I'm facing a really weird issue on Chrome 62/63, only on MacOS (el capitan and sierra at least).
I have a big sprite (14400 * 4400), that is used inside a HTML5 canvas, and the animation that I do, is really really slow on this platform. I tried on Windows (this browser version and IE/Edge), no problem, and on a lot of other browser (opera neon, vivaldi, firefox, safari), there's no problem. Even on chrome 61 and before, there's no problem. It's only working when I activate the flag "Accelerated 2D canvas" on chrome 62+
I know, that the sprite is huge, its size is already divided by 2 (28800*8800 is the initial size), so is there any way to improve the performances via canvas config or something else ? (I already read this article), I almost follow everything from this article, except the "requestAnimationFrame", because I need to do the job at a certain time. (I don't know if it's a chrome bug or just me doing the wrong thing, but I have no more idea)
Related
I discovered I had a problem when I checked one of my uploaded three.js projects. At first I thought maybe I had done something wrong. My project used r63, so I updated to r65, but that didn't solve the problem, even after clearing the cache and refreshing. I then checked a couple of the demos from the three.js site, and I found they are slow for me, too. As an example, http://carvisualizer.plus360degrees.com/threejs/ autorotates incredibly slowly in Chrome, but at normal speed in Firefox and IE 11. I also tried http://helloracer.com/webgl/ which is fine in Firefox but really choppy in Chrome. It's a disaster in IE11, by the way, but it's an older demo. My project uses OrbitControls, with autorotate enabled. The model is a 16MB JSON file (200,000 triangles), but it worked fine before and works fine in Firefox and IE11. I'm on a Windows 7 machine with a GTX Titan (work computer). Thanks!
type chrome://flags in chrome address bar, and look for settings #ignore-gpu-blacklist
this happened to me, suddenly my gpu was added to blacklist and chrome reverted to software rendering..
For me running google-chrome on linux I needed to enable Use hardware acceleration when available in chrome://settings
I'm a bit confused because my project worked yesterday but seems to no longer work correctly today. (Yes, I've checked previous versions from git.)
The problem: Some divs previously hidden with -webkit-backface-visibility: hidden; magically appeared.
I have isolated this issue into a fiddle:
http://jsfiddle.net/Js6cg/1/
The div is visible in Chrome at 23.0.1271.64 m (wrong) but hidden in 25.0.1326.0 canary (as I expected).
Can you confirm that this is indeed a bug in Chrome or am I using the CSS incorrectly somehow?
(I've updated my GPU drivers (AMD Catalyst) from 12.8 to 12.10 today, if that's important.)
Additionally, the site that demonstrates the effect I've been reproducing appears to work +- correctly at Chrome stable (except for aparrently ignoring -webkit-perspective and animating kind of choppy), while Chrome canary renders it very well and accepts the perspective. I'm confused.
OK, that is embarassing.
The story looks like: I've updated the GPU drivers but looks like I haven't actually restarted Chrome for ages. For some reason, it was unable to re-enable GPU compositing after the driver update and hence some more advanced CSS3 effects (like perspective and backface-visibility) didn't work at all, while simple transforms used a fallback CPU implementation, which also made them look choppy and on the demo site.
I've started Chrome Canary well after the driver update, so it didn't have any issues with GPU compositing. One instance worked, another didn't, but version mismatch wasn't important here at all.
Restarting Chrome fixed that issue. And I'm taking a break!
I'm trying to understand these strange rendering error boxes that are too big to be ignored. This seems to happen on Chrome in Windows 7 (my testing isn't too elaborate) and nowhere else. When I attempt to inspect, they all disappear. This could be some kind of video card issue as I'm using some pretty advanced CSS3 transitions that could mess up memory. In any case, if someone could offer advice on what I could do to fix, I'm at a loss. The site is www.crane-usa.com
Having the same issue with our site using 21.0.1180.89 and 21.0.1180.79. Problem is in Windows 7, Mac OS X latest, Ubuntu and in Chrome frame running in IE9. IE9 with Chrome frame disabled works fine. The problems are intermittent and unrepeatable. Inspect element removes the problem as you say. I tried disabling GPU compositing via chrome://flags but that didn't fix the issue.
We and our users have only been seeing these issues since approx Aug 27, 2012, 3 days ago. I took a look in crbugs.com and found that this seems to have existed for a couple of weeks already. http://code.google.com/p/chromium/issues/detail?id=143647
Sorry our site is not public so I can't post our url but you're not alone.
After upgrading to Chrome 18 on OSX 10.7.3, we started noticing fleeting rendering artifacts seemingly related to form elements:
They come and go during scrolling, it seems. I've been testing in a variety of other OS/Browser environments and have been unable to reproduce this, nor had it happened in previous versions of Chrome.
This is a mature, stable web app with fairly complicated HTML/CSS. I haven't been able to produce a simple test fiddle that reproduces this.
Anyone else seeing issues? Can you think of a possible HTML/CSS cause and not a Chrome bug?
I also noticed that Chrome 18 displays artifacts in some pages, but this only happens when that page is GPU accelerated. I asked a similar question a few days ago:
GPU acceleration crashes website
Chrome enables the hardware acceleration when it detects 3D transforms in your CSS.
You can go to chrome://flags and enable the "Composited render layer borders" option, if you see green squares all around your page thats because some CSS rule is causing your page to be hardware accelerated. You can try removing any 3D CSS rules or assign a higher z-index to these elements. It usually helps.
I'm not really sure what is causing this but in the current stable version of safari on OSX 10.7.X I'm only seeing 3-4 frames rendered. I downloaded the lastest safari beta and it seems like they improved it, but its still dropping a large amount of frames.
Here is an demo that should be viewed in Safari on Lion:
http://jsfiddle.net/JEKAh/1/
Please respond if you know why or what is going on
edit: still is a problem on mountain lion
It turns out that this bug is related to the transfer encoding of the video files. If you are sending the video with Content-Ranges you will see this issue in safari. But if you send the video using Transfer-Encoding: chunked... it will work fine
I used a simple node server to test this: https://gist.github.com/3746561/c303f84866542c4a6ec2956ecf158cb9f492a7a2
-- edit
the above is only a fix for Lion, it appears that Safari Mountain Lion is unable to render frames from a video that is sent using a chunked transfer encoding, a side affect of this is also massive safari memory leaks... I ran a video being piped for canvas for 2 mins and the Safari Web Content process shot up to 12GB of real mem used. -_-
-- edit
after additional research i have found the original issue with standard video to canvas in a recent nightly webkit 537.3 and have confirmed that currently in webkit 537.11 these issues no longer exist... so all i can do is hope that apple updates safari soon including the webkit fixes
-- edit
this is now fixed in OSX 10.9 :)
Firstly, I acknowledge that this might not be the answer you are looking for but it's something which I've just been dealing with for a client so I thought I'd throw it up here:
They reported that their site "Was no longer working well and the animation was jumpy".. (hmm..) Their site uses canvas rendered videos with some overlays for a lot of the visual elements.
So after a while we determined that they had just updated their MacBook Pro to Lion and now their site was slower and less responsive. I was a bit baffled so I got them to bring it to me. To cut to the chase:
Lion & Mountain Lion require a tonne more physical memory (RAM) than Snow Leopard did (due to the new VM architecture as I understand it), I compared their site playback to another MBP with a lower spec, with SL installed and the SL version ran smoother.
After a bit of reading on the Apple support forums which suggested adding RAM, it was all fine again, in fact it seemed smoother than ever..
Not really a programatical answer but one which I thought may be relevant..