Apple's long-rumored Mac Pro update could signal a return to NVIDIA for graphics based on claims about production progress on Tuesday. The company had reportedly been soured based on is experience with drivers and hardware failures, MIC Gadget heard. Instead, it would use NVIDIA's Kepler hardware, although which exact parts weren't mentioned.
NVIDIA took a major step towards spurring growth of its CUDA general-purpose code technology for video cards on Tuesday by posting the CUDA source code. Developers and education now have access to a variant on the LLVM compiler that will let them add new processor types and languages. The extension could see CUDA run on AMD's Radeon hardware, Intel's integrated graphics, and even use relatively old code like Fortran.
The desktop versions of Intel's Ivy Bridge processors have most of their details outlined in full through a roadmap discovered this week. The X-Bit Labs copy shows all the chips falling under the 3000 series in the same Core i3, i5, and i7 tiers, with four cores still the maximum for non-Extreme chips. Clock speeds would have a higher baseline, starting with a 2.7GHz Core i5 (3.2GHz after Turbo Boost) and peaking at a 3.5GHz Core i7 (3.9GHz).
Open Computing Language (OpenCL), the framework that allows applications to use GPU cards for non-graphical computing, has been updated to version 1.2, just 18 months after v1.1. Used on all major platforms, OpenCL 1.2 enhances parallel programming flexibility, and now includes support for device partitioning, DX9 Media Surface Sharing, DX11 Surface Sharing, enhanced image support and separate compilation and linking of objects.
Intel's Ivy Bridge processors should both provide a speed upgrade for their integrated graphics but anticipate the appearance of 'retina' displays on Lion and Windows 8 PCs. The new video core will support up to a 4K resolution at any ratio, letting it show 4,096 pixels both wide and high. The output would allow for ultra-dense professional LCDs and future home displays that emphasize print-like sharpness over sheer resolution.
AMD and MulticoreWare on Wednesday said they were working together to help optimize AMD's Fusion processors for OpenCL. Multicoreware will produce "advanced tools" to help make software that can use the graphics core woven into the processor to speed up general purpose computing. The techniques would make the most of the available hardware including both through keeping the graphics and CPU busy as well as by optimizing apps to balance them properly on multiple cores.
NVIDIA today brought its second-generation Fermi hardware into the true mid-range while simultaneously resurrecting the Ti badge not used since the GeForce 3 and 4 days. The GeForce GTX 560 Ti directly replaces the GTX 460 and is about a third faster, owing both to 384 visual effects cores (up from 332) as well as a much higher 822MHz core and 1.64GHz shader clocks (up from 675MHz and 1.35GHz). Eight hardware tessellation engines also give it a steeper performance increase for DirectX 11 and OpenGL games that support the feature.
More clues have been unearthed in the iOS 4.3 beta that have given a preview of the graphics Apple will use in 2011 iPad, iPhone and iPod updates. A driver bundle is included that makes reference to the PowerVR SGX543, a video core not used in any shipping Apple product. The MacRumors tip didn't link it to a specific device, but Apple has usually implemented a single graphics technology equally across all its iOS devices.
Intel and NVIDIA today settled their longstanding chipset dispute in a deal that heavily favored NVIDIA. The truce will see Intel pay NVIDIA $1.5 billion to license all of the patents for NVIDIA's graphics cores. NVIDIA will keep use of Intel's patents, outside of proprietary x86 processors and "certain chipsets."
NVIDIA today confirmed its second-ever GeForce 500 series chipset in a push to bring its new graphics to the mainstream. The GTX 570 has the exact same 480 cores as the old GTX 480 but, through the refined architecture, runs at a higher 732MHz main clock speed, 1.46GHz clock for each core and a 1.9GHz memory clock. It uses a narrower 320-bit memory interface (down from 384 bits) but, due to the combined improvements, has a higher texture fill rate bumped up from 42 billion to 43.9 billion pixels per second.
NVIDIA overnight quietly brought out its first 500-series GeForce notebook graphics. The GeForce GT 540M like the GTX 580 is primarily a clock speed increase with an increase in its main and effects core clock speeds to 672MHz and 1.34GHz each. It shares the GeForce GT 435M's 96 cores and 128-bit memory bandwidth.
Intel's Sandy Bridge notebook processors will include five low-power processors for ultraportables, a roadmap slip has uncovered today. The range would start with three low-voltage Core i7 chips, the 2610LM, 2620LM and 2640LM. Two ultra-low voltage chips are set to make their start, Digitimes said, including the Core i5 2530UM and Core i7 2630UM.
AMD tonight kicked off the launch of a new graphics core generation by launching the Radeon HD 6850 and 6870. The designs are roughly on par with the performance of the outgoing 5850 but, through a new architecture, are considerably cheaper at $179 (6850) and $239 (6870) while still being more power efficient, particularly at idle. The chip designer claims a performance edge of as much as 30 percent over the GeForce GTX 460; the move has already forced NVIDIA to drop the GTX 460's average price to $199 to compete.
A US patent filing published today has raised concerns as it could give Microsoft control of hardware-accelerated video encoding. A continuance of an application for "accelerated video encoding using a graphics processing unit" would cover the common technique of calculating motion for video processing with the video chipset in a computer rather than the regular, usually slower main CPU. Its techniques are broad and cover tricks like turning each frame into macroblocks to process them in parallel.
NVIDIA today confirmed its rumored entry-level video chipset, the GeForce GT 430. The design is targeted at home theater computers and other systems where video playback support is more important than raw 3D performance and now supports HDMI with passthrough audio. It can output HDMI 1.4a for full 3D video along with Dolby TrueHD and DTS-HD Master Audio surround sound.
NVIDIA chief Jen-Hsun Huang today at its GPU Technology Conference said his company would bring its CUDA general-purpose computing language directly to x86 chips. The approach developed with the Portland Group will let systems without NVIDIA cards handle the code. It will work best with multi-core processors and is seen as ideal for servers.
NVIDIA this morning released its cheapest ever graphics chipset based on its current Fermi core. The GeForce GTS 450 scales back to 192 effects cores and a 128-bit memory bus but is expected to be on par or faster than its arch-rival, AMD's Radeon HD 5750. It's pitched as an alternative to integrated graphics as a partial GF106 core is both cheaper and much more energy efficient than a full GF106 core or earlier designs.
NVIDIA's first mainstream GeForce 400 notebook chipset, the GT 425M, has surfaced in leaks earlier this month. Semi-Accurate noted that the graphics core has been listed as showing in a 17-inch ASUS notebook, but with different features. Some list it as only a DirectX 10 chip, implying that it's only a refreshed 300M chip, while others mention DirectX 11 and that it would use a version of the Fermi architecture in the GTX 480M.
The Khronos Group today published the first specification for OpenGL 4.1 in what's considered a coup for desktop graphics. The standard catches up to DirectX 11 in visual features and overtakes it in integration with other standards: it can now sync graphics with OpenCL to take advantage of video hardware's general-purpose math features. Mobile app developers also now have full compatibility with OpenGL ES 2.0, theoretically letting a developer write an app for the iPhone or Android without having to change the visual effects when porting to a computer.
NVIDIA today brought out its first more frugal GeForce 400 series chipset in the form of the GeForce GTX 460. The recently leaked hardware uses the new 40 nanometer, GF104 chipset and is actually in some areas faster than the GTX 465. It has fewer stream (visual rendering) processors, at 336 versus 352, but has more texture addresses at 56 compared to 44; it also has a faster clock speeds across the board with a 675MHz core, 1.35GHz shader (effect) clock speeds, and a 900MHz actual speed for its GDDR5 memory.
Standards body Khronos Group today rolled out OpenCL 1.1, a new version of the universal general-purpose computing format. It chiefly adds better integration with other devices: OpenCL events can be linked to those in OpenGL to have more math and graphics events start at the same time. Commands can also come from multiple destinations, and memory buffers can be spread across more than one device.
NVIDIA today launched its first mainstream graphics chipset in the Fermi family with the GeForce GTX 465. The chipset scales back from the GTX 470 with 352 visual effect (stream) processors versus 448, 32 render output processors versus 40, and slightly slower 802MHz GDDR5 memory on a 256-bit bus. It still runs at the same clock speed and carries the full feature set, which gives it DirectX 11 (and OpenGL 4) features as well as higher-performance general computing in CUDA, PhysX and OpenCL.
NVIDIA today started off its GeForce 400M notebook graphics line by launching the series' highest-end model. The GeForce GTX 480M has almost three times as many shader (effects) cores as the 285M it replaces at 352 and carries the hardware tessellation, cache and other changes that make it a generational leap. NVIDIA claims that the 480M can be as much as five times faster than ATI's Mobility Radeon 5000 series in graphics duties and 10 times faster encoding video when using technology like CUDA or OpenCL.
AMD today brought its most recent graphic core to the workstation field by launching the ATI FirePro V8800. It supports hardware tessellation and other DirectX 11 (or OpenGL 4.0) features as the Radeon HD 5000 series, but with workstation-optimized hardware and software. The core pushes up to 2.6 teraflops of computing power and becomes especially convenient for pros with Eyefinity, as even a single card has four DisplayPort outputs that can combine to form larger-resolution virtual displays.
NVIDIA at PAX East tonight finally released the first video chipsets based on its Fermi architecture. The top-end GeForce GTX 480 leads the group and is billed as the "fastest GPU in the world:" it has 480 visual processing cores, 16 geometry units and four raster units that combined should beat the ATI Radeon HD 5800 series in real-world tests. It also has major optimizations to multi-card SLI that produce a 90 percent speed boost with a second card, making the case for multiple GeForce 400 series cards in high-end systems.
The Khronos Group today outlined the official feature set for OpenGL 4.0. The new graphics standard is the first major update since the launch of OpenCL and better exploits general computing features. It can draw the output of an OpenCL calculation without having to invoke the main processor and potentially frees up the processor even more when video or a similar task is already being offloaded to the graphics core.
Smartphones will have graphics on par with a PlayStation 3 in three years, Imagination Technologies said in an interview on Wednesday. The company estimates for Gizmodo that long-term future mobile PowerVR graphics cores in development today and will not only produce much more 3D detail but will do so at 720p, whether through an HDMI output or on similar-resolution displays that should be readily available at the same time. Multi-core graphics are likely to be a central ingredient of the new technology.
NVIDIA late yesterday put up a video (viewable below) that provides some of the first official performance indicators for the GeForce GTX 480. In a synthetic test of Uniengine, a 3D engine that supports DirectX 11 (OpenGL 3.2) features, the GTX 480 was tied in maximum frame rates with the ATI Radeon HD 5870 but is much faster under high stress, often producing over 40 frames per second where the AMD-made rival produces a noticeably choppier 20. The test isn't necessarily reflective of real gameplay, and is the same demo Electronista saw at CES, but shows the card likely being faster for games that use techniques like tessellation to dynamically add detail.
NVIDIA's GeForce GTX 470 and 480 will only be available in short supply and at very high prices, a leak from within the graphics field claims. Although the company has promised a launch on March 26th at PAX East, boards based on the new architecture will reportedly only be available through the companies most loyal to NVIDIA and don't sell any ATI-branded graphics. The sources allege that "second-tier" third parties that sell both brands will have to wait until sometime in April.
NVIDIA will officially launch the GeForce GTX480 and GTX470 graphics cards next month at the upcoming PAX East gaming conference, according to a Twitter post. The components are the first to integrate the company's Fermi architecture. The technology is said to improve graphics performance while expanding support for other standards such as OpenCL and PhysX
AMD this morning launched its second new video chipset in just a matter of days and this time targeted a comparatively untapped mini PC category. The Radeon HD 5570 is a major step up in performance from the 5450 with 400 stream processors versus 80 but still occupies a single card slot and is relatively short, making it ideal for small form factor cases where even the 5600 series would be too large or use too much power.
AMD this morning launched its least expensive video chipset capable of DirectX 11 (and OpenGL 3.2) graphics. The ATI Radeon HD 5450 has all the same visual effects as the 5600 and 5800 lines but is trimmed down to 80 stream processors, 8 texture units and a 650MHz engine clock speed. Besides keeping the price low, the design allows for extremely quiet cards: it can either use a low-speed fan in a single slot or a completely fanless design in a slightly larger space.
NVIDIA in a Twitter update has confirmed the names of the first two graphics chips based on its Fermi architecture. The GeForce GTX 470 and 480 should be the first to ship. Details of their performance weren't given, but the naming scheme positions them as top-end parts for gamers.
Although it outlined some products at CES, NVIDIA has now formally detailed its GeForce 300M graphics line. The new series still uses a DirectX 10-level (OpenGL 2.1) architecture but is more efficient, in some cases using as many as 50 percent more shader (visual effect) cores. The series continues to provide full hardware video decoding and supports general-purpose computing like CUDA and OpenCL.
Imagination has used its time at CES to unveil a new PowerVR SGX graphics core that is likely to find its way into the next iPhone. The SGX545 is the only mobile phone video chip capable of both OpenGL 3.2 and DirectX 10 level visual effects, including their newer geometry and texturing features. OpenCL support is equally unique: unused processing cycles on the PowerVR chip can be used for general purpose computing and offload work from the main CPU.
AMD tonight claimed the performance crown by launching the ATI Mobility Radeon HD 5000. The new graphics core is the first to support DirectX 11 (and equivalent OpenGL 3.2) visual effects, and on most models can also handle DirectCompute and OpenCL general computing. It should also be AMD's most performance-per-watt most efficient chipset as it can push over a teraflop per second in the fastest models but, through the new design and a smaller 40nm process, consumes much less power at the same time.
AMD on Monday introduced an update to its ATI Stream SDK. The utility allows developers to utilize combined CPU and GPU power for accelerating applications. Version 2.0 now supports OpenCL 1.0 and atomic function for 32-bit integers, while adding a new Visual Studio 2008-integrated ATI Stream Profiler.
AMD on Wednesday bragged of the graphics performance crown with the launch of the ATI Radeon HD 5970. The card uses two 5800 series chipsets on one card to provide the most performance possible. While each has the speed of a 5850 with a 725MHz primary clock speed and 1GHz GDDR5 memory, the 3,200 stream (effects) processors, 160 texture units and 2GB of memory give it as much performance as two 5850s but in half the space.
NVIDIA in a low-key move today launched the GeForce GT 240. The chipset brings performance from the mid-level to sub-$100 cards and uses the newer 40 nanometer manufacturing process to make itself a reasonable fit in budget PCs: its low energy use both helps it occupy only one slot and to run entirely off the power of the PCI Express bus instead of needing a separate power connector.
NVIDIA today provided details of the first official hardware to use its upcoming Fermi architecture. The Tesla 20 series is even more optimized for general-purpose computing standards like OpenCL or NVIDIA's own CUDA and handles complex math that previously hasn't been as practical, such as ISO standard double-precision math and C++ code processing. Unlike past models, though, the card model also has a video output and works as a video card rather than just as a companion device.
The just-unveiled Fermi graphics architecture will find its way into Macs and play an important role in Mac OS X Snow Leopard, NVIDIA chief scientist Bill Dally said today. While it's expected that NVIDIA would continue to play an important part of future Macs, the researcher drew a particular connection between the new GPU design and Apple's new OS, expecting that it would provide a significant boost for those apps that implement OpenCL. Windows 7 will also get support through DirectX 11 and DirectCompute.
NVIDIA's Fermi architecture will start with three models if it ships before the end of the year as promised, one leak from Friday claims. A flagship single-chip model would have the 512 cores NVIDIA is advertising, but a second model would, like the GeForce GTX 295, have two slightly less powerful chipsets on one card that combined would be much faster. The slowdown may be necessary as Fudzilla believes the card would have a peak thermal power of 300W.
NVIDIA this evening provided an early look at the next generation of its graphics processors. Nicknamed Fermi, the architecture for future GeForce, Quadro and Tesla chipsets will jump from 240 cores to a much larger 512 and should be much faster in each core courtesy of some industry-first techniques. Fermi chips will be the first GPUs to have a real cache hierarchy, with Level 1 caches to keep specific information on hand and a single, shared Level 2 cache for larger tasks; they will also have a new GigaThread engine that can transfer data in both directions at once and handle "thousands" of tasks at once.
AMD's graphics label ATI tonight claimed to set a new benchmark for graphics with the Radeon HD 5800 series. The new video hardware is theoretically twice as fast as the 4870 and has an extremely large set of 1,600 stream (visual effect) cores -- enough to calculate 2.72 teraflops per second. Besides handling twice as many rendering tasks at once, the 5800 is also running a sixth-generation engine that shades and tessellates geometry more quickly, more GDDR5 memory bandwidth (150Gbps), and much improved techniques for antialiasing and anisotropic texture filtering.
AMD's major video chipset update this fall has had its feature set and release date uncovered thanks to a slip of information Monday. The ATI Radeon HD 5800 series should be fronted by the 5850 and 5870, both high-end cards that could be the first to support Eyefinity, a technology that not unlike Matrox cards will support three displays on a single card. They are also to be AMD's first graphics cores to support DirectX 11 and OpenGL 3.1 visual effects in hardware.
AMD is developing an integrated graphics chipset for budget PCs that could take the performance crown in the category, a leaked roadmap shows. Codenamed RS880 but likely to be badged Radeon HD 4200 when it ships, the design will have a new graphics core that should run about 15 percent faster than "anything comparable" on the market, implying that NVIDIA's GeForce 9400 or 9400M may be the primary targets. The Inquirer adds that it should support AMD's Stream general-purpose computing technology and, by extension, would eventually support OpenCL.
The sequel to NVIDIA's Ion platform could be much faster than its predecessor when it ships, leaks today would show. While the current graphics and chipset combo is based on the GeForce 9400M and has just 16 visual effects cores, Fudzilla now hears that it will have 32 cores, potentially doubling the amount of simultaneous effects it can handle at once. The difference will have the most dramatic effect on 3D but could also impact general-purpose computing tasks that need CUDA or OpenCL.
AMD this morning said it has demonstrated the world's first graphics processor capable of supporting DirectX 11 and matching features in other software standards. The part is capable of tessellating polygons in real time, or breaking them down automatically into smaller shapes. Games and other apps that use the feature can have smoother details to characters or terrain without requiring as much processing power or that artists add all the information themselves.
Apple has posted a job listing that hints at the company's future hardware direction for the iPhone, iPod touch and possible other devices. The position for a High Perform/Low Level Programmer asks for someone familiar not only for programming assembly-level code for ARM processors, which Apple already uses in its handhelds, but for the NEON vector math units used in the newer Cortex architecture for the mobile chips. Apple is especially concerned about experience with vector math and particularly values anyone with additional knowledge of vector units through general CPUs, such as Intel's SSE or the AltiVec units found on PowerPC G4 and G5 cores.
NVIDIA has released version 2.2 of the CUDA toolkit and SDK. The new version of CUDA profiles C coded applications to use the computers GPU in addition to the CPU. This can enhance the computer's overall processing speed. The update also includes the ability to zero-copy, or to read and write directly from pinned system memory rather than send it through the graphics memory system. Using the technique requires either a GeForce 9300/9400 (or 9400M) integrated graphics core or else a GeForce GT 200 series board.
