Slides from AMD’s (NASDAQ:AMD) event at the PC Cluster Consortium event in Osaka, Japan offer up details regarding the chip vendor’s roadmap for its GPUs and CPUs, with the takeaways including the launch of a 300W APU targeted at the HPC segment along with the launch of new CPU cores.

At the event, Junji Hayashi, Consumer and Commercial Business Lead at AMD Japan, shared details on the vendor’s K12 ARM as well as the x86 Zen CPU cores. AMD is looking to introduce both ARM as well as traditional x86 cores in a pin compatible platform that is codenamed SkyBridge. Aimed at the server, embedded, semi-custom and client markets, both the ARMv8 and the x86-powered cores will offer 64-bit computing and will be manufactured using a 14-nm FinFET process. SkyBridge will be launching before the end of the year, although an exact launch window was not provided.

The K12 core will feature Simultaneous Multi-Threading, which is a departure from the Clustered Multi-Thread technology that is currently utilized in AMD’s offerings. SMT will offer larger CPU cores the ability to achieve a higher throughput by allocating underutilized resources to an additional, slower, execution thread.

As for GPUs, Hayashi mentioned that AMD would be moving to a two-year release cadence cycle for updating the GPU architecture of APUs. There was also a mention of a High Performance Computing APU, which is said to be radically different from existing designs in that is features the Stacked HBM (High Bandwidth Memory) memory standard. The HPC APUs will slot in between the standard server cores and the FirePro line of cards.

The current generation HBM is nine times faster than GDDR5 memory and 128 times faster than DDR3. AMD is said to be utilizing the same standard in its Radeon 300 series, with the technology itself developed in collaboration with SK Hynix.

The post AMD Reveals Five-Year Roadmap For GPUs and CPUs appeared first on VR World.

After launching the Titan X earlier this month, it looks like Nvidia (NASDAQ:NVDA) is set to unveil a second video card based on the GM200 silicon, the GeForce GTX 980 Ti. The card is rumored to offer significant increase in better performance when compared to the GM204-based GTX 980, which is why Nvidia may decide to call the card the GTX 990.

Based on the latest leak, the GTX 980 Ti is said to offer the full complement of 3,072 CUDA cores available in the GM200, while reducing the video memory to 6GB GDDR5, down from the 12GB featured on the Titan X. Memory bus will remain the same at 384-bit, and with the GTX 980 Ti, Nvidia will allow its partners to offer custom variations of the reference design that include custom cooling solutions and factory unlocked models.

Doing so has the potential for the GTX 980 Ti to be faster than the Titan X, which is only offered in a reference design with the stock cooler. There’s no confirmation as to when we’ll see the card, with the rumor only suggesting that Nvidia is targeting its launch to coincide with that of AMD’s (NASDAQ:AMD) Radeon R9 390X. The Radeon R9 300 series will make its debut at Computex, so it is possible that we may see the GTX 980 Ti at around the same time.

The post Nvidia May Already Be Working On A GTX 980 Ti appeared first on VR World.

Futuremark has launched a new API Overhead feature test for its 3DMark benchmarking utility, with the test allowing users to test performance differences between DirectX 12, DirectX 11 and Mantle API.

Windows 10 Technical Preview users on build 10041 and the latest video drivers from Windows Update will now be able to access the test through 3DMark Advanced or Professional Edition. To test the DirectX 12 features, users must have DirectX 11-compliant hardware with at least 4GB RAM and 1GB video memory. To run Mantle tests, you need AMD hardware that works with the Mantle API.

Developed in collaboration with AMD(NASDAQ:AMD), Intel (NASDAQ:INTC), Microsoft (NASDAQ:MSFT) and Nvidia (NASDAQ:NVDA), the test’s objective is to determine the “relative performance of different APIs on a single system.” Essentially, you’ll be able to gauge how your current system performs with DirectX 11, and the differences in performance when using DirectX 12.

The API Overhead test works by sending a call to the GPU to draw an object on the screen, which is handled through the API. The more efficient the API, the more number of lines that can be drawn on the screen. The number of draw calls is increased with every iteration, with the final result calculated based on the maximum number of draw calls per second achieved by an API before the frame rate goes under 30 fps.

With DirectX 12 slated for commercial availability later this year, Futuremark’s test offers users a way to check how their current configurations will be able to handle Microsoft’s new API.

The post Futuremark 3DMark API Overhead Feature Test Lets You Benchmark DirectX 12 And Mantle appeared first on VR World.

While 4K certainly has its uses and benefits on larger displays, the potential use cases on such a small smartphone screen are negligible at best. On an average smartphone screen, which is usually around 5.5-6 inches, the resolution of QHD — let alone 4K — is too great for the eye to appreciate. Apple’s Retina display has a pixel-per inch density of 326. A QHD display on a smartphone would equal something in the range of 538.

A PPI of 538 is far beyond the maximum resolution that the human eye can recognize. There is a bit of controversy as to what the highest practical PPI is for a 5.5 inch screen held 12-18 inches away from the eye, but the consensus holds that it’s 477 PPI at 12 inches and 318 PPI at 18 inches. Of course, holding it closer than 12 inches away from your face would make it difficult to properly focus.

So this means that the iPhone’s Retina resolution introduced in 2010 is the highest practical resolution for a smartphone sized screen. Of course with bigger screens found on phablets and full-on tablets this figure jumps.

QHD and 4K smartphones are impractical

The other issue with QHD and 4K displays on smartphones is the substantial battery drain that comes as a result. Battery technology has reached a ceiling, yet the demands keep on increasing. A QHD or 4K screen requires substantially more push from the SoC to drive a screen of that resolution, which means the SoC will require more battery power for all tasks.

Before the 2015 Mobile World Congress kicked off, Huawei (SHE:002502) announced that it would not be producing 4K smartphones because it simply is not worth the battery drain.

“4K needs a lot of power, so if you use it you have to make compromises. A 4K display on a smartphone may give you half a day of battery life but a 2K display can give you maybe one day or more,” President of the Handset Product Line Kevin Ho is quoted as saying. “4K has four times the pixels of Full HD so the power consumption is maybe 4 or 8 times as much as Full HD so the smartphone has to have compromises with battery life.”

All about competition

Pushing up the resolution of smartphone displays  is not the only way to improve the quality of the screen. There’s much to be done to improve color depth, backlight quality, and the ability for the screen to display in bright conditions.

But the resolution race is a way for vendors to compete with each other. The screen resolution of a device is an easy and digestible metric to understand. But for consumers this resolution race isn’t, in the end, a value add as it does little to increase productivity or performance.   

The post Resolution Overkill: Why QHD and 4K are Too Much for Smartphones appeared first on VR World.

EVGA has launched the first GeForce GTX 980 with an all-in-one liquid and air cooler, with the card dubbed the GTX 980 Hybrid. The custom cooling option is claimed to reduce temperatures by a full 25 degrees Celsius when compared to a reference GTX 980, allowing EVGA to substantially overclock the card out of the gate.

The GTX 980 Hybrid features a base clock of 1291MHz and a boost clock of 1393MHz, with the vendor stating that there will be plenty headroom left for further overclocking. The card comes with 4GB GDDR5 memory on a 256-bit interface and a bandwidth of 7010MHz. The liquid-cooled portion directs heat away from the GPU, while the included air cooler is used to cool the memory and VRM sections of the video card.

This isn’t the first AIO liquid-cooled video card, as that distinction goes to AMD’s Radeon R9 295X2, but this is the first time we’re seeing this system used on the GTX 980. Based on Nvidia’s (NASDAQ:NVDA) new Maxwell architecture, the GTX 980 features the GM2014 GPU with 2048 CUDA cores.

The GTX 980 is listed on EVGA’s website for US $649.99/€779,00. If you’re interested in just getting the cooler, EVGA is offering it for $99.99.

The post EVGA’s GTX 980 Hybrid Comes With An AIO Liquid Cooler appeared first on VR World.

After serving up a few details regarding the card at the Game Developers Conference earlier this month, Nvidia (NASDAQ:NVDA) has officially launched its latest offering in the Titan series, the Titan X. The card will be priced at $999 and will be available in the month of May.

The specs on offer with the Titan X match what the last round of leaks predicted, with the card featuring a 28nm GM200 GPU that offers 3072 CUDA cores, 192 TMUs, 96 ROPs, 3MB of L2 cache along with 12GB of GDDR5 memory with a 384-bit memory interface. Base clock for the card is at 1002MHz, with a boost clock of 1078MHz.

Nvidia has stated that there is more than sufficient headroom available for overclocking the Titan X, and the card will reach frequencies of 1200MH without any hassles. The GM200 silicon is basically an enhanced version of the GM204 GPU that was introduced with the GTX 980. We see a 50% increase in bandwidth, CUDA cores and ROP count, with the end result being a card that features a maximum TDP of 250W.

The cooler itself is similar to what we’ve seen in earlier-generation Titan cards, although this time around the metal shroud has a black coat of paint. Nvidia is once again touting that the Titan X is the fastest single-GPU solution available in the world today, with the card billed more as a luxury gaming video card than an entry-level compute card, like the Kepler-based GTX Titan and GTX Titan Black. For that reason, its pricing at $999 may turn out to be the only sore point given that cards like the GTX 980 in SLI can outperform the Titan X.

The post Nvidia Officially Launches The GeForce GTX Titan X appeared first on VR World.

Nvidia (NASDAQ:NVDA) has announced its own computing solution aimed at deep-learning researchers at the GPU Technology Conference. Called Digits DevBox, the system includes four Titan X GPUs in quad-SLI, making it the most powerful configuration of GPU hardware in a single PC available today.

The box will feature the Digits software framework, which gives deep-learning researchers a series of tools that will enable them to configure and monitor deep neural networks, as well as process data. The system itself is powered by Ubuntu 14.04, and comes with the following frameworks and libraries installed: Caffe, Torch, Theano, BIDMach, cuDNN v2, and CUDA 7.0.

Other hardware includes an Asus X99 motherboard with an Intel Haswell-E Core i7  processor, a combination of mechanical and flash-based storage that includes 3×3 TB drives in RAID 5, as well as  M.2 SATA and SSDs along with a 1500W power supply.

The cost of the system at $15,000 is several times more than what you pay for just the hardware, so it’s clear Nvidia is also chrging a premium for its software-driven framework. The chip vendor is now taking registrations for the Digits DevBox, and there’s an online configurator as well that can be accessed by filling out this form.

The post GTC 2015: Nvidia Unveils Digits DevBox Supercomputer Aimed At Researchers appeared first on VR World.

Nvidia’s (NASDAQ: NVDA) CEO Jen Hsun Huang gave the world another look at the GPU successor to Maxwell at its GPU Technology Conference Conference (GTC 2015) in San Jose Tuesday.

Pascal was first announced as a mystery GPU between Maxwell and Volta at last year’s GTC. Tuesday’s announcement gives us the first concrete details of Pascal.

Huang said that Pascal, which is set to arrive in 2016, would have a ten-fold overall average improvement over Maxwell, and a four times boost in mixed-precision workloads. As far as performance per watt, it will offer a two-fold performance over Maxwell. However he later cautioned this was “CEO Math” and actual performance may vary.

Pascal will use a suite of new technologies, including 3D-stacked memory and NVLink (Huang says it will offer a five-fold improvement over PCI-E). It will be built on TSMC (TPE: 2330) 16nm FF+ (FinFet plus — the follow-up to FinFET)  process node. It will also use High Bandwidth Memory, allowing a three-fold improvement of bandwidth for its 32 GB of RAM.

Cards with Pascal will likely be marketed towards CUDA workstations, or perhaps as some sort of competitor to Intel’s (NASDAQ: INTC) Xeon Phi co-processors. Game developers have a tough time pushing the limits of current generation cards as it is.

Pricing and availability of Pascal-based cards will be available later this year or early next year.

The post Nvidia Teases More Pascal Details at GTC 2015 appeared first on VR World.

Although Nvidia announced the Titan X during Epic Games’ demo at the Game Developers Conference earlier this month, the chip vendor is yet to share the full specs of the video card. The card is set to be officially announced at Nvidia’s GPU Technology Conference later this week, but the full specifications of the GM200 silicon — which is what the card is based on — have been leaked.

Videocardz posted the block diagram of the GPU, with the fully-unlocked silicon featuring 6 Graphics Processing Clusters (GPCs), each holding 4 Maxwell Streaming Multiprocessor (SMMs), for a total of 24 SMMs, or 3072 CUDA cores. A total of 6 memory controllers, each with 64-bit memory means that the Titan X has a memory bus width of 384-bit. Other specs include 192 Texture Mapping Units (TMUs) and 96 Raster Operation Units (ROPs).

In terms of bandwidth, the Titan X clocks in at 1002MHz base and 1089MHz boost, and will offer an astounding 12GB of GDDR5 memory at 1753MHz. While not confirmed, the video card is set to have a TDP of 250W, with one eight-pin and six-pin PCI-Express power connectors. Display connectors include three DisplayPort, one HDMI and one DVI port. Synthetic benchmarks of the card have leaked last week, indicating significant gains in performance over the GTX 980.

Pricing of the Titan X is said to be the case as earlier-generation models, which would come out to $999. The card is slated to make its debut tomorrow, which is when we’ll have further details, as well as real-world benchmarks.

The post Nvidia Titan X Specs Detailed Ahead Of Official Launch appeared first on VR World.

Nvidia’s (NASDAQ: NVDA) upcoming GeForce Titan X is shrouded in secrecy until its official launch later this month at the GPU Technology Conference.

But with any product like this there are always leaks. Videocardz has published benchmarks on how the card stacks up to the competition.

Based on these benchmarks, the Titan X appears to have 3,072 CUDA cores, 192 TMUs, 96 ROPs, 1,002MHz core clockspeed (boost is unknown, and the core could be subject to change) and a 1,750MHz memory clock. The memory bus is 384-bit, resulting in 336 GB/s of available bandwidth. We already know that the card will ship with 12GB of VRAM.

Here’s a breakdown of the results:

• 22903 points in 3DMark 11 Performance. A 34.8% boost over a reference GeForce GTX 980
• 7427 points in 3DMark 11 Extreme. A 39.9% boost over a reference GeForce GTX 980
• 17470 points in 3DMark FireStrike. A 36% boost over a reference GTX 980.

The important thing to remember is that this performance is subject to change. The card’s drivers are likely not yet finalized, thus these numbers from synthetic benchmarks are not representative of the final performance of the card.

The card’s exact availability will be announced at GTC later this month. The big sticking point for the card will be its pricing. A competitively priced card will make it a force to be reckoned with.

The post First GeForce Titan X Benchmarks Appear Online appeared first on VR World.

In what may turn out to be a vital move for MediaTek (TPE:2454) in its bid to challenge Qualcomm (NASDAQ:QCOM) in the mobile segment, it is believed that the Taiwanese manufacturer is set to license AMD’s (NASDAQ:AMD) graphics technology for its SoCs.

There are no concrete details on the alleged partnership, which is said to have occurred at Mobile World Congress in Barcelona. Siding with a vendor such as AMD would be beneficial to both companies as it would give MediaTek leverage a significant advantage in the realm of mobile graphics, and AMD would stand to gain a foothold in the ultra low-power market.

Nvidia (NASDAQ:NVDA) already has the Tegra X1 in this segment, and even if AMD does not create an SoC by itself, siding with a manufacturer such as MediaTek would ensure that its offerings are seen in a wide variety of hardware, a feat that Nvidia has failed to achieve thus far with its Tegra chips.

MediaTek is set to up the ante this year with its range of 64-bit octa-core SoCs that offer LTE connectivity, and while its offerings are slowly gaining traction — particularly in markets like China and India — the Taiwanese vendor does not manufacture a custom graphics solution. Options in this segment include ARM’s Mali GPU and Imagination’s PowerVR offerings, both of which MediaTek has used in the past.

Qualcomm, meanwhile, holds a significant lead with its in-house Adreno GPUs, the foundation of which was built on by ATI, which AMD acquired in 2006. AMD finally sold the technology — branded Imageon — to Qualcomm for a meager $64 million in 2009. A deal with MediaTek would bring about a re-entry into the world of mobile graphic solutions for AMD.

A collaboration between the two would also lead to an increased focus on HSA/GPGPU products, considering that MediaTek is a founding member of the HSA Foundation. There’s no mention as to when we will see a MediaTek SoC powered by AMD graphics, and seeing as how Computex is where AMD is said to launch its next-generation desktop video cards, it is likely we’ll hear more about the vendor’s re-emergence into the world of low-power graphics at the event.

VR World has reached out to MediaTek for comment and will be updating this post if the company chooses to respond.

The post MediaTek’s Upcoming SoCs May Feature AMD Graphics appeared first on VR World.

Last week at the Game Developers Conference in San Francisco,  Nvidia (NASDAQ: NVDA) gave us the first glimpse at its upcoming Titan X card.

Not much is known about the card as the exact specs will be revealed during its launch at the upcoming Nvidia GPU Technology Conference taking place later this month. We do know, however, that its GPU will feature 12GB VRAM, the transistor count will be upped from 7 billion to 8 billion, and the memory bus interface is set to 384-bit.

Maximum PC was able to get its hands on the Titan X and posted a few close-up pictures of it. While the magazine wasn’t able to reveal anything more due to an NDA, we do see that the card has both 8-pin and 6-pin power connectors as well as a bevy of output ports: three DisplayPorts, one HDMI, and one DVI.
As the Titan X will be based on Nvidia’s Maxwell architecture, many are looking forward to seeing the card in benchmarking action. With specs that are sure to be much more impressive than anything the competition can offer, the card should be able to easily chew through 4K and VR gaming — both are very high-intensity applications.

Nvidia’s last attempt at a Titan, the Titan Z, was riddled with problems. Originally announced the the GTC 2014 conference, the card was long delayed. When it eventually came out it was not competitive against the Radeon R9 295 X2 on a price or performance basis. While it was really intended for CUDA programming as opposed to gaming, the Titan Z did eventually find a win inside Alienware’s Aurora gaming rig.

More details including pricing and availability will be available March 17-20 at the GPU Technology Conference.

The post Take a Good Look at the Nvidia Titan X appeared first on VR World.

After rumors suggested a launch sometime in the second quarter of 2015, it looks like AMD (NASDAQ:AMD) will unveil its R9 300 series at Computex, which is being held in the month of June in Taipei. The vendor had initially planned to launch the R9 390 series as a rebrand of the R9 280 series with improved performance, but those plans have been shelved in favor of an entirely new lineup of GPUs.

AMD’s next-gen GPU –likely the R9 390X — was on hand at the Game Developers Conference in San Francisco earlier this week, with the R9 390X powering an Oculus Rift demo. The idea of a full rebrand makes sense as an AMD engineer was quoted as saying that all R9 300 series products will come with FreeSync support, a feature that was not available in mid-tier R9 200 series cards like the R9 270X. Only the Hawaii-based GPUs such as the R9 290 and R9 290X as well as the R9 285 offer FreeSync support.

According to a series of leaks, the R9 380X and R9 390X will be powered by the Fiji XT GPU, which is said to feature 4096 stream processors as well as 4GB of High-Bandwidth Memory, offering thrice as much bandwidth as Nvidia’s GeForce GTX 980. Fiji XT is claimed to be 45% faster than the Hawaii-based R9 290X, and is said to incorporate a closed loop cooling solution, much like the R9 295X2. The Fiji XT GPU has been in testing for a few months in AMD’s driver certification division in India, and the brand itself stated that it was putting the final touches on the card.

With all the advances made by Nvidia (NASDAQ:NVDA) with the Maxwell architecture — including significant reductions in power consumption — it should be interesting to see how AMD’s next-generation offerings fare.

The post AMD R9 300 Series Said to Launch At Computex 2015 appeared first on VR World.

During Epic Games’ Unreal Engine 4 session at the Game Developers Conference, Nvidia (NASDAQ:NVDA) CEO Jen-Hsun Huang announced the vendor’s latest GPU in the Titan family, the Titan X. Huang hasn’t shared hardware details of the card, but what we do know is that the GPU will feature 12GB VRAM, twice as high as what was seen on the first-generation Titan.

Also revealed was the transistor count, which sees an increase from 7 billion to 8 billion. The memory bus interface is at 384-bit, the same as the earlier iterations in the Titan series. TDP figures were not mentioned as well, although considering that the card is based on the Maxwell architecture, it is likely we’ll see significant improvements in power draw. Nvidia is retaining the metal cooler and shroud, with the manufacturing process of the card also set to remain at 28nm.

The idea behind unveiling the Titan X during GDC was to highlight the graphic capabilities of the card, which was used to drive Weta Digital’s (the studio behind The Hobbit films) “Thief in the Shadows” VR demo running on Oculus’ VR headset.  Nvidia will reveal all the features on offer with the Titan X at its GPU Technology Conference, which is being held later this month in San Jose.

The post Nvidia’s Titan X Is The Most Powerful GPU In The World appeared first on VR World.

Unity 5 is now available across the globe, and for the first time ever users can download the game engine for free.

The news comes straight from this year’s Game Developers Conference (GDC 2015), where the developers showcased the brand new engine as well as revealed a new pricing structure to make Unity 5 more accessible to developers.

For starters the Unity 5 Personal Edition will be free of charge, but you won’t be able to earn royalties.

The Professional Edition is priced at $75 per month or $1500 lifetime, and comes with a license to Unity Cloud Build, a new addition that lets users “leverage the power of cloud computing”.

With Unity Cloud Build, “users can automate the creation and distribution” of their builds by “monitoring your team’s source control, and when changes are detected, it builds your project for multiple platforms in the cloud”.

Essentially it’s a tool that can automatically update and optimize projects for specific platforms without needing to be changed directly, and is tailored for convenience and ease of use across a team of developers.

To showcase Unity 5’s new capabilities, the team released The Blacksmith, a video entirely comprised of in-game footage that highlights everything from graphical fidelity, lighting and shading, physic engines, and more.

As for specifics, the footage shows off a variety of new features including, enhanced and rather eye-catching high-resolution real-time shadows renders and custom shaders for realistic hair and beard renders.

As for lighting, the video showcased the new real-time global illumination system for immersive indirect lighting, custom color-grading to make the scenes more dynamic, and a variety of customized motion blurs that create the illusion of movement in an easily recognizable way.

Below we also have a few other videos that demo Unity 5’s new capabilities, including the Architectural Visualization featurette and the Viking Village walkthrough.

Key features of Unity 5 include the following updates:

• New artistic toolset
• Physically-based standard shader
• Real-time global illumination
• HDR reflection probes
• New editor featuring PhysX 3.3 and an updated animator
• New Audio Mixer
  New Physics engine
• Updated animator with new features like adding State Machine Behaviors and a new Asset Creation API
• Deployment to 21 different platforms
• Web GL Preview
• Upcoming Oculus Rift Support
• iOS Metal Rendering Support
• Unity Cloud Build Toolset

Click here for a complete list of patch notes for Unity 5.

The post Unity 5 Engine Available For Free, Gets Demoed At GDC 2015 appeared first on VR World.

Microsoft’s (NASDAQ: MSFT) upcoming DirectX 12 has a ton of exciting features and optimizations that we already know about. However the most interesting new feature comes via a report from Tom’s Hardware: the ability to pool resources from multiple GPUs from different manufacturers.

The ability to work with two different GPU architectures in the same system comes from DirectX 12’s support of something called Explicit Asynchronous Multi-GPU capabilities and Split Frame Rendering (SFR). Effectively this pools all GPU resources into a bucket and allows them to be utilized as one.

SFR allows developers to manually divide data between the two GPUs to allow them to work together on each frame with the work for each frame divided between the two cards. The older method, called Alternate Frame Rendering (AFR). AFR is considered to be less efficient because it required both GPUs to have all of the data already in their frame buffers. Each GPU would render an alternate frame. The downside of this, however, is that the cards would work in parallel and not independently. If you had two cards with 4GB of memory each, you’d only have 4GB of useable space in the frame buffer in total.

The kicker is that Tom’s source said that SFR will be supported across multiple GPU architectures in the same system. It will treat both GPUs as one. Reportedly this would allow a system with an AMD (NASDAQ: AMD) APU and an Nvidia (NASDAQ: NVDA) GPU to have the two GPUs work together. The same would go for a discrete AMD and Nvidia GPU in the same system.

In the end this will be up to developers to make use of and optimize their code accordingly. AMD’s Mantle already supports a form of SFR, so in a way the precedent is already there.

More details on DirectX 12 — and a confirmation of this report — will likely be available during March’s Game Developer Conference in San Francisco.

The post Your Next Gaming PC Might Have a Radeon and Nvidia GPU appeared first on VR World.

Nvidia’s (NASDAQ: NVDA) CEO Jen-Hsun Huang has posted an open letter on his company’s blog addressing concerns over the GTX 970.

Nvidia and its add-in-board partners have faced a rough month because of the GTX 970 and its lack of advertised memory. Initially, in late January, Nvidia denied any problems with the card, however a few days later the company issued another press statement acknowledging it didn’t communicate specifications correctly and offered a driver update.

“We invented a new memory architecture in Maxwell. This new capability was created so that reduced-configurations of Maxwell can have a larger framebuffer – i.e., so that GTX 970 is not limited to 3GB, and can have an additional 1GB. GTX 970 is a 4GB card,” Huang wrote. “However, the upper 512MB of the additional 1GB is segmented and has reduced bandwidth. This is a good design because we were able to add an additional 1GB for GTX 970 and our software engineers can keep less frequently used data in the 512MB segment.”

Huang blamed a lack of proper communication with the press and reviewers for the confusion.

“Instead of being excited that we invented a way to increase memory of the GTX 970 from 3GB to 4GB, some were disappointed that we didn’t better describe the segmented nature of the architecture for that last 1GB of memory,” he wrote.

“This new feature of Maxwell should have been clearly detailed from the beginning.”

The post Nvidia CEO Comes Clean on GTX 970: ‘We’ll Do A Better Job Next Time’ appeared first on VR World.

ArrayFire found in its benchmarks that AMD’s (NASDAQ:AMD) integrated GPU in the Kaveri A10-7850K APU fares better at computational tasks than Intel’s (NASDAQ:INTC) Haswell HD 4600 series on the Core i7-4790K. The software company builds  software libraries for GPU programming, with its products working on both Nvidia CUDA-based cards as well as AMD’s OpenCL GPUs.

In its benchmarks, ArrayFire tested several GPU functions such as a bilateral filter, which is a non-linear filter that is used to smoothen out edges from images. The bilateral filter test showed AMD’s integrated GPU achieving 156 fps at 720p and 69 fps at full-HD, more than twice the score managed by the HD 4600 at 62 and 28 fps.

A similar set of results were observed in an erosion and dilution benchmark, with the Kaveri APU scoring four times as high as the HD 4600. ArrayFire also ran 2D convolution tests, which determine image processing, and found that Kaveri significantly outmatched Intel’s offering by achieving 191 fps to HD 4600’s 52 fps at full-HD resolution.

ArrayFire also ran a gamut of other computational tests, includingNxN matrix tests, sort functions as well as image resize and rotate tests, with AMD’s solution coming out on top in all benchmarks.

The post AMD’s Kaveri Holds A Significant Edge Over Intel In ArrayFire GPU Benchmarks appeared first on VR World.

According to a post on Linkedin, AMD (NASDAQ: AMD) is hiring CPU engineers to work on optimizing the CPU side of its Catalyst driver packages.

AMD is seeking someone to work on multi-threader optimizations to write code to cut through CPU bottleneck and work with software vendors to optimize drivers for their titles.

Compared to the competition, AMD’s drivers are considered to be somewhat bloated and CPU intensive. AMD’s Mantle API, which offers the ability for developers to code close-to-metal is a partial solution to cut through the driver bloat but not everyone wants to implement it.

Most likely these engineers would be working on developing the drivers for AMD’s upcoming Radeon R300 series, which has been the subject of numerous rumors both founded and unfounded. Perhaps this card will leapfrog past Nvidia (NASDAQ:NVDA) in the driver game.

The post AMD Hiring Graphics Driver Developers for CPU Optimization appeared first on VR World.

Nvidia (NASDAQ:NVDA) is under fire for incorrectly listing the specs of the GTX 970, and while the manufacturer has announced that it will provide a way for users to return the card to those looking to do so, that does not seem to extend to a majority of Nvidia’s partners around the globe.

Major retailers such as Amazon and NewEgg are offering refunds to any user looking to return the GTX 970, but in countries like India, customers are left to fend for themselves as retailers are refusing to process returns.

The issue with the GTX 970 is that Nvidia implemented a new memory segmentation in the card, which meant that the 4GB available memory was sectioned into a 3.5GB segment and a 512MB segment. The memory controllers in the 3.5GB segment have direct access to the L2 cache, and deliver an advertised bandwidth of 196GB/s, but the last 512MB has to interface with another memory controller to access the cache, leading to a drastic reduction in bandwidth to 28GB/s.

The situation is more pronounced when a user has the two or more GTX 970 cards in SLI, with a noticeable decrease in framerates when gaming at higher resolutions. Indian hardware site HardwareBBQ posted a conversation between a user facing such issues with the GTX 970 in SLI, and his failed attempts to get a refund for the card. The conversation starts off with the user talking to an Nvidia engineer on the GeForce forums, after which he is redirected to talk to the card vendor directly to initiate the refund process (which in this case is Zotac India).

However, the e-mail communique results in the technical manager for Zotac India, Swarn Singh, refusing to provide a refund as the GTX 970 was delivered “as it is written on the box with all specification.” The user was redirected to contact Nvidia, with Singh stating that a refund would be entertained only if the card suffered from any warranty-related issues. A discussion with an Nvidia engineer based out of India by the name of Rajaram also fails to yield any results, with the engineer stating that it is not “possible to suffer lag (microstutters) when the FPS is high” on the GTX 970.

Zotac India’s stubbornness in acknowledging an issue exists with the GTX 970 is alarming, considering that the brand is one of several vendors processing returns on a global level. The situation is exacerbated when taken into account the fact that video cards often carry a markup of 30 to 35% in the country.

Furthermore, Nvidia claimed that it would assist users that were unable to claim a return for their GTX 970, but the conversations with retailers and discussions on message boards in India show that the chip manufacturer is not offering a solution to users who purchased the card in the country.

With card vendors like Zotac directing users back to Nvidia and the chip vendor not providing any recourse regarding the issue, Indian users are largely left in the cold when it comes to dealing with returns or refunds for the beleaguered GTX 970.

The post GTX 970 Owners In India Left To Fend For Themselves appeared first on VR World.