Here’s Intel’s own explanation of Hyper-Threading. Note how the same number of instructions from two threads can be run on fewer total execution units.
Are there other ways in which these new 32 nm mainstream chips echo the i7 to achieve blistering desktop speeds at sub-$200 price points? You bet, and there are two key technologies at the heart of this performance: Hyper-Threading and Turbo Boost.
In essence, Hyper-Threading (also sometimes called simultaneous multi-threading, or SMT) takes idle execution units within a CPU core and puts them to work processing a second instruction thread. The result is one physical core but two virtual cores and a much more efficient use of hardware resources because Hyper-Threading yields a relative lot of performance gain for only a little more power draw.
Hyper-Threading has been around since 2002, when Intel first debuted the technology in its Xeon line, but the technology went on hiatus when Intel transitioned from the Pentium 4 into Core microarchitecture. In part, this was because Core was so much more power efficient than the Pentium 4 that Hyper-Threading wasn’t needed. But now Hyper-Threading has returned, first with the Intel Atom, then the Core i7, and now the i5 and i3 lines. In fact, the feature is so important that it’s the key differentiator between the Core i3-5xx and Pentium G series, both of which are based on the same 32 nm architecture.
Tom’s Hardware was one of many sites that measured the significant performance improvements to be had from Hyper-Threading on Intel’s latest core architecture.
Hyper-Threading isn’t the only spice in the 32 nm recipe, though. Just as Hyper-Threading separates the Core i3 from the Pentium G, Turbo Boost divides the Core i5 family from the i3. At its simplest, Turbo Boost is a form of automatic, dynamic processor core overclocking. It works from the knowledge that systems sometimes run in a state in which there are few threads but those few need every bit of speed possible.
In such cases, the operating system will request the highest performance state (P0) from the CPU. A processor with Turbo Boost will then examine several factors, including die temperature, current power consumption, and the number of active cores. If the CPU isn’t pressed against thresholds for these criteria, it will boost the core frequency in 133 MHz increments until a safe ceiling is reached. Depending on the circumstances, this might result in a small boost being applied to all cores or one core being turned off to provide more overhead for its partner. (In the Core i5-6xx series, Turbo Boost can deliver up to an extra 266 MHz.) No BIOS tweaking, no manual voltage adjustments, no special utilities—it’s all handled automatically and invisibly.
The principles of Turbo Boost apply just as much in the dual-core Core i5 chips as they do in the quad-core Core i7 series that debuted them. When one core gets shut down, another gets more headroom for added performance.
|
