Marco.org • About ▾

I’m : a programmer, writer, podcaster, geek, and coffee enthusiast.

Far Too Much Analysis Of The Alleged New Mac Pro Geekbench Score

Mac Rumors dug this up last night. For reference:

Pre-production Macs routinely show up in Geekbench. This looks legitimate1 and reveals a number of new details: we already knew that this fall’s Xeon E5 update would include up to 12 cores per socket, but this appears to specifically confirm an “E5-2697 v2” model with 12 cores at 2.7 GHz. (The current best comparable model is the Xeon E5-2690 with 8 cores at 2.9 GHz.) An uncredited Wikipedia source confirms this as the highest-clocked 12-core chip in the lineup.

As far as I can tell, this is also the first benchmark we’ve seen of any Xeon E5 v2. It compares well: it’s pulling a score of about 24,000 out of a single socket, compared to about 16,000 from the previous Xeon generation — exactly in proportion to the core increase, even at a base clock of 200 MHz less.

The Ivy Bridge-based Xeon E5 v2 appears to perform slightly better than the Sandy Bridge-based E5 series (which never came to the Mac Pro), much like the Sandy Bridge to Ivy Bridge desktop transition, and the v2 can have up to 12 cores instead of 8. No huge surprises.

John Poole, founder of Geekbench’s parent company, isn’t blown away. This is going to be a common response: the new Mac Pro can’t blow us away in Geekbench relative to the old ones because there aren’t any dual-socket models. It’s one of the biggest compromises in the new design: easily-parallelized tasks won’t be much better, and may be worse, than on the old $5,000+ dual-socket Mac Pros.

But we don’t know if many Mac Pro buyers were getting the dual-socket models.

Many pro apps, notably including Photoshop, don’t effectively use tons of cores, so the fastest Mac Pro for many people in practice has been the single-socket, 6-core, 3.33 GHz model. (This is why my photographer wife and I chose this model for ourselves.) It scores “only” about 13,500 on Geekbench, but in single-threaded tests, it appears to only be about 10–20% slower than this new E5-2697 v2.

If there’s any disappointment to be had, it’s that Intel has made so little progress in single-threaded CPU performance since 2010.

If the new Mac Pro is offered with fewer cores but a higher clock, that might be notably faster in most workloads. That Wikipedia table claims that there will also be CPUs with 6 cores at 3.5 GHz, 8 cores at 3.4 GHz, or 10 cores at 3.0 GHz. I hope Apple offers the 8-core 3.4 GHz model, because that’s probably a better choice for most buyers than this 12-core chip.2

The new Mac Pro is also extremely power-lopsided: it will initially max out at 12 cores (almost certainly this exact CPU), which is upper-midrange by Xeon standards, but it comes with a ridiculous amount of GPU power. This is overkill to just be about future desktop Retina Displays — clearly, Apple’s pushing for pro and scientific apps to shift more of the heavy lifting to OpenCL.

If they succeed, the new Mac Pro will probably crush everything else in its price range (and the rest of the Mac lineup). In the meantime, or for people who won’t use OpenCL-accelerated apps, it will probably be an incremental Mac Pro update: similar CPU increases as every other Mac Pro update, minus most of the internal expansion. So if you have a 2010 Mac Pro, there may not be much reason to upgrade.

How compelling this will be depends on two big, unanswered questions:

  1. How much will it cost? (If every model comes with dual FireGL GPUs with lots of VRAM, this might be uncomfortable.)
  2. Will there be desktop Retina Displays? If so, will any other Macs be able to drive them?3

We’ll find out soon enough.


  1. Except for one tiny detail: the CPU is reporting itself as “Intel Xeon E5-2697 v2”, with a lowercase “v”.

    In Intel’s E3-series “V2” and just-released “V3” Xeons, the “V” is always capitalized. But in Intel’s leaked roadmap slide, it’s lowercase for the E5 “v2” series. So who knows? 

  2. If the 8-core, 3.4 GHz E5-2687W v2 is too hot to be included (it has a 150W TDP while the others top out at 130W), I’d also be happy with the 6-core, 3.5 GHz E5-2643 v2. It’d probably score “only” in the 16,000 Geekbench range but be much faster than this big 12-core chip at most tasks. 

  3. Notably, we didn’t hear about Haswell updates to the Retina MacBook Pro yet. I’m hoping that they’re holding them back until the new Mac Pro launch, they’ll also include Thunderbolt 2, and they’ll join the new Mac Pro as the only Macs that can drive a new external 4K Retina Display.

    I haven’t been following the consumer desktop chips, but it’s also possible that the iMac will receive a similar update for the same reason. It would also shake things up considerably if a Retina iMac comes out this fall, which would lure many potential Mac Pro buyers.

    But this is all wishful thinking, relying on the assumption that desktop Retina Displays in some form will be out soon, which isn’t supported by much. 

Ads via The Deck