The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.
Summary
The following documents A-B testing comparing performance of 32-bit Firefox builds to 64-bit Firefox builds, utilizing the Massive benchmark.
Methodology
- Test using the same system, make note of the system profile
- Install the latest Firefox Nightly 32-bit build and start with a new profile
- Load http://kripken.github.io/Massive/ and click "Run the benchmark now!"
- Let the benchmark run
- Record the results below when the benchmark completes
- Include a screenshot of the result
- Repeat the benchmark 5 times on the same system and record the result separately
- Install the latest Firefox Nightly 64-bit and repeat the tests with a new profile
System Profile
| Operating System
|
Windows 7 64-bit
|
| Firefox
|
(user agent string with build ID) Mozilla/5.0 (Windows NT 6.1; WOW64; rv:32.0) Gecko/20100101 Firefox/32.0, 20140604030202
Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:32.0) Gecko/20100101 Firefox/32.0, 20140604030202
|
| Processor
|
(make, model, speed) AMD FX(tm)-8320 Eight-Core Processor 3.50 Ghz
|
| Memory
|
(type, speed, amount) 8 GB DDR3 1333 Mhz
|
| GPU
|
(make, model, speed, memory, driver version) AMD Radeon R7 200 Series, Core Clock 780 Mhz, 2GB DDR3 900 Mhz, Catalyst version 13.12
|
Results
Testrun #1
| main-thread-poppler-cold
|
0.452
|
0.348
|
| main-thread-poppler-warm
|
0.428
|
0.372
|
| main-thread-sqlite-cold
|
0.053 (±26%!)
|
0.062 (±25%!)
|
| main-thread-sqlite-warm
|
0.033 (±24%!)
|
0.048 (±29%!)
|
| box2d-throughput
|
6.838
|
4.882
|
| box2d-throughput-f32
|
6.316
|
4.567
|
| lua-binarytrees
|
8.067
|
7.447
|
| lua-scimark
|
10.545
|
11.730
|
| poppler-throughput
|
7.008
|
6.445
|
| sqlite-throughput
|
10.891
|
8.773
|
| poppler-cold-preparation
|
0.939
|
0.926
|
| poppler-warm-preparation
|
0.185
|
0.185
|
| sqlite-cold-preparation
|
1.084
|
1.347
|
| sqlite-warm-preparation
|
0.098
|
0.123
|
| box2d-variance
|
1.466
|
1.261
|
| poppler-variance
|
4.843
|
4.395
|
Testrun #2
| main-thread-poppler-cold
|
0.461
|
0.367
|
| main-thread-poppler-warm
|
0.427
|
0.361
|
| main-thread-sqlite-cold
|
0.065
|
0.064
|
| main-thread-sqlite-warm
|
0.040 (±30%!)
|
0.043 (±28%!)
|
| box2d-throughput
|
6.987
|
4.666
|
| box2d-throughput-f32
|
6.317
|
4.277
|
| lua-binarytrees
|
7.899
|
7.182
|
| lua-scimark
|
10.555
|
12.055
|
| poppler-throughput
|
7.023
|
5.959
|
| sqlite-throughput
|
10.733
|
8.485
|
| poppler-cold-preparation
|
0.953
|
0.846
|
| poppler-warm-preparation
|
0.186
|
0.180
|
| sqlite-cold-preparation
|
1.103
|
1.209
|
| sqlite-warm-preparation
|
0.109
|
0.115
|
| box2d-variance
|
1.758
|
1.290
|
| poppler-variance
|
4.830
|
4.036
|
Testrun #3
| main-thread-poppler-cold
|
0.456
|
0.338
|
| main-thread-poppler-warm
|
0.445
|
0.356
|
| main-thread-sqlite-cold
|
0.050
|
0.063
|
| main-thread-sqlite-warm
|
0.034 (±24%!)
|
0.042 (±28%!)
|
| box2d-throughput
|
6.806
|
4.643
|
| box2d-throughput-f32
|
6.316
|
4.284
|
| lua-binarytrees
|
8.002
|
7.239
|
| lua-scimark
|
10.250
|
12.185
|
| poppler-throughput
|
7.034
|
6.003
|
| sqlite-throughput
|
10.759
|
8.366
|
| poppler-cold-preparation
|
0.937
|
0.842
|
| poppler-warm-preparation
|
0.183
|
0.173
|
| sqlite-cold-preparation
|
1.071
|
1.203
|
| sqlite-warm-preparation
|
0.110
|
0.121
|
| box2d-variance
|
1.467
|
1.210
|
| poppler-variance
|
4.836
|
4.083
|
Testrun #4
| main-thread-poppler-cold
|
0.448
|
0.359
|
| main-thread-poppler-warm
|
0.451
|
0.362
|
| main-thread-sqlite-cold
|
0.056 (±22%!)
|
0.071 (±31%!)
|
| main-thread-sqlite-warm
|
0.033 (±22%!)
|
0.042 (±26%!)
|
| box2d-throughput
|
6.834
|
4.674
|
| box2d-throughput-f32
|
6.317
|
4.304
|
| lua-binarytrees
|
8.219
|
7.226
|
| lua-scimark
|
9.925
|
12.005
|
| poppler-throughput
|
7.021
|
5.951
|
| sqlite-throughput
|
10.768
|
8.414
|
| poppler-cold-preparation
|
0.953
|
0.840
|
| poppler-warm-preparation
|
0.184
|
0.173
|
| sqlite-cold-preparation
|
1.091
|
1.212
|
| sqlite-warm-preparation
|
0.120
|
0.140
|
| box2d-variance
|
1.546
|
2.258
|
| poppler-variance
|
4.832
|
4.067
|
Testrun #5
| main-thread-poppler-cold
|
0.453
|
0.360
|
| main-thread-poppler-warm
|
0.453
|
0.351
|
| main-thread-sqlite-cold
|
0.047 (±21%!)
|
0.060 (±27%!)
|
| main-thread-sqlite-warm
|
0.038 (±26%!)
|
0.052 (±40%!)
|
| box2d-throughput
|
7.161
|
4.666
|
| box2d-throughput-f32
|
6.589
|
4.292
|
| lua-binarytrees
|
8.364
|
7.230
|
| lua-scimark
|
9.820
|
12.195
|
| poppler-throughput
|
7.596
|
5.989
|
| sqlite-throughput
|
11.282
|
8.583
|
| poppler-cold-preparation
|
1.035
|
0.844
|
| poppler-warm-preparation
|
0.199
|
0.177
|
| sqlite-cold-preparation
|
1.232
|
1.236
|
| sqlite-warm-preparation
|
0.112
|
0.137
|
| box2d-variance
|
1.967
|
1.297
|
| poppler-variance
|
5.149
|
4.086
|
Overall
| Firefox 32.0a1 32-bit
|
3,762
|
3,967
|
3,869
|
3875.8
|
| Firefox 32.0a1 64-bit
|
4,120
|
4,323
|
4,225
|
4233.2
|
Best score is highlighted
Notes
The benchmark freezes very often at the "sqlite-warm-preparation" test. I had to right click on the page, or to select a word, open/close a new tab, or stuff like that to make it "refresh" and continue
Conclusions
- Overall, 64-bit builds seem to perform about 9% better than 32-bit builds
- 32-bit builds perform about 7% to 21% better on sqlite benchmarks