New Report Shows Importance of Evaluating Sustained Laptop Performance

We’ve known for decades that thermally stressed hardware doesn’t perform optimally. This dates back to at least the Apple III, when the system’s own heat could result in chips coming loose from their sockets and Apple’s own tech guidance involved dropping the chassis to reseat the chips. In recent years, Intel has loosened some of its OEM restrictions, giving companies more freedom to experiment with a variety of skin temperatures and throttle points. This gives companies more flexibility to design products that are always cool to the touch, for example, but it also increases the chance that manufacturers will ship products that can’t handle their own heat output.

A report from HotTech Vision and Analysis commissioned by Dell earlier this year makes this point. The firm tested systems from Lenovo, Microsoft, Asus, HP, and Dell with a simple Cinebench looped run. While looping Cinebench might not be the most comprehensive method of testing this issue, it’s also simple, taps a freely available benchmark, and is something you can do on your own hardware. Points for that.

I’m going to arbitrarily draw some lines and say that performance shifts of 5 percent or less should be considered normal (there’s always going to be a margin of error), 10 percent is noteworthy, and above 10 percent indicates a system could have issues with sustained performance under load. The good news is, most of these systems wind up below 5 percent, but Microsoft’s Surface Book is a major outlier, with performance dropping substantially in subsequent runs. The fact that the gaps are as large as they are with a simple looped Cinebench run implies they might be larger with a more inclusive, long-running set of tests.

One thing we have to stress here is that there’s no real way to generalize this data. The 2017 and 2018 variants of a system might look externally identical but have very different throttle points and skin temperature settings, to say nothing of internal cooling changes. There’s no way to generalize across product lines or families and OEMs make no guarantees on CPU clocks. No OEM currently offers any programs or marketing guarantees on clock and we’ve known since 2014 that thermal constraints can leave lower-power Intel chips outperforming higher-clocked cores due to throttling. AMD doesn’t sell into the thin-and-light ultrabook segment as much as Intel does, but the same care should be exercised with all systems, regardless of manufacturer. This test being commissioned by Dell doesn’t concern us as far as the likelihood of the findings — we have additional data points from previous system tests and Apple’s own release in the summer of 2018, as well as mobile phone data showing that sustained tests can find throttle points that short tests don’t.

This is principally an issue for those planning to invest in thinner and lighter systems, but such designs are an increasingly large part of the premium offerings from every laptop manufacturer. Be advised that when you look at performance tests, you may be seeing how two systems compare in Burst mode, but not necessarily how they’ll compare if you try to run a heavy workload. Of course, if you don’t do that kind of computing, such comparisons won’t be important — but if you transcode, edit media, or run 3D rendering applications on a regular basis, sustained performance needs to be considered as well. Hopefully, we’ll see more reviewers in the future focusing on sustained performance alongside more traditional tests.

Continue reading

5 Reasons Space Exploration Is More Important Than Ever

Continue reading

5 Reasons Space Exploration Is More Important Than Ever

5 Reasons Space Exploration Is More Important Than Ever

The 8 Most Important Changes in the Windows 10 Spring Creators Update

5 Reasons Space Exploration Is More Important Than Ever