Benchmark vs Waterfox on Waterfox Homepage

[stravinsky]

What surprised me is that the Yahoo! Mail was loading as fast as it does to me in Chromium-based browsers and Opera, astonishing, but it seems to render too fast that it actually starts to choke and instead uses pauses, noticeable in the loading bar.

??

The first thing I noticed was that Waterfox felt smoother and more responsive than Pale Moon, smoothness being the relevant point to take from this, as responsiveness was different but the gap was rather narrow. In terms of actual browsing performance, it did feel ever so slightly faster than Pale Moon

lets not compare PM12.3 and WF 14.
lets compare PM15 and WF15.

AMD CPUs have both single and multithreaded performance inferior to Intel CPUs, except on the most absurdly (special focus on absurdly) multithreaded scenarios, which exclude all consumer usage, browsers included, which is quite relevant when considering that browsers are barely dual threaded

this is dependent on the workloads. newer AMD processors are exceptionally good in integer heavy workloads. But in floating-point loads, they are worse than PhenomII's.

I thought MC did a little bit of that with moving the add-on stuff (or something) to a separate process.

Not MC, but moz devs did that in the mozilla source code. MS used that to make his PM.

Maybe MC could work on that himself since Mozilla seems to be taking their sweet old time on that. That would also further set PM apart. However, I fear that that could be a monumental undertaking.

its not monumental, its fucking impossible. . Mozilla tried to redesign firefox to be true multiprocessing, but they gave up the whole idea as too difficult and too long-term to implement. You cant really expect one dev (MC) to work on a problem that a shipfull of devs have given up on.
none of the browsers is true multiprocessing. Chrome can open each tab in a different process, but thats not true multiprocessing.
And firefox is worse than chrome,Opera and even IE in this regard.
And the devs are only working around this problem, without actually trying to implement multiprocessing. Dont expect any serious improvement for a few years.

[tribaljet]

Mozilla is working with a possibility like that on Project Snappy, and it might end up making different threads for several things.

AMD's latest offerings are still very performing. Like Intel, they actually are getting too fast for applications requirements. The only thing holding back applications was the HDD, processor is peaking too much. This is my speculation and AMD's new vision, ever since they called it quits on competing with Intel.

Is Project Snappy to browsers as Project Butter is to Android?

AMD is currently the sole manufacturer with strong CPU-IGP (with GPU-like performance) on the market, and they should focus on that, as both HTPC, AiO and, to a lesser extent, some server market usage is where they currently stand real competition.

EDIT: Intel has already made significant strides (compared to their previous endeavors) on the Intel HD 3000 contained on many mid to high end Sandy Bridge CPUs and even moreso with the Intel HD 4000 contained on mid to high end Ivy Bridge CPUs, up to the point where their systems can work entirely through integrated components and be able to handle workloads that go beyond basic browsing and domestic usage, now being able to play demanding 3D games with reasonable fps and even handle general compute tasks. So, at the time of writing (and likely for the foreseeable future), Intel CPUs will continue to have a significant upper hand over AMD CPUs, while Intel IGPs will continue to be underperforming when compared to AMD IGPs.

Curiously, I've heard some rumours (totally unsubstantiated, but entirely different from FUD) that there was some intention from ARM to eventually purchase a large chunk of AMD's shares, which could actually prove to turn around the PC consumer market quite a bit. Let's check current facts, AMD's market share is dropping significantly on the past 3 years, both due to AMD's '09 Phenom II being able to outperform their latest CPUs on several usage scenarios as well as the unexpected raw improvement of the Core i series over the Core 2 Duo/Quad architecture, with Sandy Bridge being the best example of that, quite literally mopping the floor with just about all AMD CPUs. The thing is, the IT world desperately needs AMD to release to the wild a true i7 killer, as competition is starting to get smaller and smaller, and despite what some people might think, the desktop will never disappear as mobile computer will always lack raw processing power that can only be found on desktops and beyond.

[megaman]

Is Project Snappy to browsers as Project Butter is to Android?

AMD will continue to keep their budget line with some worth and performance, it is a goal, even if it takes years. I, for one, will never stop being an AMD fanboy, regardless of facts.

Electrolysis was the first output that was being worked on but rejected later on, but Project Snappy might do a similar effect just not with many threads, just a particular few.

[tribaljet]

Is Project Snappy to browsers as Project Butter is to Android?

AMD will continue to keep their budget line with some worth and performance, it is a goal, even if it takes years. I, for one, will never stop being an AMD fanboy, regardless of facts.

Electrolysis was the first output that was being worked on but rejected later on, but Project Snappy might do a similar effect just not with many threads, just a particular few.

This? https://wiki.mozilla.org/Electrolysis
Why was it rejected?

I tend to go with higher performing parts as well as parts that abide by more stringent standards, which AMD has been lacking for years, and recently on the GPU side as well (since they acquired ATI), not adhering to standards in order to cheat themselves to obtain better compute performance, which is fine until numbers stop matching and everything falls apart. Still, if I didn't game at all and did a lot of compute tasks, I would look at current generation's Radeon GPUs. On the other hand... current generation GPUs from both major manufacturers are obscenely overpriced, not worth the asking price at all, not to mention how there's talk of manufacturers intentionally crippling performance to better adjust the market. Ok, I'm going to stop now or bleeps might surface.

[Moonchild]

.2 fps? No. 5 fps? Most definitely. You seem to consider the improvements marginal at best, even on the content type where the improvements would shine most.

A few % of 15 is in the order of 0.2, not 5.
You will not get a 25% speed increase with ICC compared to MSVC, or Waterfox would be a lot faster in its own benchmarks. (and likely you'll see a significant drop when running it on AMD).

And again, how different is the performance boost of the MSVC's Intel optimizations when compared to the Intel optimizations found on ICC?

By the looks of it, very minimal! Which is my whole point.
Apart from adding code for later SIMD revisions (which likely aren't used at all in a web browser), and a few fringe optimizations you can do for EM64T on 64-bit, all of the PC processors still have the same basic core instruction set. "x86" is a designation of an Intel architecture, after all (from the 80x86 numerical chip designation) so it's not like you're talking massively different hardware here where one build would factually have to emulate other hardware, for example.

Also, I haven't seen that much in terms of contributed builds around this project, like the Linux port some people seem to be curious about.

If you want to talk about a real minority, then you got one in Linux users (which is, what, 1.5% of PC users around the world?)
And maybe there aren't many contributed builds because Pale Moon as-published does a good job already?

[stravinsky]

You will not get a 25% speed increase with ICC compared to MSVC, or Waterfox would be a lot faster in its own benchmarks. (and likely you'll see a significant drop when running it on AMD).

@ tribaljet : you cant get 25% increase by changing compilers for browsers . you can get > 30% improvements for CPU bound applications like 3D renders or content creation programs. But depending that much on a compiler could also mean sloppy coding.

[tribaljet]

.2 fps? No. 5 fps? Most definitely. You seem to consider the improvements marginal at best, even on the content type where the improvements would shine most.

A few % of 15 is in the order of 0.2, not 5.
You will not get a 25% speed increase with ICC compared to MSVC, or Waterfox would be a lot faster in its own benchmarks. (and likely you'll see a significant drop when running it on AMD).

And again, how different is the performance boost of the MSVC's Intel optimizations when compared to the Intel optimizations found on ICC?

By the looks of it, very minimal! Which is my whole point.
Apart from adding code for later SIMD revisions (which likely aren't used at all in a web browser), and a few fringe optimizations you can do for EM64T on 64-bit, all of the PC processors still have the same basic core instruction set. "x86" is a designation of an Intel architecture, after all (from the 80x86 numerical chip designation) so it's not like you're talking massively different hardware here where one build would factually have to emulate other hardware, for example.

Also, I haven't seen that much in terms of contributed builds around this project, like the Linux port some people seem to be curious about.

If you want to talk about a real minority, then you got one in Linux users (which is, what, 1.5% of PC users around the world?)
And maybe there aren't many contributed builds because Pale Moon as-published does a good job already?

In all fairness, you're the one who mentioned 15, not me

I did consider that the rather perceptible performance improvement of Waterfox might be also due to being based on a newer Firefox build. And indeed CPUs are relatively similar nowadays, based on a core architecture that's rather shared between manufacturers. Regarding ICC's optimizations, I find it odd that other compilers are limited to SIMD revisions made almost a decade ago. Are there licensing fees associated with anything beyond SSE3? (even SSSE3)

Well, Linux users are increasing quite a bit, but while talking about solely about standard Windows users, numbers do add up in terms of x manufacturer market share. And hey, for me to be sticking with Pale Moon is indeed because it has a special brand of pixie dust that I haven't found on any other browser

In any case, I'm quite eager to see the responsiveness of the new Pale Moon, as if it shares what current Waterfox has, it will be quite pleasant to use (even moreso).

[Hoggyy]

lets not compare PM12.3 and WF 14.
lets compare PM15 and WF15.

I've seen this said a few times IIRC. What I keep on wondering is why doesn't PM implement the code that would make any speed improvements immediately available to PM, albeit just not with the ridiculous major-version hopping as with Fx lately?

And on the responsiveness issue... I've been starting to notice that PM remains pretty much just as responsive even if kept running for several days, whereas I just had to restart Wx14 because it was just so sluggish when I woke up today. The problem used to be related to All-in-one Sidebar I thought, but it seemed to have gotten fixed lately.. However, noticing the sluggishness in Wx14 just now, maybe there's still a problem there. But yet I haven't had to restart PM since I don't know when.

Restarting can be a major pain in the butt here since I have so many tabs open and Tree Style Tab takes a while to get the trees sorted out - even though I don't have it load all tabs on start up. So another +1 to PM! That makes 2 now - 1 for the sluggishness issue and 1 for the smooth scrolling niceties.

[Moonchild]

In all fairness, you're the one who mentioned 15, not me

Actually... http://forum.palemoon.org/viewtopic.php?f=15&t=1109&start=10#p6299:

The thing is that despite all browsers and respective versions currently avalable, we are still closer to 15fps rather than 100, to which a few % will make a difference.

Otherwise if you want a smoother feel to the browser, try PM12.3r2 which is up on the WIP page and which I will publish as official update soon (once I have time to mangle the website and update the browser to my mirrors). A funny side note is that the subjective smoothness and responsiveness of that browser is much improved, but at the same time any benchmarking done on it (dromaeo, peacekeeper) shows no improvement at all -- a prime example of the fact that currently available benchmarks do not, at all, give a complete picture. Note that it is a different build from the same source code with no changes in code, merely in the build process.

[tribaljet]

In all fairness, you're the one who mentioned 15, not me

Actually... http://forum.palemoon.org/viewtopic.php?f=15&t=1109&start=10#p6299:

The thing is that despite all browsers and respective versions currently avalable, we are still closer to 15fps rather than 100, to which a few % will make a difference.

Otherwise if you want a smoother feel to the browser, try PM12.3r2 which is up on the WIP page and which I will publish as official update soon (once I have time to mangle the website and update the browser to my mirrors). A funny side note is that the subjective smoothness and responsiveness of that browser is much improved, but at the same time any benchmarking done on it (dromaeo, peacekeeper) shows no improvement at all -- a prime example of the fact that currently available benchmarks do not, at all, give a complete picture. Note that it is a different build from the same source code with no changes in code, merely in the build process.

I didn't express myself correctly, as the % mentioned was related to 100, but I get what you understood

And like I said previously, I first used Waterfox and only after regular usage did I run a couple benchmarks on it, not the other way around. I don't find synthetic benchmarks to be actual performance indicators.

Are there any compatibility issues by installing 12.3r2 over 12.3 and/or will there be issues with updating from 12.3r2 to 15?

EDIT: 12.3r2 is x86 only? If it is, then I'll continue using 12.3 as I'm using x64 and have no desire to shift to x86.