Benchmarking Rails on the Retina MacBook Pro
My new Retina MacBook Pro arrived today. The display is gorgeous, but smooth text doesn’t directly make me more productive. Does the new Retina MacBook Pro provide a real workflow win over my beloved Mid-2007 2.2 Ghz Core 2 Duo?
I’m big on benchmarks, so I broke out a feature from the Cucumber suite from a current project and timed Rails.
The “hot” numbers are after running the tests a few times to warm up the unified buffer cache. The “cold” numbers are directly after a computer restart.
Here’s how long it took to start up the test suite:
| Model | Startup, Hot (sec) |
Startup, Cold (sec) |
| Mid-2007 15-inch MacBook Pro 2.2 GHz Core 2 Duo 6GB 320GB 7200rpm HD — Mac OS X 10.6 |
24.7 | 38.3 |
| Mid-2009 15-inch MacBook Pro 2.53 GHz Core 2 Duo 4GB 250GB 5400rpm HD — Mac OS X 10.7 |
19.3 | 35.9 |
| Retina, Mid-2012 15-inch MacBook Pro 2.3 GHz Core i7 (Ivy Bridge) 8GB 256MB SSD — Mac OS X 10.7 |
10.4 | 12.2 |
Here’s how long the 4 scenarios in the feature took to run, after startup:
| Model | Feature, Hot (sec) |
Feature, Cold (sec) |
| Mid-2007 15-inch MacBook Pro 2.2 GHz Core 2 Duo 6GB 320GB 7200rpm HD — Mac OS X 10.6 |
3.7 | 5.0 |
| Mid-2009 15-inch MacBook Pro 2.53 GHz Core 2 Duo 4GB 250GB 5400rpm HD — Mac OS X 10.7 |
2.9 | 5.0 |
| Retina, Mid-2012 15-inch MacBook Pro 2.3 GHz Core i7 (Ivy Bridge) 8GB 256MB SSD — Mac OS X 10.7 |
1.5 | 1.8 |
In the heat of development, the Retina MBP starts up and finishes the entire feature about 2.4 times faster than my dear Core 2 Duo MBP. That’s a significant win. I’m only left wishing that the 200GB of files I want to move over was CPU limited!
Comments
and now, what’s the result of your test for one week ? do you like it or do you prefer the other MBP without Retina ?
Would have been nice to see difference between i5 and i7 on same disk, and also running the tests with Parallel tests gem to make use of all cores and threads.
@Stephane Text on the Retina MBP is gorgeous. Terminal looks really good and you can knock the font size way down when it’s useful, like when you’re watching logs fly by.
However, with the really crisp text, I notice that the kerning of a lot of common fonts on the web feels unpleasant. Perhaps hinting isn’t handled well? I’m not sure. I just know that while it could look at nice as printed text, it doesn’t.
If I was a graphic designer for the web, I’m not sure I’d like the Retina. I would want my fonts to show up how most people would see them. While you can control some scaling, even the “most screenspace” setting is still like 1.5 Retina pixels per 1 normal pixel. (I run on the default scaling, seems fine.)
Speed wise, it’s quite a bit nicer than my Core 2, but not revolutionary. It’s only 2x as fast, not 10x.
Hi,
I have a first generation unibody macbook.
And the reason I would like to buy a new macbook pro is because the asset pipeline processing is a little bit slow. Especially when we have a lot of coffee script…
Can you post some data regarding the asset pipeline processing speed please?
@Orban, I imagine that asset pipeline is going to be CPU limited, so the speedup for the Cuke feature above (~2x over 2.2GHz Core 2) is probably the best representation of the asset pipeline speedup.
If you find or whip up a benchmark I can grab from Github, I’d be glad to give you some more numbers.
@Orban: I’ve heard people complain about asset pipeline slowness when using Compass. If you’re using that, it adds a lot of complex files to each request, and can turn a request from less than a second to tens of seconds.
If that’s the case, I recommend not using Compass, or only using the files you need.
Certainly the RMBP is faster, but I wouldn’t put much stock in those numbers – the HDs in those machines are all different. If your 3,1 MacbookPro had an SSD in it, I wonder how it would compare?