It’s been roughly 7 months since AMD released the Crimson ReLive Edition update for Radeon Software, the latest entry in their annual cadence for major driver revisions and feature additions. Today’s launch sees AMD/RTG bring the sequentia…
This afternoon, AMD announced their second quarter results for their fiscal year 2017, and the news is promising. AMD still has some work to do in order to get back to profitability, but that work has been helped tremendously by successful product launches earlier this year. Ryzen has shown exciting potential, and a diverse and strong product lineup has helped AMD’s bottom line. For the second quarter, AMD’s revenue was up 19% year-over-year to $1.22 billion, and operating income was $25 million for the quarter. Net income was still in the red with a loss of $16 million, resulting in a loss per share of $0.02 on a GAAP basis. Gross margin was 33%, hovering right around that 35% range that AMD wants to hit for profitability.
| AMD Q2 2017 Financial Results (GAAP) | |||||
| Q2’2017 | Q1’2017 | Q2’2016 | |||
| Revenue | $1220M | $984M | $1030M | ||
| Gross Margin | 33% | 34% | 31% | ||
| Operating Income | +$25M | -$29M | -$8M | ||
| Net Income | -$16M | -$73M | +$69M | ||
| Earnings Per Share | -$0.02 | -$0.08 | +$0.08 | ||
AMD also releases Non-GAAP results which exclude results such as restructuring charges, debt fees, and stock based compensation. Sometimes Non-GAAP results can help you look at an underlying business when there is restructuring charges affecting results either positively or negatively, but in this quarter for AMD, the Non-GAAP results are almost exclusively the result of not factoring in stock-based compensation which amounted to $24 million. On a Non-GAAP basis, revenue for the quarter was the same $1.22 billion, but operating income is now $49 million, compared to just $3 million a year ago. Net income was $19 million, and earnings-per-share results in $0.02.
| AMD Q1 2017 Financial Results (Non-GAAP) | |||||
| Q2’2017 | Q1’2017 | Q2’2016 | |||
| Revenue | $1220M | $984M | $1030M | ||
| Gross Margin | 33% | 34% | 31% | ||
| Operating Income | +$49M | -$6M | +$3M | ||
| Net Income | +$19M | -$38M | -$40M | ||
| Earnings Per Share | +$0.02 | -$0.04 | -$0.05 | ||
The year-over-year results may seem a bit skewed, since Q2 2016 was actually a profitable quarter for AMD, but that was due to a $150 million infusion of cash from a joint-venture with Nantong Fujitsu Microelectronics. This quarter doesn’t have any large cash deals involved, and AMD is very close to breaking even, with strong gains across its product line.
The star of the show is undoubtedly Ryzen, and the Computing and Graphics segment had a very strong quarter, with revenues of $659 million, up 51% compared to Q2 last year. AMD attributes this jump to demand for graphics and Ryzen desktop processors. Operating income for the Computing and Graphics group was $7 million, compared to an $81 million loss last year, and much of that was driven due to higher average selling prices for its processors. Although AMD is not yet able to charge the premium of Intel, it can at least charge a lot more than it did for the last generation of CPUs.
| AMD Q2 2017 Computing and Graphics | |||||
| Q2’2017 | Q1’2017 | Q2’2016 | |||
| Revenue | $659M | $593M | $435M | ||
| Operating Income | +$7M | -$15M | -$81M | ||
Enterprise, Embedded, and Semi-Custom had a 5% drop in revenue, to $563 million, mostly due to a softening in semi-custom SoC sales. This segment is where AMD’s EPYC CPU line will impact though, so the next couple of quarters should be interesting to see here, with the launch of the Xbox One X, and EPYC.
| AMD Q2 2017 Enterprise, Embedded, and Semi-Custom | |||||
| Q2’2017 | Q1’2017 | Q2’2016 | |||
| Revenue | $563M | $391M | $592M | ||
| Operating Income | $42M | $9M | $84M | ||
All Other had an operating loss of $24 million, compared with a loss of $11 million in Q2 2016, with this primarily being stock-based compensation, as well as a $7 million restructuring credit in Q2 2016 helping out that quarter.
AMD has a lot to be excited about, and they’ve delivered a strong product in Ryzen already, which will branch out to enterprise with EPYC where the higher margins are. On the GPU side, Vega has launched as well with workstation graphics cards available now. Add in the custom SoC market that they’ve worked hard to establish, and the future seems just a little bit brighter than before. For Q3, AMD is expecting a 23% increase in revenue compared to this quarter, plus or minus 3%.
Source: AMD Investor Relations
The month the Bluetooth SIG has taken the wraps off of their latest standards project: an addition to the Bluetooth specification that enables creation of large networks of devices. Dubbed “Bluetooth Mesh”, the new standard is designed for smart homes, public and manufacturing facilities. An extension of the Bluetooth LE protocol, the Bluetooth SIG hopes that the first products supporting the Bluetooth Mesh specification will be quickly available in the coming months as the new technology does not require principally new hardware.
The Bluetooth technology was originally developed in the late 1990s to enable wireless device-to-device communications. This device-to-device tech has been evolving since 1997 by improving transfer rates, extending range, and improving reliability. In the early 2000s, researchers and product developers determined that there were many devices that could benefit from short-burst wireless connectivity, but did not need a fully-fledged Bluetooth implementation due to power consumption and size concerns. To this end, developers from Nokia and other companies started to design a low-power version of Bluetooth that was first marketed under the Wibree trademark in 2006 and then became a part of the Bluetooth 4.0 spec under the Bluetooth Smart (Bluetooth LE) trademark.
Bluetooth LE introduced a one-to-many communication paradigm to the standard, enabling various new usage models and applications, such as item finding beacons or way finding beacons. Meanwhile, device-to-device and device-to-many-devices Bluetooth interconnections ultimately use a star topology, and thus such networks have limits to their range and the number of devices in a network, which can inhibit their usage models.
With the introduction of the Bluetooth Mesh standard, the Bluetooth SIG is bringing a many-to-many communication paradigm to the standard. As the name implies, Bluetooth Mesh enables building large-scale device networks with a mesh topology and thus extends the range of a single Bluetooth network virtually to infinity (mind latency and other factors though). In turn, such large networks open up new usage models for the technology, particularly in the field of IoT devices.
At a high level, Bluetooth Mesh uses the Bluetooth 4.0 LE protocol to transport data between devices (nodes). This means that from a data transmission point of view, nearly everything has already been specified, from low layer radios to encryption to application layers.
What Bluetooth Mesh adds on top of Bluetooth LE is an ability to retransmit data (a message) from one device to other devices that are in direct radio range (a single hop away) until it reaches the destination — the address it is sent to — using the so-called managed flooding technique. Managed flooding allows the receiver to determine how many hops it’s away from the sender, and thus disable message relaying further than it is needed (i.e., set the maximum number of hops over which the message is retransmitted), thus preserving power and saving the bandwidth of the whole network. This technique ensures that the message always reaches its destination potentially using various paths even if certain nodes fail. Moreover, by not using routing devices, Bluetooth Mesh networks become cheaper and more reliable — messages always get to their destinations no matter what happens to individual nodes within a network (so long as there’s a path).
Under the hood, Bluetooth Mesh networks consist of relay nodes that can retransmit messages, and low-power nodes that connect to relay nodes to periodically transmit or receive data using a mechanism called friendship (i.e., each LP node has a Friend). Low-power nodes could be various sensors or beacons that only use short-burst data transmissions with their “Friends”, which then retransmit their data to other nodes.
The Bluetooth SIG says that a single Bluetooth Mesh network can contain up to 32,767 elements (a node has at least one element, or addressable entity), but admits that in the real world such networks will typically consist of at most thousands, rather than tens of thousands of devices.
When it comes to bandwidth, the gross air data rate of the Bluetooth LE is 1 Mb/s (at maximum transmit power of 10 mW), but this figure does not account for protocol overhead. There are reports that the maximum achievable BLE data rate is around 10 KB/s depending on the devices used. Such data rates are not a problem for point-to-point communications within the contemporary star network topology. However what happens when hundreds of devices start passing numerous messages over the mesh network remains to be seen. Bandwidth, latency (6 ms per hop – the standard for the Bluetooth LE) and lack of priority attributes for data packets could be limiting factors for Bluetooth Mesh adoption in environments that generate loads of data.
It is important to note that the Bluetooth SIG made interoperability a part of the specification development process, and therefore thousands of interoperability tests have already been conducted. This is not exactly surprising as the Bluetooth Mesh spec builds up Bluetooth LE, and therefore existing Bluetooth 4/5-capable chips can support the new technology. Meanwhile, actual contemporary devices may or may not receive firmware and software upgrades to enable Bluetooth Mesh support. Some device makers interested in addressing smart home and other applications are more likely to enable the new spec on existing products, but others are more likely to qualify future products for the new tech.
At present, the Bluetooth SIG and its members are pursuing several key applications for Bluetooth Mesh: smart home, lighting, beaconing, automation, and asset tracking applications. For example, integrating relay-capable nodes into lighting devices across a home extends range of the network to the whole building. Each of the multiple relay nodes installed in every room can then connect to various low-power nodes, such as temperature sensors, thermostats, window blind controls, and so on. In a warehouse, a robot could navigate its way across the building without any network range-related constraints while enabling operators to track its whereabouts.
The Bluetooth SIG is certainly not alone with its mesh-networking standard for IoT applications, both in general and smart homes in particular. There is ZigBee that is used for various smart home appliances already, there are proprietary technologies, and Wi-Fi HaLow is incoming. A natural advantage that the Bluetooth SIG and its members have is that there are hundreds of millions of Bluetooth-enabled devices produced and sold every year, and therefore the majority of upcoming smartphones, smart TVs, notebooks, tablets, and other products will have the ability to be compatible with the Bluetooth Mesh specification. As a result, it will make a lot of sense for developers of smart home appliances to design devices compatible with Bluetooth Mesh — millions of consumers will have compatible devices in about a year from now, and this is a huge number for an emerging market. Meanwhile, when it comes to custom automation or industrial applications, it remains to be seen which technology developers prefer for their large-scale device networks.
Nowadays there are billions of Bluetooth LE-enabled devices, and in the coming years their number will grow further not only because people will buy more smartphones, but because a lot of brand new device types will emerge. The Bluetooth Mesh specification enables building large-scale device networks without the need to launch any new hardware, and thus makers of smart home appliances (and other devices) may begin quickly adopting the new tech in the coming months.
Relying on the Bluetooth LE specification and the managed flooding message transport technique ensures that Bluetooth Mesh-based large-scaled device networks will have predictable performance (RF interference withstanding), known security mechanisms, high reliability, and relatively low costs. Moreover, to a degree the usage Bluetooth LE eliminates the chicken and egg dilemma for the new standard early in its life, as supporting devices are already here. However, it is not completely clear how the number of nodes per single network affects its performance, and that will be a very important factor once large networks consisting of hundreds or thousands of nodes are built.
Keeping in mind that Bluetooth connectivity is ubiquitous nowadays, Bluetooth Mesh has a good chance to become a very popular wireless standard for smart homes and other applications, provided that it can deliver the right performance and ensure compatibility and interoperability between devices from different vendors. The Bluetooth SIG says that interoperability is nearly guaranteed for Bluetooth Mesh supporting devices, but we will have to see how that pans out ourselves.
Otherwise, while developers of smart devices can start building Bluetooth Mesh products based on silicon solutions designed by others, they will still have to complete the long-standing Bluetooth qualification and/or declaration processes in order to ensure that their products satisfy the Bluetooth license the requirements and to pay the appropriate fees. This is not going to be a problem for the established players, but may pose a small challenge for small startups.
Finally, the Bluetooth SIG is naming a number of semiconductor manufacturers, software developers, and device manufacturers as among Bluetooth Mesh’s early adopters. This include 3M, ARM, Ericsson, STMicroelectronics, Qualcomm, Toshiba, and others. While the organization does not announce any final products, it makes it clear that Bluetooth Mesh is has manufacturer support at both the silicon and on-device levels. Moreover, there are turnkey silicon and software solutions available from companies like Cypress, Silicon Labs, and Wisilica to build devices compatible with Bluetooth Mesh. Ultimately, the sky looks blue for the Bluetooth Mesh to take off, but its actual market acceptance will depend on the adoption of IoT devices for homes, offices, and production facilities.
Western Digital’s SanDisk subsidiary and Toshiba have a long history of jointly developing and manufacturing NAND flash memory. While that relationship has been strained by Toshiba’s recent financial troubles and attempts to sell of their share of the memory business, the companies are continuing to develop new flash memory technology and are still taking turns making new announcements. In recent months both companies have started sampling SSDs using their 64-layer BiCS3 TLC 3D NAND and have announced that their next generation BiCS4 3D NAND will be a 96-layer design.
Yesterday Western Digital made a small announcement about their other main strategy for increasing density: storing more bits per memory cell. Western Digital will introduce four bit per cell QLC parts built on their 64-layer BiCS3 process, with a capacity of 768Gb (96GB) per die. This is a substantial increase over the 512Gb BiCS3 TLC parts that will be hitting the market soon, and represents not only an increase in in bits stored per memory cell but an increase in the overall size of the memory array. These new 3D QLC NAND parts are clearly intended to offer the best price per GB that Western Digital can manage, but Western Digital claims performance will still be close to that of their 3D TLC NAND. Western Digital’s announcement did not mention write endurance, but Toshiba’s earlier announcement of 3D QLC NAND claimed endurance of 1000 program/erase cycles, far higher than industry expectations of 100-150 P/E cycles for 3D QLC and comparable to 3D TLC NAND.
Western Digital has not announced any specific products based on QLC NAND flash, but they will be exhibiting both removable media and SSDs using QLC NAND at Flash Memory Summit August 8-10. Western Digital’s CTO will be delivering a keynote presentation at FMS on August 8, so more details are likely to be revealed in two weeks.
Western Digital’s roadmaps also include plans for QLC parts on their 96-layer BiCS4 process, with capacities up to 1Tb (128GB) per die. BiCS4 production is scheduled to ramp up over 2018 and 2019 with the QLC parts expected to arrive later in the cycle, so Western Digital’s first-generation 3D QLC based on the BiCS3 process will probably be their highest-density flash memory in mass production for over a year.
Western Digital’s SanDisk subsidiary and Toshiba have a long history of jointly developing and manufacturing NAND flash memory. While that relationship has been strained by Toshiba’s recent financial troubles and attempts to sell of their share of the memory business, the companies are continuing to develop new flash memory technology and are still taking turns making new announcements. In recent months both companies have started sampling SSDs using their 64-layer BiCS3 TLC 3D NAND and have announced that their next generation BiCS4 3D NAND will be a 96-layer design.
Yesterday Western Digital made a small announcement about their other main strategy for increasing density: storing more bits per memory cell. Western Digital will introduce four bit per cell QLC parts built on their 64-layer BiCS3 process, with a capacity of 768Gb (96GB) per die. This is a substantial increase over the 512Gb BiCS3 TLC parts that will be hitting the market soon, and represents not only an increase in in bits stored per memory cell but an increase in the overall size of the memory array. These new 3D QLC NAND parts are clearly intended to offer the best price per GB that Western Digital can manage, but Western Digital claims performance will still be close to that of their 3D TLC NAND. Western Digital’s announcement did not mention write endurance, but Toshiba’s earlier announcement of 3D QLC NAND claimed endurance of 1000 program/erase cycles, far higher than industry expectations of 100-150 P/E cycles for 3D QLC and comparable to 3D TLC NAND.
Western Digital has not announced any specific products based on QLC NAND flash, but they will be exhibiting both removable media and SSDs using QLC NAND at Flash Memory Summit August 8-10. Western Digital’s CTO will be delivering a keynote presentation at FMS on August 8, so more details are likely to be revealed in two weeks.
Western Digital’s roadmaps also include plans for QLC parts on their 96-layer BiCS4 process, with capacities up to 1Tb (128GB) per die. BiCS4 production is scheduled to ramp up over 2018 and 2019 with the QLC parts expected to arrive later in the cycle, so Western Digital’s first-generation 3D QLC based on the BiCS3 process will probably be their highest-density flash memory in mass production for over a year.