Wired took a look at why Microsoft is dedicating a lot of resources to equip its cloud services with reprogrammable computer chips. The site explains that about four years ago, a small team of Microsoft engineers started working on Project Catapult, a dream of equipping all of Microsoft's servers with FPGAs that can be reprogrammed to perform new tasks.
After a lot of pitfalls, Project Catapult was rolled out en masse and is now powering millions of Microsoft machines across the globe, to speed up Bing, Azure and Office 365 applications. The FPGAs drive new search algorithms based on deep neural networks and execute these tasks much faster and more power efficient than ordinary processors can.
It’s a typical tangle of tech acronyms. CPUs. GPUs. TPUs. FPGAs. But it’s the subtext that matters. With cloud computing, companies like Microsoft and Google and Amazon are driving so much of the world’s technology that those alternative chips will drive the wider universe of apps and online services. [Microsoft Research vice president Peter] Lee says that Project Catapult will allow Microsoft to continue expanding the powers of its global supercomputer until the year 2030. After that, he says, the company can move toward quantum computing.
The FPGAs used by Microsoft are manufactured by Altera, and according to Intel executive vice president Diane Bryant, Microsoft is why Intel bought Altera for $16.7 billion in the summer of 2015. You can read the full piece over here.
Just a quick note that SpaceX achieved the first test firing of Raptor, the company's interplanetary transport engine. Raptor is SpaceX's first methane-fueled rocket engine, it's the successor to Merlin, the kerosene-based rocket used by the Falcon 9. The new engine will be used by SpaceX's super-heavy lift launch vehicle, which will be capable of lifting up to 100 tonnes of payload into Mars orbit.
With a production goal of a specific impulse of 382 seconds and thrust of 3MN (~310 metric tones) at 300 bar, Raptor is expected to provide over three times the trust of the current Merlin engines. Musk shared on Twitter that chamber pressure is almost three times as high as Merlin's, so the engine will be about the same size for a given area ratio.
The test firing comes just in time for Elon Musk's highly anticipated talk at the International Astronautical Congress (IAC), which is expected to cover in great detail the company's Mars exploration and colonization plans.
A livestream of the Elon Musk presentation at the IAC will be available over here. His talk will deal with making humans a multiplanetary species, it starts at 14.30 EST and could potentially be the "We choose to go to the Moon" speech of our generation. A lot of people are still very skeptic about private space exploration, but it can't be denied that firms like SpaceX and Blue Origin are changing the space industry in a very disruptive way and I sincerely hope Musk can achieve his dreams.
Over the weekend, a lot of media sites made a great buzz about the political antics of Oculus founder Palmer Luckey as it was discovered that he had donated $10,000 to Nimble America, a pro-Trump group dedicated to proving "shitposting is powerful and meme magic is real". Other than funding an anti-Clinton billboard in Pittsburgh, it's not really clear what Nimble America has actually produced.
Taken aback by the backlash, Luckey apologized on Facebook and explained he supported the organization because he thought they "had fresh ideas on how to communicate with young voters through the use of several billboards." He also stressed his actions were his own and do not represent Oculus. I guess this is another example of the danger of mixing politics with business, when you pick a side you're always going to alienate some of your (potential) customers. In fact, some small developers said they would cease supporting the Oculus Rift.
Anyway, getting back to business, DigiTimes wrote last week that Oculus is looking to enter a partnership with MSI to co-develop virtual reality backpack PCs. Earlier this month, MSI showed off its VR One, a VR backpack developed on cooperation with HTC. Now Oculus wants to get in on the action too.
Since Oculus has been cooperating with Microsoft to promote its VR products, a possible tie-up with MSI will help Oculus to expand sales in the gaming notebook segment, said the sources.
MSI has established a partnership with HTC to promote sales of its gaming notebooks bundled with the HTC Vive HMD devices as well as some VR games in Japan starting October.
Biostar introduces its very first gaming mouse. Design-wise this white mouse looks a lot like the good old IntelliMouse from Microsoft, it has an ambidextrous design but lacks any side buttons. The Biostar AM2 features an Avago 5050 sensor with up to 2400 dpi, four-step on-the-fly DPI switching, and you also get some fancy LED lights.
Overall there isn't much to say about this mouse but thanks to its suggested retail price of just $9.99 this may be a good low-budget deal.
BIOSTAR is pleased to announce the introduction of its first gaming mouse designed to offer gamers everything they need for their first-ever gaming mouse. BIOSTAR today introduces the new BIOSTAR AM2 gaming mouse featuring an Avago 5050 sensor capable of up to 2400DPI of tracking, perfect for gamers who are looking for their first gaming mouse.
As competitive gaming grows more and more, the trend in high-performance gaming mice has become a challenge for both gamers and manufacturers as it's a challenge to find the middle-ground for the various gaming genres and preferences. As BIOSTAR is well-known for its value-based proposition, it takes this into consideration as well as feedback from gamers into the design of the BIOSTAR AM2 gaming mouse. The result is a striking balance of performance and quality in a simple offering meant for any and all gamers who want the most essential features like on-the-fly DPI switching, comfortable grip and excellent tracking.
A simple style mixed with highly accessible function makes it very at-home with gamers, experience or just starting, with their first high-performance gaming mouse. Designed for comfort, performance and durability, the BIOSTAR AM2 gaming mouse understands the needs of those new to the search for the first gaming mouse and introduces an affordable balance that marries the best of this price point with BIOSTAR's engineering technology.
The BIOSTAR AM2 gaming mouse features the raw essentials needed to get started in advanced gaming peripherals. The BIOSTAR AM2 is designed with the raw feel needed to give gamers a better experience with their games especially fast-paced eSports and competitive titles. The BIOSTAR AM2 features an on-the-fly DPI toggle button allowing quick cycling between preset DPI settings for unhindered gaming and rapid adjustment so you get the best sensitivity when you need it. The built-in LED lighting switches colors to indicate the DPI setting which are as follows " 4 DPI steps: default 800 DPI starts with no color, 1200 DPI which is blue, green for 1600 DPI and blue-green for 2400 DPI. This mixes in style with function which converges both the unique gamer-style with actual function.
The HUANO switch provides excellent tactile feedback and durability lasting up to 10 million clicks for non-stop gaming.
The BIOSTAR AM2 gaming mouse is designed to be an ambidextrous mouse for use by both left- and right-handed gamers. The size is optimized for claw- and fingertip-grip for precision control even in the most intense games but provide the swift response of the grip of your choice. The high-gloss mirror UV surface makes the mouse easy to clean and doesn't absorb dust or dirt giving it that shiny, fresh look everytime especially when you bring it to LAN parties or gaming tournaments. Teflon mouse feet provide a slick and smooth glide for unhindered movement and friction reduction.
A couple of days ago, TSMC revealed some more information about its future process nodes. The company's 28+ HKMG and 16FF+ processes are in production now and 10nm will ramp this year.
Jack Sun, chief technologist and vice president of R&D at TSMC, also said yields for 256Mbit SRAM chip at 7nm are two months ahead of schedule.
Risk production on the 7nm node is expected to kick off in Q1 2016 and the company provided some details about the expected performance of its future 10nm and 7nm nodes:
The 16nm processes provide about 45% more speed and 80% less leakage than 28+ and supports four voltage levels, TSMC said. The 10nm process provides a 50% die scaling and 20% speed gain or 40% power reduction over 16FF+ and provides “the highest density in the industry today in contact pitch,” said Sun.
Compared to 10nm, TSMC’s 7nm node delivers 15-20% more speed or 35-40% less power consumption and a 1.63x better routed gate density, said Sun. An ARM Cortex-A72 core in the 7nm process could deliver 30% more performance or 56% less power consumption than in 16FFC, said Hou.
The latest rumors coming out of Asia suggest NVIDIA will roll out its GeForce GTX 1050 lineup in October, these cards will be based on the GP107 GPU.
First up will be the GeForce GTX 1050 Ti with a launch around mid-October. This card features the GP107-400 GPU with 768 CUDA cores, 48 TMUs, 32 ROPs, a 1290MHz baseclock, 1382MHz Boost and 4GB GDDR5 memory clocked at 7008MHz.
Towards the end of October we can expect the GeForce GTX 1050, that card will use the cut-down GP107-300 GPU with 640 CUDA cores. The leaked specifications indicate 40 TMUs, a 1354MHz base clock, 1455MHz Boost and 2GB GDDR5 memory.
Both cards use a 128-bit memory bus and fit in a 75W TDP. Guess we'll find out more over the coming weeks.
A couple more tidbits about AMD's future processor plans appeared online. The next big step in AMD's CPU roadmap is the launch of the first 14nm Zen processors early next year, with Zen-based APUs following later in the second half of 2017.
AMD Starship to offer up to 48 cores, 96 threads
This post takes a look at the current plans for the first 7nm processors from AMD. FUD Zilla claims this includes Starship, a 7nm processor that promises up to 48 cores and 96 threads. Starship is aimed at the high-end desktop and the server markets, and will reportedly target TDP values ranging from 35W to a massive 180W. No launch date is given but I wouldn't expect it before late 2018 at the earliest.
AMD Gray Hawk is the 7nm APU
In the second half of 2017, AMD is expected to introduce the 14nm Raven Ridge, the first Zen-based APU with Polaris-based integrated graphics. It seems Gray Hawk will be Raven Ridge's successor, this chip will be made on a 7nm process and will offer up to four cores with eight threads. Perhaps Gray Hawk will feature Navi-based graphics, but that's still unclear. The report also mentions Gray Hawk will be offered in TDPs as low as 10W and mentions it is scheduled for a 2019 launch.
As always take this information with a grain of salt, it's based on current rumors and these specifications and plans are subject to change. If the information is correct, it seems there may be a relatively big gap between the launch of the first Zen-based products and the next generation.
Earlier this week I covered some fresh rumors about AMD's Vega and Navi lineup, that news originated from VideoCardz and parts of it were confirmed by FUD Zilla. The latter now offers a couple of updates about Vega20 and Navi -- and some of the information is in conflict with what VideoCardz wrote earlier this week.
Vega10, Vega11 and Vega20 all slated for 2017?
First up, FUD Zilla writes the Vega10, Vega11 and Vega20 are all scheduled for a 2017 release. Vega10 is planned for the first half of 2017 and will be AMD's first GPU with HBM2. This is the new high-end card from AMD and its launch will be followed by Vega 11, which will be used for the mainstream segment that's currently occupied by Polaris.
The weird thing is Vega20 is expected to follow six to nine months after the launch of Vega10, the site describes it as an "updated version" but it's unclear what kind of innovation you can expect. Based on earlier rumors, it may be that Vega20 is a part that will compete against NVIDIA's flagship Titan and Tesla products. It's expected to feature double the memory and memory bandwidth as Vega10, as well as good compute characteristics.
Vega20 is believed to be a 150W+ product and may arrive with PCI Express 4.0 support. FUD Zilla claims Vega20 will be made on a 14nm FinFET process which conflicts with the VideoCardz rumor about Vega20 being a 7nm product. FUD Zilla claims Vega20 is planned for late 2017, which will likely become early 2018 after some delays, whereas VideoCardz claims Vega20 is planned for the second half of 2018.
Either way, it seems AMD's 2017/2018 roadmap will consists primarily of Vega derivatives.
Navi to be 7nm part with late 2018 or early 2019 arrival?
The second story deals with Navi. FUD Zilla claims Navi will be the first AMD GPU to be made on a 7nm process by GlobalFoundries and that this part is expected by late 2018 or early 2019. Previous AMD roadmaps indicate Navi will use an unnamed next-gen memory technology.
As always, take this information with a grain of salt.
One of the nifty new products from HP is Sprocket, a highly portable photo printer capable of printing 2 x 3" (50.8 x 76.2mm) photos on the go. The device is about the size of a smartphone, it connects to an Android or iOS based tablet or smartphone via Bluetooth and prints your photos in less than a minute.
Sprocket requires the use of HP's ZINK Zero Ink Photo Paper, which has the colors embedded on the paper itself. HP promises photos printed on the ZINK paper are smudge-proof and water resistant.
While highly convenient, printing with the HP Sprocket is quite expensive. The pocket printer itself costs $129.99 and comes bundled with 10 ZINK Photo Papers. Packs of ZINK Photo Paper with 20 sheets will be sold for $9.99.
You knew it had to happen: a miniature printer for a snap-happy generation that likes to both collect and curate the moments of their lives.
Today, HP announced Sprocket , a portable, palm-sized printer that’s small and light enough to fit in a purse or back pocket yet capable of producing 2x3” stick-able snapshots from a mobile phone or tablet via Bluetooth in less than a minute.
The vivid prints are produced using ZINK® Zero Ink® Technology, which means they are smudge-proof, water resistant and can be shared immediately.
A free HP Sprocket App for iOS and Android lets users quickly customize and print smartphone and social media photos by adding frames, text, stickers, filters, and more.
HP envisions the new device finding popularity with teenagers and millennials who are drawn to fun, stylish gadgets that might be a hit at social events. In fact, that’s precisely where HP sees Sprocket finding its biggest following – at parties and other gatherings where fun-seekers might be taking selfies or looking at social media posts that they would like to print out, share nd take home with them.
HP ZINK® Photo Paper delivers bright, glossy photos without ink or toner cartridges, -- all the color required for printing is embedded in the HP ZINK® photo paper itself. It delivers printed snaps that are vibrant smudge-proof, water-resistant, and tear-resistant. Peel off the backing, and the photos become stickers to decorate bags, scrapbooks, journals, and walls.
Aqua Tuning claims it has a world exclusive with the very first 4-pin PWM to 3-pin converter. Available via the company's website for 9.79EUR, this tiny board allows you to control 3-pin fans like any other 4-pin PWM fan. You connect the board to your motherboard or PWM fan controller and it automatically detects the PWM signal and adjusts the voltage of the 3-pin fan to the corresponding level.
By using a splitter, it's possible to control multiple fans simultaneously, as long as the current doesn't exceed a level of 1A.
Controlling 3-pin fans through a PWM signal was an unsolvable problem – until now. Phobya has tackled this problem and now presents an elegant, simple solution with the 4-pin PWM to 3-pin converter.
With the 4-pin PWM to 3-pin transformer, all 3-pin fans can now be controlled with a normal PWM signal. The converter is connected to your motherboard or PWM fan control, while the fans are connected to the converter. Now your 3-pin fan can be controlled like any other PWM fan.
The converter detects the PWM signal and controls the 3-pin fan with the corresponding voltage. This makes controlling your fan simple. The converter is capable of outputting a voltage (sic) of 1A, which equals 12W. This means you can use a Y-splitter to control several fans simultaneously through the converter.
Dimensions (LxWxH): 50 x 39 x 18mm
In: 4Pin PWM
Power connection: 4Pin plug
Mounting: Double-sided tape for attaching at/in the case