3 ms·

SunSpider – Spartan vs. Chrome 41.0.2272.101 – Number's Don't Lie

15 points by alexcasalboni 12y ago

T3RMINATED 12y ago
IE is known to lie on their numbers ... didnt they do that for IE10 lol
VeejayRampay 12y ago
Not sure we can draw conclusions from a screenshot of the numbers, especially in this early phase and for one benchmark, but it's extremely encouraging to see Microsoft's willingness to go head-to-head with its competitors, both in terms of feature range and raw performance. This is one of the many bricks that will help foster a vibrant and lively web programming ecosystem. Good on them and good on us.
gsnedders 12y ago
SunSpider is about one thing and one thing only — cold code performance. The benchmarks are all so short (the longest in Spartan is 19.7ms, albeit with a large uncertainty) that it's all about how quickly you can JIT code and get rid of the dispatch overhead of any interpreter you have, and how many optimizations you can do in that very quick first compilation. If you're interested in longer running code, where implementation quality makes a far bigger difference, look at Octane and Kraken. Chakra, Spartan's JS engine, has long been good on SunSpider.
AshleysBrain 12y ago
Chrome 41 is a real stable release browser in production use. Spartan is still a WIP. I'm not sure it's fair to compare them. How about giving Canary a spin instead?
falcolas 12y ago
Tangent Of course numbers can lie. Numbers have to be compiled and interpreted by humans, who can make them say anything they want the numbers to say. Best recent example: Gerrymandering. The majority of voters want one candidate, but the other candidate is able to draw up regions in such a way where they win, to make the numbers support their lie.
nextw33k 12y ago
In the broader benchmarks it still looks like Firefox or Chrome is the winner: http://www.winbuzzer.com/benchmark-deep-dive-microsoft-windows-10-spartan-browser-vs-ie11-vs-google-chrome-41-vs-mozilla-firefox-36/ http://www.winbuzzer.com/benchmark-deep-dive-microsoft-windo...
- mikhailt 12y ago
  Nearly two months passed since that post, I'm sure they're outdated already. It would be interesting to see them refresh it soon and compare the progress in two months.

15 points by alexcasalboni 12y ago

IE is known to lie on their numbers ... didnt they do that for IE10 lol

Not sure we can draw conclusions from a screenshot of the numbers, especially in this early phase and for one benchmark, but it's extremely encouraging to see Microsoft's willingness to go head-to-head with its competitors, both in terms of feature range and raw performance. This is one of the many bricks that will help foster a vibrant and lively web programming ecosystem. Good on them and good on us.

gsnedders 12y ago

SunSpider is about one thing and one thing only — cold code performance. The benchmarks are all so short (the longest in Spartan is 19.7ms, albeit with a large uncertainty) that it's all about how quickly you can JIT code and get rid of the dispatch overhead of any interpreter you have, and how many optimizations you can do in that very quick first compilation. If you're interested in longer running code, where implementation quality makes a far bigger difference, look at Octane and Kraken. Chakra, Spartan's JS engine, has long been good on SunSpider.

AshleysBrain 12y ago

Chrome 41 is a real stable release browser in production use. Spartan is still a WIP. I'm not sure it's fair to compare them. How about giving Canary a spin instead?

falcolas 12y ago

Tangent Of course numbers can lie. Numbers have to be compiled and interpreted by humans, who can make them say anything they want the numbers to say. Best recent example: Gerrymandering. The majority of voters want one candidate, but the other candidate is able to draw up regions in such a way where they win, to make the numbers support their lie.

nextw33k 12y ago

In the broader benchmarks it still looks like Firefox or Chrome is the winner: http://www.winbuzzer.com/benchmark-deep-dive-microsoft-windows-10-spartan-browser-vs-ie11-vs-google-chrome-41-vs-mozilla-firefox-36/ http://www.winbuzzer.com/benchmark-deep-dive-microsoft-windo...

mikhailt 12y ago
Nearly two months passed since that post, I'm sure they're outdated already. It would be interesting to see them refresh it soon and compare the progress in two months.