Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

The truth about Intel's Broadwell vs. Haswell CPU

Gordon Mah Ung | July 7, 2015
Intel's fifth-generation Broadwell CPU has been the default laptop processor of choice since its debut in January, but it's been difficult to get a real read on just how much of an improvement it really was over its Haswell predecessor.

The good news for Broadwell is that its improvements in performance close the clock speed gap. Despite that roughly 10-percent clock speed difference between the two, the actual performance gap in Cinebench R15 is closer to 5 percent.

CineBench 3D Rendering performance

What happens when you run CineBench R15 in multi-threaded mode, though, where it measures the overall performance of all the cores in a CPU? We see the Broadwell come back to edge out the Haswell chip. This is because, while the Broadwell chip is slower in overall clock speed when only one core is working, it runs a little cooler (thanks to its 14nm process). The Haswell has an advantage in the beginning, but as it heats up, it starts to throttle back on frequency so it's only about 100MHz faster by the end of the test. Combined with the better efficiency of Broadwell, it's just enough to make the 5th-gen chip barely faster here.

PCMark 8 Office performance

I also ran PCMark 8's Work test using the conventional setting. This keeps the workload to the CPU side instead of using a little of the graphics core for some tasks. PCMark 8 Work test is designed to simulate such tasks as videoconferencing, and general-office droning. 

PCMark 8's not heavily multi-threaded like CineBench and doesn't heat up the Haswell enough to make throttling a big problem. The Haswell CPU's slightly higher clock speed gives it the advantage here, but for the most part it's not enough to write home about. It's actually a fairly realistic evaluation of the low demands of mundane office tasks: The dual-core Haswell laptop scores 2,922. Consider that a tested desktop (not shown here) with a six-core Core i7-5820K scored 3,321. That tells me PCMark 8's right on the mark for office tasks and you don't need that many cores for office work. 

Handbrake Encoding performance

Looking for a heavier-duty task, I also threw in our standard encoding test, where we take a 30GB, 1080p MKV file and use Handbrake to transcode, using the Android for Tablet profile. On dual-core machines, it takes in excess of two hours to complete.

This test tells us two really useful data points. The first is how well a particular CPU performs in this heavily multi-threaded test. Short of an 18-core CPU with 1080p resolution files, it'll max out all cores. 

The second data point you get from this test is how much the design of a laptop suffers from thermal throttling. Modern CPUs are designed to slow down when they get too hot or if the PC maker determines it's heating up the entire laptop too much. For example, if you look at the Handbrake test in my recent review of the Lenovo LaVie Z, you can see how much certain PC designs will throttle back the CPU's speed--the review's Handbrake chart shows the impact of thermal throttling from the Surface Pro 3 and the Lenovo ThinkPad Carbon X1, both of which have the same CPU. If you look at the result from the HP Spectre x360 and compare it to the Dell XPS 13 that has the same CPU, you also see the performance drop-off to aid and abet the cooling capability of the laptop (or the decision by the laptop maker not to heat it up too much). 


Previous Page  1  2  3  4  Next Page 

Sign up for Computerworld eNewsletters.