{"id":2670,"date":"2013-09-12T16:22:14","date_gmt":"2013-09-12T05:22:14","guid":{"rendered":"http:\/\/blog.mozilla.org\/nnethercote\/?p=2670"},"modified":"2013-09-12T16:45:46","modified_gmt":"2013-09-12T05:45:46","slug":"bleg-for-a-new-machine-2","status":"publish","type":"post","link":"https:\/\/blog.mozilla.org\/nnethercote\/2013\/09\/12\/bleg-for-a-new-machine-2\/","title":{"rendered":"Bleg for a new machine: outcome"},"content":{"rendered":"

Recently I blegged (here<\/a> and here<\/a>) for help in designing a new machine.\u00a0 My goals:\u00a0 fast browser and JS shell builds, quietness, and a setup that wasn’t too complicated.\u00a0 I now have the new machine and have done some comparisons to the old machine.<\/p>\n

New vs old<\/h3>\n

The most important components of the new machine are:\u00a0 an Intel i7-4770 CPU (I’m using the integrated graphics), 32 GiB of RAM, a 512GB Samsung 840 Pro SSD hard disk, and a Fractal Design Define R4 case.<\/p>\n

In comparison, the equivalent components in the old machine were: an Intel i7-2600 CPU, 16 GiB of RAM, a magnetic hard disk, and an Antec Sonata III 500 case.<\/p>\n

A basic comparison<\/h3>\n

The new machine is definitely faster.\u00a0 Compile times are about 1.5x faster;\u00a0 I can do a debug browser build with clang in under 13 minutes, and one with GCC in under 17 minutes.\u00a0 (I hadn’t realized that clang was so much faster than GCC.)<\/p>\n

Furthermore, disk-intensive operations are massively faster.\u00a0 Just as importantly, disk-intensive operations vary in speed much less.\u00a0 With a magnetic disk, if you’re doing something where the data is already in the disk cache, it’ll be pretty fast;\u00a0 otherwise it’ll be horribly slow.\u00a0 The SSD doesn’t suffer that dichotomy.<\/p>\n

Finally, the new case, while not silent, is certainly quieter… maybe half as loud as the old one.\u00a0 It’s also<\/em> bigger than I expected — it’s 1–2 inches bigger in every dimension than the old one. There must be a lot of empty space inside.\u00a0 And although it has a pleasingly minimalist aesthetic — it’s about as plain a black box as you could imagine — it does have an obnoxiously bright, blue power indicator light at the top of the front panel, which I quickly covered with a small strip of black electrical tape.<\/p>\n

A detailed performance comparison<\/h3>\n

Building and testing<\/h4>\n

All builds are 64-bit debug builds.\u00a0 I used clang 3.2-1~exp9ubuntu1<\/code> and gcc-4.7.real (Ubuntu\/Linaro 4.7.3-1ubuntu1)<\/code> for compilation. I measured each operation only once, and the old machine in particular would vary in its times due to the magnetic disk.\u00a0 So don’t treat individual measurements as gospel.\u00a0 In all cases I give the old machine’s time first.<\/p>\n

    \n
  • Browser clobber build (clang): 19.7 minutes<\/strong> vs 12.7 minutes<\/strong> (1.56x faster<\/strong>).\u00a0 I didn’t measure a GCC brower build on the old machine, but on the new machine it was 16.8 minutes<\/strong> (1.32x slower<\/strong> than clang).<\/li>\n
  • Browser no-change build (clang): 48 seconds<\/strong> vs 31 seconds<\/strong> (1.55x faster<\/strong>).<\/li>\n
  • Browser clobber build, with ccache, with an empty cache (clang): 23.3 minutes<\/strong> vs 14.8 minutes<\/strong> (1.57x faster<\/strong>).\u00a0 These are 1.18x slower<\/strong> and 1.17x slower<\/strong> than the corresponding non-ccache builds.<\/li>\n
  • Browser clobber build, with ccache, with a full cache (clang): 6.2 minutes<\/strong> vs 2.6 minutes<\/strong> (2.4x faster<\/strong>).\u00a0 These are 3.18x faster<\/strong> and 4.89x faster<\/strong> than the corresponding non-ccache builds.\u00a0 Here the effect of the SSD becomes clear — the new machine gets a much bigger benefit from ccache.<\/li>\n
  • Two concurrent browser builds (clang): 45.9 & 45.4 minutes<\/strong> vs 22.5 & 22.5 minutes<\/strong> (2.03x faster<\/strong>).\u00a0 Interestingly, the amortized per-build time on the old machine (22.9 minutes<\/strong>) was 1.16x slower<\/em><\/strong> than a single build, but the amortized per-build time on the new machine (11.3 minutes<\/strong>) was 1.12x faster<\/em><\/strong> than a single build.\u00a0 The new machine, despite having the same number of cores, clearly provides more parallelism, and a single browser build doesn’t take full advantage of that parallelism.<\/li>\n
  • JS shell everything-but-ICU build (clang): 59 seconds<\/strong> vs 42 seconds<\/strong> (1.4x faster<\/strong>).\u00a0 It’s worth noting that JS shell builds spend a higher proportion of their time doing C++ compilation than browser builds.<\/li>\n
  • JS shell everything-but-ICU build (GCC): 130 seconds<\/strong> vs 81 seconds<\/strong> (1.60x faster<\/strong>).\u00a0 These are 2.20x slower<\/strong> and 1.93x slower<\/strong> than the corresponding clang builds!<\/li>\n
  • JS jit tests (compiled with clang): 179 seconds<\/strong> vs 137 seconds<\/strong> (1.31x faster<\/strong>).\u00a0 These tests are much more CPU-bound and less disk-bound than compilation, so the smaller speed up isn’t surprising.<\/li>\n
  • SunSpider: 156 ms<\/strong> vs 127 ms<\/strong> (1.23x faster<\/strong>).\u00a0 Again, CPU is the main factor.<\/li>\n<\/ul>\n

    Next, here are the times for some disk-intensive operations.\u00a0 The results here, especially for the old machine, could be highly variable.<\/p>\n

      \n
    • Delete a build directory: 10.5 seconds<\/strong> vs 1.4 seconds<\/strong> (7.5x faster<\/strong>).<\/li>\n
    • Do a local clone of mozilla-inbound: 7.7 minutes<\/strong> vs 10 seconds<\/strong> (46x faster<\/strong>).<\/li>\n
    • Recursive grep of .cpp\/.h\/.idl files in a repository, first time: 53.2 seconds<\/strong> vs 0.8 seconds<\/strong> (67x faster<\/strong>).<\/li>\n
    • The same operation, immediately again: 0.2 seconds<\/strong> vs 0.2 seconds<\/strong> (same speed<\/strong>).<\/li>\n<\/ul>\n

      Those last two comparisons really drive home the impact of the SSD, and the reduction in variability it provides. It’s hard to describe how pleasing this is.\u00a0 On the old machine I always knew when libxul.so was linking, because my whole machine would grind to a halt and trivial operations like saving a file in vim would take multiple seconds.\u00a0 I don’t have that any more!<\/p>\n

      And this is relevant to ccache, too.\u00a0 I tried ccache again recently on my old machine, and while it did speed up compilations somewhat, the extra load on the disk noticeably affected everything else — I had even more of those unpredictable pauses when doing anything other than building.\u00a0 This was annoying enough that I disabled it.\u00a0 But ccache should be much more attractive on the new machine.\u00a0 I will try it again soon, once I’ve had the new machine long enough that I will be well-attuned to its performance.<\/p>\n

      Conclusion<\/h4>\n

      The CPU is a decent improvement over the old one.\u00a0 It accounts for roughly half the improvement in build times.<\/p>\n

      The SSD is a fantastic improvement over the old one.\u00a0 It too accounts for roughly half the improvement in build times, but makes disk-intensive operations much<\/em> faster.\u00a0 It’s performance is also much less variable and thus more predictable.<\/p>\n

      clang is up to 2x faster than GCC!\u00a0 This surprised me greatly.\u00a0 I’d be interested to hear if others have seen such a large difference.<\/p>\n

      Thanks again to everybody who helped me design the new machine.\u00a0 It’s been well worth the effort!<\/p>\n","protected":false},"excerpt":{"rendered":"

      Recently I blegged (here and here) for help in designing a new machine.\u00a0 My goals:\u00a0 fast browser and JS shell builds, quietness, and a setup that wasn’t too complicated.\u00a0 I now have the new machine and have done some comparisons to the old machine. New vs old The most important components of the new machine […]<\/p>\n","protected":false},"author":139,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"_links":{"self":[{"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/posts\/2670"}],"collection":[{"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/users\/139"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/comments?post=2670"}],"version-history":[{"count":0,"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/posts\/2670\/revisions"}],"wp:attachment":[{"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/media?parent=2670"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/categories?post=2670"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.mozilla.org\/nnethercote\/wp-json\/wp\/v2\/tags?post=2670"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}