Nanotech - The Circuits Blog

Nvidia's new Fermi chip is being billed as a supercomputing chip but Nvidia doesn't want you to forget that it is also aimed at Apple's Snow Leopard and Windows 7.

The Fermi chip was announced with much fanfare on Wednesday as key silicon in a future supercomputer from Oak Ridge National Laboratory. But, wait, Fermi is also going to be great at accelerating stuff in Snow Leopard and Windows 7--not to mention a great gaming chip, according to Bill Dally, chief scientist at Nvidia who spoke during a conference call with analysts on Thursday.

The Fermi graphics processing unit (GPU)--which packs 512 processing cores--will support DirectX-11, a technology for speeding certain multimedia software in Windows 7, and also support an analogous technology in Snow Leopard, OpenCL.

"A lot of (the chip's new) features accelerate key consumer applications. Both Snow Leopard and Windows 7 enable the GPU to be used as a co-processor to accelerate third-party applications," Dally said. With a "discrete (standalone) GPU they can get very good performance on these applications," he said.

Applications that Nvidia says will be accelerated by the Fermi chip

(Credit: Nvidia)

Dally gave examples (see graphic) of consumer titles such as Adobe's Creative Suite, Motion DSP's vReveal (for fixing photographs), and Badaboom (for creating iPod video).

He offered a qualifier, however. "We are paying a bit of a compute tax in that we launched a part where a lot of the consumer compute applications haven't really taken hold yet. But over time as more consumer computer applications are developed that take advantage of our compute (consumer) features...I think it's going to give us a big leg up," he said.

And being an Nvidia chip, games are a big target market. "Fermi adds value to games by doing exactly the same kind of scientific simulations that we use to predict climate and to understand the genome and other things," according to Dally. "A great example of that is our PhysX package that basically does physical simulations to make games appear more real."

He also explained why the chip was billed as a supercomputer chip initially and not a gaming chip. "It's a zero-sum game. You have a certain amount of die (chip) area, a certain power budget. It is the case that we put a bunch of die area into double-precision floating point, a bunch of die area into ECC. And for gaming graphics applications, those give less returns than they do for the scientific applications," he said. Double-precision floating point operations are used heavily in scientific computing. ECC, or error correcting code, is a technology that can correct data errors on the fly.

And Dally explained how Fermi can be scaled down to lower-end chips used in the gaming and consumer segments. "We're not talking about other (chips) at this point in time but you can imagine that we can scale this part by having fewer than the 512 cores and by having these cores have fewer of the features, for example less double-precision," he said.

All the Fermi products, including gaming and professional workstation chips, will be announced "pretty close together." Chips are expected sometime in the coming few months.

And how does Fermi stack up against current public information about Intel's future "Larrabee" graphics chip? "We can't compare anything to Larrabee until it shows up and can actually be measured," said Jon Peddie, president of Jon Peddie Research, which tracks the graphics chip market. "But remember, Larrabee was started over two years ago and both ATI and Nvidia have had two new designs out since then," he said. "So the pressure will be on Intel to chase fast-moving ATI and Nvidia," Peddie said. ATI, which is Advanced Micro Devices' graphics chip unit, already has a chip in stores--the Radeon HD 5800-- that supports Windows DirectX-11.

Sony said that a small percentage of Vaio laptops with Nvidia graphics chips may experience problems and the company offered to provide an extended warranty to cover the cost of repair. This follows similar statements by Apple, Hewlett-Packard, and Dell.

Sony Vaio VGN-AR series laptop

(Credit: Sony)

Nvidia first disclosed the problem with its graphics chips in July 2008, saying at the time that graphics processors manufactured with a certain material set were failing in the field at a higher than normal rate.

In a Sony eSupport USA notice dated August 3, the company said: "Sony, in cooperation with Nvidia, has been looking into any possible effect to Vaio notebooks with Nvidia graphic processors. Until recently we had not identified any Vaio models that were affected by this issue."

The statement continues. "However, after closely monitoring the situation, Sony has now determined that a very small percentage of Vaio computers with the Nvidia graphics chips may experience this issue. These PCs may exhibit distorted video, duplicate images or a blank screen due to a failure of the Nvidia graphics chip."

Updated at 4:30 p.m. PDT adding Tegra, Intel, and Ion discussions.

On Tuesday, Nvidia Chief Executive Jen-Hsun Huang said at the company's analyst day that the graphics processor will be an equal partner with Intel processors, citing Apple as an early trendsetter.

On other fronts, Huang said that the ARM-based Tegra processor is expected to account for half of Nvidia's business in a few years. He also repeated claims about Intel crimping the success of its Ion processor in Netbooks.

Nvidia CEO Jen-Hsun Huang

(Credit: Nvidia)

Huang said that "CPU-GPU co-processing" is the future of computing. (CPU stands for central processing unit. GPU for graphics processing unit.)

"Apple is an early indicator," Huang said during his opening remarks that were streamed over the Web, referring to the importance that Apple is placing on the graphics processor. "The MacBook Pro to the MacBook Air has a GPU," he said. And he waxed eloquent about how the performance and power efficiency of the updated version of the Air has benefited by having co-processors: an Intel CPU and Nvidia GPU.

"Doing the right job with the right tool is more efficient," he said, referring to the Air, which Huang claims runs longer and cooler with a GPU. Typically, ultra-thin laptops like the Air don't have a discrete (separate) Nvidia or ATI graphics processor.

Apple currently uses Nvidia GPUs across its laptop product line and touts the potential for GPUs on its Web site. "OpenCL (Open Computing Language), makes it possible for developers to efficiently tap the vast gigaflops of computing power currently locked up in the graphics processing unit," according to a statement on Apple's Web site.

And at the Apple Worldwide Developers Conference, an Apple executive expanded on this theme, explaining how Mac OS X will support GPGPU--general-purpose graphics processing unit--which lets a graphics chip run some computing jobs in addition to its ordinary job displaying graphics.

Huang also addressed its Tegra chip, which is an ARM-based design that integrates an Nvidia GeForce processor. Tegra is targeted at smartphones and Netbooks. Responding to a question from an analyst, he said that in a few years Tegra may represent half of its business, with the rest divided up between the professional (Tesla, Quadro) and the consumer GeForce markets.

Huang also repeated his assertion that Intel is using pricing--what he called "subsidies"--and "MDF" (market development funds) to prevent Nvidia from selling more of it Ion processors to customers. He claimed the success of the Ion processor would be two to three times greater without Intel interference.

Nvidia said that some notebooks with its chips continue to have "failure" issues, in a filing with the Securities and Exchange Commission on Wednesday.

In the Form 10-Q filing, Nvidia stated that though it does not continue to see "abnormal failure rates" in systems using Nvidia products," some notebooks are still affected.

Specifically, Nvidia said: "We continue to not see any abnormal failure rates in any systems using Nvidia products other than certain notebook configurations. However, we are continuing to test and otherwise investigate other products," Nvidia said, adding, "there can be no assurance that we will not discover defects in other MCP or GPU products." (MCP stands for Media and Communications Processor; GPU stands for Graphics Processing Unit.)

On July 2 of last year, Nvidia announced it was planning to take a one-time charge to cover costs associated with problems with materials used in certain versions of its laptop graphics chips. Subsequently, a $196 million charge was recorded in the second quarter of its 2009 fiscal year to "cover anticipated customer warranty, repair, return, replacement and associated costs" with the problem.

In the 10-Q filing, Nvidia cited a "balance of $145.7 million associated with incremental repair and replacement costs from a weak die/packaging material set." and "$31.2 million for the three months ended April 26, 2009 in payments related to the warranty accrual associated with incremental repair and replacement costs from a weak die/packaging material set."

Nvidia paid or incurred $50.3 million against the original "warranty accrual" in its fiscal third quarter and fourth quarter 2009, such that the remaining balance of the "bump-crack accrual" (defect) was $145.7 million at the end of its fiscal fourth quarter, according to Nvidia.

Nvidia is also grappling with insurance companies over payments to PC makers for GPU failures, according to reports.

As early as 2007, Hewlett-Packard listed laptop models affected by the graphics chip glitch. In August 2008, Dell also listed affected models. And Apple said in October that it would repair faulty graphics chips.

In the 10-Q filing, Nvidia also stated (some cases were cited in previous Nvidia filings) that "in September, October and November 2008, several putative consumer class action lawsuits were filed against us, asserting various claims arising from a weak die/packaging material set in certain versions of our previous generation MCP and GPU products used in notebook systems."

Most of the lawsuits were filed in federal court in the Northern District of California, but three were filed in state court in California, in federal court in New York, and in federal court in Texas, Nvidia stated. "Those three actions have since been removed or transferred to the United States District Court for the Northern District of California, San Jose Division, where all of the actions now are currently pending."

Some of the lawsuits, such as "Inicom Networks, Inc. v. NVIDIA Corp. and Dell, Inc. and Hewlett Packard," include Dell and HP.

Updated on April 27 at 8:20 a.m. PDT with additional information about DirectX 11 and correcting for Intel comments at bottom.

Graphics chips will be tapped to accelerate more tasks in upcoming versions of Apple's and Microsoft's operating systems, according to Nvidia.

Apple's upcoming Mac OS X Snow Leopard will tap into the compute power of graphics processors

(Credit: Apple)

In an interview Friday with Sumit Gupta, product manager for Nvidia's Tesla products, Gupta described how new programming environments will tap into the latent compute horsepower of graphics processors to accelerate software in Apple's upcoming OS X Snow Leopard and Microsoft's Windows 7 operating systems.

Graphics chips aren't just for games anymore. The trend toward general-purpose graphics processing is defined by an acronym that doesn't exactly roll off the tongue: GPGPU. But the essence of General Purpose computing on Graphics Processing Units is pretty simple: use the scores--or even hundreds in higher-end chips--of processing cores inside GPUs to speed tasks that, in some cases, would be done much less efficiently by the central processing unit (CPU).

This is where OpenCL (Open Computing Language) comes in. OpenCL is a programming environment for "heterogeneous" computing. That is, computers using a mix of multicore CPUs and GPUs. Microsoft's analogous programming environment is DirectX.

Apple says this about OpenCL on its Web site. "Another powerful Snow Leopard technology, OpenCL...makes it possible for developers to efficiently tap the vast gigaflops of computing power currently locked up in the graphics processing unit."

Today, on a PC or a Mac, the CPUs made by Intel and Advanced Micro Devices are adept at handling general operating system tasks. For instance, handling the sequence of things that must happen after the user clicks on an icon to start an application on their desktop.

Nvidia aims to fortify the Mac Pro moniker with its own "Pro" graphics.

Currently, ATI Radeon HD 4870 (L) and Nvida GeForce GT 120 are offered as standard graphics chips on Mac Pro

(Credit: Apple)

The Quadro FX 4800, packing 192 processing cores, targets professional graphics customers, including scientists, engineers, and designers.

The Mac card features a standard 3-pin stereo connector for 3D stereoscopic imaging, critical to many core professional Mac applications, Nvidia said.

The 4800 also features a large 1.5GB frame buffer (used to refresh the on-screen image) and memory bandwidth up to 76.8 gigabytes per second. By comparison, the Nvidia GeForce GT 120, which is currently offered by Apple as a standard graphics card on the Mac Pro, has memory bandwidth of 25.6GB/s.

The card also integrates two Dual Link DVI Connectors and Boot Camp Support for access to native Quadro GPU accelerated professional 3D graphics on Windows.

The Nvidia Quadro FX 4800 GPU for Mac lists for $1,799 and will be available in May 2009 through Apple.com, and select Apple resellers and workstation integrators. The card will also be available from PNY Technologies, Leadteck, and Elsa (Japan).

Any gamer worth his or her salt is quick to decry gaming on Intel graphics silicon. But wait. The platform is taking off, according to Intel.

Empire: Total War

(Credit: The Creative Assembly)

"So you want to know what's so compelling about making sure your game runs on Intel integrated graphics?" Aaron Davies, a senior marketing manager in the Intel Visual Computing Software Development group, asked in a video on the Intel Software Network Web page. "Here's your answer: Mercury Research showed that in 2008, for the first time, integrated graphics chipsets outsold discrete (graphics chips), and in 2013, we expect to see integrated graphics chipsets outsell discrete by three to one," Davies said.

Intel is the leading supplier of integrated graphics--which are integrated into its chipsets--while Nvidia and Advanced Micro Devices' ATI unit are leaders in the discrete (standalone) graphics chip market. Discrete chips are the most powerful engines for running games but Intel is the leading supplier, based on market share, of graphics silicon for laptops.

Davies said he wants to help developers "capture" where the mobile-game market is going to be in 2013.

"There are games actually targeting integrated graphics chipsets at this time," according to Davies. "We found through engaging with these Triple A (AAA) game studios that within a relatively short amount of time, they can identify graphics bottlenecks in their code and resolve that to have their games running on integrated graphics chipsets."

Davies mentioned a few of the beta members: Terminal Reality, which is slated to come out with Ghostbusters later this year; Gas Powered Games, which is building Demigod, and Empire Total War, which is put out by The Creative Assembly.

On Friday, Intel engineers are detailing the inner workings of the company's first graphics chip in over a decade at the Game Developers Conference in San Francisco--sending a signal to the game industry that the world's largest chipmaker intends to be a player.

During a conference call that served as a preview to the GDC sessions, Tom Forsyth, a software and hardware architect at Intel working on the Larrabee graphics chip project, discussed the design of Larrabee, a chip aimed squarely at Nvidia and at Advanced Micro Devices' ATI unit.

And Nvidia and AMD will no doubt be watching the progress intently. Intel's extensive and deep relationships with computer makers could give it an inside track with customers and upset the graphics duopoly now enjoyed by Nvidia and AMD. In the last decade Intel has not competed in the standalone, or "discrete" graphics chip market where Nvidia and AMD dominate. Rather, it has been a supplier of integrated graphics, a low-performance technology built into its chipsets that offers only a minimal gaming experience. (In the 1990s, Intel introduced the i740 GPU which, in relative terms, was not a success.)

Forsyth said that there is not yet a Larrabee chip to work with--it's expected late this year or early next year--and that "a lot of key developers are still being consulted on the design of Larrabee." But Intel will offer ways for developers to test the processor, he said. "On the Intel Web site there will be a C++ prototype library. It doesn't have the speed of Larrabee but has the same functionality. Developers can get a feel for the language, get a feel for the power of the machine."

Beyond games, Intel is also trying to catch a building wave of applications that run on the many-core architectures inherent to graphics chips. Nvidia and AMD graphics chips pack hundreds of processing cores that can be tapped for not only accelerating sophisticated games like Crysis but for doing scientific research and high-performance computing tasks.

One of the largest test sites for Larrabee is Dreamworks, which will use Larrabee for rendering and animation. To date, Dreamworks had to wait overnight to get a rendering project completed. "Using (the) Nehalem (processor), Dreamworks can almost do it in real time and it is only going to better with Larrabee," said Nick Knupffer, an Intel spokesperson.

Larrabee is "Intel's first many-core architecture," Forsyth said. "The first product will be very much like a GPU. It will look like a GPU. You will plug it into a machine and it will display graphics," he said. (GPU stands for graphics processing unit.)

"But at its heart are processor cores, not GPU cores. So it's bringing that x86 programmable goodness to developers," Forsyth said. Larrabee will carry the DNA of Intel's x86 architecture, the most widely used PC chip design in the world.

ntel is touting the performance of Larrabee's vector unit.

(Credit: Intel)

"It's based on a lot of small, efficient in-order cores. And we put a whole bunch of them on one bit of silicon. We join them together with very high bandwidth communication so they can talk to each other very fast and they can talk to off-chip memory very fast and they can talk to other various units on the chip very fast." In-order processing cores are used, for example, in the original Pentium design and in Intel's Atom processor.

"It's the same programming model they know from multicore systems already but there's a lot more of them," he said.

The centerpiece of the chip's core is the vector unit, used to process many operations simultaneously. "The interesting part of the programming model is the SIMD (single instruction, multiple data) vector unit and the instructions that go with it," Forsyth said. "We want to show off this big new vector unit and the instruction set."

Forsyth described what the vector unit can do and how it works with the scalar unit. "(The vector unit) can do 16 floating point operations every single clock. That's a lot of horsepower. Even in just one of these cores--and we have a lot of these cores. So it's a very high-throughput unit. The good thing is that it's independent of the scalar unit. You can issue instructions on the scalar unit and vector unit at the same time. The scalar unit is extremely useful for calculating addresses, doing flow control, doing housekeeping--and keeps all those miscellaneous tasks off the real powerhouse, which is the vector unit."

At GDC, Intel is encouraging developers to experiment. "They're going to have questions about how do I find 16 things to do at once. But a lot of it is just getting in there and playing with the thing," according to Forsyth. The GDC sessions will be a tour around Larrabee's instructions--"how to actually use these new instructions," he said.

And what about markets beyond gaming? "A funny thing happened on the way to the architecture. We designed this architecture to be 100 percent graphics focused. Whatever we needed to do to get graphics good, we did. And then a year ago, we looked at what we had and said how much of this stuff is actually specific to graphics. It turns out, very little. Graphics workloads are increasingly similar to GPGPU (general-purpose graphics processor unit), increasingly similar to high-powered (high-performance) computing. So, we actually have very little that is specific to graphics. Most of the instruction set is very general-purpose."

Nvidia was hit by a Standard & Poor's ratings downgrade Wednesday, according to a Dow Jones report, adding to the graphics chipmaker's woes.

Standard & Poor's ratings services ratcheted down its outlook on Nvidia from positive to stable, according to Dow Jones. S&P cited concerns about the graphics chipmaker's sinking revenue and profitability.

The ratings agency maintains a junk-level BB- grade on the company, and S&P noted that revenues from Nvidia's recent efforts to expand into cell phones, handheld devices, and supercomputer applications is small, according to Dow Jones. Nvidia also faces new competition from Intel's upcoming Larrabee graphics chip, S&P said.

In addition, Nvidia has been grappling with graphics chip issues on laptops. Apple issued a fix Wednesday for the graphics problems suffered by some owners of the new 17-inch MacBook Pro.

Not everything was bad news for Nvidia Wednesday, however. Ironically, on the same day, Broadpoint AmTech analyst Doug Freedman upgraded Nvidia to "buy" from "neutral" saying that "our checks indicate an improving top-line allowing the company to grow from a larger revenue base...as the trough appears less deep than originally thought," he wrote in a research note.

"Although we remain cautious on gross margin mix we do believe that (operating expenditures) are ahead of expectations, potentially adding leverage to our estimates," Freedman said.