About the jerkiness on the movement...

eli0s

I have no idea why your browser should be reporting dt to that sort of accurate round number. Ashley, would you be able to shed some light on why this is happening..?? I don't even know if it's a problem, although I suspect that any events that use dt might be off by a small amount if the reported dt is inaccurate for some reason.

TiAm,

Good idea on the adjustments to the 2nd test. I've changed how it works so that now there is only one data window and the user can input the length of the test.

Here's the LINK for convenience.

Please can I ask everyone to reduce the sizes of their screen captures - it's hurting my brain....

Please can I ask everyone to reduce the sizes of their screen captures - it's hurting my brain....

Ya, will do. Sorry 'bout that

Canary updated today to Version 40.0.2209.0 canary (64-bit). Tried your new test windowed and got this:

[attachment=0:2s9nsiu9][/attachment:2s9nsiu9]

Seems like the new improvements work in window mode now.

eli0s

Give the new canary a try?

And just for interest, here's Android Chrome run on my Nexus 5.... Not bad, really...

I'll give the new Canary a go when I get home.

TiAm , unfortunately it's the same:

Canary 40.0.2209.0

The frame rate is fixed at 60, the object count fluctuates between 60 and 62 (with most frequent been 61, after that after that 62 and less often 60) and in a small browser window there is little but still noticeable frequent jerkiness in the scrolling background.

Perhaps I am cursed..? Am I the only one that has this results? Everybody else gets a graph that looks like Colludium and TiAm?

eli0s

What is your system like (cpu, gpu, os)?

The only situation in which canary gives me the result you are getting (dt's at whole number intervals) is when I run it full screen (F11). Don't know why that would be.

BTW, for anyone who runs Colludium's newest test: If you enter a custom time, you have to press 'Enter' to actually set it. Otherwise, it'll just run 30 seconds regardless.

It's really weird that browsers would show lots of timing happening at 16.0 or 17.0ms (as in, the integer values). That suggests to me some kind of incorrect integer rounding happening somewhere in the browser's timing code. Only IE11 seems to show a sensible graph of a normal-distribution curve centered on 16.7ms. This is an interesting test to report to Chrome and Firefox.

TiAm , my system has an old Intel Core2 Quad CPU (Q9550) overclocked to 3.7 GHz, 16 Gb of DDR3 ram and a nVidia GTX 660 Ti with 3 Gb ram. I know that it's not by any means a high end system, but my original concern was about the jerky movement of a single moving square

As for your full screen canary results, while they look somewhat the same as mine (having a few prominent scattered spikes), if you look closely you'll notice that there are a lot small lines evenly distributed from 13.3 ms up to around 19 ms or so. In my cursed tests I get only integer numbers or floats that have .99 in the decimal places (like 15.99, 17.99 etc).

Colludium's tests by the way look awesome and pretty professional

eli0s, your system sounds quite powerful. I wander if the over-clocking could have caused this behavior - a bit of a long shot of a guess... Is there any way you can undo the over-clocking?

Thanks for the review - I hope my embryonic game meets the same reception!!

TiAm, good point above, I've adjusted the demo so it's obvious what you should do. I've also reduced the size of the freq scan window to +/-20% of the ideal - it'll miss some of the more extreme values of dt but will draw a better curve.

I'm going to do a bit more testing and then report this to Mozilla and Google (Canary, which is much better than the current release of Chrome, still shows some sort of harmonic frequency distribution away from the ideal). If the dt is longer than the ideal it is possible that the next is forced to be shorter to compensate and hit the vsync timing, but my test doesn't resolve that level of detail from the results. I'm not convinced, though, as the data graphs appear to be classic harmonic frequency distributions.

eli0s

My system was getting similar results. My results were very stable. And seemed to drop a lot of data, the decimal values were not getting shown on graphs. for some odd reason.

Thats why I had to make the version where it could have more decimals/ move the decimal over to see what was going on.

My system:

i7 3930k (6 cores 4.6ghz)

gtx 780

32gb ram 1866mhz

I tried canary, latest download. It was terrible on my system. dt anywhere from 14 to like 18. Constantly stuttering and creating frame rate drops/ jitter

IE wont obey my refresh rates at all. 60hz gives 67fps and 120+ was giving 60 fps >.> 120hz+ had a nice dt though...if I was running 60hz

Chrome has worked the best for me. Seemingly hit or miss though. Sometimes it stays fairly close to 16.7 other times it gets closer towards the 16/17.

I have found chrome really only cares about having an average dt of 16.7 which sounds nice on paper being able to say chrome almost always has an average dt spot on...except it really does not ever have the correct dt which is useless.

I am getting results similar to eli0s in Canaray - updated to today's release. I have run the test many times, and about 75% of the time I get mostly 16 and 17 ms. I don't get exclusively whole numbers, more like what TiAm got.

But the remaining 25% of the time I do get a result that looks very much like IE results - centered right on the 16.7 ms line.

I have tried full screen, maximized, windowed... disabled my second screen, closed down all other processes, etc...

I have a Core i7, 8 GB ram, with a very old ATI Radeon HD 2600 XT (and a two year old driver). Windows 7, running on a solid state hard drive.

Ashley

This chrominium bug report you linked to a few pages back describes this very behavior. Here:

https://code.google.com/p/chromium/issu ... ?id=422000

Canary indicates that google is working on this, but the results are inconsistent. At best, it matches or exceeds IE's performance. At times, it seems to regress back to the profile of the current stable. It'll be interesting to see what state it's in when v40 goes beta.

Firefox is just an awful mess in general, but project silk indicates they are aware of this and are trying to correct it.

eli0s

Your cpu may be old, but it's powerful. If it was me I'd try stock clocks and see if the behavior changes. Your gpu is pretty good period...have you updated your graphics drivers lately? Do you notice halting when playing normal PC games?

Tylermon

Wow, now those are some spec's <img src="{SMILIES_PATH}/icon_e_smile.gif" alt=":)" title="Smile"> But your results...that's very strange. I've never had canary perform worse than chrome stable. And IE has always been rock solid. How does firefox do? Generally, it seems to be the worst of the lot.

Chrome/Firefox reminds me of this joke:

"Did you hear about the statistician who had his head in an oven and his feet in a bucket of ice? When asked how he felt, he replied, "On the average I feel just fine."

Ashley

This chrominium bug report you linked to a few pages back describes this very behavior. Here:

https://code.google.com/p/chromium/issu ... ?id=422000

Canary indicates that google is working on this, but the results are inconsistent. At best, it matches or exceeds IE's performance. At times, it seems to regress back to the profile of the current stable. It'll be interesting to see what state it's in when v40 goes beta.

Firefox is just an awful mess in general, but project silk indicates they are aware of this and are trying to correct it.

eli0s

Your cpu may be old, but it's powerful. If it was me I'd try stock clocks and see if the behavior changes. Your gpu is pretty good period...have you updated your graphics drivers lately? Do you notice halting when playing normal PC games?

Tylermon

Wow, now those are some spec's <img src="{SMILIES_PATH}/icon_e_smile.gif" alt=":)" title="Smile"> But your results...that's very strange. I've never had canary perform worse than chrome stable. And IE has always been rock solid. How does firefox do? Generally, it seems to be the worst of the lot.

Chrome/Firefox reminds me of this joke:

"Did you hear about the statistician who had his head in an oven and his feet in a bucket of ice? When asked how he felt, he replied, "On the average I feel just fine."

I dont use firefox anymore exactly because it usually performs the worst of the lot haha.

Once upon a time I loved that fiery fox. Those days are gone.

----------

It is worth mentioning I have just about every flag enabled in my chrome browser and had not done the same with canary...very possible this is why.

the url's

chrome://flags

or

chrome://chrome-urls

are fairly useful and fun to look into.

Here are the bug reports I've submitted. Fingers crossed that they create the desired effect...

https://code.google.com/p/chromium/issu ... 1415074675

https://bugzilla.mozilla.org/show_bug.cgi?id=1093454

Here are the bug reports I've submitted. Fingers crossed that they create the desired effect...

https://code.google.com/p/chromium/issu ... 1415074675

https://bugzilla.mozilla.org/show_bug.cgi?id=1093454

Looks good. Hopefully we see this addressed!

Tweeted that bug report, so hopefully it spreads around!