Geekbench 64-bit for 1.1 and 1.2

[amitdoc2b]

Not the same benchmark.
This is from the old Geek bench 2.
All others are from Geek bench 3

Mine seems even higher than his, not sure why. It's also from Geekbench 2 and its 32-bit free mode.

 

[pookitoo]

I compare with the website :

http://browser.primatelabs.com/mac-benchmarks

And it's a geek bench 3 test on this website.

On this website, i see the 1.2 rMacbook just near a 2013 i5 macbook air.

And i hope the 1.3 rMacbook near the mac mini I5 2014 1.4 ghz.

It's difficult for me to understand why every website (or some thread) say : only for web consulting and word and excel. I 've a 2013 i5 macbook air with 8 gb ram and i can do all my work (web design : sketch / pixelmator / sublime text ...) very well.

(sorry for my english

 

[Wahlstrm]

Mine seems even higher than his, not sure why. It's also from Geekbench 2 and its 32-bit free mode.

Use Geekbench 3 64-bit, that´s the standard.

 

[phas3]

so is it best to get the 1.1 or the 1.2?

 

[phas3]

About 10 64-bit Geekbench scores are up for Macbook8,1 as of noon today for M-5Y31 and M-5Y51. I plotted them in relation to the >130 M-5Y31 32-bit scores. S.D. represented as error bars and # of scores uploaded to Geekbench inlayed.

Yepp, you are wrong (or better put - not entirely correct). The everyday computing is still very asymmetrical, meaning that most of the important critical work is done by a single thread. For example, the entire UI (event processing, drawing etc.) happens on a single application thread. Browsers are essentially single-threaded etc. It is true of course that most modern application offload processing to background threads, but that is usually work that has lower priority. In order to have responsive, fluent applications, you want to maximise single thread performance.

Of course, there are applications that benefit from multiprocessing especially well. But these are usually applications that perform heavy-duty computations which can be easily split in multiple symmetric chunks — such as compression, video encoding, image processing etc. However, Core M is certainly not designed for such workflows.

To sum it up, you need to consider whether the work performed by the application can be easily split in multiple chunks of same status (priority, amount of work etc.). If yes, then you will benefit from multi-core performance more. If no, you want to allocate most of the resources to the main thread.



They do not really matter, but for a different reason Geekbench is a synthetic benchmark and should be taken very likely in this context. First of all, it measures sustained performance (which is irrelevant for average everyday tasks). Second, it only measures certain aspects of performance. This is akin to judging worth of a family car by how fast it can carry a ton of bricks up a very steep hill at full throttle.

Most real world application operate very differently — you have a very long cycles of waiting for user input, were literally nothing happens. Then the application has to 'burst'-process that input as quickly as possible and update its UI/data etc. to maintain the illusion of fluid interaction. These bursts are usually very short, we are speaking milliseconds here. Geekbench does not measure things like that. It also does not seem to benchmark other things that are vital for everyday applications such as indirect branches, branch prediction, cache behaviour etc. All in all, Geekbench is a reasonable measure if you want to judge how well the CPU will perform in a long-lasting computationally intensive task. For 'casual' (=home/office) usage though, you are probably better of looking at real-world browser tests and similar.

Good post Leman.

And I absolutely agree - Assume that practically all software you will run on that machine is not optimized for multiple threads. Also, we need real world results to see how good the machines actually are, especially in the light of the Anandtech article that showed major differences solely based on the ability to dissapate heat with different notebook architectures. Hopefully the aluminium chassis of the rMB will prove to be excellent at that.

So if you guys were to choose 1.1 or 1.2 strictly based on CPU performance?