Question How fast is your browser?

Page 12 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
231
106
This benchmark simulates user actions for adding, completing, and removing to-do items using multiple examples in TodoMVC. Each example in TodoMVC implements the same todo application using DOM APIs in different ways. Some call DOM APIs directly from ECMAScript 5 (ES5), ECMASCript 2015 (ES6), ES6 transpiled to ES5, and Elm transpiled to ES5. Others use one of eleven popular JavaScript frameworks: React, React with Redux, Ember.js, Backbone.js, AngularJS, (new) Angular, Vue.js, jQuery, Preact, Inferno, and Flight. Many of these frameworks are used on the most popular websites in the world, such as Facebook and Twitter. The performance of these types of operations depends on the speed of the DOM APIs, the JavaScript engine, CSS style resolution, layout, and other technologies.

Core M-5Y10c @ 52.02

Images aren't necessary, but please state your cpu speed. The web browser of your choice. Mine is Edge 83. Thank you.

speed.pngspeed2.png
 
Last edited:
  • Like
Reactions: SPBHM

moinmoin

Diamond Member
Jun 1, 2017
4,944
7,656
136
Never posted my old iPad 2 (the original one, running the latest iOS 9), just for giggles now:

img_0743mqc47.png
 

Doug S

Platinum Member
Feb 8, 2020
2,254
3,492
136
That's why you want to see results for older stuff using the latest OS/browser to insure you are capturing all possible software side optimizations including compiler improvements. So I think it is valuable to see ikjadoon's iPhone 12 result because it would allow a direct comparison to A17P in an iPhone 15 Pro as well as the M3 result posted above.

Which makes me really curious to see an iPhone 15 Pro result, because seeing an M3 iMac (with a very slightly older version of Safari) scoring 2.4x better than an iPhone 12 is kind of shocking. Is the M3 that much faster than an iPhone 15 Pro that's basically the same hardware modulo 10% higher clock rates, more cores and memory bandwidth? Or has Apple made some strides in browser performance (at least as measured by Speedometer, and I have no idea exactly what it is measuring) between A14 and A17P far greater than more general benchmarks indicate? If so, I wonder what might account for it.
 
Last edited by a moderator:

sallymander

Junior Member
Nov 20, 2020
12
30
61
I got the latest software and did a re-test on my M3 iMac:

Version 17.2 (19617.1.17.11.9)

Screenshot 2023-12-19 at 10.59.32.png

I've also got an iPhone 15 Pro Max (on iOS 17.2):

Screenshot 2023-12-19 at 10.59.50.png

I guess there could be some difference in thermal throttling? It doesn't seem to be a test with any pauses in it, so perhaps the active cooling on the iMac can sustain higher clocks for the whole test.
 

Doug S

Platinum Member
Feb 8, 2020
2,254
3,492
136
I guess there could be some difference in thermal throttling? It doesn't seem to be a test with any pauses in it, so perhaps the active cooling on the iMac can sustain higher clocks for the whole test.

How long does the test take? I don't know how long it would take an iPhone to begin throttling but I suppose it would have to be a fairly quick test to avoid it.

Regardless it is impressive that from A14 -> A17P performance almost doubled with both running iOS 17.2, versus Geekbench only increasing by around 50%. Given how much time people spend on the web that's a pretty significant improvement (assuming that Speedometer is representative of general web use, something I have no idea whether is true or not...)

EDIT: I found it (I was looking in the app store, duh...) and tried it on my 14 Pro Max. Took about 18 seconds and got 442. Phone is barely warm to the touch, hard to imagine it was throttling. Maybe I'll toss it in the fridge when I go to the gym and see if I get a better score running at 40F when I get back lol
 
Last edited:

Panino Manino

Senior member
Jan 28, 2017
820
1,022
136
On my Skylake Dual Core Mobile, I tested the Vivaldi Snapshot I was using because of video acceleration, I think it's version 5.8 and the current Stable.

Went from 56 to 80, a noticeable diference.
 

Doug S

Platinum Member
Feb 8, 2020
2,254
3,492
136
Tried it again with my 14 Pro Max after sitting in the fridge on top of some ground beef (to maximize heat transfer) and got 1 point higher. I wonder if iPhone SoCs will lower frequencies based on core runtime (i.e. always lower frequency to x after 2 seconds, to y after 15 seconds, etc.) regardless of measured temperature. Though I suppose even though my phone started at around 40F that doesn't preclude the possibility that a small area of a core heats up unacceptably even though it felt ice cold even after running the benchmark 10 times in a row as fast as I could.
 
  • Like
Reactions: lightmanek

moinmoin

Diamond Member
Jun 1, 2017
4,944
7,656
136
When M1 first launched we discussed the possibility that Apple's chips don't have the kind of boost modes we have grown accustomed to on PC. I think it's likely that the chips have little more than hardcoded boost tables which are furthermore adapted to their use case (so significantly lower in phones).
 

Doug S

Platinum Member
Feb 8, 2020
2,254
3,492
136
Or the iPhone SoC is artificially prevented from hitting max perf to avoid draining the battery and also to prevent it from overwhelming the cooling solution.

"Max perf" is whatever the maximum clock rate observed for it. Apple doesn't publish this, people figure it out running i.e. GB6 and seeing what clock rate it reports. Unless there's something funky about the Speedometer test, I would assume it reaches the same maximum clock rate that GB6 observes.

If the battery falls below 20% it will go into a low power mode, but absent that it should perform the same whether at 25% or 100%.
 

ikjadoon

Member
Sep 4, 2006
113
155
126
This is an interesting discussion. I have some more data points with my M1 MacBook Air.

M1 (4+4) vs A14 (2+4):
  • 75% to 100% larger GPU
  • Doubled memory controllers
  • 50% larger L2 cache
  • Doubled P-cores
  • Slightly higher clocks
  • Separate RAM packaging
  • 50% less system cache (some say equal, some say 25% less)

Device & OSBrowser & VersionScoreVariation +/-
iPhone 12 Pro
(iOS 17.2)
Safari 17.225015
M1 MacBook Air
(macOS 14.2.1)
Safari 14.2.144031
M1 MacBook Air
(macOS 14.2.1)
Arc 1.21.1 (44329)25418
M1 MacBook Air
(macOS 14.2.1)
Edge 120.0.2210.913117.7
M1 MacBook Air
(macOS 14.2.1)
Firefox 121.024011

Safari scoring really well on macOS seems to imply strong browser optimizations (either in the browser alone or the browser exploiting M1 / Apple Silicon) or maybe improved L2 cache, as IIRC, Speedometer 2.0 isn't very multi-threaded (at least beyond a certain #; eg here with Threadripper 16C vs 32C).

The scores can bounce around; Edge was first 29x something, but I forgot to add the version number. Re-running it (only difference was closing a second Edge window, IIRC) and it jumped to 311, so the variation estimate provided on the site may not be comprehensive.

Comparing like for like (M1 vs M3 on macOS Safari 17.2.x) shows a somewhat more reasonable leap (+41%) which aligns much closer to the Geekbench 6 1T jump (something like +30%).
 

ikjadoon

Member
Sep 4, 2006
113
155
126
Two notes:

  • I've just realized there is a Speedometer 2.1 (direct link) and Speedometer 2.0 (linked in the OP & Google). Because 2.0 is so prevalent, I think virtually everyone has tested with 2.0, but it'd be great if Speedometer added a version # to the final results. I can't find the differences of 2.1 vs 2.0, but if people clicked the link in the Speedometer GitHub repo, they may inadvertently be on the 2.1 version.
  • I forgot: Speedometer 3.0 is in active development, though it's been 1+ year, so it still may be a while longer. Google claimed they'd share updates "in the coming months", but I guess 3.0 needs more time to bake (you can test it).
 
  • Like
Reactions: igor_kavinski

ikjadoon

Member
Sep 4, 2006
113
155
126
On the second PC. I don't know what all is important to Speedometer, but the hopefully enough: i7-12700K (MCE enabled, no OC), 32GB DDR5-6400 CL32, and a 2TB 990 PRO.

Device & OSBrowser & VersionScoreVariation +/-
i7-12700K system
Win 11 Pro 23H2 / 22631.2861
Microsoft Edge 120.0.2210.913106.5
i7-12700K system
Win 11 Pro 23H2 / 22631.2861
Google Chrome 120.6099.13039111
i7-12700K system
Win 11 Pro 23H2 / 22631.2861
Mozilla Firefox 121.024014

Comparing the best scores, the i7-12700K on Chrome is ~11% slower than the M1 Air on Safari and 56% faster than the A14 on Safari.