Primate Labs Blog

After a lot of nudging (sorry it took so long!) I’ve updated Primate Labs’ Mac Benchmark chart. The updated chart adds the following Macs:

• iMac (Late 2009)
• iMac (27-inch Late 2009)
• MacBook (Late 2009)

The Mac Benchmark chart is generated by gathering results from the Geekbench Result Browser and averaging scores for each Mac model and processor combination. However, not every combination is included; for a combination to be included it has to meet a number of criteria which includes the total number of results for that combination. If the number is below a certain threshold the combination isn’t included.

Which leads me to something surprising; I was hoping to include Apple’s latest Mac Pro update, the single-processor Mac Pro @ 3.33GHz, but there are only two results for it in the Geekbench Result Browser. I was expecting more results for this particular model.

There are, however, a large number of dual-processor Mac Pros @ 3.33GHz in the Geekbench REsult Browser even though Apple doesn’t offer this configuration for sale. Apparently adding a second processor (or replacing both processors) in a Mac Pro is a popular upgrade. It’s a powerful upgrade, too, as one of these upgraded Mac Pros currently has the sixth highest score in the Geekbench Result Browser.

Late last week Geekbench results from what appeared to be a yet-unreleased MacBook Pro appeared on the Geekbench Result Browser. These results generated a lot of excitement, along with a lot of speculation; people wanted to know if these results were legitimate.

I spoke with Wired’s Brian Chen earlier today and shared my thoughts regarding the legitimacy of these results:

Geekbench creator John Poole told Wired.com that Geekbench results can be faked, but he believes the Core i7 MacBook Pro benchmark is real.

“If the system information for this system has been faked, it’s a convincing fake; there aren’t any inconsistencies in the system information, the Mac OS X build number matches Apple’s numbering scheme for new hardware builds, and the processor matches people’s expectations for what will appear in the refreshed MacBook Pro,” Poole said. “In other words, I believe this result is authentic.”

I have a feeling we’ll find out sooner rather than later whether this result is real or not!

The redesigned Geekbench Result Browser is online:

While it’s not a major departure from the old design, the new design is much cleaner and fits in with the other Primate Labs websites. Also, the new design lays some important groundwork for some new features and functionality I’m hoping to release later this year.

Earlier this month Apple announced a lot of new hardware. While not all of it is available yet (in particular the Core i5 and Core i7 iMacs) most of it is and folks have been asking about how the new hardware performs compared to the old hardware. So, I’ve gathered Geekbench results for the new Mac models and compared them with results for the older Mac models (all results were gathered from the Geekbench Result Browser).

Geekbench is Primate Labs’ processor and memory benchmark where a score of 1000 means a system has the same performance as a PowerMac G5 @ 1.6GHz. Higher scores are better.

iMac Benchmarks

MacBook Benchmarks

Mac mini Benchmarks

Conclusions

The results are as expected; faster processors (of the same architecture) mean higher Geekbench scores. This shouldn’t be surprising since Geekbench is predominantly a CPU benchmark. What I’m interested in seeing, though, is how the Core i5 and Core i7 iMacs stack up; not only do they have a new processor architecture they also have double the number of cores. I wonder if the high-end iMac i7 will (finally!) be comparable to a low-end Mac Pro?

I’ve switched the comment system on the Primate Labs Blog from WordPress’ built-in comments to Disqus. Disqus is a hosted comment and discussion service that offers a lot of nice features; you can “log in” and leave comments with your Twitter, Facebook, or OpenID account (you can also sign up for a Disqus account).

Disqus also enables threaded discussions instead of a flat list of comments, making it easier to reply to one person instead of an entire comment thread. Disqus also (optionally) notifies you when someone responds to your comments, making it easier to follow discussions.

I really hope Disqus makes it easier to have meaningful discussions on the Primate Labs Blog. If you’ve got any questions or comments, you can let me know by posting a comment!

While it’s not something I talk about a lot, many of you already know that Primate Labs is not my full-time job. I have a full-time job (as a software developer) during the day and I work on Primate Labs during my evenings and weekends. Although I lose much of my leisure time, my day job gives me access to amazing technology and Primate labs constantly forces me to deal with new challenges. It is a “double life” that has made me a better developer.

You might wonder why any of this matters to you, my users. I have been fortunate that the companies that I have worked for in the past, including most recently RapidMind, have been willing to allow me to pursue Primate Labs projects. A few weeks ago RapidMind was acquired by Intel. This acquisition has forced me to question what I should do as the developer of a vendor-neutral cross-platform processor benchmark AND as an employee of the world’s largest semiconductor company. Can I remain neutral when it comes to my Geekbench work?

I have considered many options, including selling or open-sourcing Geekbench. Abandoning Geekbench was not an option since Geekbench has over 10,000 registered users, over 170,000 results submitted to the Geekbench Result Browser, and companies all over the world in different industries use Geekbench.

I have decided to continue to work on Geekbench. I will strive to provide a benchmark that remains vendor-neutral in spite of my daytime employment. It is just another challenge, but one that you as Geekbench users should be aware of and understand.

If you have any questions, comments or concerns about this please let me know. You can reach me via email at email hidden; JavaScript is required or by posting a comment below.

I’ve added a page to the Primate Labs website that lists the experimental builds (along with the official releases) of Geekbench. These experimental builds are, well, experimental (meaning things can and probably will break) and come with absolutely no support.

Right now there are experimental builds of Geekbench 2.1 for Solaris and Haiku available on the download page. I’ll update the blog whenever additional builds are available.

Snow Leopard is out today! While I’ve been running the developer versions for a while now, I went out and picked up a retail copy at my friendly neighborhood Apple store and installed it on my MacBook Pro (the laptop where I do most of my Mac development).

Now, what’s interesting about Snow Leopard is that unlike most new versions of operating systems (or most new versions of software in general) Apple didn’t add a lot of new features to Snow Leopard. Instead, Apple focused on making Snow Leopard faster and more stable than Leopard.

While some of the improvements, like Grand Central Dispatch and OpenCL, will only benefit new (or rewritten) applications, do the other improvements help existing applications run faster?

To find out just how much faster existing applications run under Snow Leopard, I ran Geekbench on my MacBook Pro under Leopard and Snow Leopard. Geekbench doesn’t take advantage of Grand Central Dispatch or OpenCL so it’s a good way to determine how much of a performance boost existing applications will receive under Snow Leopard.

Setup

Here’s the configuration of the MacBook Pro I used:

• MacBook Pro (Late 2008)
  • Intel Core 2 Duo P8600 @ 2.40GHz
  • 2.00 GB 1067 MHz DDR3
  • Mac OS X 10.5.8 or
  • Mac OS X 10.6

If you’re not familiar with Geekbench, higher scores are better.

Results

Overall Performance

Integer Performance

Floating Point Performance

Memory Performance

Stream Performance

Conclusions

While the performance improvement is small, it is there — Geekbench runs between 2% and 3% faster under Snow Leopard than under Leopard. While this might not seem impressive at first keep in mind that Geekbench was slower under Leopard than Tiger. Having a new operating system improve performance, even if it’s a small improvement, is still something to get excited about.

One thing worth mentioning that isn’t captured in the Geekbench results above is that Snow Leopard feels faster and smoother than Leopard; the increased responsiveness of Snow Leopard makes it a joy to use.

Geekbench 2.1.4 has been released! Geekbench 2.1.4 features the following change:

• fixed a crash on Macs running Mac OS X 10.4.11

Geekbench 2.1.4 is a free update for all registered Geekbench 2 users.

Geekbench 2.1.3 has been released! Geekbench 2.1.3 features the following changes:

• added support for MacBook Pro (Mid 2009).
• added support for the latest AMD and Intel processors.
• added experimental support for Windows 7.

Geekbench 2.1.3 is a free update for all registered Geekbench 2 users.