AMD indicated that the official Ryzen launch date will be sometime before March.
While they haven’t specifically given an exact date, a talk to be given by AMD at the annual Game Developer Conference (GDC) says the following: “Join AMD Game Engineering team members for an introduction to the recently-launched AMD Ryzen CPU followed by advanced optimisation topics.”
Obviously for this to be the subject of the talk Ryzen would have had to be recently launched which means that it is probably timed for that week.
GDC event runs from 27 February to 3 March and has not been put on the schedule yet and it could appear any day during the event.
AMD has not disclosed an exact date either, launching the new set of Ryzen CPUs right in the middle of both GDC and Mobile World Congress would be insane as the news would end up being buried under other GDC and smartphone announcements.
It would make sense to do it the week before all that, if not two.
MediaTek, the fabless semiconductor company from Taiwan that provides SoCs for HDTVs, Blu-ray players and wireless products, saw its revenues jump by 29.2 percent year-over-year to a record high of $8.6 billion (¥$275.51 billion), according to the latest industry reports.
Deemed one of the fastest growing chip companies in 2016, MediaTek’s upswing in performance last year is attributed to a larger share of the worldwide smartphone SoC market, along with higher sales in local China and Taiwan markets. In Q4 2016, revenues totaled $2.18 billion (¥68.68 billion) which is down 12.4 percent over the previous quarter, but still falls within the company’s projection of $2.11 and $2.31 billion (¥66.6 to 72.9 billion).
In Q3 2016, revenues totaled $2.49 billion (¥78.4 billion), an increase of 8.1 percent over Q2 and a 37.6 percent increase over the previous year. Net profits also rose to $248.4 million (¥7.83 billion) in Q3, an increase of 18.8 percent over the previous quarter but down 1.6 percent over the previous year.
Going forward into 2017, company officials now want to shift its focus from increasing market share to improving gross margins and profitability. This will include an effort to market its high-end Helio X30 and X35 mobile processors more effectively to compete against the likes of Qualcomm’s Snapdragon 830 and 835 and Samsung’s Exynos 8895, as all three companies are now using ARM cores with 10-nanometer designs.
MediaTek MT5597 supports Dolby Vision and HLG
The Taiwanese chip designer was the first to develop an 4K Ultra HD-capable SoC for Android TVs with the introduction of the MT5595 for Android TV 5.0 back at CES 2015. It followed up a year later with the MT5996 for Android TV 6.0, another world’s first featuring four 64-bit CPU cores based on the Cortex A53 design.
Now in 2017, the company is releasing its third-generation Ultra HD SoC for Android TV 7.0, the MT5597. This chip also features a quad-core Cortex A53 design but now includes support for Dolby Vision HDR and Hybrid Log Gamma (HLG), the standard expected to be used in UHD television broadcasts when providers are ready to roll out HDR terrestrial and satellite services.
Last week at CES 2017 we noticed that Intel felt so unsecure about itself that it used a big curtain to hide the Qualcomm logo from the booth next door. This really happened and we were lucky to document it.
Intel and Qualcomm had booths next to each other at CES 2017 and this has been going on for a few CES shows. The difference this year, was that Intel ordered a huge curtain to block the view to Qualcomm’s logo standing on the side of the booth. Qualcomm didn’t.
We are not sure how the CES 2017 organizers were OK with it, but it definitely felt at least a bit tasteless to do this to a company you are directly competing against. We took the picture from both Intel’s and Qualcomm’s booth.
This of course has a big background and deserves a bit of an explantation. Intel won part of the iPhone 7 modem deal and clearly stated that it wants to compete even more on 5G, picking on Qualcomm’s core business more actively. Intel chose CES 2017 to announce its 5G modem, while Qualcomm had already announced the Snapdragon X50 5G SoC. Qualcomm is expected to deploy Snapdragon X50 for testing in the second half of 2017 and is expected to do 5G trials with AT&T, SK Telekom and Verizon.
At the same time, Microsoft shocked the IT world by confirming that the Snapdragon 800 series of chips have full support for Windows 10. We saw a demo of a Snapdragon 820 machine and there will definitely be Snapdragon 835 notebooks supporting Windows 10 coming to the market later this year. Needless to say, the companies developed a big rivalry in the last two years.
Last year, Intel let go some 12,000 employees in April while Qualcomm let go off some 4,500 people in order to restructure the company. In the meantime Qualcomm has acquired an automotive SoC giant NXP for $47 billion preparing the company to be a big influence in the automotive industry.
Fudzilla also noticed that there have been quite a few people crossing over from Intel to Qualcomm and in the other direction, implying that Qualcomm can use some help in competing wotj Intel on the PC side of the industry.
Intel on the other hand has its hands full. Its big rival AMD is just weeks from releasing the Summit Ridge codenamed Zen-architecture based 14nm eight core Ryzen processor. The first impression is that the CPU looks really competitive to Intel’s high end Extreme Edition desktop line and AMD announced more than a dozen systems that are expected to launch in Q1 2017 and onward.
Later in 2017, both AMD and even Qualcomm are expected to announce their server solution and increase the pressure on Intel in this heavily profitable market. Back in 2013 Intel stated that for every 400 smartphones you need one server. Since the amount of traffic grew massively between 2013 and today and the fact that video is now 55 percent of all internet traffic, we believe that you need one server for a few hundred phones, definitely less than 400 these days.
Qualcomm was the first to announce and showcase the 10nm SoC, and we saw a live demo of the Snapdragon 835 based prototype of a phone. As a few executives mentioned the other day, there is a high expectation that Snapdragon 835 might end up being even more successful than Snapdragon 820, and we already pointed out that Snapdragon 820 had more than 200 design wins. The performance, footprint and 25 percent better battery life compared to Snapdragon 820 have every chance to make a three billion transistor Snapdragon 835 an instant success.
Intel has completely abandoned smartphones after billions of investments and a failure to make a dent in this huge market. Intel’s X86 based mobile SoCs simply failed to compete with ARM based chips.
So far, Intel has tried and failed to launch 10nm processors, and this is happening to a company that owns some of the world’s most advanced and biggest FABs in the world. After decade of tick-tock execution, the recent launch of Kaby Lake, the third generation 14nm processors for desktop and notebooks definitely confirmed there’s trouble in paradise. One can only hope that Intel can get the 10nm processors out before the end of 2017 but it has become increasingly hard to migrate from one to another manufacturing node.
Tim Sweeney has urged virtual reality platform holders to ensure their devices and marketplaces are open to other providers in order to avoid the same monopoly that affects a significant portion of the smartphone market.
Speaking to Glixel, the co-founder of Epic Games said he say “a lot going on that’s wrong” in other tech sectors, and points to Apple as a prime example. While he thinks its fine for the firm to be the only distributor of its hardware, he disagrees with their monopoly on distributing software and collecting in-app revenue.
His added that virtual reality pioneer Oculus seems to be operating in a similar way to Apple, adding that this is “the wrong model” for virtual reality and something Sweeney “argued passionately against”.
“When you install the Oculus drivers, by default you can only use the Oculus store,” he said. “You have to rummage through the menu and turn that off if you want to run Steam. Which everybody does. It’s just alienating and sends the wrong message to developers. It’s telling developers: ‘You’re on notice here. We’re going to dominate this thing. And your freedom is going to expire at some point.’ It’s a terrible precedent to set.”
Sweeney believes that ultimately the open platforms will win as they will have a better selection of software. He praised HTC Vive for being such a platform and noted that the device is currently outselling Oculus two-to-one around the world, a trend he expects to continue.
The Epic Games founder acknowledged that his firm is making a closed-platform game for Oculus in the form of Robo Recall, a title that stemmed from Unreal Engine’s Bullet Train VR demo. However, he attributed this to the fact that the game is funded by Oculus and could never have been built on a budget based purely on sales.
“The Oculus store… is an awesome store [but] should run on all PC and VR devices,” he said. “Oculus would do best if they tried to bring users into their store by supporting HTC Vive and Oculus Rift and any other PC hardware that comes out. I think if they don’t do that, they’re going to pretty quickly fail, because you’re not going to want to buy a multiplayer game that you can’t play with half of your VR friends.”
Sweeney previously spoke out about the need for open virtual reality platforms during his keynote at the latest Steam Dev Day, saying: “It would be really tragic if we let the future metaverse, that binds all humanity together into shared online environments, were a closed platform controlled by a giant corporation.
“As always, they’d use it to spam you with advertising, they’d use it to gather information about your private life and sell it to the highest bidder, and they’d act as the universal intermediary between all users, content creators, and transactions, ensuring that everything has to be approved by them.”
A set of new AMD Ryzen slides that have surfaced recently has confirmed that the entire Ryzen CPU lineup will have an unlocked multiplier and be ready for launch by the end of Q1 2017.
Although AMD has mostly been talking about its high-end 8-core/16-thread Ryzen SKU, there will be a full line of Zen-based Ryzen CPUs, including the rumored 6-core/12-thread and 4-core/8-thread version, probably coming with different clock speeds.
The newest set of slides confirms that every AMD Ryzen CPU will actually be unlocked, which means an unlocked multiplier, which should also make them quite overclocking friendly. Of course, the actual overclocking potential is still unknown but we hope there will be plenty of it.
The slide also reveals that AMD will actually have all-new lineups of both system integrators and OEM systems as well as an extensive lineup of 3rd party cooling solutions.
As mentioned by Robert Hallock, AMD is not targeting the very last day of Q1 2017 as the launch date so hopefully, these will come pretty soon.
As we wrote earlier, AMD’s CTO, Mark Papermaster, has confirmed that the company expects Ryzen to have a four-year lifespan, which is quite common for a new architecture.
While we have said that Intel’s new Kaby Lake are a bit of snooze a high-end custom PC builder has taken advantage of the one thing it is good at – overclocking.
For $75 extra Origin will boost Kaby Lake desktops chips over the 5GHz barrier. Intel’s processors usually stop at about 4.5GHz. Origin can’t guarantee a set frequency as that depends on the chip Intel flogs them. But they do promise that “Origin PC’s award winning system integrators will overclock your processor and squeeze out every last megahertz” with every overclock “stringently tested and benchmarked for ensured stability.”
However, the outfit does seem confident in mentioning the 5GHz number in its marketing bump.
It might breathe some life into Kaby Lake and make it workwhile. All it would take is a few PC makers in Europe to get on the bandwagon and there could be a good product out there. Of course you could wait until Zen arrives, but that is a bit like waiting for Godot.
Intel is starting to talk up its new, 10-nanometer chip-manufacturing technology and claims that it is going to prop up Moore’s Law.
According to IEEE Spectrum claims its new 10nm transistors will be cheaper than those that came before.
This contradicts talk that transistor-production costs have already sunk as low as they will go. Intel plans to make further improvements to the design of these transistors and optimise its manufacturing technology to accommodate other companies that wish to use Intel’s facilities to produce chips based on ARM architecture.
Intel senior fellow Mark Bohr said that transistors on Intel’s 10-nm generation will still be denser than those on today’s 14-nm chips—as well as other companies’ 10-nm offerings.
The minimum gate pitch will go from 70 nm to 54 nm. And logic cells will be less than 46 percent the size of those built on 14-nm technology.
Bohr said that this is a more aggressive level of miniaturisation than before.
“One important message is that this node, and the products that we’ll be making on it, will hopefully dispute some of the concerns of the industry that Moore’s Law is slowing down,” Bohr says.
While the cost of producing a 10nm wafer will be higher at 10 nm than at 14, Intel says the cost per transistor will be lower.
The 10-nm chips are also expected to deliver improvements to switching speed and energy consumption. As has been the case for years already, clock speed isn’t liable to increase, though. “It’s really power reduction or energy efficiency that’s the primary goal on these new generations, besides or in addition to transistor cost reduction,” Bohr says.
Improved compactness and efficiency will make it more attractive to add more cores to server chips and more execution units onto GPUs, he says.
“It’s really power reduction or energy efficiency that’s the primary goal on these new generations”
Intel used to upgrade its transistors once every couple of years, with the introduction of a new manufacturing generation. At 14 nm it just issued another improvement before 10 nm appeared. At 10 nm, the company will to introduce two 10 nm+ and 10 nm++ before it moves to 7 nm.
Bohr said that if it is going to take longer to get to 10-nm or to the 7-nm generation, then the next best thing to do is to find ways to enhance what you have today and have new products annually.
Intel says the first 10-nm chip to ship will be one of its own processors, with chips for others to follow, but perhaps not until 2018.
For the past three years, production of TSMC’s leading-edge process nodes – 20nm, 16nm and 16FF+ – has always ramped up following orders of a new iPhone SoC – the Apple A8, A9 and A10. In 2017, it appears this will not be the case as the researchers claim that TSMC’s 10nm yields are currently still below 50 percent.
Apple A10 undergoing small pilot runs
In response, TSMC is now scheduled to be doing small pilot runs of Apple’s A11 SoC right now in Q4 2016 and will instead begin ramping up production in late April. The report suggests that the foundry’s 10nm process should be able to exceed 20,000 wafers per month in Q1 2017 following the small pilot runs.
Back in May, we wrote that TSMC had finished taping out the company’s A11 chips, with certification coming in Q1 2017. and that they would likely start generating revenues for TSMC in the third quarter of 2017.
Apple A10X production for next iPad Pro will also ramp slowly
Meanwhile, TSMC’s 10nm volumes for Apple’s A10X chip slated for the next iPad Pro models are reported to be roughly half of MediaTek’s 10nm Helio X30 and X35 SoC volumes, with just 7,500 wafers per month expected from January to March.
As Ashraf Eassa from The Motley Fool notes in his own report, however, Apple’s A10X and MediaTek’s X30 and X35 are still relatively low-volume products compared to Apple’s A11 for the next-gen iPhone. Even if TSMC’s 10nm yields are below 50 percent, it would suggest that shipment volumes should be relatively modest up until the end of April, without TSMC’s overall gross profit margins being diluted too much.
Given the estimated $120 million design cost for developing 10nm chips, according to Gartner at least, TSMC really does not want to have any issues with low yields once production begins to ramp up in just a few months from now.
A recent report from Chinese-based Economic Daily News claimed that MediaTek would be slashing orders for its upcoming ten-core, 10-nanometer chips by at least 50 percent. Now, the company has publicly dismissed the rumor, saying it was not the source.
The report originally claimed that MediaTek had revised its 10nm chip outlook for 2017, which would affect its overall competitiveness with TSMC’s other customers including Apple, Hisilicon and Spreadtrum.
The company’s upcoming deca-core Helio X30 and X35 chips are expected to compete with Qualcomm’s Snapdragon 835 by using a tri-cluster configuration – two Cortex A73 cores at 2.8GHz, four Cortex A53 cores at 2.2GHz, and four Cortex A35 cores at 2GHz.
In September, the company announced its partnership with TSMC for volume production of the Helio X30-series SoCs for the high-end market, along with X35 SoCs for the lower-end market. The latter are expected to use a lower standard of the company’s 10nm process design. Production on both chips is scheduled to have already begun this month through January, while sales revenue is expected to begin in the first quarter of next year.
AMD’s CEO Lisa Su said that Summit Ridge, the processor that we will know as Ryzen, will arrive in Q1 2017 and she mentioned that Naples Zen for server will appear in Q2 and Raven Ridge, Zen APU will deliver in the second half of 2017.
It looks like 2017 might be a good year for AMD as the company plans to fill its void in the desktop, server and even the notebook market. Ryzen is an eight core CPU clocked at 3.4GHz or higher – it is the first one and AMD told us to expect it in Q1 2017.
The more interesting part is codenamed Naples and Fudzilla exclusively posted the news about this, including a diagram.
Since AMD said that Ryzen has 20MB of total cache it is easy to work out that there are eight cores each with 512KB L2 cache and 16GB of L3 cache. We posted these details a few months ago. This means that Summit Ridge and an eight core Snowy Owl server will be based on the same core.
AMD has two separate codenames in the server market, Snowy Owl for 8 and 16 core server chip while Naples is reserved for 24 and 32 core versions. They are all based on Zeppelin blocks of eight Zen cores Fudzilla mentioned earlier this year.
The information we gathered lists 16 core, 24 core and 32 core version of Naples server architecture core too. Obviously an eight-core version of Naples shares the DNA of Summit Ridge. The obvious reason is that the desktop version uses the AM4 socket while the server versions will use the SP3 and SP4 sockets.
One can only hope that Ryzen’s desktop part is going to ship to customers in Q1 2017 while the server Naples could launch in Q2 2017, in customer designs of course. One obvious reason to firstly launch a desktop part and follow up with servers is that desktops need less qualifications and all AMD needs is a few motherboards and some OEMs to launch the products.
Launching Naples is a bigger deal, as it means higher margins, higher revenues and an attack on the highly profitable server market. This is a market where Intel controls 99 percent of sales. This advantage is going to go, and AMD will win some server market back. Intel doesn’t have a 32-core server based part to compete with Zen.
The Skylake-SP processor that is supposed to launch in Q2 2017 supports up to 28 cores indicating that AMD will put pressure on Intel’s server market cash cow. Again the Skylake-SP processor might as well be a 32 core processor that has 28 cores enabled (due to yield problems) which could likely mean Intel will have a 32-core response soon. If such a core is possible, we would expect limited availability but again, maybe Intel didn’t want to show all its cards, while waiting for AMD to show its Naples hand first.
Intel will unveil its new Baby Canyon NUC platform in the first quarter of 2017 to succeed its existing Rock Canyon and Swift Canyon platforms, using Kaby Lake or entry-level Apollo Lake based processors.
A Core i3 version of the Baby Canyon platform will arrive in January, followed by a Core i5 version in February and a Core i7 in March. This means that Intel will offer five CPUs for its new NUC systems including Core i7-7567U, Core i5-7200U and Core i3-7100U.
The company’s existing Skull Canyon platform with Core i7 processors for the gaming market is expected to die off in the first half of 2018.
Intel does have competition in the NUC market from Gigabyte, Elitegroup and ASRock, Asustek, Zotac and Shuttle. Gigabyte and ECS have also been promoting their NUC devices for enterprises.
AMD noted that Intel’s Core i7-6900K is currently the only eight core/16-thread CPU on the market so it was impressive to see it compared to AMD’s Ryzen SKU, core to core and thread to thread. While these two might be matched in terms of cores and threads, AMD’s Ryzen SKU was running at a 3.4GHz base clock rate without Boost while Intel’s Core i7-6900K was left at the starting gate with a 3.2GHz base and a 3.7GHz Turbo CPU frequency.
AMD picked two CPU “intense benchmarks”, Blender and Handbrake tests, both which ought to scale well with multi-core and multi-thread CPUs.
In Blender, both AMD Ryzen SKU and Intel Core i7-6900K were neck to neck, although running at different clocks, which is quite impressive. In the Handbrake test, an open-source video transcoder, AMD’s Ryzen was actually faster in transcoding the Ryzen video to AppleTV 3 preset.
Analyst at Canalys think that there will be more than two million VR units sold in 2016 and the figure will grow to 20 million by 2020.
The bulk of 2016 shipments are basic VR headsets that rely on other devices, generally being tethered by cable to a desktop PC.
Shipments of smart VR headsets, which can function independently, will only be 100,000 units.
These estimates only include VR headsets with integrated displays and exclude viewers, such as Samsung’s Gear VR and Google’s Daydream View. These are shipping in the millions.
Sony is the VR market leader, with its affordable PlayStation VR catering to the vast PlayStation 4 installed base. Canalys expects over 800,000 shipments in less than three months on the market.
HTC will ship around half a million units in 2016, putting the Vive in second place. Facebook’s Oculus Rift, meanwhile which has been boosted from its long-awaited Touch motion controllers and will reach almost 400,000 shipments in 2016.
More than 300,000 VR headsets are estimated to ship in Greater China in 2016, with HTC leading the charge, while local vendors, such as Deepoon, Idealens, 3Glasses and ANTVR, have provided their own unique innovations and localization efforts that are vital for the massive China market, Canalys said.
It has been said for a while now that Intel’s glorious empire, on which the sun never sets, could be overthrown by barbarians, but so far the Chipmaker has managed to thrive.
However next year there are few milestones coming up which must be filling Intel executives with dread – competition in a market which is shrinking.
Firstly there will be the much awaited arrival AMD’s Zen. Early rumours suggest not only will the chips be better than anything Intel will have, they will also be cheaper. Unlike many AMD fanboys we don’t underestimate the fabless chipmaker’s ability to snatch defeat from the jaws of victory. AMD might still stuff up what could be an open goal and let Intel off the hook. AMD depends on getting enough chips to market at the right time, something historically it has always managed to stuff up. Still on paper at least, Intel is going to bat as the underdog.
To make matters worse, the PC market is still shrinking, so any war with AMD will be taking business away from Intel just when it does not need it.
Last Wednesday Qualcomm entered the PC and server markets with its Centriq 2400 server chips, which started shipping to test customers. The chips have the backing of Microsoft which said that PCs based on Qualcomm’s Snapdragon 835 chip would come next year.
Qualcomm is onto a winner here because the chip can also be used in high-end smartphones, giving a cross platform performance which Intel can’t match – particularly after it pulled out of the mobile market.
The jury is still out on whether Qualcomm will succeed, but the company is the first legitimate competitor to Intel in the server, PC, and IoT chip markets.
So now Intel must face both AMD and Qualcomm on its home turf and in a shrinking market. It would need to come up with something new and better fast. It tends to say the mantra Internet of Things a lot, but so far that gold mine is yet to provide any ore.
Chipzilla has tried its hand at telecommunications and modems. It seemed to be starting a bold new era in that direction when it flogged mobile modems for the Apple iPhone 7. However, that experiment was not particularly successful. Apple had to throttle its faster Qualcomm modems so that they did not show up the weaknesses of Intel’s modems to its customers. Apple wanted to slap Qualcomm over the nose with a rolled-up newspaper for daring to charge it money for its modems, but it ended up slowing down the iPhone 7. It is a mistake it probably will not make again.
Our bet is that Intel will not react to the new pressure. There is a certain arrogance which the outfit has. It has seen off similar problems in the past and might just do so again. However, never has Intel looked so shaky and we predict it will start to suffer next year.
The divination team from Pacific Crest have been shuffling their tarot cards and are seeing some dark clouds forming for AMD and Nvidia.
Chief diviner Michael McConnell claims that his “dark clouds forming” note was after listening to comments from Asia’s supply chain and not the whispering demons who spoke to him after chewing on laurel leaves.
McConnell, writes that he was surprised by the sudden negative tone about “desktop” graphics cards:
“Our specific findings were as follows: High-end NVIDIA GeForce GTX 1080 and 1070 card inventory levels have risen to two to 2.5 months in the channel versus targeted levels of one to 1.5 months due to weaker-than-expected sell-through in late October and November,” he said.
Two weeks ago, desktop graphics card manufacturers began to experience order pushouts and cancellations of GTX 1080 and 1070 cards from channel customers ahead of the holiday season.
Given the excess supply, GeForce GTX 1080 pricing dropped 10 per cent in the channel, with desktop card manufacturers now unwilling to order product due to perceived working capital risk.
“Given the weaker-than-expected sell-through of higher ASP GeForce GTX 1080/1070 product and higher sales mix of mainstream GeForce GTX 1060/1050, not one desktop graphics card manufacturer we surveyed is expecting sequential revenue growth in calendar Q4, with forecasted sales declines of five to 15 per cent,” he wrote.
This contrasts with Nvidia’s guidance for sequential revenue growth in its gaming segment (62 per cent of sales) in the fourth quarter, after record-high sales in the third.
Sales of AMD’s desktop Radeon 480/470 graphics cards were also disappointing, he added.
“Given weaker-than-expected desktop graphics card sell-through and oversupply, we believe Nvidia and AMD have now implemented inventory controls to channel card manufacturers. Desktop graphics card manufacturers believe that sell-through trends over the Chinese New Year holiday are likely to determine whether channel card inventory can be reduced at end customers in January, or if this sales correction will continue into February and the remainder of calendar 1Q17 given weaker seasonality,” he said.