antec_468x60
Open login

Functionality to Improve Performance

Written by John Martinez    PDF Print E-mail
die256k250
The latest lineup of desktop CPUs from Intel are based on the Nehalem architecure first introduced earlier this year with the Xeon 5500 series.
When we think about forward progress in technology, performance seems to be the most sensible yardstick. But measuring the speed of a processor isn’t always as straightforward as running a handful of benchmarks. And now that power consumption is also being scrutinized, what a given platform is doing when it’s idle is almost as important as how well it runs under a heavy load.

Fortunately, Intel’s LGA 1156-based Core i5 and Core i7 CPUs include functionality to improve performance, regardless of the software workload being applied. Moreover, when the chip is idle, granular power management ensures that energy efficiency gets maximized.

Turbo Boost Improves Performance

Earlier this decade, processor vendors discovered that there were some very hard-to-overcome challenges associated with continually cranking up the clock rate on single-core processors.

 
Turbo Boost technology can increase the performance of the new lineup of desktop CPUs from Intel by as much as 5 bins.

So, they started pouring resources into parallelism—accomplishing several tasks concurrently, rather than tackling one at a time. Dual-core CPUs emerged, and then quad-core chips surfaced.

The hope was that software developers would, whenever possible, optimize their applications to take advantage of multiple cores, realizing a performance boost via parallel execution. Since then, a great many applications have, in fact, been written to utilize many cores. But others haven’t. And when a customer runs one of those single-threaded titles on a multi-core CPU, the rest of the processor does little work—hardly ideal when performance is needed.

Enter Intel’s Turbo Boost technology—one of the most significant highlights of the Core i5 and Core i7 families. All i5s and i7s are quad-core processors. Intelligence built into the CPU knows when one, two, or four cores are being used, and that logic keeps tabs on temperature, power consumption, and current flow.

If a given workload is only taxing one core, for instance, and all of those monitored attributes are under hard-wired limits, Turbo Boost automatically ramps up clock rate—by more than 600 MHz in some cases. As a result, single-threaded performance jumps substantially.

When all four cores are active, either by one well-optimized app or multi-tasking with several apps, clock rates slow down. That’s quite alright though, since the utilization of four cores outstrips the benefit of higher frequencies. Thus, your customers don’t have to worry about whether software is threaded or not; Intel’s Core i5 and Core i7 CPUs continually monitor use and improve performance in single- and multi-threaded titles through Turbo Boost.

Getting More Out Of Threading

If you’ve stared at the LGA 1156-based Core i5 and Core i7 specifications, wondering how the two processor families are different, Hyper-Threading likely popped out as the most significant differentiator.

 

Hyperthreading is Intel's simultaneous multi-threading implementation that allows CPUs to address as many as eight threads at a time.
As mentioned, both lineups center on a quad-core design. However, the Core i7 processors also include Hyper-Threading. With the feature enabled, each physical core appears as two logical cores to the operating system. The resulting performance gains come from maximizing underutilized execution resources, and can be quite significant in threaded apps.

With a fundamental knowledge of each processor family’s capabilities, it becomes much easier to sell into the right environment. All Core i5 and Core i7 CPUs compatible with Intel’s LGA 1156 interface sport Turbo Boost technology, enabling a great balance between single- and multi-threaded performance. But the customers running well-threaded software or heavily multi-tasking will want to step up to Core i7, which adds Hyper-Threading to execute as many as eight threads concurrently.

 

Comments (0)add comment

Write comment

busy