For PC users on the go, machines with a long battery life are essential, but so is performance. Balancing these contradictory requirements lies at the heart of Intel's new CPU architecture, code-named Haswell, which is expected to appear in shipping tablets, Ultrabooks, and other computers in 2013.
At the 2012 Intel Developer Forum earlier this week, Intel dove deeper into what makes Haswell tick.
Or maybe "tock" would be a more appropriate verb, since Haswell represents the "tock" in Intel's CPU development program. Intel uses "tick/tock" to denote its CPU development strategy. A "tick" occurs when Intel tweaks an existing CPU design relatively little but takes advantage of all the efficiencies of a new manufacturing process. Ivy Bridge, for example, was a tick, improving on Sandy Bridge only incrementally, but moving to 22nm. Intel always builds new architectures, like Haswell, on proven manufacturing processes--and Intel's 22nm manufacturing process has been well shaken out, thanks to the company's Ivy Bridge line. So Haswell constitutes a "tock."
Haswell isn't just another Intel PC processor, though. Intel is talking about producing Haswell-based processors across a range from dual-core chips that run at less than 10 watts (making them suitable for tablets) to quad-core desktop CPUs that can outperform the fastest Ivy Bridge processors.
How did Intel hit its aggressive target of a 20-fold improvement in power efficiency? In answering this question, we need to looking at Haswell's power management technology before diving into the CPU proper.
Modern CPU power-management technology cuts power to large chunks of the processor when doing so makes sense. A special processing unit--which the company calls it the Power Control Unit--built directly inside the main CPU manages power almost to the transistor level. The PCU looks at which parts of the processor are idle, and turns individual parts on and off as needed.
The problem isn't putting parts of the CPU to sleep, though; it's waking the processor up quickly enough for the napping power capabilities to be useful. After all, if you had to wait a minute every time your laptop always took a minute to wake up after it went to sleep, you'd soon be tempted to throw it against the wall in frustration. Intel CPUs prior to Haswell have two main states: active and sleep. (The technological details are more complicated than that, but that's the general idea.) Over the years, Intel has steadily decreased the amount of time a sleeping CPU takes to wake up. The current Ivy Bridge processor takes several seconds to rouse itself from a deep state of slumber--but several seconds is still not quite "instant on."
Haswell's solution is to add a third power state--something Intel designers call "Active Idle" (aka "SOix"). Active Idle is an extremely low-power active state that uses one-twentieth the power that Ivy Bridge uses. In that state, the PC system considers itself awake, but the CPU is mostly asleep. This trick translates into wake times of no more than a few tenths of a second. A worst-case wake-up time of half a second is considerably better from the user's perspective than the several seconds that today's CPUs take to wake up. Haswell is almost always in this "instant resume" state while running. Much of the technology came directly from Intel's Atom processor power management scheme.
Intel adopted a few other tricks in building Haswell. The new CPU's sleep and Active Idle states are actually divided into multiple smaller states. Each mini-state (known as a "C-state") defines exactly what part of the CPU is turned off. The new C-states permit more-granular power management, which in turn yields longer battery life, since your CPU won't constantly be waking up parts of the CPU that it doesn't need in order to wake up some part of the CPU that it does need.
Intel also examined the way CPU power usage interacts with a system's display. LCD panels take a relatively long time to wake up in today's systems, so Haswell processors will include panel self-refresh. For example, if you're just sitting and staring at your screen, a Haswell CPU will go to sleep, with only a tiny part remaining awake to refresh the monitor. As soon as you move the mouse or press a key, however, the CPU will wake up. You won't notice the wake-up time involved, because the display never went to sleep.
Now that we've discussed how Intel achieved better power efficiency, let's look at the architectural enhancements in the Haswell CPU.
Next: Performance and power efficiency
Improved performance, better power efficiency
During one of the IDF technical sessions, Intel Senior Principal Engineer Ronak Singhal noted several times that Intel added no new features to the CPU that would have imposed a power penalty. Even so, CPU designers have a number of options for improving performance while accommodating the need for greater power efficiency.
One trick is branch prediction, which lets the CPU anticipate instructions that are likely to be executed in the near future. If the CPU knows which instructions will be coming through the pipeline next, it can allocate CPU resources much more efficiently, turning on only the parts of the CPU that the new task requires. So Intel tweaked Haswell's architectural elements to improve branch prediction, enlarging the internal buffers and the out-of-order windows.
The more work a CPU can do in a single cycle, the better its performance will be at the same level of power usage. So Intel added the ability to run two floating-point multiply-add operations every clock cycle, doubling the performance throughput over Ivy Bridge for the same power usage. L1 and L2 cache throughput is better, too, reducing how long the CPU must wait for data to arrive.
Of course, none of this good stuff comes for free. Though power efficiency has improved, Haswell pays a price in chip real estate. Given that Haswell will still be built on the 22nm CMOS process, the chips themselves are likely to be larger than Ivy Bridge CPUs.
The chip size will likely increase for another reason as well: graphics.
High-end PC gaming on tablets: Haswell graphics
Haswell builds on the existing Intel HD graphics core in Sandy Bridge, adding refinements and improving power efficiency. Haswell now offers three different integrated graphics options for Intel CPUs (called GT1, GT2, and GT3), as opposed to the two options (Intel HD 2500 and HD 4000) available with Ivy Bridge.
From a performance perspective, GT3 is the most interesting new graphics option. GT3 doubles graphics performance over what was possible with the older HD 4000 GPU, simply by doubling the number of execution units. Execution units act as the GPU's core computation engine, handling graphics shader and GPU compute tasks. These execution units are built into a common modular unit, which Intel calls a "slice common."
The slice common contains a number of other key components for real-time graphics, such as the raster engines and cache. To double the number of compute engines over the HD 4000, Intel added a second slice common to GT3. This additional slice takes up some chip space, but it saves power because the GPU doesn't need to enter turbo mode for additional performance.
Other tweaks to the GPU include improved texture samplers, enhanced overall bandwidth, and additional circuitry to handle tasks that the driver deals with in the HD 4000.
All of these features improve performance without increasing power consumption. According to Intel, an 8W Haswell unit could conceivably integrate a full GT3 GPU, though the company offered no specifics on product versions. Intel showed two different applications running: Unigine Heaven, a synthetic graphics benchmark; and Bethesda's Skyrim, a PC RPG with demanding graphics requirements. Haswell ran both tests at double the performance of Ivy Bridge, permitting a much smoother visual experience.
Historically, Intel has been late to the party in adding software support for the latest programming interfaces. Haswell breaks with this tradition this by implementing all of the latest standards: DirectX 11.1 for Windows 7 and Windows 8, OpenCL 1.2 for GPU compute, and OpenGL 4.0. Intel has been good about driver support, supplying both Windows and Linux drivers.
Though Haswell's 3D graphics engine substantially improves on previous Intel efforts, desktop PC users will probably still want to use a high-end discrete graphics card for best PC game performance. But Haswell's graphics core will make even extremely thin Ultrabooks respectable gaming platforms, and the new GPU opens up possibilities for modern PC games that run on Haswell-based tablets.
Next: The video engine
The video engine
Ivy Bridge introduced the QuickSync video block--a dedicated, fixed-function video unit built into the GPU. A fixed-function video engine permits much faster video encode and decode performance. Unfortunately, it's not programmable, so if some hot new HD codec came along, the video engine wouldn't be able to handle it. But given that video codecs are fairly standardized, that's not likely.
Intel did, however, add additional codec support to the Haswell video engine. Motion JPEG (MJPEG) is important for videoconferencing. SVC (scalable video codec) is useful in mobile environments, where the quality of the video may change depending on the speed of the connection, so the video quality can scale gracefully. When network bandwidth degrades, SVC enables users to continue to see good frame rates instead of experiencing jerky or dropped frames. Intel also added support for 4K video for the upcoming generation of ultra-high-definition panels.
Bottom line: Bigger chip, better performance, lower power
Haswell will take up more die space than Ivy Bridge does, and that change translates into a higher materials cost per CPU for Intel, since it won't be able to manufacture as many Haswell CPUs on a single wafer. Haswell's modularity mitigates this disadvantage somewhat, however: Intel can build many different Haswell products, targeting lower-power niches with smaller versions. Even better, Haswell is well positioned for Intel's next-generation 14nm manufacturing process.
Most of the performance tweaks to the main CPU are evolutionary advances over Ivy Bridge's capabilities. At the same clock frequency, users may see performance improvements of up to 10%. The big gains are in power efficiency and graphics performance. Improvements in power efficiency will enable laptops finally hitting all-day battery life in Ultrabooks, while GPU improvements will give mobile users reasonably robust game experiences on the go.
Haswell may have more impact on Intel's bottom line than any CPU in years. The wide range of products supported by the new CPU's modular nature and power efficiency will likely encourage manufacturers to develop a wide array of Haswell products. Users will benefit from having more choices than ever in mobile and desktop designs, with improved performance and longer battery life.
This story, "Inside Intel's Haswell CPU: Better performance, all-day battery" was originally published by PCWorld.