Category: Graphics cards

Never mind AI, Intel has Nervana

3d2a123ddc312423225755a14fe7db2dChipzilla’s billion dollar investment in Nervana might be the key to making its server chips more intelligent.

Intel is laying out its roadmap to advance artificial intelligence performance across the board and Nervana technology appears to be everywhere.

The high-performance silicon market is dominated by GPUs. However, with Nervana inside, Intel hopes its new corporate tech with its a fully-optimized software and hardware stack will give that business model a good kicking.

Nervana hardware will initially be available as an add-in card that plugs into a PCIe slot. The first Nervana silicon, codenamed Lake Crest, will make its way to select Intel customers in H1 2017.

Intel is also talking about Knights Mill, which is the next generation of the Xeon Phi processor family. Intel said that Knights Mill will deliver a 4x increase in deep learning performance compared to existing Xeon Phi processors and the combined solution with Nervana will offer orders of magnitude gains in deep learning performance.

Diane Bryant, Executive VP of Intel’s Data Center Group said that the Intel Nervana platform to produce breakthrough performance and dramatic reductions in the time to train complex neural networks.

Intel CEO Brian Krzanich said that Nervana’s technologies will produce a 100-fold increase in performance in the next three years to train complex neural networks, enabling data scientists to solve their biggest AI challenges faster.

Japan broadcasts 8K

old-school-tvThe Land of the Rising Sun began the world’s first regular 8K television broadcasts this week, five days ahead of the opening of the Olympic Games so that you will be able to watch people being infected with a Zita virus in super high res.

8K are broadcasts with a resolution of 7,680 x 4,320 pixels. That’s 16 times the resolution of today’s full high-definition (FHD) broadcasts and four times that of the 4K standard, which is only just emerging in many other countries.

NHK is broadcasting what it calls  “Super Hi-Vision,” also features 22.2-channel surround sound across a satellite channel that will broadcast a mix of 8K and 4K content as it prepares to launch full-scale 8K transmissions in time for the Tokyo Olympics in 2020.

The channel will be on air daily from 10am until 5pm, with extended hours during the Rio Olympics.

A look at Monday’s schedule reveals programming focused on the arts, sports, music and documentaries — not dissimilar to early HDTV channels. Highlights of this year’s Rio Carnival, the opening ceremony of the 2012 London Olympics, and Japan’s Aomori Nebuta festival were among the 8K programs. A J-pop concert and program on Japanese botanical drawings was also broadcast.

NHK was the broadcaster behind the development of high-definition TV, starting broadcasts in an early analogue format in 1989. It has been working on 8K for two decades. It showed off its 8K in 2002, when many homes hadn’t even transitioned to high-def TV.

It is pretty hard work. Cameras, mixers, recording equipment, monitors and other studio gear must be built to handle uncompressed 8K video in real time, at data rates that can easily reach 100Gbps.

Compression equipment then needs to take that signal and encode it into more efficient streams for broadcast, again in real time. The satellite signal is several tens of megabits per second, and consumer receivers and televisions are required to decode it.


Polaris finally shines today

polarisAMD finally starts flogging its new RX 480 GPU today.

For those who came in late, the RX 480 uses the company’s latest Polaris architecture which is built lt on 14nm FinFET process technology.

The starting price is $199 for the 4GB model and $239 for the 8GB and has some interesting performance characteristics. Compared to the GeForce GTX 970 which sells for $280, the RX 480 performs is about five to ten percent better. But when it comes to DX12 games like Gears of War, Hitman and Rise of the Tomb Raider it is about 40 per cent faster.

Compared to previous AMD products, the RX 480 is as fast as a Radeon R9 390 but uses just 150 watts compared to 275 watts for the previous generation.

Rivals Nvidia are expected to have a competing product based on Pascal available sometime in July, so AMD’s advantage may be short-lived; but in the meantime, the Radeon RX 480 is clearly the best GPU for $200.

Of course the world is also waiting to see AMD’s entry into the CPU league tables with the much touted never seen Zen chip, which should be in the shops in December.

AMD shows off Polaris-based Radeon RX 470 and RX 460

4528082378_4d5b9fb99e_zAMD has been showing off its latest Polaris based GPUs at E3 2016.

For those who came in late, Polaris is AMD’s bright new hope in the GPU world – a bit like Zen is for the CPU, only it appears to exist whereas Zen doesn’t.

THe Radeon RX 470 and RX 460  join the recently announced RX 480 as part of the company’s new Polaris family. Polaris is AMD’s newest GPU micro-architecture, which is based on the 14nm FinFet production process.

AMD is not telling us the prices of its new GPU, but it is possible to have stab at it. The  RX 480 is made for 1440p gaming, and the RX 470 will focus on delivering a “refined, power-efficient HD gaming” experience. The RX 460 will offer a “cool and efficient solution for the ultimate e-sports gaming experience.”

The 4GB version of the RX 480 will start out at $200, it’s safe to assume that these two other cards will launch at lower price points.

AMD says the chips are extremely thin, offering a very low Z-height, and will fit into thin and light gaming notebooks.

The entire new RX line will also support a wide variety of features that include DX12, Vulkan, HDR, HDMI 2.0b, DisplayPort 1.3/1.4, and H.265 encoding/decoding.

There is no release day  but since the RX 480 is scheduled to launch on June 29 the other two should be soon after. AMD is claiming that card outperforms $500 graphics cards in VR.

Nvidia talks up Pascal

nvidiaNvidia CEO Jen-Hsun Huang has been telling the Nvidia Computex 2016 press conference about his latest Pascal-based GPUs.

He claimed they were being broadly adopted in applications including data centers, car-use electronics and deep-learning platforms.

He said Pascal GPUs will continue to use TSMC 16nm FinFET manufacturing process to provide enhanced performance than previous-generation products.

He said next-generation servers built by Nvidia’s Tesla P100 GPUs can compete against servers that are equipped with several hundreds of CPUs. Nvidia also showcased a server system manufactured by Quanta Computer.

He also showed the Pascal-based GeForce GTX 1080 reference design, displayed at the conference, the board is manufactured by Foxconn.

Nvidia has been pushing its Pascal GPUs into industry sectors, including virtual reality, car-use electronics and deep-learning. As for mobile devices such as smartphones and tablets, Huang noted that Nvidia has no interest in smartphones.

Nvidia gets under the bonnet of Volvo

volvo-adNvidia says that it has teamed up with the Swedish inventor of seat belts to place its chips under the bonnet of its cars.

Nvidia unveiled a new, lunchbox-size super-computer for self-driving cars and said Volvo Car Group will be the new device’s first customer.

CEO Jen-Hsung Huang said that his new Drive PX 2 has computing power equivalent to 150 PCs and can deliver up to 24 trillion “deep learning” operations – allowing the computer to use artificial intelligence to program itself to recognise driving situations – per second.

Partnerships between automakers and Silicon Valley companies on self-driving technologies are taking centre stage at this year’s show.

General Motors has announced a $500 million investment in ride-sharing service outfit Lyft.
Huang didn’t say how much his company is going to profit on the Drive PX 2, but automotive is the fastest-growing business segment for Nvidia. Its bread and butter is still video game GPUs.

AMD users see Crimson after new driver burns their cards

AMD logoAMD’s new Crimson drivers, which killed off the Catalyst name and software, have a serious problem – they make cards overheat.

There are widespread reports of cards overheating and dying. The new driver is setting the video card fans to 20 percent and then leaving them there.

Fan speed should increase as the GPU temperature goes up, but Crimson sticks them there at 20 percent, even during games and intensive workloads.

This means that GPU temperatures  climb to more than 90° C which causes high temperatures, thermal throttling, graphical glitches and crashes.

Some users are reporting permanent hardware damage. Although the GPU itself throttles when it overheats, there’s speculation that other components on the cards, such as the VRMs, can still be damaged.

AMD has acknowledged the fan speed problem and says that a “hot fix” should be rolled out from today.

It seems AMD did not learn from Nvidia which had a similar problem in 2010 and 2013.

Nvidia releases Jetsons for drones and robots

what-you-can-learn-from-the-jetsons-about-home-automation-image-0Graphics card maker Nvidia has lifted the kimono on its Jetson TX1 developer kit which it hopes will encourage people to build drones and robots.

The credit card sized TX1 kit is the size of a credit card but has a 1 teraflop of horsepower.

Jesse Clayton, product manager at Nvidia said that robots and drones require autonomous and smoother navigation capabilities, and the TX1 will help.

The TX1 has 256 graphics cores to process images so Robots can recognize objects and avoid collisions using “deep-learning” algorithms and image processing engines.

Clayton said Nvidia was also providing a software development kit for theTX1, including a debugger, compiler, libraries and other tools. The SDK will help programmers load applications that allow robots and drones to be truly autonomous.

The software uses Nvidia’s CUDA parallel programming framework, and taps into technologies such as OpenCV, OpenVX and Nvidia’s VisionWorks for image recognition. The board also supports OpenGL and OpenGL ES graphics standards.

The board can connect to more powerful cloud services for post-processing of images, Clayton said.


The TX1 is three times faster than last year’s original Jetson board, which delivered 300 gigaflops of horsepower. It uses the Tegra X1 chip which Nvidia is putting under the bonnets of cars and tablets. They use 64-bit ARM CPUs.Additional specifications include 4GB of DDR4 memory, 16GB of storage, Ethernet, 802.11ac Wi-Fi and Bluetooth.

The developer board will be available starting on November 16 for $599 through online retailers like Amazon and Newegg, it will be in the shops next year.

Intel minions create fast open source graphics

minionsChipmaker Intel has set its minions the task of creating a high-performance software rasteriser for the Linux Mesa 3D Graphics Library.

Mesa currently uses swrast, LLVMpipe, and Softpipe drivers as software rasterisers that run OpenGL on the CPU rather than any dedicated GPU. But apparently Intel’s minions have been developing a new, high-performance software rasteriser.

The minions hope to upstream their new “OpenSWR” project into Mesa as offering fast, CPU-rendered graphics.

For those who came in late, OpenSWR is the newly-announced high performance software rasterizer that’s developed at Intel by a different team of minions.

That group was looking at software-defined visualisations and scientific visualisations. Chipzilla already had developed a high-performance software rasteriser internally and then later they decided to engage in this project and work on upstream Mesa3D support.

Writing in its blog, a spokesMinion said that for high geometry workloads the software was faster than llvmpipe.

“This is to be expected, as llvmpipe only threads the fragment processing and not the geometry frontend. The linked slide below shows some performance numbers from a benchmark dataset and application.”

On a 36 total core dual E5-2699v3 there is a performance 29 times to 51 times that of llvmpipe.

“While our current performance is quite good, we know there is more potential in this architecture. When we switched from a prototype OpenGL driver to Mesa we regressed performance severely, some due to interface issues that need tuning, some differences in shader code generation, and some due to conformance and feature additions to the core swr. We are looking to recovering most of this performance back,” the spokesMinion said. Whatever any of that means.

The new rasteriser is being put out under the Mesa MIT license. Intel is making it open-source to and making it easier to deploy. Unlike their Intel i965 Mesa driver, this rasterizer builds atop Gallium3D. Additionally, OpenSWR makes use of LLVM.

The rasteriser should work with AMD CPUs, if it has AVX/AVX2 support. Intel plans on adding AVX512 support as well.

AMD confirms closed source Linux Vulkan

spokeAMD has confirmed that it has a Linux driver for its Vulkan card and have one prototype already – the only problem is that it is closed source.

Those who want their Linux to be totally open saucy will be completely disappointed. In fact, there is a lot about the project, which is closed source.

AMD said that OpenCL support will initially be closed and then opened later. They already have some basic OpenCL open-source support via the Clover Gallium3D driver; this is referring to OpenCL 2.1+ support with SPIR-V alongside Vulkan or the OpenCL Catalyst code.

Vulkan driver communicating with libdrm that interfaces with the AMDGPU kernel driver. AMD slides do not mention whether they intend to support Vulkan with the current Radeon DRM driver for HD 7000 through Rx 300 (non-Tonga/Carrizo/Fiji) GPUs. Instead, they only talk about the AMDGPU kernel driver just for the very newest AMD GPUs like the R9 Fury and Carrizo APUs.

However, AMD said that this is all transitory and Vulkan will eventually become completely open sauce. In fact, AMD has promised to focus more on open-source than closed-source in the future.
The problem is that closed-source software is needed for some workstation features, OpenGL 4.5, and OpenCL.

Basic support for AMD’s closed-source (Catalyst) OpenGL and OpenCL support riding atop the rest of the AMDGPU driver stack is complete. Their initial Vulkan driver is using DRI3.