Syber Group
Toll Free : 855-568-TSTG(8784)
Subscribe To : Envelop Twitter Facebook Feed linkedin

nVidia NVLINK 2.0 Going In IBM Servers

August 31, 2016 by  
Filed under Computing

Comments Off on nVidia NVLINK 2.0 Going In IBM Servers

On Monday, PCWorld reported that the first servers expected to use Nvidia’s second-generation NVLINK 2.0 technology will be arriving sometime next year using IBM’s upcoming Power9 chip family.

IBM launched its Power8 lineup of superscalar symmetric multiprocessors back in August 2013 at the Hot Chips conference, and the first systems became available in August 2014. The announcement was significant because it signaled the beginning of a continuing partnership between IBM and Nvidia to develop GPU-accelerated IBM server systems, beginning with the Tesla K40 GPU.

The result was an HPC “tag-team” where IBM’s Power8 architecture, a 12-core chip with 96MB of embedded memory, would eventually go on to power Nvidia’s next-generation Pascal architecture which debuted in April 2016 at the company’s GPU Technology Conference.

NVLINK, first announced in March 2014, uses a proprietary High-Speed Signaling interconnect (NVHS) developed by Nvidia. The company says NVHS transmits data over a differential pair running at up to 20Gbps, so eight of these differential 20Gbps connections will form a 160Gbps “Sub-Link” that sends data in one direction. Two sub-links—one for each direction—will form a 320Gbps, or 40GB/s bi-directional “Link” that connects processors together in a mesh framework (GPU-to-GPU or GPU-to-CPU).

NVLINK lanes upgrade from 20Gbps to 25Gbps

IBM is projecting its Power9 servers to be available beginning in the middle of 2017, with PCWorld reporting that the new processor lineup will include support for NVLINK 2.0 technology. Each NVLINK lane will communicate at 25Gbps, up from 20Gbps in the first iteration. With eight differential lanes, this translates to a 400Gbps (50GB/s) bi-directional link between CPUs and GPUs, or about 25 percent more performance if the information is correct.

NVLINK 2.0 capable servers arriving next year

Meanwhile, Nvidia has yet to release any NVLINK 2.0-capable GPUs, but a company presentation slide in Korean language suggests that the technology will first appear in Volta GPUs which are also scheduled for release sometime next year. We were originally under the impression that the new GPU architecture would release in 2018, as per Nvidia’s roadmap. But a source hinted last month that Volta would be getting 16nm FinFET treatment and may show up in roughly the same timeframe as AMD’s HBM 2.0-powered Vega sometime in 2017. After all, it is easier for Nvidia to launch sooner if the new architecture is built on the same node as the Pascal lineup.

Still ahead of PCI-Express 4.0

Nvidia claims that PCI-Express 3.0 (32GB/s with x16 bandwidth) significantly limits a GPU’s ability to access a CPU’s memory system and is about “four to five times slower” than its proprietary standard. Even PCI-Express 4.0, releasing later in 2017, is limited to 64GB/s on a slot with x16 bandwidth.

To put this in perspective, Nvidia’s Tesla P100 Accelerator uses four 40GB/s NVLINK ports to connect clusters of GPUs and CPUs, for a total of 160GB/s of bandwidth.

With a generational NVLINK upgrade from 40GB/s to 50GB/s bi-directional links, the company could release a future Volta-based GPU with four 50GB/s NVLINK ports totaling of 200GB/s of bandwidth, well above and beyond the specifications of the new PCI-Express standard.


Intel To Acquire Deep Learning Company Nervana

August 19, 2016 by  
Filed under Computing

Comments Off on Intel To Acquire Deep Learning Company Nervana

Intel is acquiring deep-learning startup Nervana Systems in a deal that could help it make up for lost ground in the increasingly hot area of artificial intelligence.

Founded in 2014, California-based Nervana offers a hosted platform for deep learning that’s optimized “from algorithms down to silicon” to solve machine-learning problems, the startup says.

Businesses can use its Nervana cloud service to build and deploy applications that make use of deep learning, a branch of AI used for tasks like image recognition and uncovering patterns in large amounts of data.

Also of interest to Intel, Nervana is developing a specialty processor, known as an ASIC, that’s custom built for deep learning.

Financial terms of the deal were not disclosed, but one estimate put the value above $350 million.

“We will apply Nervana’s software expertise to further optimize the Intel Math Kernel Library and its integration into industry standard frameworks,” Diane Bryant, head of Intel’s Data Center Group, said in a blog post. Nervana’s expertise “will advance Intel’s AI portfolio and enhance the deep-learning performance and TCO of our Intel Xeon and Intel Xeon Phi processors.”

Though Intel also acquired AI firm Saffron late last year, the Nervana acquisition “clearly defines the start of Intel’s AI portfolio,” said Paul Teich, principal analyst with Tirias Research.

“Intel has been chasing high-performance computing very effectively, but their hardware-design teams missed the convolutional neural network transition a few years ago,” Teich said. CNNs are what’s fueling the current surge in artificial intelligence, deep learning and machine learning.

As part of Intel, Nervana will continue to operate out of its San Diego headquarters, cofounder and CEO Naveen Rao said in a blog post.

The startup’s 48-person team will join Intel’s Data Center Group after the deal’s close, which is expected “very soon,” Intel said.


AMD Goes 16 Core Snowy Owl

July 22, 2016 by  
Filed under Computing

Comments Off on AMD Goes 16 Core Snowy Owl

Naples is a 32 Zen core Opteron with 64 threads. The 16 core Zen version with a BGA socket is codenamed Snowy Owl. AMD thinks that Snowy Owl will be a great match for the communication and network markets that needs a high performance 64-bit X86 CPU.

Snowy Owl has 16 cores and 32 threads, all based on 14nm FinFET Zen transistors. The processor supports up to 32MB of shared L3 cache. We also mentioned a processor cluster codenamed Zeppelin. This seems to be the key to the Zen architecture as more Zeppelin clusters are creating more core Opterons.

Each Zeppelin has eight Zen cores and each Zen core has 512KB dedicated L2 cache memory. Four Zen cores share 8MB of L3 memory making the total L3 cache size  16MB.  Zeppelin (ZP) comes with PCIe Gen 3, SATA 3, 10GbE, sever controller Hub, AMD secure processor as well as the DDR4 Memory controller. AMD is using a super-fast coherent interconnect to create more than one Zeppelin core.

One Zeppelin cluster would make an 8 core, 16 thread CPU with 4MB L2 and 16MB L3 cache and in our case product codenamed Snowy owl has 16 cores, 32 threads 8MB of L2 (512KB x 16) and 32MB L3 (4x8MB).

The Snowy Owl with 16 cores uses a SP4 Multi Chip Module (MCM) BGA socket, while the Naples uses MCM based SP3. These two are not pin compatible but 16 and 8 core Zen based Opterons will fit in the same socket.

Snowy Owl has four independent memory channels and up to 64 lanes of PCIe Gen3. When it comes to storage, it supports up to 16 SATA or NVME storage channels and 8x10GbE for some super-fast networking solutions.

As you see, there will be plenty of Zen based Opteron possibilities and most of them will start showing up by mid-2017.  The TDP Range for Snowy Owl is sub 100W and capable of sinking the TDP down to 35W. Yes, we do mean that there may well be a quad core Zen Opteron too.


AMD Goes After Intel’s Skylake With Bristol Ridge

June 9, 2016 by  
Filed under Computing

Comments Off on AMD Goes After Intel’s Skylake With Bristol Ridge

AMD has revealed the firm’s seventh-generation system-on-a-chip accelerated processing units (APUs).

Bristol Ridge and Stoney Ridge sound a little like locations in a Somerset version of Game of Thrones, but they both feature AMD’s Excavator x86 processor cores and Radeon R7 graphics, which AMD sees powering e-sports gaming on laptops.

Bristol Ridge is the more powerful of the two coming in 35W and 15W versions of AMD FX, A12 and A10 processors, offering up to 3.7GHz of processing power. The former two processors are paired with up to eight Graphics Core Next (GCN) cores in the R7 to provide a decent pool of graphics processing power.

Stoney Bridge offers less in the way of processor power, topping out at 3.5GHz, and versions include 15W A9, A6 and E2 processor configurations coupled with lower powered graphics accelerators.

AMD claimed that the new APUs offer a 50 per cent hike in performance over the previous generation Carrizo APUs. However, this rise is over APUs from the early part of Carrizo’s lifecycle, so performance gains over the most recent Carrizo APUs are likely to be 10 to 20 per cent.

AMD also said that its silicon is faster than rival chips from Intel, including the i3-6100U found in several ultraportable laptops.

Many of these tests are subjective and depend on how a hardware manufacture configures and sets up the APUs in a laptop or tablet, but AMD does have its graphics tech to draw on, such as the GCN architecture, which could give it the edge over Intel’s chips when it comes to pushing pixels.

The APUs will be aimed primarily at slim laptops that need low-power consumption chips, much like Intel’s Skylake line.

Bristol Ridge is currently available to end users only in the form of HP’s latest Envy laptop. But now that AMD has debuted the full range of the seventh-generation APUs we can expect to see them in other ultraportable machines before too long.



Oracle Goes Deeper Into The Cloud

May 13, 2016 by  
Filed under Computing

Comments Off on Oracle Goes Deeper Into The Cloud

Right on the heels of a similar acquisition last week, Oracle has announced it will pay $532 million to buy Opower, a provider of cloud services to the utilities industry.

Once a die-hard cloud holdout, Oracle has been making up for lost time by buying a foothold in specific industries through acquisitions such as this one. Last week’s Textura buy gave it a leg up in engineering and construction.

“It’s a good move on Oracle’s part, and it definitely strengthens Oracle’s cloud story,” said Frank Scavo, president of Computer Economics.

Opower’s big-data platform helps utilities improve customer service, reduce costs and meet regulatory requirements. It currently stores and analyzes more than 600 billion meter readings from 60 million end customers. Opower claims more than 100 global utilities among its clients, including PG&E, Exelon and National Grid.

Opower will continue to operate independently until the transaction closes, which is expected later this year. The union will create the largest provider of mission-critical cloud services to an industry that’s worth $2.3 trillion, Oracle said.

Oracle’s Utilities business delivers applications and cloud services that automate core operational processes and enable compliance for global electric, gas and water utilities.

“Oracle’s industry organizations maintain unique domain knowledge, specialized expertise and focused product investments,” said Rodger Smith, a senior vice president who leads the Utilities global business unit, in a letter to customers and partners. “This model has proven highly successful across several industries, and we look forward to bringing these same benefits to the customers of Opower.”


Is nVidia Taking Qualcomm To Court?

April 21, 2016 by  
Filed under Computing

Comments Off on Is nVidia Taking Qualcomm To Court?

Nvidia has dragged Qualcomm into court for allegedly crushing a $352 million chipset deal.

Nvidia claims it was forced to wind down its cellular mobile broadband chipset business, including its Icera unit just four years after buying it, because of Qualcomm’s anti-trust antics.

Qualcomm’s alleged tactics led to “unexplained delays in customer orders, reductions in demand volumes and contracts never being entered into, even after a customer or mobile network cooperating with a prospective customer has agreed or expressed a strong intention to purchase” Nvidia’s chipsets, the company moaned.

The claim for cash comes as European Union regulators step up antitrust investigations into Qualcomm sales tactics that officials said thwarted other designers of mobile-phone chip technology. This could result in fines or an EU order forcing a company to change its behaviour.

The EU thinks Qualcomm may have charged below-cost fees for chips used in mobile Internet modems known as dongles from 2009 to 2011 to thwart smaller competitor Icera. Regulators are separately probing what they say are exclusivity payments Qualcomm paid to a phone and tablet manufacturer for using its designs.

Qualcomm is “confident” it would prevail in both the EU investigation and the lawsuit.

Nvidia is seeking a declaration from the judge that Qualcomm’s conduct was an abuse of a dominant position, compensation, and an account of the profits it says Qualcomm gained from unlawful conduct, according to the court filings.


Do Carriers Want To Abandon Google?

April 14, 2016 by  
Filed under Consumer Electronics

Comments Off on Do Carriers Want To Abandon Google?

Carrier dissatisfaction with the Android maker Google is growing as more of them look to alternatives to curb what they perceive as the search engine outfit’s inflexibility.

AT&T has publically mentioned it is looking at flogging a smartphone powered by an alternative version of Android. If true, the move is a deliberate slap in the face to Google.

US carriers are a little perturbed about the amount of control has over its products and are looking to rivals such as Cyanogen, which distributes a version of Android that’s only partially controlled by Google.

ZTE had been in discussions to make the device, these people say. But mysteriously its involvement was put in jeopardy when the US government suddenly imposed trade sanctions on the company – of course this is nothing to do with Google.

The big idea is to do something like Amazon and create new flavor of Android based on Google’s source code but controlled entirely by AT&T. It would also give AT&T sole responsibility for maintaining the OS going forward.

It would bugger up Google’s because changes to the Android system might be difficult to incorporate into AT&T’s new version, and some might not make it over at all. However AT&T would be able to integrate phones more deeply into its existing infrastructure and issue updates when it wants.

One likely possibility would be an OS-level integration with AT&T’s DirectTV service which is tricky under Google’s rules. It is not clear if AT&T is serious, or if it is just a move to force Google to pull finger.


Microsoft Surprises And Goes Ubuntu

April 13, 2016 by  
Filed under Computing

Comments Off on Microsoft Surprises And Goes Ubuntu

Microsoft has announced a partnership with Canonical which means it is possible to install Canonical’s Ubuntu on Windows 10.

The software is available to all through the Developer Mode on Windows Settings and it is not a virtual machine. Microsoft will allow native ELF binaries, written for Linux, to run under Windows through a translation layer. It is a bit like the WINE project, which runs native Windows binaries on Linux.

Normally you have to recompile Linux software under Cygwin, or run a Linux virtual machine to get it to run in Windows.

Microsoft claims the new feature offers a considerable advantage in performance and storage space. It also includes the bulk of Ubuntu’s packages, installed via the apt package manager directly from Canonical’s own repositories.

The big question is why. Redmond does not appear to be targeting the server market with this launch but desktop and laptop users. It appears to be mainly of use to developers, who need access to Linux software but for whatever reason wish to keep Windows 10 as their main OS.

Canonical’s Dustin Kirkland said the Windows Subsystem for Linux nearly has equivalent performance to running the software natively under Linux. The only downside is the software is free, but not open source.

General release is scheduled for later this year as part of the Windows 10 Anniversary Update, which will also include support for running Windows Universal Apps on the Xbox One, turning any Xbox One into a development system, the ability to disable V-sync for games installed through the Windows software storefront, ad-blocking support by default in Microsoft Edge, and improved stylus support.


AMD’s Vegas GPU Details Spotted

April 11, 2016 by  
Filed under Computing

Comments Off on AMD’s Vegas GPU Details Spotted

Details of AMD’s Vega GPU were leaked and then taken down from AMD R&D Manager’s LinkedIn profile page over Easter.

Hexus spotted Yu Zheng, an R&D Manager at AMD Shanghai had listed the work on Project Greenland as a work experience highlight on his LinkedIn profile page. Greenland is described as “A leading chip of the first graphics IP v9.0 generation, it has full capacity of 4096 shader processor along with whole new SOC v15 architecture.”

The Graphics IP v9.0 designation is thought to signify a Vega GPU in the making. Zheng mentions this is an SOC, but then Hawaii and Fiji chips were described the same. Fiji is part of the graphics IP v8.0 family, as will be Polaris.

Vega following after Polaris, and designated as a ‘HMB2′ GPU by AMD, it looks like Vega based graphics cards will be the successors to the HBM equipped Fiji range such as the Radeon Fury and Nano. Fiju can manage 4096 stream processors,  but with an upgrade to HBM2, 14nm process and other optimisation it is estimated that a Greenland/Vega GPU based graphics cards will offer 20 to 30 per cent better performance.

So with Greenland/Vega sporting HBM2 memory Hexus thinks that Polaris packing graphics cards will therefore feature GDDR5/X memory.


Cisco Fixes Major Flaw

March 23, 2016 by  
Filed under Computing

Comments Off on Cisco Fixes Major Flaw

Cisco has patched high-impact vulnerabilities in several of its cable modem and residential gateway devices which are popular among those distributed by ISPs to their customers.

The embedded Web server in the Cisco Cable Modem with Digital Voice models DPC2203 and EPC2203 contains a buffer overflow vulnerability that can be exploited remotely without authentication.  Apparently all you need to do is send a crafted HTTP requests to the Web server and you could see some arbitrary code execution.

Cisco said that its customers should contact their service providers to ensure that the software version installed on their devices includes the patch for this issue.

The Web-based administration interfaces of the Cisco DPC3941 Wireless Residential Gateway with Digital Voice and Cisco DPC3939B Wireless Residential Voice Gateway are affected by a vulnerability that could lead to information disclosure. An unauthenticated, remote attacker could exploit the flaw by sending a specially crafted HTTP request to an affected device in order to obtain sensitive information from it.

The Cisco Model DPQ3925 8×4 DOCSIS 3.0 Wireless Residential Gateway with EDVA is affected by a separate vulnerability, also triggered by malicious HTTP requests, that could lead to a denial-of-service attack.

Hackers have been hitting modems, routers and other gateway devices, hard lately – especially those distributed by ISPs to their customers. By compromising such devices, attackers can snoop on, hijack or disrupt network traffic or can attack other devices inside local networks.


Next Page »