Nvidia Takes Deep Learning to School

Release time:2017-05-10
source:EE Times

Deep learning is "transforming computing" is the message that Nvidia hopes to hammer home at its GPU Tech conference. On that theme, Nvidia has styled itself as a firebrand, catalyst and deep learning enabler and — in the long run — a deep profiteer.

Among the telltale signs that Nvidia is betting its future on this branch of artificial intelligence (AI) is the recent launch of its “Deep Learning Institute,” with plans increase the number of developers to 100,000 this year. Nvidia trained 10,000 developers in deep learning in 2016.

Over the last few years, AI has made inroads into “all parts of science,” said Greg Estes, Nvidia’s vice president responsible for developer programs. AI is becoming an integral component of “all applications ranging from cancer research, robotics, manufacturing to financial services, fraud detection and intelligent video analysis,” he noted. 

Deep learning transforms computing (Source: Nvidia)
Deep learning transforms computing (Source: Nvidia)

Nvidia wants to be known as the first resort for developers creating apps that use AI as a key component, said Estes.

“Deep learning” is in the computer science curriculum at many schools, but few universities offer a degree specifically in AI.

At its Deep Learning Institute, Nvidia plans to deliver “hands-on” training to “industry, government and academia,” according to Estes, with a  mission to “help developers, data scientists and engineers to get started in training, optimizing, and deploying neural networks to solve the real world problems in diverse disciplines.”

How can you fit AI into apps?
Kevin Krewell, principal analyst at Tirias Research, told EE Times, “The challenge of Deep Learning is that it’s just hard to get started and figuring out how to fit it into traditional app development.”

He noted, “I think Nvidia is trying to get a wider set of developers trained on how to fit machine learning (ML) into existing development programs. Unlike traditional programs where algorithms are employed to perform a task, ML is a two-stage process with a training phase and a deployment phase.”

Nvidia’s edge is that “Machine learning performs better with an accelerator like a GPU, rather than relying just on the CPU,” Krewell added.

As Nvidia readies its Deep Learning Institute, the company is also entering a host of partnership deals with AI “framework” communities and universities. Among them are Facebook, Amazon, Google (Tensor Flow), the Mayo Clinic, Stanford University and Udacity.

Such collaborations with framework vendors are critical, because every developer working on AI apps needs to have cloud and deep learning resources.

As Jim McGregor, principal analyst at Tirias Research, told us, “The most difficult thing for app developers are the cloud resources and large data sets. As an example, the mobile suppliers are promoting machine learning on their devices, but to develop apps for those devices you need cloud/deep learning resources and the data sets to train those resources, which the mobile players are not providing.”

Nvidia can provide the hardware resources and a mature software model, but “developers still need the service provider and data sets,” McGregor added.

According to Nvidia, the company is also working with Microsoft Azure, IBM Power and IBM Cloud teams to port lab content to their cloud solutions. 

Online messageinquiry

Nvidia stock fell as much as 19 percent Thursday after the company reported earnings for the third quarter of its 2019 fiscal year, which ended on Oct. 28.Here's how the company did:Earnings: $1.84 per share, excluding certain items, vs. $1.71 per share as expected by analysts, according to Refinitiv.Revenue: $3.18 billion, vs. $3.24 billion as expected by analysts, according to Refinitiv.With respect to guidance, Nvidia said it's expecting $2.70 billion in revenue in the fiscal fourth quarter, plus us minus 2 percent, excluding certain items. That's below the Refinitiv consensus estimate of $3.40 billion.Overall, in the fiscal third quarter, Nvidia's revenue rose 21 percent year over year, according to its earnings statement.In its fiscal second-quarter earnings, the chipmaker fell short of analyst expectations on guidance despite beating on earnings and revenue estimates. The company's cryptocurrency mining products suffered a hefty decline in that quarter, and the trend continued in the fiscal third quarter.It has become less profitable to use graphics processing units, or GPUs, for mining, according to a recent analysis by Susquehanna. To mine cryptocurrency, computers compete to solve complex math problems in exchange for a specific amount of bitcoin or ethereum. But as both currencies have sunk in value, so too has this segment of revenue for Nvidia."Our near-term results reflect excess channel inventory post the crypto-currency boom, which will be corrected," Nvidia CEO Jensen Huang is quoted as saying in a Thursday press release. In the fiscal third quarter Nvidia's revenue from original equipment manufacturers and intellectual property totaled $148 million, which was down 23 percent year over year but above the FactSet consensus estimate of $102 million. Nvidia chocked up the decline to "the absence of cryptocurrency mining" in its earnings statement.In the quarter Nvidia had a $57 million charge related to older products because of the decrease in demand for cryptocurrency mining."Our Q4 outlook for gaming reflects very little shipment in the midrange Pascal segment to allow channel inventory to normalize," Nvidia's chief financial officer, Colette Kress, told analysts on a conference call after the company announced its results.It will take one to two quarters to go through the extra inventory, Huang said on the call."This is surely a setback, and I wish we had seen it earlier," he said.Inventory issues also affect other brands, Huang said. AMD stock fell 5 percent in extended trading on Thursday.Nvidia's gaming business segment generated $1.76 billion in revenue in the quarter, below the $1.89 billion FactSet consensus estimate.Nvidia's data center segment came in at $792 million in revenue, lower than the $821 million estimate.Revenue for the company's professional visualization business segment was $305 million, surpassing the $284 million estimate.Nvidia, like most other tech stocks, was hit hard in October, which was the worst month for the Nasdaq Composite Index since 2008. The stock is now up 4 percent since the beginning of the year.
2018-11-16 00:00 reading:375
Nvidia is in Munich this week to declare war that it is coming after the advanced driver assistance system (ADAS) market. The GPU company is now pushing its AI-based Nvidia Drive AGX Xavier System — originally designed for Level 4 autonomous vehicles — down to Level 2+ cars.In a competitive landscape already crowded with ADAS solutions provided by rival chip vendors such as NXP, Renesas, and Intel/Mobileye, Nvidia is boasting that its GPU-based automotive SoC isn’t just a “development platform” for OEMs to prototype their self-driving vehicles.At the company’s own GPU Technology Conference (GTC) in Europe, Nvidia announced that Volvo cars will be using the Nvidia Drive AGX Xavier for its next generation of ADAS vehicles, with production starting in the early 2020s.NVIDIA's Drive AGX Xavier will be designed into Volvo's ADAS L2+ vehicles. Henrik Green (left), head of R&D of Volvo Cars, with Nvidia CEO Jensen Huang on stage at GTC Europe in Munich. (Photo: Nvidia)Danny Shapiro, senior director of automotive at Nvidia, told us, “Volvo isn’t doing just traditional ADAS. They will be delivering wide-ranging features of ‘Level 2+’ automated driving.”By Level 2+, Shapiro means that Volvo will be integrating “360° surround perception and a driver monitoring system” in addition to a conventional adaptive cruise control (ACC) system and automated emergency braking (AEB) system.Nvidia added that its platform will enable Volvo to “implement new connectivity services, energy management technology, in-car personalization options, and autonomous drive technology.”It remains unclear if car OEMs designing ADAS vehicles are all that eager for AI-based Drive AGX Xavier, which is hardly cheap. Shapiro said that if any car OEMs or Tier Ones are serious about developing autonomous vehicles, taking an approach that “unifies ADAS and autonomous vehicle development” makes sense. The move allows carmakers to develop software algorithms on a single platform. “They will end up saving cost,” he said.Phil Magney, founder and principal at VSI Labs, agreed. “The key here is that this is the architecture that can be applied to any level of automation.” He said, “The processes involved in L2 and L4 applications are largely the same. The difference is that L4 would require more sensors, more redundancy, and more software to assure that the system is safe enough even for robo-taxis, where you don’t have a driver to pass control to when the vehicle encounters a scenario that it cannot handle.”Better than discrete ECUsAnother argument for the use of AGX for L2+ is that the alternative requires the use of multiple discrete ECUs. Magney said, “An active ADAS system (such as lane keeping, adaptive cruise, or automatic emergency braking) requires a number of cores fundamental to automation. Each of these tasks requires a pretty sophisticated hardware/software stack.” He asked, “Why not consolidate them instead of having discrete ECUs for each function?”Scalability is another factor. Magney rationalized, “A developer could choose AGX Xavier to handle all these applications. On the other hand, if you want to develop a robo-taxi, you need more sensors, more software, more redundancy, and higher processor performance … so you could choose AGX Pegasus for this.”Is AGX Xavier safer?Shapiro also brought up safety issues.He told us, “Recent safety reports show that many L2 systems aren’t doing what they say they would do.” Indeed, in August, the Insurance Institute for Highway Safety (IIHS) exposed “a large variability of Level 2 vehicle performance under a host of different scenarios.” An EE Times story entitled “Not All ADAS Vehicles Created Equal” reported that some [L2] systems can fail under any number of circumstances. In some cases, certain models equipped with ADAS are apparently blind to stopped vehicles and could even steer directly into a crash.Nvidia’s Shapiro implied that by “integrating more sensors and adding more computing power” that runs robust AI algorithms, Volvo can make their L2+ cars “safer.”On the topic of safety, Magney didn’t necessarily agree. “More computing power doesn’t necessarily mean that it is safer.” He noted, “It all depends on how it is designed.”Lane keeping, adaptive cruise, and emergency braking for L2 could rely on a few sensors and associated algorithms while a driver at the wheel manages events beyond the system’s capabilities.However, the story is different with a robo-taxi, explained Magney. “You are going to need a lot more … more sensors, more algorithms, some lock-step processing, and localization against a precision map.” He said, “For example, if you go from a 16-channel LiDAR to a 128-channel LiDAR for localization, you are working with eight times the amount of data for both your localization layer as well as your environmental model.”Competitive landscapeBut really, what does Nvidia have that competing automotive SoC chip suppliers don’t?Magney, speaking from his firm VSI Labs’ own experience, said, “The Nvidia Drive development package has the most comprehensive tools for developing AV applications.”He added, “This is not to suggest that Nvidia is complete and a developer could just plug and play. To the contrary, there is a ton of organic codework necessary to program, tune, and optimize the performance of AV applications.”However, he concluded that, in the end, “you are going to be able to develop faster with Nvidia’s hardware/software stack because you don’t have to start from scratch. Furthermore, you have DRIVE Constellation for your hardware-in-loop simulations where you can vastly accelerate your simulation testing, and this is vital for testing and validation.”
2018-10-11 00:00 reading:344
Nvidia cracked the list of top 10 semiconductor vendors by sales for the first time in 2017, joining Qualcomm as the only other strictly fabless chip supplier to attain that distinction last year, according to market research firm IHS Markit.   Nividia's 2017 sales total of $8.57 billion was good enough for the company to secure the 10th position among chip vendors last year, IHS said. Ironically, Nvidia edged out fellow fabless chip supplier MediaTek of Taiwan to crack the top 10, according to Len Jelinek, director and chief analyst for technology, media and telecommunications at IHS.Qualcomm, Nvidia and MediaTek are the only strictly fabless chip vendors to ever crack the top 10 list of chip suppliers in a calendar year. MediaTek was among the top 10 chip vendors in 2014 and again in 2016. Qualcomm first cracked the top 10 list in 2007 and has remained a fixture on the list ever since.Overall, global semiconductor sales rose 21.7 percent in 2017 to reach a record $429.1 billion. It was the highest year-over-year growth for the industry in 14 years.Most of the industry's sales gains were driven by a blockbuster year for memory chip sales, which increased by 60.8 percent from 2016. Outside of memory, the rest of the semiconductor industry grew by 9.9 percent last year, due largely to what IHS called solid unit-sales growth and strong demand across all applications, regions and technologies.Craig Stice, senior director for memory and storage at IHS, said through a press statement that NAND prices — which increased throughout 2017 — are expected to decline this year."Entering 2018, the 3D NAND transition is now almost three-quarters of the total bit percent of production, and it is projected to provide supply relief for the strong demand coming from the SSD and mobile markets," Stice said. "Prices are expected to begin to decline aggressively, but 2018 could still be a record revenue year for the NAND market."
2018-04-13 00:00 reading:430
 At its annual GTC event, Nvidia announced system-level enhancements to boost the performance of its GPUs in training neural networks and a partnership with ARM to spread its technology into inference jobs.Nvidia offered no details of its roadmap, presumably for 7-nm graphics processors in 2019 or later. It has some breathing room, given that AMD is just getting started in this space, Intel is not expected to ship its Nervana accelerator until next year, and Graphcore — a leading startup — has gone quiet. A few months ago, both Intel and Graphcore were expected to release production silicon this year.The high-end Tesla V100 GPU from Nvidia is now available with 32-GBytes memory, twice the HBM2 stacks of DRAM that it supported when launched last May. In addition, the company announced NVSwitch, a 100-W chip made in a TSMC 12nm FinFET process. It sports 18 NVLink 2.0 ports that can link 16 GPUs to shared memory.Nvidia became the first company to make the muscular training systems expected to draw 10 kW of power and deliver up to 2 petaflops of performance. Its DGX-2 will pack 12 NVSwitch chips and 16 GPUs in a 10U chassis that can support two Intel Xeon hosts, Infiniband, or Ethernet networks and up to 60 solid-state drives.Cray, Hewlett Packard Enterprise, IBM, Lenovo, Supermicro, and Tyan said that they will start shipping systems with the 32-GB chips by June. Oracle plans to use the chip in a cloud service later in the year.Claims of performance increases using the memory, interconnect, and software optimizations ranged widely. Nvidia said that it trained a FAIRSeq translation model in two days, an eight-fold increase from a test in September using eight GPUs with 16-GBytes memory each. Separately, SAP said that it eked out a 10% gain in image recognition using a ResNet-152 model.Intel aims to leapfrog Nvidia next year with a production Nervana chip sporting 12 100-Gbit/s links compared to six 25-Gbit/s NVLinks on Nvidia’s Volta. The non-coherent memory of the Nervana chip will allow more flexibility in creating large clusters of accelerators, including torus networks, although it will be more difficult to program.To ease the coding job, Intel has released as open source its Ngraph compiler. It aims to turn software from third-party AI frameworks like Google’s TensorFlow into code that can run on Intel’s Xeon, Nervana, and eventually FPGA chips.The code, running on a prototype accelerator, is being fine-tuned by Intel and a handful of data center partners. The company aims to announce details of its plans at a developer conference in late May, though production chips are not expected until next year. At that point, Nvidia will be under pressure to field a next-generation part to keep pace with an Intel roadmap that calls for annual accelerator upgrades.”The existing Nervana product will really be a software development vehicle. It was built on 28nm process before Intel bought the company and it's not competitive with Nvidia's 12nm Volta design,” said Kevin Krewell, a senior analyst with Tirias Research.Volta’s added memory and NVSwitch “keeps Nvidia ahead of the competition. We're all looking forward to the next process shrink, but, as far as production shipping silicon goes, Volta still has no peer,” he added.Among startups, Wave Computing is expected to ship this year its first training systems for data centers and developers. New players are still emerging.Startup SambaNova Systems debuted last week with $56 million from investors, including Google’s parent Alphabet. Co-founder Kunle Olukotun’s last startup, Afara Websystems, designed what became the Niagara server processor of Sun Microsystems, now Oracle.Nvidia's DGX-2 will sell for $399,000 starting in the third quarter. Source: EE TimesNvidia currently dominates the training of neural network models in data centers, but it is a relative newcomer to the broader area of inference jobs at the edge of the network. To bolster its position, Nvidia and ARM agreed to collaborate on making Nvidia’s open-source hardware for inferencing available as part of ARM’s planned machine-learning products.Nvidia announced last year that it would open-source IP from its Xavier inference accelerator. It has made multiple RTL releases to date. The blocks compete with AI accelerators offered byCadence, Ceva, and Synopsys, among others.Just what Nvidia blocks that ARM will make available when remains unclear. So far, ARM has only sketched out its plans for AI chips as part of a broad Project Trillium. An ARM representative would only say that ARM aims to port its emerging neural net software to the Nvidia IP.Deepu Talla, general manager of Nvidia’s group overseeing Xavier, said that he is aware of multiple chips being designed using the free, modular IP. However, so far, none have been announced.Nvidia hopes that the inference effort spreads use of its machine-learning software also used in training AI models. To that end, the company announced several efforts to update its code and integrate it into third-party AI frameworks.TensorRT 4, the latest version of Nvidia’s runtime software, boosts support for inferencing jobs and is being integrated into version 1.7 of Google’s TensorFlow framework. Nvidia is also integrating the runtime with the Kaldi speech framework, Windows ML, and Matlab, among others.Separately, the company announced that the RTX software for ray tracing that it announced last week is now available on V100-based Quadro GV100 chips, sporting 32-GBytes memory and two NVLinks.The software enables faster, more realistic rendering for games, movies, and design models. It runs on Nvidia proprietary APIs as well as Microsoft’s DirectX for ray tracing and will support Vulkan in the future.The software delivers 10x to 100x improvements compared to CPU-based rendering that dominates a market that forecasts to be larger than $2 billion by 2020, said Bob Pette, vice president of Nvidia’s professional visualization group.Nvidia CEO Jensen Huang claims GPUs are setting a new direction above Moore's law for performance on AI and other jobs. Chart: Nvidia.
2018-03-28 00:00 reading:507
  • Week of hot material
  • Material in short supply seckilling
model brand Quote
CD74HC4051QPWRQ1 Texas Instruments
PCA9306DCUR Texas Instruments
TPS5430DDAR Texas Instruments
TXB0108PWR Texas Instruments
TPS61021ADSGR Texas Instruments
TPIC6C595DR Texas Instruments
model brand To snap up
TPS61021ADSGR Texas Instruments
ULQ2003AQDRQ1 Texas Instruments
TPS61256YFFR Texas Instruments
TPS5430DDAR Texas Instruments
TPS63050YFFR Texas Instruments
TXS0104EPWR Texas Instruments
Hot labels
Information leaderboard
  • Week of ranking
  • Month ranking
About us

Qr code of ameya360 official account

Identify TWO-DIMENSIONAL code, you can pay attention to

AMEYA360 mall (www.ameya360.com) was launched in 2011. Now there are more than 3,500 high-quality suppliers, including 6 million product model data, and more than 1 million component stocks for purchase. Products cover MCU+ memory + power chip +IGBT+MOS tube + op amp + RF Bluetooth + sensor + resistor capacitance inductor + connector and other fields. main business of platform covers spot sales of electronic components, BOM distribution and product supporting materials, providing one-stop purchasing and sales services for our customers.