ARM To Focus On 64-bit SoC
ARM announced its first 64-bit cores a while ago and SoC makers have already rolled out several 64-bit designs. However, apart from Apple nobody has consumer oriented 64-bit ARM devices on the market just yet. They are slowly starting to show up and ARM says the transition to 64-bit parts is accelerating. However, the first wave of 64-bit ARM parts is not going after the high-end market.
Is 64-bit support on entry-level SoCs just a gimmick?
This trend raises a rather obvious question – are low end ARMv8 parts just a marketing gimmick, or do they really offer a significant performance gain? There is no straight answer at this point. It will depend on Google and chipmakers themselves, as well as phonemakers.
Qualcomm announced its first 64-bit part late last year. The Snapdragon 410 won’t turn many heads. It is going after $150 phones and it is based on Cortex A53 cores. It also has LTE, which makes it rather interesting.
MediaTek is taking a similar approach. Its quad-core MT6732 and octa-core MT6752 parts are Cortex A53 designs, too. Both sport LTE connectivity.
Qualcomm and MediaTek appear to be going after the same market – $100 to $150 phones with LTE and quad-core 64-bit stickers on the box. Marketers should like the idea, as they’re getting a few good buzzwords for entry-level gear.
However, we still don’t know much about their real-world performance. Don’t expect anything spectacular. The Cortex A53 is basically the 64-bit successor to the frugal Cortex A7. The A53 has a bit more cache, 40-bit physical addresses and it ends up a bit faster than the A7, but not by much. ARM says the A7 delivers 1.9DMIPS/MHz per core, while the A53 churns out 2.3DMIPS/MHz. That puts it in the ballpark of the good old Cortex A9. The first consumer oriented quad-core Cortex A9 part was Nvidia’s Tegra 3, so in theory a Cortex A53 quad-core could be as fast as a Tegra 3 clock-for-clock, but at 28nm we should see somewhat higher clocks, along with better graphics.
That’s not bad for $100 to $150 devices. LTE support is just the icing on the cake. Keep in mind that the Cortex A7 is ARM’s most efficient 32-bit core, hence we expect nothing less from the Cortex A53.
The Cortex A57 conundrum
Speaking to CNET’s Brooke Crothers, ARM executive vice president of corporate strategy Tom Lantzsch said the company was surprised by strong demand for 64-bit designs.
“Certainly, we’ve had big uptick in demand for mobile 64-bit products. We’ve seen this with our [Cortex] A53, a high-performance 64-bit mobile processor,” Lantzch told CNET.
He said ARM has been surprised by the pace of 64-bit adoption, with mobile parts coming from Qualcomm, MediaTek and Marvell. He said he hopes to see 64-bit phones by Christmas, although we suspect the first entry-level products will appear much sooner.
Lantzsch points out that even 32-bit code will run more efficiently on 64-bit ARMv8 parts. As software support improves, the performance gains will become more evident.
But where does this leave the Cortex A57? It is supposed to replace the Cortex A15, which had a few teething problems. Like the A15 it is a relatively big core. The A15 was simply too big and impractical on the 32nm node. On 28nm it’s better, but not perfect. It is still a huge core and its market success has been limited.
As a result, it’s highly unlikely that we will see any 28nm Cortex A57 parts. Qualcomm’s upcoming Snapdragon 810 is the first consumer oriented A57 SoC. It is a 20nm design and it is coming later this year, just in time for Christmas as ARM puts it. However, although the Snapdragon 810 will be ready by the end of the year, the first phones based on the new chip are expected to ship in early 2015.
While we will be able to buy 64-bit Android (and possibly Windows Phone) devices before Christmas, most if not all of them will be based on the A53. That’s not necessarily a bad thing. Consumers won’t have to spend $500 to get a 64-bit ARM device, so the user base could start growing long before high-end parts start shipping, thus forcing developers and Google to speed up 64-bit development.
If rumors are to be believed, Google is doing just that and it is not shying away from small 64-bit cores. The search giant is reportedly developing a $100 Nexus phone for emerging markets. It is said to be based on MediaTek’s MT6732 clocked at 1.5GHz. Sounds interesting, provided the rumour turns out to be true.
Can AMD Grow
AMD posted some rather encouraging Q1 numbers last night, but slow PC sales are still hurting the company, along with the rest of the sector.
When asked about the PC market slump, AMD CEO Rory Read confirmed that the PC market was down sequentially 7 percent. This was a bit better than the company predicted, as the original forecast was that the PC market would decline 7 to 10 percent.
Rory pointed out that AMD can grow in the PC market as there is a lot of ground that can be taken from the competition. The commercial market did better than expected and Rory claims that AMD’s diversification strategy is taking off. AMD is trying to win market share in desktop and commercial segments, hence AMD sees an opportunity to grown PC revenue in the coming quarters. Rory also expects that tablets will continue to cannibalize the PC market. This is not going to change soon.
Kaveri and Kabini will definitely help this effort as both are solid parts priced quite aggressively. Kabini is also available in AMD’s new AM1 platform and we believe it is an interesting concept with plenty of mass market potential. Desktop and Notebook ASPs are flat which is something that the financial community really appreciated. It would not be so unusual that average selling prices were down since the global PC market was down.
Kaveri did well in the desktop high-end market in Q1 2014 and there will be some interesting announcements in the mobile market in Q2 2014 and beyond.
Can DirectX-12 Give Mobile A Boot?
Microsoft announced DirectX 12 just a few days ago and for the first time Redmond’s API is relevant beyond the PC space. Some DirectX 12 tech will end up in phones and of course Windows tablets.
Qualcomm likes the idea, along with Nvidia. Qualcomm published an blog post on the potential impact of DirectX 12 on the mobile industry and the takeaway is very positive indeed.
DirectX 12 equals less overhead, more battery life
Qualcomm says it has worked closely with Microsoft to optimise “Windows mobile operating systems” and make the most of Adreno graphics. The chipmaker points out that current Snapdragon chipsets already support DirectX 9.3 and DirectX 11. However, the transition to DirectX 12 will make a huge difference.
“DirectX 12 will turbocharge gaming on Snapdragon enabled devices in many ways. Just a few years ago, our Snapdragon processors featured one CPU core, now most Snapdragon processors offer four. The new libraries and API’s in DirectX 12 make more efficient use of these multiple cores to deliver better performance,” Qualcomm said.
DirectX 12 will also allow the GPU to be used more efficiently, delivering superior performance per watt.
“That means games will look better and deliver longer gameplay longer on a single charge,” Qualcomm’s gaming and graphics director Jim Merrick added.
What about eye candy?
Any improvement in efficiency also tends to have a positive effect on overall quality. Developers can get more out of existing hardware, they will have more resources at their disposal, simple as that.
Qualcomm also points out that DirectX 12 is also the first version to launch on Microsoft’s mobile operating systems at the same time as its desktop and console counterparts.
The company believes this emphasizes the growing shift and consumer demand for mobile gaming. However, it will also make it easier to port desktop and console games to mobile platforms.
Of course, this does not mean that we’ll be able to play Titanfall on a Nokia Lumia, or that similarly demanding titles can be ported. However, it will speed up development and allow developers and publishers to recycle resources used in console and PC games. Since Windows Phone isn’t exactly the biggest mobile platform out there, this might be very helpful and it might attract more developers.
AMD, Intel & nVidia Go OpenGL
AMD, Intel and Nvidia teamed up to tout the advantages of the OpenGL multi-platform application programming interface (API) at this year’s Game Developers Conference (GDC).
Sharing a stage at the event in San Francisco, the three major chip designers explained how, with a little tuning, OpenGL can offer developers between seven and 15 times better performance as opposed to the more widely recognised increases of 1.3 times.
AMD manager of software development Graham Sellers, Intel graphics software engineer Tim Foley and Nvidia OpenGL engineer Cass Everitt and senior software engineer John McDonald presented their OpenGL techniques on real-world devices to demonstrate how these techniques are suitable for use across multiple platforms.
During the presentation, Intel’s Foley talked up three techniques that can help OpenGL increase performance and reduce driver overhead: persistent-mapped buffers for faster streaming of dynamic geometry, integrating Multidrawindirect (MDI) for faster submission of many draw calls, and packing 2D textures into arrays, so texture changes no longer break batches.
They also mentioned during their presentation that with proper implementations of these high-level OpenGL techniques, driver overhead could be reduced to almost zero. This is something that Nvidia’s software engineers have already claimed is impossible with Direct3D and only possible with OpenGL (see video below).
Nvidia’s VP of game content and technology, Ashu Rege, blogged his account of the GDC joint session on the Nvidia blog.
“The techniques presented apply to all major vendors and are suitable for use across multiple platforms,” Rege wrote.
“OpenGL can cut through the driver overhead that has been a frustrating reality for game developers since the beginning of the PC game industry. On desktop systems, driver overhead can decrease frame rate. On mobile devices, however, driver overhead is even more insidious, robbing both battery life and frame rate.”
The slides from the talk, entitled Approaching Zero Driver Overhead, are embedded below.
At the Game Developers Conference (GDC), Microsoft also unveiled the latest version of its graphics API, Directx 12, with Direct3D 12 for more efficient gaming.
Showing off the new Directx 12 API during a demo of Xbox One racing game Forza 5 running on a PC with an Nvidia Geforce Titan Black graphics card, Microsoft said Directx 12 gives applications the ability to directly manage resources to perform synchronisation. As a result, developers of advanced applications can control the GPU to develop games that run more efficiently.
What Do Smaller Controllers Mean?
If you want a wearable Internet of Things, the electronics have to be as tiny and as energy efficient as possible. That’s why a new microcontroller by Freescale Semiconductor is noteworthy.
The company has produced the Kinetis KLO3 MCU, a 32-bit ARM system that is 15% smaller than its previous iteration but with a 10% power improvement.
Internet of Things is a buzzword for the trend toward network-connected sensors incorporated into devices that in the past were standalone appliances. These devices use sensors to capture things like temperatures in thermostats, pressure, accelerometers, gyroscopes and other types of MEMS sensors. A microcontroller unit gives intelligence and limited computational capability to these devices, but is not a general purpose processor. One of the roles of the microcontroller is to connect the data with more sophisticated computational power.
The Kinetis KLO3 runs a lightweight embedded operating system to connect the data to other devices, such as an app that uses a more general purpose processor.
Kathleen Jachimiak, product launch manager at Freescale, said the new microcontroller will “enable further miniaturization” in connected devices. This MCU is capable of having up to 32 KB of flash memory and 2 KB of RAM.
Consumers want devices that are light, small and smart. They also want to be able to store their information and send it to an application that’s either on a phone or a PC, Jachimiak said.
This microcontroller, at 1.6 x 2.0 mm, is smaller than the dimple on a golf ball, and uses a relatively new process in its manufacturing, called wafer level chip scale packaging. The process involves building the integrated package while the die is still part of a wafer. It’s a more efficient process and produces the smallest possible package, for a given die size.
Did Intel Kill Bay Trail?
Intel has decided that some of its budget Bay Trail parts have been out evolved and flung them into a tar pit. According to CPU World the parts first appeared in September. Intel released budget Bay Trail systems on a chip for mobile and desktop markets, under Celeron and Pentium brands.
They were manufactured on 22nm technology, and featured such enhancements as greater number of CPU cores, higher clock speeds, beefed up graphics unit, not to mention an out-of-order microarchitecture, that improved per-clock CPU performance by up to 30 per cent faster compared to their predecessors. With this performance goodness it is a little surprising the Intel has decided that all the all Bay Trail SoCs will be discontinued in a matter of a few months. Details of the planned discontinuation were published this week by Intel in several Product Change Notification documents.
The Desktop Pentium J2850, along with mobile Celeron N2810 and Pentium N3510 are already End of Lifed and its last orders will be in two weeks, on February 11. The chips will ship until April 25, 2014. Also retired are mobile Celeron N2806, N2815, N2820, N2920, and Pentium N3520. Their EOL date is April 11, 2014, and they will ship until May 30, 2014. On August 22, 2014, Intel is going to discontinue Celeron J1750, J1850, N2805 and N2910. The “J” models are desktop processors, and the “N” are mobile ones. There is no word on Z-series Bay Trail-T parts, none appear to be EOL’d at this time.
Furthermore, on the same date Intel will retire Core i7-3940XM Extreme Edition, and boxed and tray versions of Core i7-3840QM and i7-3740QM CPUs. The last shipment date for the Celerons and Core i7s is February 6, 2015.
AMD Changes Kaveri
Since AMD officially launched its 4th generation A-Series Kaveri APUs and lifted the NDA veil from all press materials, we noticed that it has started to use a new term to define the structure of its new Kaveri APUs. As we reported last week, AMD is now talking about Compute Cores, which practically puts CPU and GPU cores on an equal footing, suggesting that there should not be any difference between them and that some tasks, previously limited to the CPU, can be done by the GPU as well.
If you take a look at the official AMD slide below which details the three new Kaveri APUs, the A10-7850K, A10-7700K and the A8-7600, you will notice that AMD lists the flagship as the APU with 12 Compute Cores or simply four CPU and eight GPU cores. Since the Kaveri APU is actually the first APU with HSA (Heterogeneous System Architecture) support, with hUMA, or equal memory access by both CPU and the GPU, heterogeneous queuing, which allows the GPU and CPU to have equal flexibility to create/dispatch work and an ability to talk about APU GFLOPS, or combined compute power of the entire APU, it makes sense for AMD to also talk about Compute Cores.
Of course, there are still some application specific tasks where the CPU or the GPU are much better, but, according to AMD, Kaveri is the first true APU, where the GPU is not just for gaming, it can actually do much more.
AMD Senior Manager Sasa Marinkovic, Technology lead for the Client Business Unit, said: “At AMD, we recognize that our customers often think of processors (CPUs) and graphics cards (GPUs) in terms of the number of cores that each product has. We have established a definition of the term “Compute Core” so that we are taking a consistent and transparent approach to describing the number of cores in our HSA-enabled APUs. A Compute Core can be either a CPU core or GPU core i.e. Kaveri can have up to 12 Compute Cores (4 CPU and 8 GPU).”
Although it does sound like a marketing gimmick, but actually is not due to HSA, it will definitely mark a new way for AMD to market/sell its APUs and it will definitely simplify the shopping experience for many casual buyers, more Compute Cores, more performance.
Did Intel Have an IPad In Y2K?
Intel apparently built an IPAD ten years before Steve Jobs though of the tablet and the name. It was in the days when sticking an I in front of anything meant it was Intel rather than Apple and the Intel Pad, or IPAD for short, could browse the Internet, play music and videos, and even act as a digital picture frame.
Intel scrapped the IPAD before consumers could get their hands on it as its move into Tablets was seen as one of the outfit’s biggest blunders. According to CNET in the late 1990s and early 2000s, Intel wanted to diversify its operations beyond the PC. The IPAD came from one of several small teams within its research arm tasked with exploring new business opportunities. The IPAD, which included a touch screen and stylus, would not run entirely on its own but connected to a computer to browse the Internet through an Intel wireless technology.
Intel thought that “mobility” meant moving around your home or business and the IPAD was to be a portable device you could take around your house. The reason that they never thought of connecting it to the phone network was because Intel wanted to tie it all back to its core PC chip business. After several years of development on the Intel Web Tablet, then-CEO Craig Barrett unveiled the device at the Consumer Electronics Show in January 2001. The company planned to sell the tablet to consumers later that year.
Sadly though it miffed Intel’s PC partners, which didn’t want a product that could potentially compete with them and Intel caved in and cancelled the project.
Intel’s Cherry Trail Forthcoming
The next generation desktop and mobile Atom is Cherry Trail in 14nm and the first parts are expected in late 2014. Intel has been working hard to accelerate the introduction of Atom parts based on the new architecture and in 2014 it will finally ship Broadwell notebook chips and Cherry Trail Atoms in the same year, both using the new 14nm node.
The Cherry View is a notebook SoC version of a chip based on new Airmont core, while Cherry Trail is the part meant for tablets. The phone version is based on Moorefield architecture and they are all expected to show up in late 2014, most likely very late Q3 2014.
The TDP should go down compared to Bay Trail platform as the new 14nm needs less voltage to hit the same speed and should produce less heat at the same time. With the 14nm shrink Intel’s new Atoms will be able to get more fanless design wins.
The significance of 14nm products for mobile phones and tablets will be in the fact that ARM alliance lead by Qualcomm, Samsung, MediaTek and a few other players will be struggling to get 20nm designs out of the door in 2014, and Intel can already get to a 14nm.
However, Intel still has to integrate LTE inside its mobile phone SoCs, which has traditionally been proven to be a tough task. At this time only Qualcomm has on-die LTE and its LTE enabled SoCs are under the bonnet of almost every significant ARM based high-end phone out there.
Only time will tell how successful Intel’s mobile push will be. Even with these 14nm parts, once they show up roughly a year from now, it might be really tough for Intel to get some high-volume design wins in the phone space, despite the transition to 14nm.
Will Intel’s Tablet Gamble Work?
Earlier this year Intel caused quite a stir when CEO Brian Krzanich announced the company’s ultimate goal is to make $99 tablets a reality. So far Intel has failed to gain much market share in the tablet space, dominated by cheaper ARM application processors and Android.
However, Bay Trail-T has a good chance to turn things around. The new chip can easily take on high-end ARM parts and in most cases, wipe the floor with them. Since it’s an x86 part, it can also be used in Windows 8.1 tablets. However, the price was a problem. Intel’s official Bay Trail-T prices range from $32 to $37, making the chips significantly more expensive than mid-range and low-end ARM parts. However, many vendors are said to be getting discounts and paying a bit less, in the $20 to $30 range.
Things may be about to change. According to Digitimes, Intel is planning to spend up to $1 billion on tablet chip subsides. The cash should sweeten the deal for vendors willing to give Intel SoCs a go. Since we are talking about relatively low average selling prices, Intel could use the cash to practically halve the prices and offer Bay Trail-T parts for as little as $10. This would make them competitive overnight, as high-end ARM SoCs like the Exynos 5 and Tegra 4 are estimated to cost well over $20.
Intel has a long tradition of overspending on marketing. A few years ago it showed Ultrabook vendors with $300 million worth of market development funding and it has a huge Core marketing program. Intel recently announced that it would start treating Atom and Core equally, hence the move would make sense. Since Core lifecycles are getting longer, Intel could simply shift some of the funding to Atom products, namely tablet parts like Bay Trail-T.
The only problem? Well the report comes from Digitimes and the site’s hit and miss track record has been on the “miss” side lately, so take it with a grain of salt.