Benchmark vs Waterfox on Waterfox Homepage

[stravinsky]

who gives a shit about benchmarks ?
99.99% of my time using a browser i dont benchmark. I use to load and view real world, ACTUAL pages. .

Increasingly, i use PM over other FF clones becuase :

1. PM is not a typical clone. it has real hard work done in the direction i like, and which moz dont care about.

2. I can get "out of schedule" patches in the current browser version. Something which moz cant/wont do (read: hierarchy).

3. I fucking dont want a homepage/UI change every 3 versions.

4.MC is a real dev. anybody can compile code with moar switches and PGO. But only a dev can add/remove features and modify code. And i havent yet seen a clone with selective patches. Lets see that MrAlex do that.

No offense MC, but subjectively, all browser based on moz code appear similar to me in speed.
But i prefer PM.
Its not about only speed anymore.

/Rant.

[tribaljet]

I should say that once I tried Pale Moon, I never looked back at Firefox again, but knowing that performance could be far better but isn't due to politically correct manufacturer balance isn't that great of an idea IMHO, but I do respect each person's views on the matter.

Honestly, I do understand Moonchild not wanting to to use ICC, as that would most likely show significant performance improvements on Intel CPUs rather than AMD CPUs, but while that balances out performance between both manufacturers, it certainly holds back higher performance that could be obtained when running specific Intel optimizations. Now, I'm not saying it is a correct attitude from Intel in terms of locking down additional optimizations when the GenuineIntel flag is missing, but now that the 3.6.x branch is reaching EOL status, maybe separate Intel and AMD packages could be made available.

Also, is there an alternative to ICC with the same level of optimizations, and that can bring AMD CPUs to the same performance level of Intel CPUs?

[Moonchild]

... it certainly holds back higher performance that could be obtained when running specific Intel optimizations. ...

You must understand that MSVC also uses specific Intel CPU optimizations, but ICC actively cripples execution on hardware that does not have the "GenuineIntel" identifier in the CPU. It's not a matter of being politically correct or even allowing for more advanced SIMD instruction sets, it's the fact that there is an unfair play on the market that makes me not want to use the compiler.

A second point is that Pale Moon is not "held back" in higher performance considering what is used is very well optimized for Intel CPUs as well as any other manufacturer - the instruction sets used are common for all CPUs that are not over a decade old, and what little extra ICC uses in Waterfox is not going to make much of a difference for a web browser.

not wanting to to use ICC, as that would most likely show significant performance improvements on Intel CPUs rather than AMD CPUs

I wouldn't call the scores posted "significant performance improvements" - Waterfox 14 only just edges ahead of an older code version of Pale Moon.

In fact, there are numerous reports about stability issues with the most recent versions of Waterfox; and in that respect I'd prefer Pale Moon to score just slightly lower in synthetic benchmarks on the specific hardware Waterfox is compiled for, but otherwise being stable and smooth.

[tribaljet]

It's clearly wrong doing on Intel's part, that's a given. My question is, do the Intel CPU optimizations present on MSVC give the same level of performance improvement as the ones present on ICC (for Intel CPUs, I mean)? I ask if you have tested it yourself, as basing on performance numbers of Waterfox or other builds isn't that relevant to me, as you have done quite a lot of work on Pale Moon, and that is what I base myself on for performance considerations.

I don't care for benchmarks, but real world usage, and that is something that I felt Pale Moon to improve over stock Firefox.

[Hoggyy]

Now, I'm not saying it is a correct attitude from Intel in terms of locking down additional optimizations when the GenuineIntel flag is missing, but now that the 3.6.x branch is reaching EOL status, maybe separate Intel and AMD packages could be made available.

Also, is there an alternative to ICC with the same level of optimizations, and that can bring AMD CPUs to the same performance level of Intel CPUs?

I would LOVE the availability of separate Intel and AMD packages. That would be so wonderful - to be able to get the most out of my new generation AMD laptop. (Now that I finally have a machine that ISN'T ~10 years old added to my bunch! )
I would also say using a compiler that actively cripples non-Intel processors would be a bad idea - on principle ALONE.

But then again, I don't know how much is involved in producing separate versions (I'm not a programmer)... I have no idea if it's just different compiler optimizations or also changes in the code itself. If it's just different compiler optimizations, I would say why not have 2 different versions???

As to the conversation about the benchmarking results, I would also be in the camp that doesn't care about them. I just care about real world performance in areas that would be my preferences alone. The differences between PM and Wx can often be a little hard to tell. And I simply love the way PM uses smooth scrolling - it feels so natural compared to Fx/Wx.

[Ryrynz]

I would say why not have 2 different versions???

it's the fact that there is an unfair play on the market that makes me not want to use the compiler.

There's your answer

who gives a shit about benchmarks ?

I do.

[Moonchild]

But then again, I don't know how much is involved in producing separate versions (I'm not a programmer)... I have no idea if it's just different compiler optimizations or also changes in the code itself. If it's just different compiler optimizations, I would say why not have 2 different versions???

Separate versions means twice the amount of testing, more testing and building hardware required (to do it properly you need to build AMD builds on an AMD machine and Intel builds on an Intel machine, because of the way PGO works), twice the amount of web maintenance, and the inherent confusion among less technical people who don't know which version they want or need.
On top, it would be too fine-grained building for what will be, maybe, a few % difference in speed (Hard to tell without actually building and testing).

In short: it would be a LOT of overhead for very little result, and less clarity for the user. I refuse to go down the "overclocker path" where people boast about 5 fps difference in a game that already gives them over 100 - you won't see it.

[tribaljet]

In short: it would be a LOT of overhead for very little result, and less clarity for the user. I refuse to go down the "overclocker path" where people boast about 5 fps difference in a game that already gives them over 100 - you won't see it.

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. And considering the 3.6.x version won't be available anymore, the system requirements for Pale Moon will technically increase, namely in terms of SIMD instructions, making things relatively simple, as a computer either has an AMD or an Intel CPU (currently, that's a single bit of information that's rather accessible to 99% of users), downloading the version that matches the manufacturer's name.

Would two separate branches, Intel and AMD, require more testing than when having 3.6.x and 12.x currently? Also, would it be possible for you to create an experimental build for a few testers to try and post results according to your own specified testing guidelines?

[Moonchild]

we are still closer to 15fps rather than 100, to which a few % will make a difference.

Really now? So you are going to notice the difference between 15 fps and 15.2 fps when playing a game?
Seriously though, I simply don't have the resources in time & hardware (see above). And making an "experimental" build is simply not going help the overall Pale Moon project in any way. If I'm going to invest in a separate build, I need to be able to actually have it matter. I won't start making "private builds" for specific hardware combinations - if you're that serious about getting the few % more specifically for your hardware, then you can grab the source code and build it yourself!

This is probably an area where contributors can look into submitting a contributed binary to officially have released with Pale Moon branding, and go from there if it's really significant.

[tribaljet]

we are still closer to 15fps rather than 100, to which a few % will make a difference.

Really now? So you are going to notice the difference between 15 fps and 15.2 fps when playing a game?
Seriously though, I simply don't have the resources in time & hardware (see above). And making an "experimental" build is simply not going help the overall Pale Moon project in any way. If I'm going to invest in a separate build, I need to be able to actually have it matter. I won't start making "private builds" for specific hardware combinations - if you're that serious about getting the few % more specifically for your hardware, then you can grab the source code and build it yourself!

This is probably an area where contributors can look into submitting a contributed binary to officially have released with Pale Moon branding, and go from there if it's really significant.

.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.

I say experimental build as in beta build for a future stable build for the majority of users, as AMD users are a minority.
EDIT: Beta build for an Intel build, as for all intents and purposes, the AMD build would effectively be the regular build currently available.

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

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.

Out of curiosity, what are your system specs?