Intel conducted the first-ever demonstration of its massively-parallel Larrabee GPU yesterday at the Intel Developer Forum.

Larrabee is Intel’s first foray into discrete, or standalone, GPUs in about 10 years. The new design is said to use an array of x86-compatible execution engines which are harnessed to crunch data in parallel. Intel hopes to leverage this architecture to compete directly with ATI and NVIDIA in gaming, and later in other markets.

At hand, the IDF demonstration called upon their GPU to render Enemy Territory: Quake Wars with ray tracing. Ray tracing is a rendering technique that accurately imitates in 3D the real world interaction between light and objects. This gives scenes rendered via ray tracing a very high level of photorealism, but it has traditionally come with a heavy performance penalty.

The water in the movie, says Intel senior research scientist Bill Mark, was done in real time with just ten lines of C++ code. The same effect with today’s rendering technique would require a lot of  “cheating” in the form of textures, pre-rendered elements and lighting tricks.

Ray tracing is still too slow for 30-60 FPS gaming, but empowering developers with a platform that emphasizes it might make it viable in a few years. Should that happen, the days of depending on the skill of a game engine’s developer to produce outstanding lighting just may be numbered.

Products based on the Larrabee architecture are on schedule for a launch in 2010.

After months of speculation, AMD has pressed the big red “go” button on the Radeon HD 5000 series. The reviews are pouring in–and we’ll get to those–but we want to focus on what improvements have been made to make for a superior end-user experience.

Raw horsepower

One of the big initiatives over the last few years has been the field of “stream computing.” This fledgling industry leverages the power of today’s GPUs–which are more like CPUs than ever before–to handle tasks like physics, video coding, and game AI. In essence, GPUs are now crunching the numbers that once required a CPU.

Both ATI and NVIDIA have worked to develop the stream computing initiative. Between NVIDIA’s CUDA language and ATI’s Stream Technology, both major industry players now have a solution to do stream computing. But just like the processor, heavier workloads and higher performance call for faster chips, and the HD 5000 series is no slouch in that department.

According to AMD, the new Radeon HD 5870 offers 544 double-precision GFLOPS of processing power. The FLOP stands for FLoating point Operations Per Second. A floating point operation is a basic calculation used by the CPU to process code, especially “scientific” ones like computer AI, video encoding and physics. Double-precision FLOPS ensure a high degree of accuracy in these calculations, which translates to more accurate rendering or encoding. The Radeon HD 5870 can perform 544 billion such double-precision calculations every second; in comparison the Intel Core i7 975 XE chip can only perform 70 billion of them.

(more…)

What conclusion can you draw when a GPU made with a brand new architecture appears on Newegg on the eve of large press event held by its maker? The launch is here!

A Radeon HD 5870 (”Cypress XT”) has appeared on Newegg for a preorder price of $379.99, which is right in line with the price we expected. Made by Sapphire, the card appears to be a reference design with 1GB of 1200MHz (4.8Gbps) GDDR5, an 850MHz core clock, 1600 shaders, 2 DVI ports, 1 HDMI port, and a single DisplayPort.

Readers interested in taking a peak at the SAPPHIRE 100281SR Radeon HD 5870 can pop on over to the product page for a closer look.

NVIDIA’s boisterous CEO Jen-Hsun Huang will purportedly meet with Taiwan Semiconductor Manufacturing Corporation in October to discuss 40nm prices.

As a so-called fabless company, NVIDIA has long relied on the manufacturing capabilities of foundries like TSMC to realize the designs its boffins dream up. The most recent designs, such as those for the mythical GT300 GPU, happen to require 40nm.

Indeed, receiving word that Huang is prepping to chat TSMC up regarding 40nm supports prior speculation that the GT300 is on the November-January train. Were the GT300 being prepped for September or October, 40nm talks would have happened in July or August.

Better three months late than never, we guess.

If reports from ATI-Forum.de can be believed, we now have official pricing, specs, and launch dates for AMD’s upcoming Radeon HD 5000 series GPUs.

Radeon HD 5870

Launch date: September 22
Core Frequency: 850MHz
Memory: 1600MHz GDDR5
Shaders: 1600
Consumption: 188/27w max/idle
Price: < $400.00 USD

According to AMD benchmarks (grab your salt), the GPU is roughly equivalent to NVIDIA’s reference GeForce GTX 295 adapter which sandwiches an SLI configuration into a single card. Put another way, the HD 5870 is allegedly some fifty percent faster than the single-GPU GeForce GTX 285.

Radeon HD 5850

Launch date: September 22
Core Frequency: 725MHz
Memory: 1000MHz GDDR5
Shaders: Unknown
Consumption: 155/20w max/idle (est.)
Price: < $300.00 USD

Radeon HD 5870 X2

Unfortunately for extreme GPU enthusiasts everywhere, the least amount is known about the 5870 X2. We know for certain that the card is a dual-GPU solution consisting of twin HD 5870 cores, but rumors allege that the switch to 40nm production has allowed for both GPUs to fit on the same PCB. The same body of rumors also claim that the card will fall just shy of $600 USD and, more reasonably, say that it will be a dual-slot solution.

If indeed AMD’s performance figures regarding a single 5870 can be believed, CrossFire’s traditional scalability tells us that a 5870 X2 should deliver another 40-60% on the GTX 295. If our numbers are right, that will make the 5870 X2 fully twice as fast as the current crop of single-GPU cards.

Sadly, we know that this card will not in fact launch on September 22 with the rest of its family. The 5870 X2’s launch is set for a later date, and we’re not quite certain when that is.

ATI has taken the wraps off of the newest driver revision for all Radeon 2000, 3000 and 4000-series GPUs. Weighing in at 75-100MB, the WHQL-certified Windows XP/Vista/7 driver delivers the following changes:

• Removed a noticeable delay when alt+tabbing between a 3D-accelerated application and another app in Windows Vista.
• OpenGL apps will no longer cause the second monitor to blank on some GPUs in Windows 7.
• Several AVIVO/HDMI fixes specifically for Windows XP.
• Several AVIVO/HDMI fixes for all Windows OSes.

There is also a huge chunk of bug fixes for both Vista and XP, so be sure to check the release notes (PDF).

Downloads

Windows 7 x86-64: Here
Windows 7 x86-32: Here
Windows Vista x86-64: Here
Windows Vista x86-32: Here
Windows XP x86-64: Here
Windows XP x86-32: Here

For a peek at the power of AMD’s upcoming Radeon HD 5000 series, the firm recently demonstrated one GPU powering six 30″ monitors for a combined resolution of 7680×3200.

AMD calls it Eyefinity, and it’s a technology that seamlessly combines multiple monitors into what AMD calls the SLS, or Single Large Surface. Rigging up the monitors into a single mighty display is often awkward if left to Windows, even Windows 7, but Radeon HD 5800 products will smoothly take care of it at the driver level.

The technology isn’t limited to just six 30″ monitors, either. AMD promises that Eyefinity will work on fewer displays just as well. It’s all a matter of personal preference, of course, but $600 in 22″ displays may be a fair sight better to some than a single larger monitor for the same price.

AMD has made a series of demo videos available to make their point. AMD’s Simon Solotko presents a more “reasonable” 3×2 array of 1920×1200 monitors in his brief presentation:

AMD also claims it has been working with Samsung to produce displays with razor-thin bezels for a more uninterrupted experience. Surely this is to remedy the fact that a shooter’s crosshair would get divided amongst several displays if more than three are used.

(more…)

Windows 7 is less than forty days away from its retail debut, and it carries the new DirectX 11 specification in tow. AMD will soon fire the first shots with a suite of cards, but what are you really going to get? Today we’ll be talking about the changes DirectX 11 brings to the table and what they mean for you.

DirectX evolved

DirectX 11 is the next iteration of the Microsoft DirectX architecture. DirectX is a collection of common application programming interfaces, or APIs, that standardize the way code for a GPU is written and executed. As a result of DirectX, and its “competitor” OpenGL, users across the globe can run any game on any GPU from any company. Whether or not that game runs well depends on the hardware’s horsepower, but DirectX has helped to avoid a standards war not unlike the one that divided Blu-ray from HD-DVD.

As the eleventh major entry into the DirectX annals it’s not quite as drastic as DirectX 10 that came before it. While DirectX 10 was all about dramatic improvements in lighting and texturing, DX11 is a more subtle movement that emphasizes GPU-accelerated apps and refined scene detail.

DirectX accomplishes its more transparent emphasis in three distinct ways.

(more…)

NVIDIA has taken the wraps off of the newest driver revision for GeForce 100, 200, 6000, 7000, 8000, 9000 GPUs. Weighing in at roughly 80-120MB, this WHQL Windows XP/Vista/7 driver delivers the following changes:

• Adds support for OpenGL 3.1
• Adds support for CUDA 2.3 for improved performance in GPU Computing applications. See CUDA Zone for more details.
• Adds a new user-controlled power management setting for select GeForce 9-series and later graphics cards (only available on cards that already support more than one power state). This option allows users to set a performance level for each DirectX or OpenGL application.
• Includes several new control panel features and numerous bug fixes. More information can be found in the release documentation.

There is nothing earth shattering in this release, but it’s always a cozy feeling to have the newest drivers, so get your fix below.

Download

Windows Vista & 7 x86-32: Here
Windows Vista & 7 x86-64: Here
Windows XP x86-32: Here
Windows XP x86-64: Here

PC Magazine writes that PCI SIG, the overseer of all things PCI, has decided to delay the release of the finalized PCI Express 3.0 specification until 2010.

The delay will push products based on the PCIe 3.0 spec into 2011, given the average delay between design and implementation. The change in scheduling is an obvious departure from the original spec that called for a 2009 release, with products coming the following year.

The delay, says PCI SIG president Al Yanes, is owed to extra effort being done to assure that current PCIe 1.0 and 2.0 products will work in 3.0 slots.

The new PCI Express 3.0 interfaces uses 8GHz signaling, compared PCIe 1.0’s 2.5GHz and PCIe 2.0’s 5GHz. The new bus also switches from 8-bit and 10-bit data encoding to 128- and 130-bit encoding. Yanes confirmed that extra effort is under way to verify that each frequency’s electrical model can handle the new encoding scheme.

Yanes said that PCI SIG members were largely happy with the decision to delay the specification for the sake of backwards compatibility testing.

NVIDIA has just announced that Intel and major motherboard manufacturers have all grabbed an SLI license for motherboards based on the upcoming LGA1156 P55 chipset. It’s not any particular surprise, but it’s a win for the budget conscious.

SANTA CLARA, CA—AUGUST 10, 2009—NVIDIA Corporation today announced that Intel Corporation, and the world’s other leading motherboard manufacturers, including ASUS, EVGA, Gigabyte, and MSI, have all licensed NVIDIA® SLI® technology for inclusion on their Intel® P55 Express Chipset-based motherboards designed for the upcoming Intel® Core™ i7 and i5 processor in the LGA1156 socket.

Based on a modern contemporary of the C language called C99, OpenCL is a framework that rapidly and easily assists developers in the creation of scalable applications. Imagine a single body of code that could just as easily run on a CPU, GPU, DSP, or any other execution engine; that is the promise of OpenCL. That promise has inched a little bit closer to reality as AMD yesterday announced that it has submitted the industry’s first x86 OpenCL development platform for certification with OpenCL’s overseers, the Khronos Working Group.

The inclusion of an x86 OpenCL platform in AMD’s Stream SDK 2.0 is significant as it will allow a developer to code against ATI’s graphics API for execution on any x86 CPU. Using the Stream SDK, a developer can freely choose to execute code on the GPU or CPU, without committing exclusively to one or the other. The processing workload could even be divided between the parts, which would give a single body of code more computational horsepower than ever before.

“By supporting multi-core CPUs and GPUs with our OpenCL environment, AMD gives developers easy access to both processing resources, so they can efficiently write cross-platform applications for heterogeneous architectures with a single programming interface,” said SVP and GM of the AMD Products Group Rick Bergman. “AMD is supporting OpenCL with our ATI Stream SDK as an enabler of wider GPGPU adoption among developers and users.”

For a rough example of OpenCL’s power, we need only look back to a demonstration conducted at SIGGRAPH ‘08: