by Anton Shilov
07/01/2009 | 07:12 PM
In this editorial we are taking a look back on the top ten news-stories of the first half of 2009. We picked up ten stories that were the most read by our readers and which we consider to be important for the future of the computing industry going forward.
Since we are mostly focused on consumer technology, events like Oracle’s acquisition of Sun Microsystems or SGI takeover by Rackable are not covered in this issue. The happenings in the enterprise world are definitely important, however, this particular editorial is about what our readers read more often and what we consider to be important.
In addition to the facts themselves, we decided to add some comments to the information and tried to outline some trends and end results of the covered events. So, let’s proceed to the top 10 news-stories of the first half of 2009 and begin from the 10th. Remember, if you want the full original story, just click on the links.
ATI, graphics business unit of Advanced Micro Devices, in early June demonstrated the world’s first graphics processing units (GPUs) code-named Evergreen that support the next-generation DirectX 11 application programming interface (API). While AMD’s graphics business unit did not reveal any technical details regarding the graphics chip that was demonstrated, it did say that the novelty is made using 40nm process technology at Taiwan Semiconductor Manufacturing Company. Besides, it demonstrated an actual wafer with DirectX 11 chips.
Perhaps, the news that ATI is the first to demonstrate DirectX 11-compatible graphics processing unit is the most important thing that happened in the first half of 2009. However, we put this event on the 10th position just because it happened a month ago and it was not read by truly massive number of readers.
ATI DirectX 11 technology demo
ATI Technologies was the first company in the industry to unveil DirectX 9.0-compatible ATI R300 graphics processing unit in July, 2002. At that time its arch-rival Nvidia Corp. had a huge market share and, more importantly, mind-share, which meant that consumers did not expect much from ATI Radeon 9700 graphics chip. However, Nvidia was seven months late with its NV30 and the first graphics board on its base – Nvidia GeForce FX 5800 Ultra – appeared to be slower compared to the Radeon 9700 Pro. Even the successor of the graphics processor, the NV35/GeForce 5900, could not outperform ATI Radeon 9700 Pro and eventually 9800 Pro/9800 XT in pixel shader intensive games since the GeForce FX architecture proved to be very inefficient in execution of pixel shaders.
Nvidia has learned its lesson: it was the first company to unveil DirectX 9 shader model 3.0 graphics chip as well as the first with the DirectX 10 graphics processing unit, slowly but surely capturing market share from ATI. But it looks like ATI is back: it was the first with DirectX 10.1 chip and now it can demonstrate DirectX 11 GPU, whereas Nvidia is only trying hard to push its proprietary CUDA, GeForce 3D Vision and PhysX technologies.
ATI DirectX 11 GPU silicon
ATI/AMD openly claims that it will be the only DirectX 11 graphics processors supplier in 2009. Besides, according to the graphics chip developer, the first DirectX 11-supporting games will emerge on the market already in 2009 with Codemasters, Phenomic/Electronics Arts, Emergent and Rebellion already developing titles.
In late May Intel Corp. officially unwrapped its next-generation microprocessor designed for high-end multi-processor servers. The new chip not only boasts with eight processing cores with Hyper-Threading technology, but the new multi-processor (MP) server platform also features functionality traditionally found in the company’s Itanium processor family.
Intel has been offering the fastest processors on different market ever since it introduced its Core 2 micro-architecture into the appropriate segments. The latest achievement of Intel is six-core Intel Xeon “Dunnington” chip that the giant began to ship in September ’08, eight or nine months ahead of competing offering from Advanced Micro Devices.
But the octa-core code-named Beckton (Nehalem-EX) is not exactly a success story. When Intel showcased its code-named Nehalem microprocessor at Intel Developer Forum Fall 2007, the company indicated plans to ship its eight-core Xeon microprocessors in 2008.
“At the largest configuration that we'll ship in 2008, they'll be an eight-core product. Eight core on one die, and each core will have two threads. So, each eight-core die will be supporting 16 threads,” said Paul Otellini, chief executive officer and president of Intel Corp., at that time.
However, this year Intel said it would produce its Xeon MP “Nehalem” processors in late 2009 and servers on their base will be launched in 2010. Intel already has over 15 design wins from 8 OEMs with its new MP platform.
Even though eight-core Xeon MP processors will be available considerably later than Intel originally claimed, the new chips will offer something that no Xeon processor ever provided. In fact, Intel claims that the next-generation Xeon MP platform will be able to compete against proprietary RISC processor-based systems, the market that Intel traditionally targeted with its Itanium microprocessors.
For example, thanks to point-to-point Quick-Path interconnection (QPI), for the first time in history Intel plans to offer its customers eight-socket platform, which is capable of processing up to 128 threads simultaneously in case of eight-core processors with HyperThreading enabled. Additional scalability options including greater sockets counts will be possible with third-party solutions. Moreover, the forthcoming Nehalem-EX will add new reliability, availability and serviceability (RAS) features traditionally found in Intel Itanium processor family, such as machine check architecture (MCA) recovery.
So, while the new chips contain unprecedented number of x86 processing engines on the single piece of silicon, they will offer much more than that. The only question now is how successful will Intel’s eight-core Xeon MP chips be against AMD’s twelve-core Opteron microprocessors due in Q1 2010.
For months numerous web-sites have been claiming that ATI, graphics business unit of Advanced Micro Devices, planned to release a major update for the Radeon HD 4800-series family. The code-named ATI RV790 was supposed to feature higher number of shader processors and performance considerably higher compared to the Radeon HD 4870 graphics card. Well, despite of rumours floating around the Internet, sources familiar with the RV790 said in mid-February that ATI’s approaching update for the high-performance graphics segment would not bring in considerable performance improvements as the new product would just feature improved clock-speeds over existing graphics boards.
ATI Radeon HD 4890 graphics card
Just like its predecessor, ATI RV790 graphics processing unit features 800 stream processors, 40 texture units, 16 render back ends as well as 256-bit memory bus compatible with GDDR3, GDDR4 and GDDR5 memory types. The RV790 features 850MHz clock-speed, but add-in-card manufacturers may clock the new chips at speeds up to 1000MHz. ATI recommends its partners to clock 1GB of GDDR5 memory to 3900MHz. While the RV790 chip resembles specifications of the RV770 and the chip is still made using 55nm general purpose process technology, the novelty is not hand-picked RV770 with high clock-speed potential, but a reworked graphics processor. In fact, the RV790 even has higher transistor count: it features 959 million transistors, 3 million higher compared to RV770.
ATI Radeon HD 4890/RV790 scheme
In order to enable considerably higher clock-speeds of the RV790, ATI had to reconfigure the chip: to re-do the power distribution of the ASIC, re-time it and also add larger de-coupling capacitors on the ASIC to clean up noise and increase signal integrity (this latter element is the primary cause of the increased transistor count).
It was not logical to expect the RV790 to be an enhanced version of the RV770 in terms of the number of stream processors or other elements. The recent history of graphics processing units does not know examples when a company introduced a brand-new GPU with new internal design that would only be 20% - 30% faster compared to predecessor in the same price-range. Moreover, it would not make much sense to unveil a considerably faster graphics chip several months before the release of the DirectX 11-compatible GPUs. Finally, considering current 40nm yields at TSMC's fab, it would hardly make much sense to start transition to 40nm process technology from a relatively big chip, at the same time, it would not make much sense for ATI to make the "enhanced" RV790 using 55nm fabrication process.
We do not hear a lot about Fujitsu when it comes to microprocessors these days, but in mid-May Fujitsu disclosed some additional details regarding its Venus processor. Unfortunately, the company is still unsure about availability timeframe for the chip as well as systems on its base. However, in the light of the fact that Sun Microsystems reportedly scrapped development of its sixteen-core Rock processor, Fujitsu's Venus may easily become the highest-performance SPARC processor available.
Fujitsu’s SPARC64 VIIIfx (Venus) processor. Image by PC Watch web-site
Fujitsu’s SPARC64 VIIIfx (Venus) is made using 45nm low-power process and sports eight processing cores. According to Fujitsu, the chip has 128GFLOPs peak performance, which is 2.5 times higher compared to Intel’s top-of-the-range dual-core Itanium 2, however, even with built-in memory controller Venus consumes only 33% of what Itanium 2 uses, hence, about 35W. By contrast, the Sun UltraSPARC RK (the official name of the Sun Rock processor) with16 processing cores each capable of processing two threads at once was to be made using 65nm process technology and with maximal frequency of about 2.30GHz the chip could dissipate 250W, according to some sources.
Since at present all the server makers are trying hard to make "green" machines, Fujitsu's processor might be a better solution for Sun than its own processor with truly extreme power consumption. As a result, it looks like Rock may be the last microprocessor that Sun Microsystems was developing. Moreover, as soon as Oracle takes over Sun, it is highly likely that the latter will simply cease to develop microprocessors at all.
Advanced Micro Devices said in early March that the company had reset plans for the launch if its new micro-architecture code-named Bulldozer and the first truly next-generation processors will only emerge in 2011. This contradicted the optimism caused by comments of the company’s chief executive officer, who promised to “ramp up” the first chips produced at 32nm node in mid-2010.
Since Intel released its microprocessors powered by the Core 2 micro-architecture, there have been rather tough times for AMD. The delay and eventual disappointment with performance of quad-core AMD Opteron "Barcelona" chips as well as errata that stopped their sales for another six months were a catastrophe for AMD. Even though the quad-core Shanghai as well as six-core Istanbul processors allowed AMD to offer better performance, it still cannot catch up with Intel in many terms, particularly on the desktop. As a result, many now hope that microprocessors based on Bulldozer micro-architecture will help AMD to finally return itself performance crown. However, the Bulldozer chips will not be available in 2010, but only sometimes in 2011.
Apparently, the first 32nm chip of AMD will be code-named Llano and will be the first chip with the Fusion concept that AMD unveiled back in 2006. Llano will feature up to four Shanghai/Phenom II-class cores; 4MB of L3 cache; PC3-12800 (DDR3 1600MHz) memory controller, possibly, with some tweaks to better serve x86 and graphics engines; DirectX 11 graphics core with third-generation universal video decoder; PCI Express 2.0 bus for external graphics cards. What is even more interesting is that Llano appears to be a monolithic chip combining both x86 and graphics cores. Fortunately, it seems that AMD will ramp Llano in mid-2010, at least, according to Derrick Meyer, chief executive officer of the company.
But what about Bulldozer? Well, it seems that the chip code-named Orochi with six or eight processing engines will truly be only available sometimes in 2011. Before that, AMD will have to rely on binning of its quad-core and eventually six-core processors for both desktops and servers in order to stay competitive against Intel's high-end quad-core and six-core Core i7 central processing units.
It was definitely the sensation of February that Microsoft Corp. had decided to open its own retail stores to sell its products. Besides getting new sales channels, own chain of retail stores will help the company to create deeper engagement with consumers and continue to learn firsthand about what they want and how they buy. Moreover, analysts claim that own retail stores will help Microsoft to improve its image among consumers.
It is not a secret that Microsoft is not a purely software company: the firm competes with Apple for the market of portable digital media players and supporting services, the company rivals Sony and Nintendo in video game console business, Microsoft also produces keyboards and pointing devices and thus competes with Logitech. It is very likely that in future the lineup of hardware from Microsoft will expand: there are rumours about own-brand smartphone from Microsoft.
Microsoft department in a retail store. Image by Dvice web-site
Own retail stores will help Microsoft to be more competitive against its rivals in the hardware business who have their own retail outlets. This is especially important in the light of the fact that major retailers in the U.S. have become victims of economic crisis.
“If Microsoft intends for retail stores to drive revenue growth they will likely be disappointed, but leveraging stores to improve customer perception and as marketing tools could be successful,” said Allan Krans, an analyst with Technology Business Research.
It is a well known fact that stores do not draw consumers to products, but innovative products bring consumers into stores. Apple’s retail store rollout coincided with the introduction of the iPod in 2001, which gave a very compelling reason for consumers to visit its locations, reckons TBR. Microsoft’s main products – Windows operating systems and Office suites – are available by default on the vast majority of personal computers sold on the planet, hence, newer versions are hardly considered as breakthroughs that attract consumers to the stores, according to the analyst.
But besides software, Microsoft also sells its keyboards and mice, Xbox video game consoles, Zune personal digital media players and is rumoured to enter the smartphone business shortly. Moreover, latest technologies that Microsoft implements do not reach retail store shelves widely overnight: average consumers still hardly know about the benefits that DirectX 10 provides; with the Windows 7 release it will be necessary to demonstrate advantages of touch-interface (it is clear that touch-screen sensitive displays are unlikely to reach retailers truly widely); once motion-sensing Natal controller for Xbox 360 is launched, Microsoft will need a lot of physical space in retail stores to show off the benefits of “controller-less gaming”. Therefore, there are a lot of reasons to open Microsoft’s retail stores to show the latest technologies of hardware and software.
But many of computer/electronics/software stores may not like Microsoft’s initiative: with Office, Windows and video games available directly from Microsoft, many consumers will start visiting Microsoft’s stores, especially for large purchases, instead of traditional outlets.
To sum up, while Microsoft will be able to make itself look closer to the final user, it may have certain issues with relationships with its traditional partners in retail. The result of the move will only be seen in the next few years.
Intel Corp. in mid-February said that it had filed a lawsuit against Nvidia Corp. in order to ban Nvidia from making and selling its nForce or GeForce chipsets compatible with Intel’s processors featuring integrated memory controller. Intel also accused Nvidia of breaching their mutual agreement by claiming that it does have license to make chipsets compatible with Intel Core i7 microprocessors that use quick-path interconnect (QPI) bus to connect to other components of the system as well as for other similar processors.
“Intel has filed suit against Nvidia seeking a declaratory judgment over rights associated with two agreements between the companies. The suit seeks to have the court declare that Nvidia is not licensed to produce chipsets that are compatible with any Intel processor that has integrated memory controller functionality, such as Intel's Nehalem microprocessors,” said Daniel Snyder, a spokesman for Intel.
“At the heart of this issue is that the CPU has run its course and the soul of the PC is shifting quickly to the GPU. This is clearly an attempt to stifle innovation to protect a decaying CPU business,” claimed Jen-Hsun Huang.
Nvidia has been providing advanced chipsets for Intel platform for about five years now and in many cases its chipsets were better for overclocking of microprocessors. The actual reason of their popularity was support for Nvidia multi-GPU SLI technology and the only reason why Nvidia SLI was not available on other platforms was because Nvidia blocked it from the drivers. But it was not chipset business that was the motive of the conflict between Intel and Nvidia. The reason why both companies initiated the dispute is graphics processing units and rasterization.
In early April, 2008, Patrick Gelsinger, senior vice president and general manager of Digital Enterprise Group at Intel said that current rasterization-based 3D graphics rendering technology was no longer suitable for new challenges and that CPUs could be faster.
“First, graphics that we have all come to know and love today, I have news for you. It’s coming to an end. Our multi-decade old 3D graphics rendering architecture that’s based on a rasterization approach is no longer scaleable and suitable for the demands of the future. And we can do much better,” said Mr. Gelsinger in front of a huge auditory at Intel Developer Forum in Shanghai, China. Mr. Gelsinger referred to the fact that Intel's Larrabee graphics processors can do both rasterization as well as ray-tracing, another method of rendering that cannot be efficiently handled by modern GPUs and CPUs.
Patrick Gelsinger holding a wafer with Larrabee GPUs at IDF 2009
While Mr. Gelsinger did not state that there will be no need for a separate graphics accelerator in a PC, his colleague Ron Fosner, a graphics and gaming technologist at Intel, said that in several years time people “probably won’t” need discrete graphics cards.
Nvidia was quick to respond its much larger rival with promises made by chief executive Jen-Hsun Huang to open up “a can of whoop ass” on Intel and then state that the central processing unit was “dead”.
Jen-Hsun Huang, Image by AP/Daylife.com
“Basically the CPU is dead. Yes, that processor you see advertised everywhere from Intel. It runs out of steam. The fact is that it no longer makes anything run faster. You don’t need a fast one anymore. This is why AMD is in trouble and it’s why Intel are panicking. They are panicking so much that they have started attacking us. This is because you do still [need] one chip to get faster and faster – the GPU. That GeForce chip,” said the head of Nvidia to analysts.
A year after Nvidia said the x86 microprocessor was dead it nearly lost AMD CPU-compatible chipset business and is about to lose Intel CPU-compatible core-logic business. In fact, chipset revenue still represents about 20% of Nvidia’s earnings and are over $100 of million. As a result, by now rhetoric of Jen-Hsun Huang and Nvidia has changed dramatically even from the claims made several months ago.
“CPU-GPU co-processing is the future of computing,” Mr. Huang said at a meeting with financial analysts in June, 2009.
“The core-processing is the present and the future of PC world,” he added in another interview.
Perhaps, after Intel filed a lawsuit, Nvidia still hopes that it can find a way to capitalize somehow on Intel platforms?
Nvidia Corp. claimed in late March that it could terminate its cross-licensing agreement with Intel Corp., which would prevent the latter from launching its first discrete graphics processor in ten years. Besides, claims Nvidia, Intel utilizes Nvidia’s intellectual property in every integrated graphics core it ships.
Nvidia believes that Intel is using Nvidia IP in currently shipping IGP products and the company also believes that Intel would not be able to develop its Larrabee graphics chips without making use of Nvidia's IP portfolio. Nvidia may ask the courts to terminate Intel's rights to these IPs in the event that it is found that Intel is in breach of contract,” Drew Henry, the head of Nvidia’s core-logic business unit, is reported to have said in an interview.
Nobody, except those who work at Intel Corp. or certain software developers, knows what Intel’s code-named Larrabee discrete graphics processing unit actually is, but it may seem that Nvidia is quite concerned about it.
In fact, Nvidia is presently more concerned about its own chipset business: it cannot develop core-logic sets for Intel’s new-generation microprocessors with integrated memory controllers (at least now), AMD is gradually forcing Nvidia out of the AMD-compatible chipset market and Intel is about to start to progressively replace Intel Core 2 Duo and derivatives with new processors that have built-in memory controllers and/or graphics cores.
In Q4 2009 even Intel’s low-cost Atom platform will start to integrate memory controller and graphics/video processors onto the CPU itself, which makes Nvidia Ion and even Ion 2 practically useless for the vast majority of end-users and netbook manufacturers. Moreover, Intel’s aggressive ramp of desktop code-named Clarkdale processor with integrated graphics (potentially, nothing prevents Intel from quick ramp of similar Arrandale chip) will quickly leave Nvidia without business in the low-cost desktops. For Nvidia, all this means that it will have to cease development of core-logic sets. Therefore, the attack on Larrabee and cross-licensing agreement could be a way to slowdown Intel’s expansion onto the desktop/laptop graphics business.
What Nvidia needs to do, and what the company seemingly is doing, is to develop ultra low-cost graphics processing units that will enable graphics cards for $30 - $40 in retail. This will allow Nvidia to claim huge performance benefit over Intel's integrated graphics and potentially minimize the loss of market share.
It is unlikely that Nvidia will terminate cross-licensing agreement with Intel and it is even more unlikely that such a termination would cause Intel to stop selling its integrated graphics cores or Larrabee. After Nvidia terminates the agreement, it will have to sue Intel and then wait for years till the court orders the chip giant to stop using Nvidia's intellectual property. In several years time the products that may infringe Nvidia's patents will most likely be absent from Intel's product range. Moreover, Intel may file a counter-lawsuit against Nvidia, which result might be another cross-licensing agreement between the two companies.
All-in-all, Nvidia will hardly terminate the cross-licensing agreement with Intel, but threatening Intel with this may allow Nvidia to remain in Intel-compatible business.
Acer Group, the world’s third largest maker of personal computers, has revealed that Microsoft Corp.’s Windows 7 next-generation operating system (OS) will be released officially on the 23rd of October, 2009, if there are no changes in the plans.
“23rd October is the date the Windows 7 will be available. There is a 30 day upgrade time so that customers don’t wait to buy a new computer, so if you buy during that 30 day period, you’ll get a free upgrade to Windows 7,” said Robert Watkins, managing director of Acer UK, in an interview. Mr. Watkins was almost correct: Microsoft will release its Windows 7 on the 22nd October this year.
The fact that the news-story about availability date of Microsoft Windows 7 has become so popular speaks for itself: people are interested in that information because they are interested in transition to Windows 7. There is no secret that many are tired of instabilities and other issues of Windows Vista operating system and want to install Windows 7 as soon as possible. However, the vast majority of people eager to install Windows 7 are those, who just want something new, which is a good sign: despite of the global recession there are still those, who want to invest into new technologies.
The estimated retail prices for full packaged retail product of Windows 7 in the U.S. are:
Microsoft has been working hard to improve user interface as well as performance in Windows 7 and make the new OS much more user friendly overall. Many reviewers and early adopters agree that Microsoft has succeeded in improving its desktop OS dramatically. Microsoft Windows 7 will also bring numerous further improvements, including DirectX 11 application programming interface, as well as support for touch-based input.
It has been a long time since the world saw a breakthrough from Advanced Micro Devices, but it looks like AMD did a breakthrough in late April by demonstrating its code-named Magny-Cours central processing unit with twelve cores. Needless to say that this sensation attracted more readers than any other story published this year.
AMD Opteron 6000-series "Magny-Cours" with twelve processing engines will be released in the first quarter of 2010, according to some reports. However, already now AMD can demonstrate the chip to public and customers to show off its benefits compared to competing platforms and technologies.
Twelve-core AMD Opteron processor
AMD Magny-Cours processor will be the first chip for the AMD G34 “Maranello” platform designed for Opteron processors 6000-series with up to 16 cores, quad-channel memory interface, 2 or 4 sockets, up to 12 memory modules per socket and some server and enterprise-specific functionality. Thanks to multi-chip-module nature of the code-named Magny-Cours processor, creation of such chip was a relatively easy challenge. It should be noted that AMD G34 will the company’s first own server platform in seven years after the launch of AMD 8000 platform in 2003.
The demonstration of Maranello prototype as well as twelve-core processors is indisputably a notable even for AMD. The company has had bad luck with its quad-core processors in many ways and the showing of the next-generation platform about six to nine months before its release should inspire optimism both among server makers as well as investors. In addition, AMD promises to deliver its first high-end server code-named Interlagos chips based on Bulldozer architecture with up to 16 cores in 2011.
AMD’s twelve-core Opteron “Magny-Cours” will compete against Intel’s eight-core Xeon MP “Nehalem-EX”. The battle between the two leading designers of microprocessors in the very high-end space promises to be rather exciting. With the introduction of QuickPath interconnect by Intel, AMD’s platforms no longer have the advantage of superior performance scalability, meanwhile, Intel’s next-generation multi-processor platform supports up to 16 memory modules per socket, up to eight sockets per system as well as various high-end RAS features. Still, AMD will have more processing engines, and, provided that it can ensure high enough clock-speeds, potentially greater performance in case of four socket servers.