Tuesday, March 17, 2009

GSI Technology Is Synonymous With Profits

Today's slight increase in the market is what we expected for the entire month. Get a little gain, give a little of the gain back. While the folks at tech:stocker would have liked a little more strength in today's market up tick, small gains are still better than losses. We were also happy to see the management changes at Fannie Mae (FNM:NYSE) and Freddie Mac (FRE:NYSE). Hopefully this will give investors a little more confidence in the financial services industry as these giants try to steady themselves after their recent fall. Until the financial stocks like AIG (AIG:NYSE), Washington Mutual (WM:NYSE), and Merrill Lynch (MER:NYSE) prove they can return to their profitable ways, investors are going to show some restraint towards many companies in this industry.

Today's stock grabbing our attention, GSI Technology (GSIT:NASDAQ) shouldn't worry potential investors as it has many similar attributes of tech companies that we have recently discussed. Devoting any time discussing the spartan Web site, minimalist press releases, and perfunctory management team would only detract from what this company does extremely well: posting profits.

According to their latest earnings release, GSI is on a profit roll that many other tech companies can only dream about. For 19 consecutive quarters, GSI has delivered positive earnings. The last time they failed to deliver profits for their shareholders, Red Sox fans were cursing Aaron Boone and wondering if the curse of the Bambino would ever be broken.

As a manufacturer of SRAM, GSI is a one-trick pony. If SRAM were to be replaced by a better technology, GSI's profitability streak would come to screeching halt. We're not too worried about this in the immediate future as companies like Cisco Systems (CSCO:NASDAQ) are increasing their orders of GSI's SRAM. Because Cisco is one of our favorite picks in part because we think they will do well in the third and fourth quarter of this year unless the economy hits a brick wall. Even if that happens, we doubt it will affect their spending with GSI substantially.

Over the next 12 months, we expect GSI to climb to 5 giving investors a nice little profit as the company continues their profitable quarter streak.. As with any micro cap, we do recommend keeping a close eye on this investment but there's no need to add GSI to your iPhone along with the rest of your financial stocks.

The EEMBC® consortium is the only independent source for certified embedded compiler performance comparisons

EEMBC, the Embedded Microprocessor Benchmark Consortium, develops and certifies real-world benchmarks and benchmark scores to help designers select the right embedded processors for their systems. Every processor submitted for EEMBC® benchmarking is tested for up to 46 different parameters, each representing a different workload and capability in telecom, networking, consumer, office automation, and automotive industrial applications. With members including more than 40 of the world's leading semiconductor, intellectual property, and compiler companies, EEMBC establishes benchmark standards and provides certified benchmarking results through the EEMBC Certification Labs (ECL) in Texas and California." -www.eembc.com

Green Hills Optimizing Compilers Lead the Market
Green Hills Software has been the leader in embedded optimizing compiler technology for the last twenty-five years. Using the world's most advanced compiler technology, Green Hills Software has dominated the EEMBC benchmarks from their introduction in 1999. We implement all of the most advanced optimization techniques used by competing compilers. But in addition, our optimizing compilers offer hundreds of unique optimizations that we have developed as a result of twenty-five years of research into how to extract the highest possible performance and the smallest possible memory size from thousands of embedded programs.

Using the Green Hills optimizing compilers included with the MULTI software development environment most programs can be made to run at least 20% faster than if compiled with a "free" GNU compiler. Green Hills optimizing compilers consistently beat all other compilers on the EEMBC benchmarks.
embedded compilers, fastest compilers, optimizing embedded fast c compilers, compilers, dualcore, Global Optimizer, Target-Specific Optimizer, Code Generator
March 21, 2007:
EEMBC Adopts DoubleCheck™ For Its Industry-Standard Processor Benchmarks - Green Hills Software’s Static Analysis Tool Increases Code Quality
“DoubleCheck has performed as well as other significantly more expensive source code analyzer products on the market,” commented Shay Gal-On, chief software engineer, EEMBC. “We were impressed with DoubleCheck’s ability to detect a wide range of software flaws, its low false positive rate, its execution time performance, and the intuitive format and usability of its reports.” (click here to read more)
July 11, 2005:
EEMBC Announces Benchmark Scores for AMCC’s PowerPC® 440EP Processor
"When you combine the performance of AMCC's 440EP processor with the efficiency of Green Hills compilers, the results are impressive," said Markus Levy, EEMBC president. "These scores will provide designers with additional objective data to help in selecting the optimal processor for their given application." (Click here to read more)
April 26, 2004:
Freescale Semiconductor Selects Green Hills Software’s Compilers to Certify Their Highest PowerPC Scores Ever
Green Hills compilers surpassed competitors' compilers on benchmarks conducted by IBMR and certified by EEMBCR Certification LabsT (ECL). The certified scores show Green Hills Software's C/C++ compilers lead the pack on 14 out of 15 performance benchmarks, some by 20%, for the IBM PowerPCR 405GPr and 440GP processors.

Green Hills Software's compilers produced faster code than compilers from GNU, Wind River Systems (Diab) and Metaware. Moreover, these "out of the box" scores were produced by production-level Green Hills products shipping today to customers.

"With Green Hills Software's compilers, our PowerPC 440GP processor exceeded all other System-on-Chip processors on four of the five EEMBC benchmark suites," said Kalpesh Gala, PowerPC strategic marketing Manager at IBM Microelectronics. "EEMBC benchmarks are based on real-world code that indicates how our PowerPC 405GPr and 440GP processors work in our customers' applications. Add these to the latest results on our 405GPr processor, and Green Hills Software's compilers produced the best results on 14 out of 15 of our EEMBC benchmarks."
Freescale Semiconductor chose the Green Hills C/C++ compilers to achieve record-setting performance with the PowerPC MPC7447A processor families on benchmarks. With the publication of these results certified by EEMBC Certification Laboratories, the Green Hills C/C++ compilers gain the distinction of producing the highest out-of-box scores in every EEMBC category. (Click here to read more)
Feb 18, 2003:
Green Hills Software's PowerPC Compiler Performance Dominates EEMBC Field


What is EEMBC?

EEMBC (Embedded Microprocessor Benchmark Consortium, pronounced “embassy”) is a non-profit consortium, funded by over 40 member companies. The goal of EEMBC is to provide independently certified benchmark scores relevant to deeply embedded processor applications. Members of the consortium include: ARM, AMD, IBM, Intel, LSI Logic, MIPS, Motorola, National Semi, NEC, TI, Toshiba and more. Tensilica is a member of EEMBC.

What are the EEMBC Benchmarks?

The are five different EEMBC benchmark suites:

• Consumer
• Networking
• Telecom
• Office Automation
• Automotive

Each suite is comprised of a range (five to sixteen) of individual benchmark tasks representative of that product category. Example: the consumer benchmark suite includes algorithms for image compression, image filtering, color conversion, and other tasks common to consumer digital imaging products. Scores for EEMBC are as an aggregated geometric mean score normalized to other processors – known as the ConsumerMARK, TeleMARK and NetMARK scores. Individual benchmark test reports detail the number of iterations of a benchmark routine in 1 second of operating time.

Two Metrics: Out-of-box -vs- Optimized

EEMBC supports two kinds of performance scores for certified results. The Out-of-the Box score represents performance results from taking the EEMBC C code “as is”, compiling, and measuring the performance. No code optimization or assembly coding allowed. Optimized – or “Full-Fury”– results allow the use of C coding changes and assembly coding to accelerate performance.

Optimizing Conventional Processors

For conventional embedded processors the only good avenue available for optimization is to laboriously hand-tune assembly code — not a task for the faint of heart, or the programmer unfamiliar with the micro-architecture of the processor. This is, in essence, changing the code to fit the processor.

Optimizing with Xtensa Cores

To get optimal performance in an application using the Xtensa processor, you don’t start by changing your application code — you begin by changing the Xtensa processor to fit the application. Using the same development path used by our customers, Tensilica used the Xtensa processor generator, the code profiling tools, and the TIE Compiler to create three optimal Xtensa configurations – one each for Consumer, Telecom and Network – to ideally suit the EEMBC benchmarks. Tensilica has certified EEMBC results for the Xtensa processor for the Consumer, Telecom and Networking benchmarks in both optimized and out-of-the-box categories, and out-of-the-box results for the Office Automation suite.

Illustrations conceptual: see EEMBC report for more details. Clock rates: post-layout extracted data, worst case conditions, 0.13 micron

More About EEMBC

EEMBC, the Embedded Microprocessor Benchmark Consortium, develops and certifies real-world benchmarks and benchmark scores to help designers select the right embedded processors for their systems. Every processor submitted for EEMBC® benchmarking is tested for parameters representing different workloads and capabilities in communications, networking, consumer, office automation, automotive/industrial, embedded Java, and microcontroller-related applications. With members including leading semiconductor, intellectual property, and compiler companies, EEMBC establishes benchmark standards and provides certified benchmarking results through the EEMBC Certification Labs (ECL) in Texas and California.

EEMBC is a registered trademark of the Embedded Microprocessor Benchmark Consortium.

Intel's Net Plunges as Demand Dries Up

Intel Corp.'s fourth-quarter profit plunged 90% and the semiconductor giant predicted even weaker conditions ahead, the latest evidence that the recession is rocking all segments of the high-tech industry.

Intel's report, which serves as a leading indicator of consumer and business spending on technology, kicks off what investors expect to be a parade of dismal technology earnings reports.

Slumping demand and shrinking profits, which triggered layoffs last year at struggling Silicon Valley companies such as Yahoo Inc., Applied Materials Inc. and Sun Microsystems Inc., have recently spread to companies with strong positions in healthier markets.

Big tech companies, including Oracle Corp., SAP AG and Google Inc., in the past week have made rare trims to their work forces. Microsoft Corp., meanwhile, is considering significant job cuts that could be announced when it reports results next week.

[tech troubles]European Press Agency (Intel); Dell, Microsoft, Nokia

Intel said it is suffering from a rapid deceleration in computer demand and moves by PC makers to cut inventories in anticipation of a weak 2009. Intel's quarterly profit was also hurt by a $1 billion writeoff in the value of its investment in Clearwire Corp., a provider of wireless service.

Intel, which twice lowered its fourth-quarter sales forecast, reported quarterly sales fell 23% from a year earlier to $8.23 billion. Citing uncertain economic conditions, it declined to issue a revenue forecast for the current first quarter, though it disclosed a figure for internal planning purposes that is 28% lower than revenue it reported for the year-earlier period.

Paul Otellini, Intel's chief executive officer, said in a conference call that the fourth quarter marked only the second time in 20 years that Intel's revenue was lower than in the third period.

Despite the gloomy outlook, he indicated that Intel's past expense cuts should be sufficient and said it will not slow spending on research and development. "We've always believed that the best way to successfully emerge from recessions is with tomorrow's products, not by standing still with today's," he said.

Bill Kreher, an analyst at Edward Jones, said Intel's unwillingness to provide a revenue range for the current period underscores the fact that even the largest companies have a hard time forecasting in the current environment. "It's a very uncertain time right now, and visibility is extremely limited," he said.

Demand for personal computers was healthy through the third quarter of this year. But the picture changed dramatically in the typically strong fourth period, as negative economic news caused consumers and companies to slow their purchases. Research firm IDC on Wednesday said unit shipments declined 0.4%, the first year-over-year decline in shipments in six years.

Intel, which supplies roughly four-fifths of the chips that serve as calculating engines in PCs, experienced a particularly sharp reversal in its fortunes. While consumers have pulled back on computer purchases, PC makers have pulled back even more sharply on their purchases from Intel, holding off purchases for new systems they might build in early 2009.

Many tech companies had problems well before the economic slump took hold. Makers of memory chips, in particular, have seen pricing plunge because of competitive forces that led them to add too much production capacity and make too many chips. Now demand has softened, as people slow purchases of computers, cellphones and other products that use the components.

Micron Technology Inc., one of the biggest makers of chips known as dynamic random-access memory, or DRAM, last month reported a $706 million loss for its fiscal quarter ended Dec. 4. The company said average selling prices for its DRAM chips fell 24% in the period, while prices on other popular chips known as flash memory fell 34%.

"A downturn in the general economy is sufficient to cause a downturn in the semiconductor industry, but the semiconductor industry has at least as many downturns that are caused by overcapacity and excess inventory situations," said Walden Rhines, chief executive officer of Mentor Graphics Corp., which makes software used in designing chips.

The story is similar in disk drives, where a glut of the devices was already squeezing profits before demand problems began adding to the industry's woes this fall. Seagate Technology Inc., the biggest disk drive maker by sales, this week replaced its CEO, cut management salaries and said it would lay off 2,950 employees, or about 6% of its work force.

Intel said Thursday that its closely watched gross profit margin -- which it forecast in October at 59% in the fourth quarter -- hit 53.1% in the period. More surprisingly, the company said margins are likely to descend to percentages in the "low 40s," largely because of charges it takes when its factories are not full and costs associated with moving to a new production process.

One bright spot for the company in the quarter was Atom, a new chip for low-end portable PCs called netbooks. The microprocessor, and accessory chips that work with it, generated $300 million in revenue in the fourth period, up 30% from the third period, said Intel Chief Financial Officer Stacy Smith. Atom has a much lower price than other Intel chips; excluding the product, the company's average selling prices were up.

But Leslie Fiering, a research vice president at Gartner Inc., said Atom and the netbook craze are causing buyers to look for lower-priced systems. "It's putting downward pressure on all PC prices across the board," she said, a trend that tends to put pressure on Intel's prices.

Intel reported that net income for the period ended Dec. 27 declined to $234 million, or four cents a share, from $2.27 billion, or 38 cents, in the year-earlier period. The company's stock at 4 p.m. was up 21 cents on Nasdaq at $13.29, and rose to $13.51 in after-hours trading following the announcements.

iPhone “Net Applications” Market Share Still Booming

Confession: We don’t have any idea where Net Applications really gets their numbers from either. Ars Technica says they:

Collect [web metrics] data from an “exclusive on-demand network of live stats customers” compiled from some “160 million visitors per month”

But they could just as soon be reading runes or casting the bones. However they get them, if we can assume it’s a consistent measure, then iPhone is still on the rise, going from just under 0.2 in July to hitting 0.3 in August. Doesn’t sound particularly big? Well, it confirms the numbers Casey cited last week, and its out of all operating systems everywhere, and it’s for tiny little MobileSafari.

In other words, in the very big pond, our tiny 3G fish is putting on some serious weight.

January 2009 Search Engine Market Share | Net Applications

All engines show minor losses in market share except for MSN which actually showed a slight gain for January. ON the other hand, every engine but Google posted lower market share numbers from this time last year except Google which had a little over a two point growth.

Net Applications Search Engine Market Share January 2009

January 2009: 54.31%
December: 55.85%
Change: -1.54

January 2008: 52.14%
Change: +2.17

January 2009: 8.76%
December: 9.02%
Change: -0.26

January 2008: 10.94%
Change: -2.18

January 2009: 4.47%
December: 4.43%
Change: +0.04

January 2008: 5.78%
Change: -1.31

January 2009: 1.69%
December: 1.74%
Change: -0.05

January 2008: 1.86%
Change: -0.17

January 2009: .74%
December: .77%
Change: -0.03

January 2008: 1.09%
Change: -0.36

Data scoring techniques tend to change over time making past data inaccurate. As always, information here is for entertainment purposes only.

Creating resource Pools

VMKernel manages all memory, except the memory that is allocated to Service console. VM will only power on when Server is sure to allocate that VM’s it’s reservations. This is calculated by Server by checking how much unreserved resources are available and if that meets the reservation of the VM.

Let’s learn how to create resource pools.Resource pool can be created on individual hosts, only if hosts are not clustered. In order to create resource pool, you need to have following information in hand.

  1. Name: Name of the resource pool
  2. Shares: Number of shares to be allocated to every VM
  3. Reservation: Minimum resources guranteed to VM’s
  4. Expandable reservation: There are two options YES/NO, if say YES, if reservation of existing resource pools are not available they can be used from Parent resource pool, If select NO, then host won;t be able to power ON the VM.
  5. Limit: Maximum resources any VM, under this resource pool would get.

Similiar option you would get for memory configuration. Once this step is completed you might get yellow triangle, which suggest something is incorrect.

Now resource pool is created, let’s play around with it.

Select resource pool(left hand-side), select on summary tab (right-side), you see below mentioned screen

I’ve underline things which could be of worth observation when you see them in Virtual center.

Now select resource allocation, resource allocation gives what reservation have been used, what is available, In order understand this clearly, let take look at picture below

You can see currently there are no reservation defined in this resource pool, but you still see CPU/Memory reservation used,

because VM-Marketing has been allocated reservation which is consumed from total reservation available

Other important information you get from here is the unreserved status, this unreserved status helps you in allocating resources in future. You also see resources are set to unlimited field, there is custom shares defined for particular VM.

Of course it is true, you can certainly change values under resource pools by editing it after you create resource pools

Lets discuss another important topic, How to add VM to existing Resource pool or move vm to another resource pool. You can vm to existing pools in two ways

  • When you create VM, you can assign VM to any specific resource pool
  • You can also drag and drop VM to any resource pool even after it is created

NB: you don;t need to power off VM in order to just move it to different resource pool.

When you move drag & drop VM to resource pool, its Memory/CPU/Shares/Limit don’t change, this change is reflected only when the VM is powered ON, in power off state resource pool’s resources are not affected by it or not being allocated to this new VM. Though you will observe in resource allocation tab Reservations for both CPU/Memory will reflect the changes in value. Also a point to be noted if Resource pool is not in a position to allocate the reservation for this VM, it generate error and move will fail.

Sunday, March 15, 2009

Intel to deliver dual-core Atom chip next month

SAN FRANCISCO--Intel plans to bring its first dual-core Atom to market next month, it was revealed here Monday during the Intel Developer Forum. The chipmaker also disclosed more details of the Nehalem processor.

The power-efficient processor will be targeted at Atom-based desktops called nettops. Currently, Intel offers the Atom N230 processor for nettops. This chip has a slightly higher power envelope than the Atom processors built exclusively for mobile devices.

That news was revealed to this reporter by an Intel employee as senior vice president Pat Gelsinger was delivering his IDF keynote, which included more specifics about Nehalem, the family of chips the company plans to begin rolling out in the fourth quarter. Gelsinger, the general manager of Intel's Digital Enterprise Group, showed the first wafer holding individual eight-core processors, detailed the power-saving features of the Nehalem processors, and confirmed future mobile Nehalem processors.

Intel Nehalem processor lineup as shown at IDF 2008

Intel Nehalem processor lineup as shown at IDF 2008

(Credit: Intel)

Also due in September is the six-core Dunnington server processor, the final member of Intel's 45-nanometer "Penryn" family, which will ship to customers next month, Gelsinger said.

Most of his keynote centered on Nehalem, and one of the features Intel was pushing hard at IDF was a technology called Turbo mode.

Turbo mode is essentially a switch that turns off unused processor cores and then uses the remaining active cores more efficiently. This kind of sophisticated power-management technology will be used in both Nehalem-based laptops and servers, according to Gelsinger, and will become increasingly necessary as Intel brings out chips with more cores like the eight-core Nehalem processor due next year.

In short, in multi-core processors, cores not doing much can still use power. So, it's better to use, for example, a couple of cores more efficiently than four cores inefficiently.

The power saving technology is enabled by "an integrated microcontroller which only works on power management," said Rajesh Kumar, an Intel Fellow, who spoke during Gelsinger's keynote. There are about 1 million transistors dedicated solely to power management, Kumar said.

The feature "requires no operating system intervention. It is fully detected and managed by the hardware. If it has detected an idle core, it is able to reallocate that power budget to the other cores," Gelsinger said in an interview after his keynote.

On another front, Intel showed the first eight-core Nehalem chip. "This is the first showing of the eight-core Nehalem-EX," Gelsinger said in his keynote. He said the chip is a monolithic design, meaning that all eight cores are on one piece of silicon.

Nehalem-EP, or Nehalem Efficient Performance, will be a quad-core chip for mainstream servers and workstations. What Intel traditionally calls two-socket servers, Gelsinger said.

The mainstream desktop will be the Core i7. "With the i7 we have high-end desktop and extreme," Gelsinger said. The extreme edition is for overclockers, he said. Enthusiast gamers often overclock processors (ratcheting up clock speed beyond the rated speed) to gain extra performance.

"Turbo Mode" is a linchpin Nehalem technology

"Turbo Mode" is a linchpin Nehalem technology

(Credit: Intel)

But there will be more pedestrian dual-core versions of Nehalem too. "There will be versions for the desktop that will be dual-core as well," Gelsinger said.

Gelsinger also talked about Intel's plans to put graphics directly onto the same piece of silicon as the processor. This will be a first for Intel.

He described why Intel is putting graphics right next to the processor. "There's a big sucking sound near the CPU. It keeps pulling things closer to it. This is uniquely enabled by Moore's Law...and as things get closer together I'm able to drive down thermal envelopes (i.e., heat) and decrease physical form factors (i.e., enable smaller computer designs)," Gelsinger said.

Gelsinger broke down the future processor lineup--with graphics on the processor die and without--as follows. "Lynnfield and Clarksfield are the versions without graphics. Havendale and Auburndale are the versions with integrated graphics." (Even Intel executives occasionally get confused by all the code names and it took two tries for Gelsinger to get this right.)

Peer into Intel’s Future: 2008 Roadmaps

We don't often peer into the ether here at PC Perspective to give our readers an outlook very far into the future, simply because that has traditionally not been our style. Taking rumors from other sources and then spouting off about our predictions for various technologies is more the news hounds of the world rather than the actual hardware reviewers. Besides, most of the information we get from Intel, AMD and NVIDIA is protected under those damn NDAs.

Every once in a while though some data comes across our desk that isn't from the companies directly and we like to share that whenever we can. A large amount of Intel roadmaps and documentation was recently found that gives us a clearer view of how the chip leader is going to execute for the remainder of 2007 and well into 2008.

Intel CPU Releases

This below diagram details the processor releases from Intel start in Q4 2004 (!!) all the way up to the end of 2008.

Click to Enlarge

It is really interesting to see a layout of all the CPUs over this long of a period, but we are going to focus on the end of 2007 and 2008 years are they are displayed. In the high end market, which shows up as "Extreme" in the diagram, after the actual release of the QX9770 processor, which we recently previewed, there are no more releases scheduled until Bloomfield arrives. In the DP (dual processor) segment we have the QX9775 coming which will use the LGA771 socket rather than the LGA775 and be used pretty much exclusively for the super-mega-awesome Skulltrail system.

You can see as well that the additional Yorkfield processors are going to arrive starting in Q1 of next year, the most appealing of which might be the Q9450 that we simulated in our recent AMD Phenom review. There will be a Q9300 that uses the 1066 MHz FSB and a Q9550 on the 1333 MHz FSB to surround the Q9450 and even a pair of dual-core Yorkfield CPU will be released in Q1 at 3.0 GHz and 3.16 GHz clock speeds.

Bloomfield is listed in Q4 of 2008 and is the codename for the Nehalem processors that have talked about in quite a bit detail in the past.

Click to Enlarge

This large table lists all the upcoming and current Intel processors and the various features and specifications for them. Had we posted this up a few days ago, the news of seeing a 1600 MHz FSB part for the single processor market would have been a surprise. The Core 2 Duo, Quad and Extreme specs are all what we would expect, but the Bloomfield and Gainestown parts are new -- and in fact do not have model names associated with them yet.

Bloomfield and Gainestown are Nehalem processor names with the later referring to upcoming dual processor configuration models. Interestingly, even though we were told about 8-core Nehalem parts at this year's IDF they are only showing as quad-core parts on this sheet. Looks like four cores is going to be the top level through 2008. The on-die cache is sized at 8MB, less than the 12MB currently on the Yorkfield cores, but there is no level attached to them (L1, L2, L3) indicating that Nehalem might have a new memory structure that Intel hasn't decided how to market.

Instead of a front-side bus speed, the Bloomfield and Gainestown have QPI links -- quick path interconnects -- Intel's answer to AMD's HyperTransport technology. Gainestown will have two QPI interconnects for dual processor configurations while the Bloomfield will have just a single connection for communicating with the chipset.

As we had expected, Nehalem will have an integrated DDR3-1333 memory controller, again similar to AMD's Athlon 64 cores and beyond, and will use a new LGA1366 package. That's a full 75% more pins than current Core 2 products indicating a lot of bandwidth for the QPI connections. The die of Nehalem will be monolithic (AMD will lose its "only true quad core" campaign then) and will be based on the very robust 45nm process technology that the Yorkfield cores use. The TDP of the top parts is listed at 130 watts: quite a bit higher than current Core 2 Extreme CPUs.

Click to Enlarge

In regards to the die size of the upcoming Nehalem core, this diagram attempts to shed some more light on the subject by comparing die shots of Penryn and Nehalem. Since both are based on the 45nm process, the die shots can be compared to make guesses about transistor counts - following the math shows that each new Nehalem core could be around 30M transistors with the whole monolithic die coming it at 730M transistors or so. Current quad-core Yorkfield CPUs are composed of two 410M transistor dies for a total of 810M transistors making Nehalem SMALLER in terms of actual wafer area.

Click to Enlarge

This diagram is estimated CPU shipments for Intel's desktop products based on a percentage of total CPUs shipped. Yorkfield quad-core parts are only supposed to gain market share of the desktop CPUs up to 85; that's somewhat less than I expected though the dual-core Wolfdale (45nm shrink of Penryn, half of Yorkfield) will expand quickly. Also note that on this graph, Intel is predicting the death of the NetBurst architecture by the end of 2007!

Intel Penryn: first 45-nm processors

High-k dielectrics, metal gates and other novelties of the 45-nm process technology

Intel Penryn. 45 nm SRAM chip

Intel Penryn. 45 nm transistor
45 nm transistor

Introducing new process technologies, Intel is keeping pauses for about a year between the first demonstration of chips of new process technologies (normally, SRAM chips) and the demonstration of the first CPU specimens. That was the last time (65 nm) and this has happened again now.

Intel Penryn

In a year after demonstration of the first 45-nm SRAM memory wafers, there have been demonstrated the first Penryn CPUs, with more details on the used materials and technologies provided. First of all, we should note that transition to new more precise process technologies entails the need for solving the issues of reducing the leakage currents in transistors, which directly affects the power consumption and heat emission of the chip in general, with all that being implemented under increased performance and complication of the CPU architecture and topology.

Intel Penryn

While moving to the 45-nm process technology to create transistor gates of low leakage currents, engineers at Intel had to use a new material for the dielectric - the so-called high-k dielectric combined with a new material of the transistor gate electrode made as a combination of metals. The thing is, the silicon dioxide (SiO2, traditionally used as the dielectric to create transistor gates for forty years, has simply run out of steam. In developing the previous 65-nm process technology, engineers at Intel succeeded in creation of a layer of dielectric made of silicon dioxide with alloy additions as thick as 1.2 nm – or as many as five atom layers!

Intel Penryn

Intel Penryn

However, further reduction in the thickness of the layer results in the direct tunneling effect and a sharp rise of leakage current through the material of the gate dielectric – in fact, silicon dioxide stops being an obstacle for a free electron drift which under these conditions demonstrate the properties of not only particles but waves, so there is no possibility of guaranteed control over the state of such a transistor.

A different type of dielectric helped solve this critical issue. To this end, silicon dioxide was replaced with a thin layer of material on the base of hafnium salts (rare-earth metals) having a high level of dielectric permeability k (high-k), which helped to reduce the leakage current by almost ten times as compared to the traditional silicon dioxide.

Intel Penryn

However, not all is as smooth as it looks. It's all right with physics, but chemistry of the new miraculous gate of high-k dielectric turned out to be incompatible to silicon traditionally used for the manufacture of gate electrodes, so the issue proved to be another nut to crack for Intel engineers when moving towards the 45-nm production process. There was a need to develop a new metal gate compatible to the new dielectric. Years were spent so as not only to find the suitable material for gate electrodes, but also to develop technologies of its use for gates of various conductivity - NMOS and PMOS.

Intel Penryn

Intel Penryn

By the way, we should mention that the number of materials and chemical elements involved ion the production of modern chips is rising enormously. There used to be no more than ten to twenty quite recently, but now they count over half the Mendeleev periodic table!

Intel Penryn

Intel Penryn

Intel's 45-nm process technology is codenamed P1266, and the lithography used for the manufacture of Penryn is the same as that in the 65-nm process technology. Despite the new design of photo resists and new generation of photo masks, use of the same 193 nm lithographic tools allowed for substantial reduction of costs while moving to the 45-nm process technology.

Intel Penryn

Intel Penryn

Intel's new 45-nm process technology implies smaller size of transistors with more dense positioning of these transistors on the wafer – almost twice as dense than in the previous 65 nm generation. The reduced dimensions of transistors have resulted in 30% drop of the current required for their switching, the switching speed of transistors has gone up by over 20%, and leakage currents in the "source-drain" channel have gone down by more than 5 times. Also, the leakage currents in the gate dielectric have dropped by over ten times. Some specialists call the introduction of high-k dielectrics and metal materials when creating the gate electrode a more complex and effective task than transition to a new precision process technology. It is also interesting to note that the next Intel's process technology P1268, 32-nm, is also quite likely to aim at using 193 nm lithographic tools.

Intel Penryn

Currently, Intel's new 45 process technology using 300 mm wafers is being introduced at Hillsborough, Oregon, at D1D factory. Among the most forthcoming plans of the company is the launch of 45-nm process at the new 300-mm Fab 32 in Ocotillo, Arizona (for mass production to be started in 2007) and Fab 28 in Israel (start of the mass production in the first half of 2008).

Intel Penryn. Fab 32 in Ocotillo, Arizona
Fab 32 in Ocotillo, Arizona, under construction

Intel's scientific research center in Hillsborough, Oregon, D1D
Intel's scientific research center in Hillsborough, Oregon, D1D

Practical implementation: Penryn in our motherboards

Despite the details of announcement, some details of the architecture and practical implementation of Penryn processors are not yet sufficient. For instance, we are in for telling about the new SSE4 instruction set which was first announced at the Fall IDF 2006.

For now it is still not clear which FSB speeds will be supported by the new processors, what clock speeds, and what motherboards will support them. There is information that the Penryn family of chips for desktop PCs will be originally presented in the familiar Socket 775, so it is quite possible that to make use of the existing motherboards it would suffice to update the BIOS, however, this info is still to be verified.

Lastly, closer to the mid-2007 Intel is about to plan presentation of new platforms with support for the new-generation DDR3 memory, and of course Penryn processors in these platforms are to play the first fiddle. I am absolutely sure that in the most immediate future we are going to know more and more details of the new design of Penryn processors once more and more briefings by Intel are held, and it is quite probable that the major "stratum" is still to be announced during the forthcoming IDFs in Taipei and Beijing.

For today, that is about all the info on new processors, so we only bring in a photo of Penryn chips taken in the premises of Fab28 in Kiryat Gat, Israel, currently under construction (follow the link to an article in Hebrew. You are unlikely to read much without knowledge of the language but at the end of the page there is a comical movie on the construction of the factory). For today, we are closing the topic of Penryn chips and promise to come back to it once more details appear.

Intel Penryn

Intel Penryn

Intel Penryn: first 45-nm processors

Intel Penryn

For the first time in the history of semiconductor industry, Intel today has officially introduced processors made following the 45-nm process technology. Merely a year has passed since Intel's demonstration of the first engineering samples of 45-nm SRAM memory, and now we see a new milestone to come: processors of the working name Penryn which comprise hundreds of millions of transistor gates made following the complex and "delicate" 45-nm process technology where the width of each transistor gate is as little as 35 nm.

Following the tradition, the announcement of CPU prototypes means that the production lines are in theory ready for wholesale manufacture - what is left is only a few verification arrangements and polishing of the equipment to the optimum chip yield. Afterwards, later to the summer a production of next-generation multicore Intel Core 2 Duo, Intel Core 2 Quad and Xeon processors will start, and their mass deliveries are planned for the second half of 2007.

Photo of 45-nm CPU Intel Penryn
Photo of 45-nm CPU Intel Penryn

Intel's roadmap of CPU micro architectures and CPU designs

Today's even should be viewed from all the sides at a time due to a number of reasons. First, the 45-nm process technology was used in practice for the release of so complicated products. There have been of course serious modifications to the production technology as well as entirely new tools and materials at almost all the stages.

Secondly, transition to the so highly precise process technology with the evident change in physical and chemical properties of the circuit components called for use of absolutely new materials and technologies to handle the substrates, creation of transistor junctions, their electrodes etc. – whatever to attain lower leakage currents and more economical operation of the chip without the drop of performance per watt of energy spent.

Thirdly, the Penryn architecture of the CPU core has changed although they still are representatives of the Core (Merom) micro architecture. To avoid the mess of terms, we should recall that generations of CPU micro architectures change one another less frequently than various designs of cores (and even less frequently than new process technologies). For example, the NetBurst micro architecture which in 2000 came to replace the P6 micro architecture (the latest representatives of P6 are 0.18 mk Coppermine and the final 0.13 nm Tualatin in 2001) was in different times represented by a number of different core designs: Willamette (2000, 0.18 mk, 0.13 mk), Northwood (2002, 0.13 mk), Prescott/Smithfield (2004, 90 nm) and Cedar Mill/Presler (2006, 65 nm) – four generations of core design and six years on the same micro architecture!

At the same time, the Banias micro architecture presented in 2003 and which was progressing in parallel also contained a few generations of core architecture generations – Banias (2003, 0.13 mk), Dothan (2004, 90 nm), and Yonah (2006, 65 nm).

In fact, the emergence of new CPU core designs can be regarded as a gradual process of improvement within the same micro architecture with the addition of new features, redesign and scaling to a new process technology. At the same time, it turns out that Intel, due to two master scientific research centers working at the processor development – in Oregon and Israel, is developing and improving at least two micro architectures in parallel thus making use of the best practices of colleagues and getting rid of repeated errors.

The modern Core architecture which came in 2006 to replace the NetBurst and Banias, along with absolutely new engineering ideas and technologies, has inherited the best of the predecessors' experience. As a result, the very first design of the core having the working name Merom (2006, 65 nm) based on the Core micro architecture proved really successful in terms of performance and efficiency.

Today's novelty, the design of core of the working name Penryn is nothing more than the improved design of Merom with a simultaneous transition to the 45-nm process technology. The family of 45-nm Penryn chips will be the framework for new Core dual-core Wolfdale chips and 4-core Yorkfield chips which come to replace today's 65-nm Core 2 Duo and Core 2 Quad, respectively. Therefore, the processors of Penryn family will underlie the new generation of the mobile platform Intel Centrino codenamed Montevina. The new mobile platform Montevina which is expected in the first half of 2008 will replace the Santa Rosa version.

Intel Penryn

To sort out with the micro architectures of Intel processors and CPU core designs released on their base, let's look into the future. It is expected that the newly presented design of Penryn along with the Core micro architecture will survive for quite a long time – perhaps until 2010, and it is quite possible the processors with the Penryn core will be produced following the next, 32-nm, process technology.

Intel Penryn

For now, we are not talking about the new versions of core designs based on the Core micro architecture, since next year (2008) we expect the announcement of an entirely new micro architecture of Intel's processors – Nehalem, based on which 45-nm processors of the same name Nehalem will be released closer to year 2009 , as well as Nehalem-C chips of the modified design. Further forecasts are still vague, but if everything goes as planned, then by 2010 we will see the first prototypes of chips built on the absolutely new micro architecture Gesher, and we only have to guess which process technology – 32 nm or 25 nm - will be used.

Penryn CPU architectural specifics

Anyway, let's come back to Penryn processors based on the Core micro architecture: according to the official statement of Intel's representatives, the company is currently offering five products of earlier steppings aimed at various sectors of the market- over 15 chips planned for the forthcoming release following the 45-nm process technology. According to the company's statements, Penryn chips are already under trial runs in systems powered by Windows Vista, Mac OS X, Windows XP, and Linux.

As per the practical experience (or maybe the tradition?) acquired during the release of first chips of the Core micro architecture, the Penryn family will include desktop and mobile PCs, workstations and corporate systems.

We'll be considering the technology changes that touched upon the manufacturing process and materials used later, but for now let's list the innovations implemented in the Penryn processors. Among them is the increased number of transistors – over 410 mln for the dual-core design (291 mln transistors in 65-nm dual-core Conroe) and over 820 mln for the quad-core Yorkfield, with the chip area reduced to 110 mm2 (from 143 mm2 in Conroe). That's where we see preservation of the Moore's Law according to which the quantity of transistors is doubled once every two years, with the unit cost of manufacture per single transistor going down and the performance going up, and in the near future - I am sure - we can ascertain the doubling of CPU cores on a chip, why not?

Along with that, the Penryn family of CPUs will support for up to 50 new Intel SSE4 instructions aimed at the increase of capabilities and performance of handling the multimedia content. In this regard, it is interesting to note that support for a number of new SSE instructions was announced still for Conroe processors, however at IDF Fall 2006 the support for SSE4 was reserved for the next generation of micro architecture, Nehalem. As is stated in the press release, Penryn will support for the new Intel SSE4 instruction set.

Among the new chips there will be versions of L2 cache size up to 12 MB, and in general the whole family will differ in increased performance and advanced capabilities of controlling the power consumption modes. As to the power consumption of new 45-nm Core 2 Duo, Core 2 Extreme, Core 2 Quad, and Xeon processors based on the Penryn core, it is already known that the TDP will be approximately matching to today's realities – about 35 W in chips for notebooks, about 65 W in the mainstream chips for desktop PCs, about 80 W in 4-core server processors and chips for extreme gamers, under increased performance.

Inside Intel's Harpertown Processor

Intel’s quad-core Xeon 5400-series “Harpertown” processors run at up to 3.2GHz and are based on the new 45-nm Intel Core microarchitecture. The new chips are known for high performance and energy efficiency.

Apple’s previous Mac Pros were equipped with either a) two dual-Core Intel Xeon 5100 “Woodcrest” processors running at 2.0, 2.66, or 3.0GHz, or b) two quad-core Intel Xeon 5300 “Clovertown” processors running 3.0GHz in an “8-core” configuration. More on the whole Xeon family is here.

The biggest benefit over the previous generation Xeons are Harpertown’s 64-bit 1.6GHz dual independent frontside buses (up from 1.33GHz). These buses deliver processor bandwidth up to 25.6GB per second (up from 21.3GB/s). Then there’s 12MB per processor of L2 cache with 6MB shared between pairs of processor cores (up from 8MB per with 4MB shared).

Full benchmarks of the 3.2GHz Harpertown versus previous generation Intel Macs (like the Power Mac G5, Mac Pro Quad-core 2.66GHz and even the previous generation Mac Pro 8-core 3.0GHz) are available on Apple’s Mac Pro performance page.

Inside Intel's Penryn Processor

The next generation transistors, used in the new Penryn processors (Intel Core 2 Duo, Intel Core 2 Quad and Xeon family) feature “new high-k + metal gate material that will allow for higher frequencies, lower resistance, less leakage and lower power consumption.”

Aside from the simple fact that the new 45nm process will allow for double transistor density over current generation processors, Intel also promises that the new microarchitecture will offer a significant performance-per-watt increase over current 65nm technology. Other benefits of the new 45nm process and high-k and metal gate combination will be:

* Significant reduction in transistor switching power
* Improvement in transistor switching speed
* Reduction in source-drain leakage power
* Reduction in gate oxide leakage power
* Greater energy efficiency

According to AppleInsider Penryn-based Core 2 notebook chips boost battery performance by as much as 16 percent and encoding operations by as much 40 percent over Merom processors in mobile benchmark tests.

Penryn is the successor to the fourth-generation Intel (Santa Rosa) Centrino notebook processor currently shipping in the current MacBook and MacBook Pro. Penryn chips have already started shipping PC notebooks from HP, Toshiba and Fujitsu.

The MacBook Pro was last updated on 5 June 2007 with the Santa Rosa processor and new NVIDIA GeForce 8600M GT 250 days ago. The MacBook was last updated just 72 days ago (on 31 October 2007) with Santa Rosa and Intel GMA X3100 video.

While the MacBook is still pretty current the MacBook Pro is due for a refresh. I’d love to see a new quad-core Penryn MBP announced at MWSF - ideally in an entirely new enclosure.

Intel® G31 Express Chipset

Desktop PC platforms with the Intel® G31 Express Chipset, combined with either the Intel® Duo or the Intel® Core™ Quad processor, deliver new technologies and innovating capabilities for consumers.

The Intel® G31 Express Chipset supports Intel's upcoming 45nm processors, enabling a scalable platform for value conscious consumers. With a 1333MHz system bus, DDR2 memory technology and support for Windows Vista* Premium, the Intel G31 Express chipset enables scalability and performance for everyday computing

Features and benefits

1333/1066/800 MHz System Bus Supports the Intel® Core™ Duo and the Intel® Core™ Quad processors
PCI Express* 1.1 Interface The PCI Express 1.1 provides 8GB/s bandwidth for platform graphics.
Intel® Fast Memory Access Updated Graphics Memory Controller Hub (GMCH) backbone architecture that improves system performance by optimizing the use of available memory bandwidth and reducing the latency of the memory accesses.
Dual-Channel DDR2 Memory Support Delivers up to 12.8GB/s (DDR2 800 dual 6.4GB/s) of bandwidth and 4GB memory addressability for faster system responsiveness and support of 64-bit computing.
Intel® Flex Memory Technology Facilitates easier upgrades by allowing different memory sizes to be populated and remain in dual-channel mode.
Intel® Graphics Media Accelerator 3100 3D enhancements enable greater flexibility and scalability and improved realism with support for Microsoft DirectX* 9.0c Shader Model 2.0, OpenGL* 1.4. Intel® Graphics also support the highest levels of the Windows Vista* Aero experience.
Intel® High Definition Audio1 Integrated audio support enables premium digital surround sound and delivers advanced features such as multiple audio streams and jack re-tasking.
Intel® Matrix Storage2 With a second hard drive added, provides quicker access to digital photo, video and data files with RAID 0, 5, and 10, and greater data protection against a hard disk drive failure with RAID 1, 5, and 10.

Intel® Core™2 Processor with Viiv™ Technology

When you want a sensational high-definition (HD) experience, look no further than PCs with Intel® Core™2 processor with Viiv™ technology. Built on the latest hafnium-infused 45nm dual- and quad-core processors, these systems give you performance for HD audio and video and your most demanding applications.

Key components

The cutting edge is now. Every PC with an Intel Core 2 processor with Viiv technology is powered by an Intel Core 2 Duo or Intel Core 2 Quad processor to give you the performance to run demanding applications and manage your HD entertainment. To see the list of all processors available for the Intel Core 2 processor with Viiv technology, see the pc requiremnets

PCs built from the ground up for sensational high-definition experiences. The Intel Core 2 processor with Viiv technology includes the latest Intel chipsets with most everything you need to build a multimedia PC for HD content. Experience stunning sound from movies and music with up to 7.1 surround sound capabilities enabled by Intel® High Definition Audio (Intel® HD Audio). Get full 1080P video playback for movie clips, media streams, and the latest generation of HD video cameras with optional Intel® Clear Video Technology. And explore the Internet with 1GB-capable broadband Ethernet to quickly access high-definition content. The possibilities are endless.

Operating system

Microsoft Vista Home Premium* makes it easy to control and enjoy your digital media.

Features and benefits

For systems based on Intel® Core™2 Duo processor Advanced dual-core performance opens up the world of high definition
  • Get incredible performance from up to 6 MB of cache (a 50 percent increase), higher frequencies, and 1333 MHz FSB
  • Get up to 30 percent faster performance for applications like photo-editing, retouching, and publishing¹
For systems based on Intel® Core™2 Quad processor Extraordinary quad-core performance lets you get the most of HD multimedia
  • Enjoy incredible quad-core performance with 12 MB of cache (a 50 percent increase), higher frequency and 1333 MHz FSB
  • Four processing cores and Intel's hafnium-infused circuitry delivers high-performance for multimedia applications
Intel® HD Boost Improved multimedia performance on 45nm dual and quad core processors when editing and encoding video or working with photos
  • Intel Core 2 Duo processors with Intel HD Boost give you up to 70 percent more performance when encoding high-definition memories from your HD video camera²
Intel® High Definition Audio (Intel® HD Audio) Enjoy superior sound.
  • Experience high-quality audio for stunning sound from movies and music and up to 7.1 surround sound capabilities
Intel® G35/G33 Express Chipset with (optional) Intel® Clear Video Technology Crystal clear high definition.
  • Get full 1080P video playback for movie clips, media streams, and the latest generation of HD video cameras
Intel® PRO Client Wired LAN Broadband ready.
  • Explore the Internet with 1GB-capable broadband Ethernet to quickly access high-definition content
Optional Intel® Matrix Storage Technology Protect your digital memories.
  • Improved storage performance and help protect digital photos, video, and music with automated back-up and restore on systems with optional dual disk drives
Optional surround sound speakers Hear all of your music and movies
  • Take advantage of surround sound with up to eight high-quality speakers with Intel HD Audio
Optional TV tuner card TV on your terms.
  • Watch, record, and pause live TV

Intel® X38 Express Chipset

Desktop PC platforms based on the Intel® X38 Express Chipset, combined with either the Intel® Core™2 Duo, Intel® Core™2 Quad or Intel® Core™2 Extreme processors establish a new standard for performance. Designed with headroom and engineering passion, the Intel X38 Express Chipset supports new dual-channel DDR3 memory technology, next generation dualX16 PCI Express 2.0 and Intel® Extreme Tuning to unleash exceptional performance in today's gaming applications

With the growing imbalance between CPU and memory performance, it is critical to optimize the memory controller design to obtain the maximum possible performance from the memory subsystem. The redesigned Intel X38 Express Chipset Memory Controller Hub (MCH) architecture significantly increases overall system performance through the optimization of available bandwidth with the new 1333 MHz system bus and reduction of memory access latency with Intel® Fast Memory Access. These technology breakthroughs result in optimized system architecture with built-in intelligence, greatly improving system memory performance. The updated MCH also includes support for Intel 45nm next-generation Intel Core 2 processor family and wider internal data buses that support dual-channel DDR3 memory technology at 1333 MHz.

Intel finally launches X38 Express chipset

Intel finally launches X38 Express chipset with ASUS, MSI and Gigabyte mainboards ready.The chipset was previously known by its codename, Bearlake, and first surfaced last year.

Desktop PC platforms based on the Intel X38 Express Chipset, combined with either the Intel Core 2 Duo, Intel Core 2 Quad or Intel Core2 Extreme processors establish a new standard for performance. Designed with headroom and engineering passion, the Intel X38 Express Chipset supports new dual-channel DDR3 memory technology, next generation dualX16 PCI Express 2.0 and Intel® Extreme Tuning to unleash exceptional performance in today’s gaming applications.


Engineers at the time cited problems with vendor microcode support as major contributors to Intel’s decision to slide the date back even further. The launch date for the X38 Express chipset was then changed to October 10th, and today Intel was finally able to hit its launch target.

ASUS, Gigabyte and MSI all have X38 Express-based offerings. ASUS announced today its P5E3 Deluxe, which comes with the option of an embedded install of Linux. MSI launched the X38 Diamond, which features a new Circu-Pipe cooler design; and Gigabyte launched its GA-X38-DQ6 (DDR2) and GA-X38T-DQ6 (DDR3).

Intel® X38 Express Chipset Features

  • 1333/1066/800 MHz System Bus
  • PCI Express* 2.0 Interface
  • Intel® Fast Memory Access
  • Dual-Channel DDR3 memory support
  • Dual-Channel DDR2 memory support
  • Intel® Flex Memory Technology
  • Intel® High Definition Audio²
  • Intel® Matrix Storage Technology³
  • Intel® Rapid Recover Technology
  • Serial ATA* (SATA) 3 Gb/s
  • Intel Qual-Core Extreme QX6800 Announced - Clocked at 2.93GHz!

    Intel on Monday debuted its fastest Core 2 Extreme Quad-Core processor yet, further distancing itself from rival AMD. The 2.93GHz QX6800 will be produced using the company’s 65-nm process and carry a pricetag of $1,199 USD.

    The Santa Clara, Calif. chipmaker says the processor would be ideal for the next-generation of PC games, which are now beginning to take full advantage of what multicore and multithread processors offer. Intel has partnered with several game developers to assist them in rewriting code.

    Some of the first games that will be able to take advantage of the new processor include Crytek’s Crisis, Gas Powered Games’ Supreme Commander, Flagship’s Hellgate London, and Microsoft’s Flight Simulator X with Service Pack 1 applied.

    The QX6800 is not the first quad-core Extreme model from Intel - the first one was shipped back in November and ran at speeds of 2.66GHz. Meanwhile, AMD’s quad-core processor, the Opteron “Barcelona”, is not expected to be available until mid-year.

    Would you run a server using Intel's Atom processor?

    Intel desperately wants its Atom netbook processor inside almost every computing device on the market — smartphones, automobiles, robots, regular desktops and laptops and of course, netbooks.

    But would you run an Intel Atom server?

    In a Pc World article, Intel reveals that it wants to push Atom into everything except the server market, the market that serves as the chip manufacturer’s cash cow, allowing the company to sell its newest, fastest, and (thus) most expensive processors at the highest profit margins.

    Take the Xeon family of server processors, which are available for as little as $200 but carry price tags as high as $3,000 for those intended for four-way or eight-way servers. They may be fast, but they’re power hogs even in idle state, which tests the patience of any efficiency-minded IT pro.

    Consider this: would you build a server around Intel’s Atom CPU, for the low, low price of just $29?

    Just where do you draw the line between cost (upfront and upkeep) and performance?

    Is an Atom server really worth considering?

    According to the article, it is. For one, UK-based Tranquil PC Ltd. is selling Windows Home Servers running Atom; a Chicago hosting provider, SingleHop is leasing dual-core Atom servers to small business customers.

    Is the Atom welcome in the server saloon? Or should it just beat it, and don’t let the door hit you on the way out?

    One analyst says there is potential in the low-end server space for a ‘data pusher’ server that doesn’t need to compute everything, all the time.

    But Microsoft Research is going one step further, and testing Atom for large-scale data centers.

    In the article, an Intel spokesman clarifies the situation:

    “Lots of companies are experimenting with various usages including Microsoft. While some may experiment with servers, the current Atom is not the right fit for these opportunities. However, as we enter the many-core era and more devices and machines add Internet access, our Xeon, Core and Atom opportunities will be almost endless.”

    But with a global economic downturn, the Atom’s low power consumption, fast sleep/wake features and low price take it from being the butt of every joke to a serious consideration.

    According to the article, Microsoft Research is developing software that can put Atom servers to sleep — a particular power-saving strength of the Atom — waiting to be awakened on demand.

    So: total cost, including energy cost over time, divided by the processing work actually delivered.

    Forget maximum potential performance. Can slow and steady win the race?

    Or are virtualization or rewritten applications better bets, much less avoiding the massive footprint of a veritable army of Atom servers?

    Would you run a server using Intel’s Atom?