the QX9770 to be a powerful enthusiast chip in the $1,000-range.

POWER WITH PENRYN
Intel’s S7000FC4UR server is a 4P powerhouse that shines brightest when fueled with 45nm-based Xeon 7300 processors. The system is based on Intel’s latest 7300 chipset, which enables the Penryn compatibility.

Supporting the Core 2 Extreme QX9770 is another matter entirely because none of Intel’s current desktop chipsets officially feature the 1600 MHz bus setting. Even X38, the top enthusiast chipset, isn’t supposed to do 1600 MHz. Of course, that doesn’t stop third-party partners or Intel board-owning enthusiasts from pushing their X38 solutions an extra 266 MHz to enable the setting.

Don’t expect any major changes jumping from X38 to X48. PCI Express 2.0 is still there, as is CrossFire (and perhaps SLI) compatibility, DDR3 memory support, and the ICH9’s impressive connectivity list. The most significant addition is that official 1600 MHz bus setting, and there still aren’t many CPUs that can take advantage of it. If a customer has to have the QX9770 today, go with a third-party X38 board.


Penryn Hits The Road

Once Intel has let loose with its Xeons and released all of the desktop Core 2 processors, focus will shift to the mobile market. David Perlmutter, senior vice president and general manager of Intel’s mobility group, announced at this year’s fall IDF that the Santa Rosa platform would be refreshed in January of 2008 with Penryn mobile processors. The move is particularly significant because of the impact of smaller process technology. Core 2 Duo processors in the Penryn family are expected to feature 6MB of cache, run on an 800 MHz bus, and ship at speeds of up to 2.6 GHz. Despite sporting more transistors than ever, Penryn maintains the same TDP clock for clock.

Mobile CPUs centering on the 45nm technology also incorporate some interesting new energy efficiency functionality. For instance, Deep Power Down Technology further minimizes power consumption at idle. Single-threaded performance also increases, thanks to Enhanced Dynamic Acceleration Technology, which takes advantage of the fact that one core is idle to turbocharge the core in use.

With the addition of Penryn, Santa Rosa notebooks will run faster and eat up less battery power. And because today’s whitebook shells are already based on Mobile 965 Express chipsets, you shouldn’t run into any trouble dropping the new CPUs right in.

PENRYN PART 2
Intel’s “tick-tock” manufacturing dictates that the fab shrink of this year will become the foundation of next year’s architecture overhaul. Here, Intel execs Pat Gelsinger and Jim Brayton show off a wafer built with 2008’s 45nm “Nehalem” design.

Alphabet Soup

Penryn, Harpertown Wolfdale DP, Yorkfield, Wolfdale—-who wouldn’t mix up that long list of names? But what’s the alternative? When your customer comes in and asks if you have the new Xeon, you don’t want to have to ask which one. Confusing though they may be, code names are a great way to keep the latest technologies straight without mixing up product families or models.

Take the time to know what tomorrow’s Xeons do for your whitebox servers beyond the ones you sell today, why your business customer wants 6MB of cache in his dual-core workstations, and explain that the notebooks he buys at the beginning of 2008 will last longer on the same battery. The small business owner might not care about 45nm manufacturing, but he does want better database performance, a lower energy bill, and an office that doesn’t sound like a construction site. Intel’s Penryn processors are a means to those ends, and you, the reseller, can help your customer pick the right processor with the right technologies for each job.
 
         
    Back to top    
   
Copyright © 2007 RAM Magazine. All rights reserved.
Do not duplicate or redistribute in any form.