[Most Recent Entries] [Calendar View]
Wednesday, January 11th, 2017
| Time | Event | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 7:45a | AMD Set to Launch Ryzen Before March 3rd, Meeting Q1 Target
Thanks to some sleuthing from various readers, AMD has accidentally let the cat out of the bag with regards to the official Ryzen launch date. While they haven’t specifically given an exact date, the 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 optimization topics.
The GDC event runs from February 27th to March 3rd, and currently the AMD talk is not on the exact schedule yet, so it could appear any day during the event (so be wary if anyone says Feb 27th). At this time AMD has not disclosed an exact date either, but it would be an interesting time to announce the new set of Ryzen CPUs right in the middle of both GDC and Mobile World Congress which is also during that week. It would mean that Ryzen news may end up being buried under other GDC and smartphone announcements.
Then again, the launch could easily be anytime during February – this March 3rd date only really puts an end-point on the potential range. AMD has stated many times, as far back as August, that Q1 is the intended date for launch to consumers in volume. When we spoke with AMD at CES, nothing was set in stone so to speak, especially clock speeds and pricing, but we are expecting a full launch, not just something official on paper. Ryan will be at GDC to cover this exact talk, and I’ll be at MWC covering that event. Either way, we want to make sure that we are front of the queue when it comes time to disclosing as much information as we can get our hands on ahead of time. Stay tuned! | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 9:30a | Corsair’s Bulldog 2.0 Gets Kaby Lake-Compatible Z270 Motherboard, New Cooler
Corsair introduced its new Bulldog 2.0 small form-factor HTPC console-like barebones kit at CES. The new Bulldog 2.0 received a new motherboard based on the Intel Z270 PCH with improved features, as well as a new CPU liquid cooling system that is said to be quieter compared to the predecessor. At the same time, the kit retained its visual design and a relatively moderate price point. Corsair’s Bulldog case blends enthusiast-class performance and features with a living room aesthetics, which is a rather rare combination. The Bulldog chassis can accommodate a mini-ITX motherboard, a full-height graphics card (which is not longer than 300 mm and is not thicker than 90 mm), two liquid cooling systems, a 3.5” HDD, up to three 2.5” storage devices, multiple fans as well as an SFX power supply. At CES Corsair demonstrated its new Bulldog 2.0 barebones kit featuring MSI’s Z270I Gaming Pro Carbon AC motherboard, its own new Hydro H6 SF low-profile cooler as well as the SF600 PSU.
Since the Bulldog 2.0 uses almost the same chassis as the initial product, the key differentiators of the new barebones kit is the mainboard and the LCS. The latter is not yet available separately and the company even has not published its specs. The only thing that Corsair says about the H6 SF is that it is quiet even when it has to cool down an overclocked CPU, which is not really a detailed description. As for the Z270I Gaming Pro Carbon AC motherboard, it comes with an LGA1151 socket supporting both Kaby Lake-S and Skylake-S processors, two DIMM slots for up to 32 GB of DDR4 memory, a PCIe 3.0 x4/NVMe M.2-2280 slot for SSDs and a PCIe x16 slot for graphics cards. The motherboard is equipped with the new ASMedia ASM2142 controller (uses PCIe 3.0 x2 interface and thus provides up to 16 Gbps of bandwidth to two USB 3.1 Gen 2 ports) powering USB 3.1 Gen 2 Type-A/C headers, Intel’s dual-band Wireless-AC 8265 module (Wi-Fi 802.11ac + BT 4.2), Intel’s I219-V Gigabit Ethernet controller, a 7.1-channel Realtek ALC1220-based audio sub-system, SATA connectors and so on.
The motherboard looks to be more advanced than the one installed into the first-gen Bulldog as it is based on the latest Intel Z270 PCH, supports Optane Memory caching, a newer audio codec and an improved USB 3.1 (10 Gbps) controller. If the H6 SF LCS is really quieter than the predecessor, then the Bulldog 2.0 has a nice set of improvements over the first version.
The refined Corsair Bulldog 2.0 barebones kit will be available shortly for $399.99, the price point of the first-gen product. In addition, select PC makers and retailers will offer their PCs based on the Bulldog 2.0 that will cost according to their specifications: The higher-end models will use MSI’s liquid-cooled Hydro GFX GTX 1080 graphics cards along with Intel's K-processors, whereas more affordable builds will use something less extreme for an SFF system. Related Reading: | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 11:00a | SK Hynix Announces 8 GB LPDDR4X-4266 DRAM Packages
SK Hynix on Monday officially announced the industry’s first 8 GB LPDDR4X (LP4X) packages for next-generation mobile devices. The new memory chips not only increase DRAM performance but also reduce its power consumption due to lower I/O voltages (and come in a smaller form-factor). Interested parties have already obtained samples of SK Hynix’s LPDDR4X ICs and the first devices featuring the new type of memory are expected to hit the market in the coming months. The LPDDR4X is a new mobile DRAM standard that is an extension of the original LPDDR4, and is expected to reduce power consumption of the DRAM sub-system by 18~20% according to developers (everything else remains the same: a 200~266 MHz internal memory array frequency, 16n prefetch, etc.). To do that, LPDDR4X cuts output driver power (I/O VDDQ voltage) by 45%, from 1.1 V to 0.6 V. LPDDR4X is supported by a number of mobile SoC developers. The first application processor to support the new type of memory is MediaTek’s Helio P20 that was announced nearly a year ago and the initial devices powered by the chip are likely to hit the market in 1H 2017. Another notable SoC to support LPDDR4X is Qualcomm’s new flagship Snapdragon 835, which was announced in November and detailed earlier this month. Smartphones featuring this chip will not show up for a while, but MWC just around the corner which lends nicely to various handset announcements.
The 8 GB (64 Gb) LPDDR4X package stacks four 16 Gb DRAM parts that feature a 4266 MT/s data transfer rate and provide up to 34.1 GB/s of bandwidth when connected to an application processor using a 64-bit memory bus. For its 8 GB LPDDR4X solution SK Hynix uses a new 12 mm × 12.7 mm BGA package, which is 30% smaller compared to standard LPDDR4 stacks that come in 15 mm × 15 mm form-factor. SK Hynix’s 8 GB LPDDR4X solution has a thickness of less than 1 mm to enable PoP stacking with a mobile application processor or a UFS NAND storage device.
SK Hynix did not announce exact power consumption figures for its LP4X parts, but confirmed that the reduction of I/O voltage by 45% reduces power consumption of the whole memory sub-system by around 20% versus a hypothetical LPDDR4 memory sub-system running at the frequency in the same conditions. This is not exactly a good description because SK Hynix’s LPDDR4 offerings top at 3733 MT/s. Assuming that the manufacturer did not optimize the design of its LPDDR4X DRAM arrays to reduce power consumption, but only reduced VDDQ to 0.6 V, a memory sub-system based on the new 8 GB LP4X-4266 part should consume less than a similar sub-system running the company’s 8 GB LP4-3733 stack, but the exact figure is unknown.
To make its 16 Gb LPDDR4X memory ICs, SK Hynix uses its 21 nm fabrication process, which is also used to manufacture 16 Gb LPDDR4 ICs. So, from manufacturing technology standpoint, SK Hynix’s LP4X chips are similar to its LP4 chips. Initially, SK Hynix will offer only 8 GB LPDDR4X packages with 4266 MT/s data transfer rate based on its 16 Gb DRAM ICs. Eventually, the company intends to expand the lineup with 6 GB/8 GB LPDDR4X-3733 (these are already listed in the company's Q1 databook) and LPDDR4X-3200 solutions as well as parts based on 8 Gb LPDDR4X ICs (these are not listed in the official documents, but are mentioned in the company's official blog post). The latter make a lot of sense as far not all mobile are going to use 8 GB of DRAM this year. SK Hynix quotes researchers from IHS Markit, who believe that an average high-end smartphone this year is going to integrate 3.5 GB of memory on average (a mix of 3GB, 4GB, 6GB and 8GB solutions on Android). Meanwhile, keep in mind that DRAM requirements for Apple’s iOS and Google’s Android are different, which is why smartphones running the latter need more memory and handsets featuring 4 GB of Mobile DRAM are going to become mainstream in 2017. By contrast, Apple’s iPhone 7 and iPhone 7 Plus have 2 GB and 3 GB of DRAM, respectively. SK Hynix said that its 8 GB LPDDR4X-4266 packages are already in mass production. Mobile devices based on the new memory are expected to arrive in the coming months and it is highly likely that select manufacturers may demonstrate their MediaTek Helio P20- and LPDDR4X-based products at MWC next month. Related Reading: | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 1:00p | Intel Compute Card: A Universal Compute Form-Factor for Different Kinds of Devices
At CES 2017, Intel introduced a new form-factor for computing platforms in order to enable easy development, configuration, maintenance, repair and upgrade of various devices. Intel’s Compute Card is as small as a credit card, but, packs everything needed for computing, including the CPU, DRAM, storage, communications and I/O. The first cards are set to be introduced in mid-2017. Computing has become so ubiquitous nowadays that almost every more or less sophisticated piece of hardware has a microprocessor inside. Many such devices are designed to operate for years, but, since computer chips inside them get outdated, it is almost impossible to upgrade their functionality (e.g., add new security capabilities, speed up processing, etc.) without replacing the whole unit, or a significant part of it. Alternatively, if a CPU or a memory IC fails, repair of such a device may cost a lot in terms of money and effort. This translates to downtime and lost revenue. Such things happen because of various reasons, but the main two are proprietary platforms with tight integration and no upgrade path, or complex architectures that do not allow for a quick replacement of faulty components. The list of such devices includes everything from business PCs to point-of-sale kiosks and from smart TVs to commercial equipment.
The Intel Compute Card has been designed to be a universal computing platform for different kinds of devices, including those that do not exist yet. The ultimate goal is to simplify the way companies develop equipment, use, maintain, repair, and upgrade it. Creators of actual devices have to design a standard Intel Compute Card slot into their product and then choose an Intel Compute Card that meets their requirements in terms of feature-set and price. For example, PC makers could create systems in all-in-one or clamshell form-factors and then use Compute Cards instead of motherboards. For corporate customers that need to provide a lot of flexibility (and, perhaps, solve some security concerns too) - every employee has a card and can switch between PCs. In other markets such as automated retail kiosks, vendors can easily provide upgrades to deliver better functionality as Intel releases new Compute Cards in the future.
From a technology standpoint, Intel’s Compute Card resembles the company’s Compute Stick PC. However, its purpose is much wider: it is a small device that packs an Intel SoC or SiP (including Kaby Lake-based Core processors with vPro and other technologies), DRAM, NAND flash storage, a wireless module and so on into a small enclosure. Nonetheless, there are a number of important differences between the Compute Card and the Compute Stick. The Compute Card is a sealed system with “flexible I/O” in the form of a USB Type-C and an extension connector. The “flexible I/O” is not Thunderbolt (obviously, due to power consumption concerns), but it handles USB, PCIe, HDMI, DisplayPort connectivity and has some extra pins for future/proprietary use.
Intel is currently working with a number of partners to enable the Compute Cards ecosystem. The list of global players includes Dell, HP, Lenovo and Sharp. There are also regional partners interested in the new form-factor, including Seneca Data, InFocus, DTx, TabletKiosk and Pasuntech.
At the moment, Intel and its partners are not discussing their Compute Card-based projects, which is understandable. Moreover, do not expect all these companies to release their Compute Card hardware simultaneously because different equipment has different design and validation cycles.
Speaking of the form-factor itself, this is by far not the first form-factor the size of a credit card or something close (e.g., Mobile-ITX, Pico-ITX were announced, but never took off). However, this one seems easy to integrate and it is backed by Intel, which gives it credibility. There are a number of applications (and usage scenarios) that could take advantage of the Compute Card right away (e.g., corporate PCs, smart TVs, digital point-of-sales, emerging devices, etc.). However, there are also many embedded applications that just require uninterrupted operation without any need to upgrade. For those, traditional industrial PCs and boards will continue to be the mainstay, and the credit card form-factor will not bring any clear advantages. The other interesting aspect here is the future of the Compute Stick form factor. Given that the ARM-based HDMI sticks are not a popular form factor any more, it is not surprising that Intel has also not decided to update the Compute Stick lineup with Kaby Lake. Intel indicated that they would be evaluating the future of the Compute Stick in 2018, and decide if it warrants an update with the latest processors at that time. Our opinion is that the Compute Stick form factor has reached the end of its life, and it is for the Compute Card to carry on the miniaturization revolution. The Compute Card has much more flexibility in terms of the differentiation from the vendors' side, and it is not encumbered by an active cooling mechanism. Obviously, the ability to just plug the device into a HDMI port is not there, but, the Compute Card, by itself, is light enough to just hang directly off a display's HDMI port. Therefore, it is possible that some vendors can deliver a Compute Stick-like platform with the Compute Card also (albeit, with a slightly different form factor). We expect to hear more about Compute Card related projects in Q3, either during Computex or Intel's Developer Forum. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3:00p | GIGABYTE Exhibits an Aquantia AQC107 based 10G Ethernet PCIe Card
During December, Aquantia announced that it will be launching two multi-gigabit NICs into the market, offering 2.5G/5G capability on both and one with 10GBase-T as well. We were told that industry partners would announce solutions with the chips in due course, and on Kaby Lake launch day we saw ASRock announce a pair of motherboards to come with the chips. GIGABYTE is also joining the fray, with a PCIe card to be potentially bundled with future motherboards or offered as a standalone product. The GIGABYTE solution is a PCIe 3.0 card featuring a single 10G port, which allows for half-height and full-height PCIe slots. Only an AQC107 version was at the show for 10G, and it wasn’t clear if a 2.5G/5G version using the AQC108 would be inbound, but at this point in time GIGABYTE is keeping its cards close to its chest.
Aside from showing it exists, not much else was given – if it will be sold standalone, or what the extra price will be. The interesting thing for us to determine is the BOM cost (bill of materials) for the Aquantia chips – Aquantia has mentioned that they want to undercut 10G solutions significantly, and help drive multi-gigabit ethernet to both the PC and the backhaul of a home or business network. Having PCIe cards you can slot in certainly helps, and we mentioned to GIGABYTE that if this card hits the market in the $80-$100 range for a single port, that would help (and any cheaper would mean it will fly off the shelves).
A big question with multi-gigabit ethernet, especially 2.5G/5G, is the availability of consumer-grade switches and hubs. We might have to wait another 12-18 months for those to come through, and again, pricing is a concern here. Aquantia has said they are working with the major players in that space, but it will be up to them to announce products. I’ve told GIGABYTE that when the cards are available, I will take a few for testing. I’m slowly building up a sizable stack of 10GBase-T controllers, and we might start looking into relevant networking tests for them for comparisons. Any suggestions, please let us know. Related Reading: | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 5:00p | CES 2017: GIGABYTE Shows Passive Apollo Lake BRIX in Embedded UCFF
Despite the candid approach that Intel has had to its Apollo Lake platform, we are now starting to see fully developed platforms coming out to play from the major vendors. (I personally have some APL motherboards for desktops in for testing sometime during Q1 as well.) One of these platforms was GIGABYTE’s embedded BRIX line, more focused at commercial deployments than consumer, but interesting nonetheless.
The main unit we saw was this: the GB-EAPD-4200 BRIX. Using an Apollo Lake SoC, it looks to be in a passive case and sports three WiFi antenna. The chassis is 0.46L, with support for 2x DDR3L-1866 SO-DIMM memory modules, one M.2 (2280) SSD slot and a mini-PCIe slot for a 3G WLAN module. The WLAN accounts for one antenna, and the other two come via an Intel 802.11ac card inside, most likely the 8265 or 8260 depending on how GBT’s design cycle was in play. Using Apollo Lake means getting HD Graphics 505, although given this system it is more than adequate for video playback or casual office use.
This unit has plenty of IO to keep the buyer satisfied – dual gigabit network ports, dual HDMI 1.4 display ports, four USB 3.0 ports and support for a microSD card. VESA support for 75mm and 100mm brackets is included.\
This unit is to be sold as a barebones kit, requiring DRAM and storage, most likely for system integrators to add their own hardware solutions depending on how local customers will want them configured (small DRAM counts and low storage, or something more robust). Interested parties will need to contact their local distribution partners for more information.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| << Previous Day |
2017/01/11 [Calendar] |
Next Day >> |