Qualcomm formally launches the Snapdragon 820 SoC: Here’s what you need to know

Today, Qualcomm formally launched the Snapdragon 820, ending the slow drip of announcements of piecemeal revelations that we’ve received over the last few months. This SoC is arriving at a critical time for Qualcomm. From 2012 to 2014, the company rose to command a significant chunk of the smartphone SoC market. By Q3 2014, Qualcomm was measured as holding 42% of the market for personal mobile devices, as measured by Jon Peddie Research.

2015, however, brought fresh headwinds. The Snapdragon 810 was widely believed to have an overheating problem. The actual situation was more complex than that; at least some Snapdragon 810 devices weren’t profiled properly and didn’t manage their own thermals well as a result. Later point iterations on the chip may also have improved the situation. Either way, Qualcomm took a serious revenue hit when Samsung opted to use its own silicon in the Galaxy S6 family, even overall Galaxy sales failed to meet expectations.

Details! Or not. Click to enlarge.

All of which is to say, Qualcomm needs this chip to be a hit. The “Kryo” CPU core is a custom chip built on Samsung’s second-generation 14nm technology (14 LPP). Qualcomm is advertising the CPU as offering up to 2x performance and 2x efficiency compared to previous generation parts, which is rather frustratingly vague. At a guess, this refers to the chip’s theoretical maximum burst performance for a very short period of time and its overall efficiency when measured at idle, or under specific workloads. Intel, AMD, and other vendors excel at this kind of particular argument — actual power consumption improvements and performance boosts tend to be highly workload dependent. Qualcomm is encouraging the use of its DSP and GPU as computation offloads to improve overall CPU efficiency.

Qualcomm does offer some additional color around overall power consumption, though they don’t disclose their test environments or conditions.

Power consumption in “real-life” workloads.

While we don’t know the criteria that Qualcomm used to arrive at these conclusions, the 65% of Snapdragon 801 power metric makes sense given what we know about estimated improvements for 14nm compared to TSMC’s 28nm planar silicon. It’s still sobering to note that this improvement arrives roughly three years since Qualcomm announced the Snapdragon S4 family, which featured the then-new custom Krait CPU core. Even if we take Qualcomm at its word about the relative level of advancement in CPU performance and power consumption, it’s taken three years and two node shrinks to deliver the kind of improvements we used to see in 18 months and a single node shrink — and as we’ve already noted, we fully expect that Qualcomm’s improvement claims will be situational, just as AMD’s and Intel’s are. “Real-life” workloads simply vary too much to be easily captured in a single metric.

Qualcomm’s Tim McDonough took to Twitter to answer questions about the new chip, but he kept his answers light. Qualcomm claims more than 60 device wins for the new 820 SOC, including cameras, cars, and drones. Nvidia has been the SoC manufacturer making noise about the automotive business of late, so this could imply that QC wants to challenge NV on turf Team Green is hoping to stake out for itself.

Without much concrete detail on Kryo, we’re forced to make some educated guesses about the core. Qualcomm has already revealed that the chip is targeting a 2.2GHz clock frequency. The Exynos 7420, built on Samsung’s 14nm LPE process, uses a 1.5GHz frequency base and a 2.1GHz boost clock. What’s interesting to me is that Qualcomm has walked back the 810’s commitment to a big.Little approach and instead opted for a single unified processor block. This suggests that the Kryo CPU may emphasize instructions-per-cycle efficiency over clock speed.

Our guess is that Qualcomm has staked out a midpoint between Apple’s A9 processor, which tends to pound the rest of the ARM family when it comes to single-thread performance, and scaling up to 8 or more cores, the way Samsung and companies like MediaTek plan to do. A higher-efficiency quad-core may represent the sweet spot between these two extremes. Presumably, Qualcomm is using DVFS to control its clock frequencies and power gating, though there’s always an outside chance that it adopted a technique like the AVFS AMD is using for Carrizo.

Putting it all together

Qualcomm’s comprehensive spec sheet on the Snapdragon 820 is shown below. The company is claiming a bevy of improvements for its Adreno 530 GPU (up to 40% more performance and 40% greater power efficiency than previous-generation hardware), new hardware decode and encode engines, an X12-class modem, and a variety of enhancements to image processing and security.

The full spec sheet. Click to enlarge.

We’ve also heard tell that while Qualcomm’s SoC design for its upcoming server family is quite different from Snapdragon 820, the CPU may derive from Kryo. This makes sense, if true — building a common architecture and leveraging it across multiple contexts is how both AMD and Intel have historically done things, to good effect (Bulldozer notwithstanding). If Qualcomm is serious about competing in the server market, it will want to leverage economies of scale.

If the Snapdragon 820 does well, we’ll see that reflected in phone design wins at CES and Mobile World Congress in early 2016. The Snapdragon 810 may not have won the accolades that the Snapdragon 800 family received, but the 808 and 810 combined have sold well and shipped in a number of high-profile designs. At the same time, however, Qualcomm faces increased competition from mainland Chinese and lower-cost Taiwanese companies. It needs the Snapdragon 820 to sell well if it wants to retain its de facto leadership of the Android market — a position Samsung arguably usurped with its own Exynos 7420 in 2015.

Devices based on the Snapdragon 820 are expected to ship in the first half of next year, probably starting in March or April.

Qualcomm’s new Hexagon 680 DSP: Fast, efficient, shipping with Snapdragon 820

The annual engineering and technical conference known as Hot Chips kicked off yesterday, and Qualcomm was out in front to detail its new DSP, the Hexagon 680. Digital Signal Processors (DSPs) aren’t something we’ve discussed much at ExtremeTech, and Qualcomm is putting a major marketing push behind their DSP technology for the first time. How does the chip work, what makes it an integral part of Snapdragon 820, and how does it advance heterogeneous computing?

DSPs are specialized processors dedicated to digital signal processing. Like GPUs, DSPs are designed to exploit parallelism. Like CPUs, they often make use of SIMD (single instruction, multiple data) and VLIW processing to boost throughput and total performance per watt. Also like GPUs, DSPs are designed to perform a very specific subset of tasks. CPUs can handle these tasks (and sometimes do), but DSPs offer better performance than general processors, and more flexibility than a traditional ASIC. This relationship is captured in the slide below:

Qualcomm’s Hexagon 680 DSP

Qualcomm’s Hexagon 680 is designed to accelerate certain workloads at performance efficiencies well above anything a modern CPU can offer. The Hexagon 680 is a VLIW (Very Long Instruction Word) processor, meaning it’s designed to extract maximum parallelism per clock cycle and to spread workloads across a wide set of execution units.

The 680 DSP offers four parallel scalar threads, each with 4-way VLIW support and a shared L1/L2. Each of these scalar groups is clocked at 500MHz for a maximum throughput of 2GHz-equivalent worth of processing. On the vector side of the equation, the 680 has 32 1024-bit vector registers. Each instruction can address up to four of these per cycle, for a maximum output of 4096 bits per cycle per instruction. It also includes support for Qualcomm’s new Hexagon Vector Instructions, or HVX. The HVX registers can be controlled by any two of the scalar registers.

Here’s what this means in aggregate: The Hexagon 680 is designed to allow for extensive threading and to share data across the L1 and L2 caches. There’s no penalty to using the HVX units and the scalar units simultaneously, provided that the workload is designed for it. The vector processors don’t have access to L1, but treat L2 as their first level of memory. L1 and L2 are kept coherent and data can be streamed into L2 from DDR memory at up to 1.2Gpixels/s. This supports some of the advanced capabilities of the Hexagon 680 (we’ll talk about these below).

According to Qualcomm, the performance advantages of these new features is enormous. While this data is provided by the company and should be taken with a grain of salt, there’s nothing outlandish here. These kinds of accelerations are typical when moving to a high-end dedicated chip as opposed to executing code on a general-purpose CPU.

Qualcomm believes that the programming model for the Hexagon 680 is similar enough to CPU models to allow programmers to use the hardware effectively, but with significant overall improvements.

Power consumption should also be much reduced, thanks to the simpler nature of the VLIW model and use of L2 for vector processing rather than both the L1 and L2. The company also notes that by adopting its DSP for low frequencies, it can cut leakage current and reduce overall power consumption.

Applications and heterogeneous computing

The best application processor on Earth isn’t worth much without applications to run on it, but the Hexagon 680 DSP delivers on this front as well. Qualcomm claims that the new chip is fully heterogeneous, meaning it can share data between CPU, GPU, and the DSP. Qualcomm is also a founding member of AMD’s HSA consortium, and while it isn’t calling its heterogeneous compute model by that name, we expect the two to be similar on a conceptual level. The DSP inside the Snapdragon 820 can be used to render AR or VR, tapped for better video playback and encoding, or used by the camera for extensive improvements in low-light photography. Alternately, HVX can be used to enhance detail in standard photos, as shown below.

Qualcomm has stated that the Hexagon 680 can perform low light enhancement 3x faster than a Krait SoC, while using 1/10 as much power. Programmers will be able to use the DSP and write applications to run on it, which could give the Snapdragon 820 platform a substantial leg up over the competition. DSPs have shipped on SoCs for a long time, but few companies spend as much time talking up their solutions as part of a heterogeneous compute platform as Qualcomm has.

In the past, a component like the DSP would be invisible, buried under interest in the CPU and GPU. Qualcomm’s decision to talk about the chip is a sign of the times. As visual processing, augmented reality, and virtual reality take the stage, more and more consumers expect advanced capabilities from their smartphones. For lower-tech users, that means high quality photos and video, while gamers and enthusiasts want cutting-edge performance and better battery life. The Hexagon 680 DSP is meant to speak to all these needs, with power efficiency that will beat even the upcoming Kryo CPU, flexibility and heterogeneous compute capability to whet the appetites of programmers and application developers, and performance that appeals to enthusiasts, gamers, and the general public.

After these disclosures, the Kryo is the last piece of the puzzle still to drop into place. Hopefully we’ll have details on the CPU core sooner rather than later.

HTC O2 rumored for 2016 with Snapdragon 820, 4GB of RAM, and Android 6.0

HTC was the original Android phone maker. It manufactured the first retailAndroid device, the T-Mobile G1, as well as the prototype devices Google used to develop the platform. After a solid run, the last few years haven’t been particularly kind to HTC. After reportedly disastrous performance for the HTC One M9, the Taiwanese company is rumored to be planning a vastly different flagship phone for early 2016 called the HTC O2 — the “One” name is apparently gone.

We’re still months out from any official announcement of HTC’s 2015 Flagship device, but the details we’re hearing sound very plausible. The O2 will reportedly have a Snapdragon 820 system-on-a-chip (SoC), replacing the embattled Snapdragon 810 in the One M9. This chip has run hot in all the phones that run it, but the M9 in particular suffered from severe thermal throttling that hurt performance.

The Snapdragon 820 will have a completely different architecture consisting of four custom 64-bit ARM cores (probably called Hydra) designed by Qualcomm. That’s what the company has traditionally used in its SoCs, but the 810 runs reference Cortex ARM cores in an octa-core configuration. The custom solution simply wasn’t ready in time. The device will also reportedly have 4GB of RAM, 32GB or 128GB of storage, and a 3500mAh battery.

Rumors also point to a QHD (2560×1440) screen, which would be a step up from the M9’s 1080p. A 1440p screen might not be necessary strictly speaking, there are situations where it can look better than a 1080p screen. And shipping a 1080p LCD for the third year in a row made HTC look like it as lagging behind other OEMs anyway. The resolution bump is necessary in 2016.

The HTC O2 (or whatever they call it) will be running Android 6.0 Marshmallow, but that’s not much of a surprise. Any flagship phone released in 2016 will be running the latest version of Android. Software hasn’t really been HTC’s problem lately, though. It’s all about that boring hardware.

The M9 under-performed largely because it looked and felt like the same phone HTC made in 2014. The design was only slightly changed from the M8 to the M9, and many of the specs stayed the same. To turn things around in 2016, the O2 ought to shake up the design — we know HTC is capable of making great hardware and the M7, M8, and M9 are all very well-made. The camera also needs some serious attention, and that’s not mentioned in the leaks. The M9’s 20MP camera was atrocious.

HTC recently saw its market cap drop after another in a series of disappointing quarterly reports. The company is now valued lower than the value of the cash it holds, which does not usually end well. HTC does still have almost $1.5 billion on-hand, which should get it through another product cycle to make the O2 a reality.