Major performance issues with R192 of Construct?

Hello everyone, I noticed recently Construct 2 has horrid performance on mobile devices. Any simple game I have thrown at my new device seems to do poorly in fps tests.

So I decided to run a simple, blank project, with just text object, and a "every tick" condition to set the text object to 'fps' and to my disappointment, it averages 50 fps even at the most bare bone project you can physically achieve. There is definitely something wrong, I am using the chrome browser at the newest revision.

My mobile device is the LG G3, with a Snapdragon 801 clocked at 2.5 Ghz with 3Gigs of RAM and the Adreno 330 GPU, there has to be an issue occurring. either at Construct, or Chromium. Anyone notice a dip in performance lately?

It seems that you can get good framerate only on deprecated CocoonJS (but it does not support text object and native ads)

Umm...yeah, something is certainly wrong.

FYI, the text object is pretty demanding (I know that sounds bizarre, but it's true). If you are updating it every tick, use a spritefont instead. From the man himself (search for 'text'):

https://www.scirra.com/manual/134/performance-tips

What about trying a real game with an exporter? Have you tried a test export of space blaster? Throw an fps/cpu meter on it (remember, use a spritefont) and export it. You could also print out the renderer; if it says canvas2D, then something is amiss.

For comparison, I have a much lower-end device -- LG Ultimate 2, Snap 205 1.2ghz, Adreno 305 -- and get very playable framerates in space blaster, even at it's most intense. My current project is using a couple webGL shaders and up to 50 onscreen objects, and hits 60fps consistently.

Also, I'm not seeing a great deal of difference between CocoonJS and crosswalk, but that may just be my projects.

Chrome could be causing a problem too...I haven't tested Chrome mobile extensively, but on desktop v38 had poor rendering performance that led to unstable framerates and 'janking'. This may have affected mobile as well.

Umm...yeah, something is certainly wrong.

FYI, the text object is pretty demanding (I know that sounds bizarre, but it's true). If you are updating it every tick, use a spritefont instead. From the man himself (search for 'text'):

https://www.scirra.com/manual/134/performance-tips

What about trying a real game with an exporter? Have you tried a test export of space blaster? Throw an fps/cpu meter on it (remember, use a spritefont) and export it. You could also print out the renderer; if it says canvas2D, then something is amiss.

For comparison, I have a much lower-end device -- LG Ultimate 2, Snap 205 1.2ghz, Adreno 305 -- and get very playable framerates in space blaster, even at it's most intense. My current project is using a couple webGL shaders and up to 50 onscreen objects, and hits 60fps consistently.

Also, I'm not seeing a great deal of difference between CocoonJS and crosswalk, but that may just be my projects.

Chrome could be causing a problem too...I haven't tested Chrome mobile extensively, but on desktop v38 had poor rendering performance that led to unstable framerates and 'janking'. This may have affected mobile as well.

Just tried the space blaster, and I was getting 50 fps on intense gameplay, thats so weird. As simple as one text object changing every tick getting the same frame rate?

Might be something else going on. Your device is way more powerful than mine, but I'd estimate I was getting around 30-40fps in space blaster myself.

My guess is that it's your device resolution. 2560x1440 is a crazy number of pixels to push around, even on a desktop, much less a mobile. If you are using high quality scaling, you are wasting a ton of GPU rendering pixels you'd need a microscope to see. Try:

Fullscreen in browser: Letterbox scaling

Scaling Quality: Low Quality

Game Resolution: Leave at 960x720. If you want it sharper, try 1280x960, but IMHO it would be a waste.

Sick phone, BTW.

"As simple as one text object changing every tick getting the same frame rate?"

IIRC drawing a text object is not an easy task at all, it has to recalculate entirely how to render it for each character, if you try with a spritefont instead, you might (should) not have that issue.

I haven't noticed any issues at all. sbperftest gets a great score on my new Moto X (2nd gen) which sounds like it has the same hardware as the G3.

Ashley when the player dies in sbperftest, it seems it restarts with a very large amount of ennemies and projectiles, is tgat a part of the test, or should that be ignored?

Ashley

His phone has a resolution of 2560x1440. Space Blaster defaults to high quality scaling.

He's trying to push around almost 2x the number of pixels as you are on a Moto X. Fillrate issue.

Aphrodite - it's just a quirk of the test, you can ignore it.

TiAm - on a high end phone it should have no problems at all with the fillrate.

TiAm - the change in quality and resolution helped a little bit, not very much, about 2 fps. The phone is pretty sweet, thank you haha.

Ashley - I should be getting a constant 60 fps on a high end phone when I first start space blaster, or close to 60 fps right? I just cannot grab 60 fps at all on the test at any given time.

As you can see here, it drops a lot of frames in my screenshot.

I tried this same test on my Samsung Galaxy Tab Pro 8.4 with a Snapdragon 800 2.3 Ghz, 2GIG RAM and with the Adreno 330 GPU as well and a resolution with 1600x 2560.... and I can constantly hit 60 fps no problem on the tests, with MINIMAL frame drops.

Is there some problem with fill rate maybe with higher resoltuon devices? I would hope not, because higher resolution devices are the future. The LG G3, along with the Samsung Note 4 and the Droid Turbo are all sporting Quad HD displays (about 2k display) already, and the trend will continue.

Ashley

Come to think of it, you're right. For some reason I thought space blaster had a couple webgl effects attached, but I guess it doesn't.

GameThirsty

Not fillrate; space blaster uses no webgl effects or large, transparent objects. My mistake. However, if you were piling on some webGL effects/large transparent objects, I'd watch out. But you are having some other problem.

GameThirsty

Can you post a screenshot running that on your Galaxy Tab Pro 8.4? Their CPUs are very similar, it would be interesting to see the CPU utilization.

Because if a game is sub 60 fps, the CPU is normally averaging much higher loads than your 36%.

Here's from my LG G-Pad 8.3, with Snapdragon S600, its a 1080p device. FPS fluctuates from 50 to 60, it appears smooth enough even with some skipped frames.

[attachment=0:3ewfqc78][/attachment:3ewfqc78]

GameThirsty

Can you post a screenshot running that on your Galaxy Tab Pro 8.4? Their CPUs are very similar, it would be interesting to see the CPU utilization.

Because if a game is sub 60 fps, the CPU is normally averaging much higher loads than your 36%.

Here's from my LG G-Pad 8.3, with Snapdragon S600, its a 1080p device. FPS fluctuates from 50 to 60, it appears smooth enough even with some skipped frames.

[attachment=0:1nyayd55][/attachment:1nyayd55]

—

Here is the screenshot of the Galaxy Tab Pro 8.4