Hacker News new | past | comments | ask | show | jobs | submit login
Laggy phones and misleading benchmarks (glek.net)
36 points by tarasglek on Nov 20, 2016 | hide | past | favorite | 6 comments

Hi! I saw your post on Reddit, and had some reactions—I'm curious to see what you think.

So you obviously have domain expertise and credibility in IO optimization and your article is one of the better researched articles to discuss the topic of Android UI "jank". I feel like there's more to the story than your thesis and is ultimately not the whole conclusion.

Primarily—aside from citing some academic articles on SQLite IO impact on Android performance, what I think is missing is a real-world connection from the theory that IO is related to UI jank. And I think most particularly, you cite the iPhone 7 and Pixel as having subpar random 4K write performance, even though there is broad anecdotal consensus that those two flagship phones have some of the smoothest, jank-free UI of the industry. I'm not suggesting that this consensus is 100% correct, but it does provide a baseline for a gutcheck and skeptical inquiry (side note: I really wish there was a way to objectively test jank... no one's seemed to be able to get Google's WALT device working...)

And interestingly, you note some midrange phones that appear to have better measured IO performance... but what I think what is missing is evidence (even anecdotal) to suggest that the measured higher IO performance of the Motorola G4 or ZTE ZMax Pro actually translates to smoother performance, compared to flagships with more poorly performing NAND. In fact, the AnandTech review the author cites to for Motorola G4 4K write speed has this to say about performance: "The Moto G4's performance hasn't impressed me in any respect. ... The performance issues aren't just limited to benchmarks. Motorola's previous phones provided a smooth experience despite their limited specifications, and the Moto G4 and G4 Plus are significantly jankier throughout the UI, with even the simplest of ListViews stuttering at times."

If the Motorola G4 in fact does have weak UI performance and jank while having " 7x better than the flagships" random 4k write performance, then I'm not sure the known evidence supports your ultimate thesis yet. It seems more investigation is warranted. Thoughts?

I agree, proof would be nice. I don't know Android well enough to connect UI pauses to IO. Suggestions are welcome

Ah man that is horrible. My Samsung S4 mini is only nearly 3 years old and gets laggier and laggier.

The only reason right now to buy a new phone is that lag.

Would be nice if i wouldn't need to buy a new phone every 3 years because of broken NAND and/or better io performance.

I suspect if you buy G4 and better Motorola phones with good random write perf and reserve 10-20% of space, they should last a long time.

This could be verified via benchmarks.

Point of my blog is that if people focus on the right benchmarks they can select phones that will last longer before the lag.

> I found a great paper on how SQLite accounts for 90% of Android IO

So with only marginally better memory and cache management in SQLite we could get old tired phones to fly?

Injured old phones would be less laggy with slightly different SQLite setup. I think one could further optimize android/sqlite interaction to make it more efficient too.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact