Monday, October 20, 2008

AMD Intends to Produce “Some” of Its GPUs at The Foundry Company Using 32nm Process Tech

Chief executive officer of Advanced Micro Devices officially outlined plans to utilize manufacturing capacities of The Foundry Company, a joint venture between AMD and Advanced Technology Investment Company, for manufacturing of ATI’s graphics processing units (GPUs). Apparently, TFC will only be able to produce GPUs using 32nm process technology, not earlier.

“We will be ramping Fab 38 and we will be installing into Fab 38 a 32nm bulk process for the purpose of enabling the foundry business. We will also be moving some GPU product into the Foundry Company in order to help the Foundry Company get off the ground relative to that foundry type business. Clearly, as a result of having two really good suppliers on the GPU side we got an opportunity to improve our costs there and capability,” said Dirk Meyer, chief executive and officer of AMD, during a conference call with financial analysts.

AMD has already confirmed that it plans to use capacities of the Foundry Co. to manufacture GPUs by ATI, graphics products group of AMD. However, this is the first time when AMD reveals details: it intends to use 32nm process technology, which may mean that the first GPUs made at TFC will be available sometime in 2010.

AMD has reported third quarter 2008 revenue from continuing operations of $1.776 billion, including process technology license revenue of $191 million. Third quarter 2008 revenue increased 32%compared to the second quarter of 2008 and 14% compared to the third quarter of 2007.

In the third quarter of 2008, AMD reported a net loss of $67 million, or $0.11 per share. For continuing operations, third quarter 2008 income was $41 million, or $0.07 per share, including the process technology license revenue of $191 million, or $0.31 a share. Third quarter 2008 operating income was $131 million. Loss from discontinued operations was $108 million, or $0.18 a share.

Intel Takes Over NetEffect, An Ethernet and Server Computer Cluster Specialist

Intel Corp. has announced its acquisition of the assets of NetEffect, a network connectivity solutions company specializing in Ethernet products and technologies for server compute clusters.

Intel purchased NetEffect's assets for $8 million, which include the company’s Ethernet product portfolio, intellectual property and technology. NetEffect was a provider of solutions incorporating iWARP, an Ethernet alternative to InfiniBand. NetEffect's product portfolio includes 1Gb and 10Gb Ethernet (GbE) adapters for servers and blade configurations as well as 10GbE application specific integrated circuits (ASICs).

“NetEffect’s role as a data communications solution provider and the company’s technology will enhance Intel's current Ethernet efforts. The combination of Intel and NetEffect technology will allow Intel to address our customers’ most important 10GbE needs, including server virtualization, convergence of network and storage traffic, and server compute clusters,” said Tom Swinford, general manager, Intel LAN access division.

Designed for multi-core processor-based servers and optimized for virtualization, Intel's current portfolio of 10GbE server adapters includes single and dual port versions for both copper and fiber implementations. The NetEffect acquisition provides complementary High Performance Network Interface Card (NIC) products to Intel's Ethernet portfolio.

NetEffect was founded in 1998 as Banderacom, a company focused on InfiniBand adapters and 16-port IB switch, and was recapitalized in 2004 as NetEffect. Thirty employees, primarily engineers, have joined the Intel team from NetEffect and will continue to be based in Austin, Texas.

AMD Preps More Dual-Core Microprocessors for Q1 2009 – Rumours

Advanced Micro Devices reportedly readies two new dual-core microprocessors based on the K10 micro-architecture. The new chips were not in the plan several months ago, but the company apparently decided to introduce the new chips in order to stabilize pricing of its dual-core processors aimed at mainstream market.

According to a news-story at Expreview web-site, AMD intends to release Athlon X2 7550 and 7750 chips sometime in Q1 2009. The chips will operate at 2.50GHz and 2.70GHz, feature 1MB of L2 cache [512KB per core], 2MB L3 cache as well as dual-channel DDR2 memory controller. The new products will have thermal design power of 95W, hence, should be compatible with the vast majority of AM2+ mainboards.

The new dual-core AMD Athlon X2 processors are based on the code-named Kuma core, the same one that is already featured in AMD Athlon X2 6500 processor that operates at 2.30GHz.

It is unclear why AMD decided to introduce a new model number sequence for its new dual-core central processing units despite of the fact that the novelties hardly provide tangible performance benefits compared to predecessors.

AMD did not comment on the news-story.

Intel Initiates Shipments of Solid-State Drives for Enterprise Systems

As expected, Intel Corp. said this week it had begun shipments of solid-state drives (SSDs) aimed at enterprise systems. The new SSDs are based on 50nm single-level cell (SLC) NAND flash memory and are projected to feature low power consumption, high reliability and extreme performance thanks to new controller and firmware.

“Hard disk drive performance has not kept pace with Moore's Law. Intel’s high-performance SSDs unleash the full performance of the latest Intel Xeon processor-based systems while increasing reliability and lowering the total cost of ownership for a broad range of server and storage workloads,” said Kirk Skaugen, general manager of Intel server platforms group.

Intel X-25E Extreme Serial ATA SSD, aimed at server, workstation and storage systems. The product is based on 50nm single-level cell (SLC) NAND flash memory. Systems equipped with these drives will not suffer from the performance bottlenecks associated with conventional drives.

The product was designed for intense computing workloads which benefit primarily from high random read and write performance, as measured in input/output per second (IOPS). Key technical performance specifications of the 32GB Intel X-25E SATA SSD include 35 000 IOPS (4KB Random Read), 3 300 IOPS (4KB Random Write) and 75 microsecond read latency. This performance, combined with low active power of 2.4Ws, delivers up to 14 000 IOPS per watt for optimal performance/power output. The product also achieves up to 250MB/s sequential read speeds and up to 170MB/s sequential write speeds, considerably higher compared to Intel’s consumer-oriented SSDs, all in 2.5” form-factor.

Intel achieves this breakthrough performance through innovations such as 10-channel NAND architecture with native command queuing (NCQ), proprietary controller and firmware efficient in advanced wear-leveling and low write amplification. The 32GB X25-E is capable of writing up to 4PB of data over three-year period, or 3.7 TB/day, and double that for the 64GB version – delivering outstanding data reliability.

The 32GB capacity drive is in production and priced at $695 for quantities up to 1000. The 64GB version is expected to sample in the fourth quarter with production estimated for the first quarter of 2009.

“Solid-state drive technology will change the economics of enterprise data centers. SSDs, along with our systems and Solaris ZFS with hybrid storage pools, are important components of the Open Storage initiative. Sun expects to offer enterprise storage solutions that will exploit the breakthrough performance of Intel's high performance solid-state drives and deliver significant performance gains while consuming a fraction of the energy of traditional spinning disk arrays,” said John Fowler, executive vice president of systems group at Sun Microsystems.

Another Dual-Chip ATI Radeon HD Graphics Card Finally Approaches the Market

ATI Radeon HD 4850 X2 graphics card, which was formally announced back in August and is still yet unavailable, will finally reach stores in the coming weeks, according to ATI, graphics product group of Advanced Micro Devices.

The ATI Radeon HD 4850 X2 graphics cards that feature two ATI Radeon HD 4850 graphics chips along with 2GB of GDDR3 memory (1GB per chip) are now “validated” by graphics cards makers, said Sasha Marinkovic, senior product marketing manager at AMD’s graphics products group at a news-conference on Thursday.

According to Mr. Marinkovic, ATI provided reference design of the Radeon HD 4850 X2 graphics cards to manufacturers and expects to see the new graphics cards on the market shortly. It is unclear why did makers of graphics boards decided to delay the model HD 4850 X2 by months when compared to mid-August release time-frame for the top-of-the-range model HD 4870 X2.

The dual-chip graphics card based on ATI Radeon HD 4850 graphics chips is expected to cost between $350 and $399, according to AMD’s ATI unit. The new product will compete directly with Nvidia GeForce GTX 280 graphics board, currently the highest-performance single-chip 3D graphics accelerator for video games on the market, which retails for $449 - $499.

While ATI Radeon HD 4850 X2 may offer higher performance than that of the GeForce GTX 280, it may not be as popular as the Radeon HD 4870 X2 despite of lower price. All multi-GPU configurations depend on driver and application optimizations, which is why dual-chip graphics cards may not deliver horsepower of two chips in all game titles, meaning that performance may drop to the level of a single-chip graphics board that costs two times less. Since the Radeon HD 4870 X2 graphics card is very high-performance graphics card itself, many gamers may not get worried that in certain apps only one chip functions. Since the model HD 4850 belongs to performance-mainstream class of accelerators, the speed one of such cards delivers may be unacceptable for those, who buy the HD 4850 X2 model for $399.

Intel Vows to Ship Enough Atom Processors by Year End

Intel Corp. admitted during the most recent conference call with financial analysts that the demand for Intel Atom chips is higher than expectations and that the company could not meet it in the third quarter. The company promised to deliver enough central processing units by the end of the year. Besides, the company revealed that Atom bring higher profits compared to traditional entry-level chips.

“We did not meet demand [for Intel Atom] in Q3 for the product, even though we were up substantially from the second quarter. We are up again substantially in the fourth quarter. Our expectation is that we will meet demand by the end of the year but not the early part of the quarter,” Paul Otellini, chief executive officer at Intel told financial analysts during quarterly conference call.

Still, the head of Intel says that sales of Atom processors were on the high-levels. Moreover, since the CPUs are really cheap to manufacture, even their low prices allow Intel to get profits similar to those it earns on the traditional low-end and mainstream markets.

“The Atom family is off to a very good start, with Atom microprocessor and related chipset revenues approximately $200 million this quarter. Total microprocessor ASP was lower than Q2 but was approximately flat without Atom, reflecting strength in the core business. […] When I look at the product margin, it is a nice [and] healthy product margin and on a dollar basis it is equivalent to what we see in Celeron and on a product margin percent, it is higher. So, if you look at it kind of relative to the low-end of our mainstream stack, it’s generating nice profit characteristics,” said Mr. Otellini.

Considering the fact that despite of low price Intel Atom processors bring more profits to Intel than higher-performance Celeron chips, over time Intel may become more interested in promoting the Atom microprocessors rather than more traditional products. This may indeed change the game on the whole PC market as the competition between computer vendors will get much more aggressive on the netbook side, where margins are even thinner than those on desktop or laptop markets.

Sony Has No Plans to Reduce PlayStation 3 Pricing for Holidays

Sony Computer Entertainment Inc. has no plans to respond to Microsoft’s aggressive price-cuts of the Xbox 360 video game console in the European Union and the United Stations with similar moves concerning the PlayStation 3 this holiday season, the head of SCEI said in an interview.

“The answer is ‘yes’, if you’re asking, ‘Are these the prices we’re going with this Christmas?’” said Kazuo Hirai, president and group chief executive of Sony Computer Entertainment, in an interview with the Financial Times.

Microsoft recently reduced the price of its Xbox 360 Arcade game console to $199 in the U.S. in response to lowering demand and financial crisis. The company has also consistently cut the cost of the Xbox 360 consoles in other regions, including Asia, Europe and Japan this year in a bid to attract gamers to its latest video game systems.

Obviously, the least expensive Xbox 360 Arcade does not offer certain features that are available on Sony PlayStation 3 ($399) and more expensive Xbox 360 Pro 60GB ($299) game console since the latter have hard disk drives, whereas the PS3 even has Blu-ray disc optical drive that transforms game console into a high-definition video player. In fact, a number of exclusive titles and value proposition have already made Sony PlayStation 3 nearly as popular as the Xbox 360 in some regions and allowed Sony to even leave the rival behind in some others. Still, neither of the consoles is as successful as Nintendo Wii, which usually outsells both PS3 and X360 combined in the USA.

However, for the majority of gamers the X360 for $199 is likely to be an option they will be satisfied with. As a result, there are reasons for Sony to worry about: more Xbox 360 game consoles on the market means that game developers start to give the systems higher priority compared to others, which ultimately leads to higher amount of higher quality titles, which further popularizes the platform.

In common with the rest of the video game industry, Mr Hirai expects video games to be resilient in the economic downturn because consumers buy video games rather than go out.

“As long as we can generate excitement, then we will be less affected than other industries,” the head of SCEI told the FT.

Nvidia Formally Unveils New Intel-Compatible Core-Logic with Built-In Graphics Core

Nvidia Corp. has unveiled its latest core-logic set compatible with Intel Corp.’s latest central processing units that sports DirectX 10-class graphics processing unit (GPU). The novelty boasts with decent feature-set and will replace the aging predecessors.

Nvidia recently started to brand its core-logic chips with integrated graphics processing cores (IGPs) as GeForce motherboard GPUs (mGPUs), which indicates that the company does not see a lot of value of its nForce IGP brand on the entry-level market, where consumers are concerned about the price and trendy features, but not rich enthusiast-class functionality the nForce is primarily known for these days. As a result, the new single-chip core-logic from Nvidia is called GeForce 9300 mGPU, even though it more than a GPU.

Nvidia’s GeForce 9300 mGPU is compatible with Intel Core 2 processors that utilize up to 1333MHz processor system bus (PSB) and also sports dual-channel DDR3 memory controller (up to PC3-16000, 1333MHz), PCI Express 2.0 controller (PCIe x16 + 4x PCIe x1), Serial ATA controller, high-definition audio controller, PCI controller, Gigabit Ethernet support and so on. Eventually Nvidia also plans to release GeForce 9400 mGPU with a little bit higher-performing IGP thanks to boosted clock-speeds: 580MHz/1400MHz (core/shader domain) vs. 450MHz/1200MHz for the 9300.

The integrated GeForce 9300/9400 graphics core features 16 stream processing units, Nvidia PureVideo HD engine for decoding and post-processing of high-definition video as well as dual-link DVI, HDMI and DisplayPort output support. Performance of such integrated graphics cores is likely to be similar to low-end graphics cards, such as GeForce 8400 GS or 8500 GT. End-users will be able to utilize an entry-level discrete GPU in multi-GPU mode with GeForce 9300/9400 mGPUs.

Nvidia did not reveal when GeForce 9300/9400 mGPU-based mainboards are set to become available on the market and at what price-point.