Hackintosh Performance Hardware Options

Recently I described my latest Hackintosh build, a mini-ITX system based on an Intel Core i3-3220 CPU, an ASUS P8H77-i motherboard, and an ASUS GTX660 video card. I had to work with what our vendors sent us for the project, but that doesn't mean that you do. In this follow-up article I'll explore the performance and price of some different CPU and video hardware options.

As Apple moves aggressively down the road towards making all of their products sealed, non-expandable, and non-serviceable computing appliances, the allure of the Hackintosh only grows. Apple started this with the original Macbook Air's soldered-in memory, and the iPhone's non-removable battery and lack of a card slot carried onwards to the iPad. The iMac is the latest product to be a glued-together black box, and I'm sure the next refresh of the Mac Mini will be a sealed slab of aluminum, too. Only the very expensive Mac Pro retains any semblance of expansion and service features.

For readers of this web site, the appeal of a Hackintosh is that you can design a Mac work-alike machine exactly as you want it, and can easily upgrade it to boot, since it's just a PC hardware-wise. I've written three previous Hackintosh articles, covering the original X58-based machine in October, 2010, and the updated Sandy Bridge version in August, 2011. In between these two articles I covered the Hackintosh Experience in November, 2010.

In my most recent article, I showed you how to build a mini-ITX based system with a modern Ivy Bridge CPU and NVIDIA GTX650 video card. However, since I had a Core i7-3770K CPU available, and ASUS had generously supplied one of their hot GTX660 TI-DC2O-2GD5 video cards, I thought it would be interesting to benchmark some alternative component configurations.

Hackintosh Components

The two components that have the most effect on your Hackintosh's performance are the CPU and GPU. Which specific ones you choose depends on your needs and your budget. Today I'll be testing combinations of two different CPUs and three different GPUs. The CPUs are:

The GPUs are:

Intel's HD4000 GPU is surprisingly capable; in fact, Apple uses it across a wide range of Mac Minis and Powerbooks. It has more than enough power to handle the OS X interface effects and animations, and you'll never notice any artifacts, pauses, or stuttering in non-game applications. But if you are going to game, you're going to want a discrete video card, even with the relative paucity of Mac games as compared to PC games.

Choosing a CPU is a decision you'll want to look at carefully. The two Ivy Bridge processors I have on hand represent the low and high ends of the range, and neither would be my first choice for a Hackintosh. If you wanted a small, quiet, non-gaming system, the Core i3-3225 ( $144.99 at Newegg) is identical to the Core i3-3220, except that it includes a full HD4000 iGPU, allowing you to dispense with a graphics card. If you want more performance, especially for virtual machines, rendering, or other heavier workloads, the Core i5-3570 CPU ($214.99 at Newegg) will provide all the real-world performance of the 50%-more-expensive 3770K, unless you have heavily-threaded applications that would benefit from the latter's Hyper Threading, or you want to overclock.

On the GPU front, I can finally compare the performance of Intel's touted HD4000 integrated graphics with the NVIDIA GTX250 I used in my original article, as well as the ASUS-provided GTX660.

Let's take a look at the CPU performance first.

Hackintosh CPU Performance

Benchmarking a Hackintosh is more difficult than benchmarking a Windows PC simply because there are so few benchmarking applications. For CPU and system benchmarking, the current standard is Primate Labs' Geekbench. Geekbench "...provides a comprehensive set of benchmarks engineered to quickly and accurately measure processor and memory performance." It is multi-threaded and multi-core aware and tests integer performance, floating point performance, memory bandwidth, and "stream" operations; each section consists of 5 to 12 sub-tests whose scores are combined to summarize the performance in each category. Finally, an overall "Geekbench" score is computed from the results. Geekbench offers both 32-bit and 64-bit tests; I ran the 64-bit tests exclusively.

For this benchmark I tested the Core i7-3770K using its integrated GPU as well as a separate video card to see if there was any noticeable performance hit. As you can see from the chart below, there was, with the use of the integrated GPU dropping scores by about 8.5%.

Although I didn't have any "real" desktop Macs to compare these scores against, Geekbench allows you to upload your test results and compare them against the results uploaded by other users. Using this results browser, the 8413 score achieved with the Core i3-3220 CPU compares well with the 8442 score of the mid-2011 21" iMac, with came with a Core i5-2500S CPU running at 2.7GHz. The 13,349 scored by the 3770K with a discrete video card is very close to the score of 13,672 achieved by the 2013 Macbook Pro with Retina Display, which is running a quad-core Core i7-3840QM CPU. The highest score posted by a non-Mac Pro was the 13,998 returned by the latest 27" iMac, which has a Core i7-3770 CPU. So with the right CPU, our Hackintosh is very close to the performance of the fastest non-Pro Mac you can buy.

The next CPU test I ran was CINEBENCH R11.5, which renders a complex ray-traced scene using either a single core or all the available cores on the CPU. Here, the iGPU had virtually no effect on the 3770K results, likely because there was no animation or video going on. The Ivy Bridge cores of the Core i3-3220 are only a little over 6% slower than the individual cores of the 3770K in this test, which explains why the much less expensive CPU can hold its on in single-threaded tasks-- the incremental increase in clock speed and going from 3M to 8M of L2 cache just don't make that much different under most use cases.

When you load all the cores up, though, the 3770K surges ahead. Although both CPUs are Hyper Threaded, the 3770K has twice the cores and more than twice the cache, and the results in more than twice the performance...in this benchmark at least. If you're doing things like running virtual machines or video transcoding, more cores is obviously the way to go.

Next, let's take a look at video performance.

Hackintosh Video Performance

Configuring your Hackintosh to use the Intel HD4000 integrated CPU is pretty simple: you need to set the video memory used by the iGPU in the ASUS BIOS, and make one tweak to the org.chameleon.Boot.plist file.

First, specify your monitor's resolution by adding the following key/string pair to org.chameleon.Boot.plist (located in the Extra folder at the root level of your Hackintosh hard disk):

<key>Graphics Mode</key>
<string>1920x1080x32</string>

I'm using a 1080P monitor, so I set 1920x1080x32; substitute the horizontal and vertical resolution of your monitor for the first two numbers and use the "x32" on the end. If you don't set this, your desktop will appear interleaved and scrambled, so get it right! You can leave this setting even if you later switch to a discrete video card.

Next, reboot, go into the BIOS of the ASUS P8H77-I, and set the iGPU Memory to 96M as shown below. And although this screen shots shows the Primary Display set to iGPU, you can leave it at Auto if you wish. Note that you'll have to use either the DVI or HDMI ports for integrated video; the VGA port of the ASUS P8H77-i motherboard isn't supported.

For whatever reason, the Chameleon boot loader requires the iGPU memory to be set at 96M as shown above. However, once OS X boots, the iGPU memory is automatically set to 512M, as a quick dive into the system report shows:

So how does the iGPU perform? Well, just doing "stuff"-- web surfing, playing back DVDs, and such-- there's no visible difference at all in the performance of the iGPU and the two video cards. However, differences do appear one we start testing. CINEBENCH's OpenGL test shows, oddly, virtually no difference between the GTX650 and the much more powerful GTX660, although either card (when paired with the 3770K) is over 50% faster than the iGPU.

The Unigine Heaven 4.0 benchmark, though, shows huge differences. The GTX660 spits out more than 70% more frames than the GTX650, and the Intel HD4000...well, that's really just a slide show. It is interesting to note here that the performance of the GTX650 is identical with both processors, while even the 660GTX is only 1.53% faster with the 3770K.

Actual gaming performance is more difficult to assess. There's nothing like FRAPS available for OS X, and none of the games I have access to have in-game benchmarks. So as a meager substitute, I turned on the on-screen FPS counter in Serious Sam BFE and played the same section of the game with the same settings (1080P, CPU Ultra, GPU High, GPU Memory High) while making notes of the displayed frame rates.

The GPU makes a huge difference here, as you'd expect. The HD4000 is unplayable at 1080P with the above settings, although you can get frame rates into the mid-20s if you let the game automatically choose "optimum" settings...but the visuals are pretty horrible. Dropping from the quad-core 3770 to the dual-core 3220 takes brings about a 15% frame rate hit in this game with either NVIDIA card, and the GTX660 is roughly 50% faster than the GTX650 on either CPU.

So what does all this mean for your Hackintosh? Join me in the Final Thoughts section.

Final Thoughts

As with any computer, the components you choose will depends on your needs and budget, and the results will reflect those components. The most interesting part of these experiments for me was how good the performance of the Core i3-3220 was. I had originally asked Intel for a quad-core Core i5 CPU for this latest Hackintosh project, and was initially disappointed when all they had available was the Core i3. Its excellent single-core performance was a real surprise!

Unless you're running programs that leverage more than two cores, or like to do a lot of video transcoding, there's little reason to spend more money on a CPU...although you might want to pony up the extra $15 for a Core i3-3225 to get the Intel HD4000 graphics if you're not a gamer.

Apple's support for the Quick Sync video transcoding feature built into Intel's HD4000 iGPU is spotty. FaceTime and Airplay reportedly use it, but none of the various video programs (Handbrake, Final Cut Pro X, etc.) seem to use it. In contrast, many programs (including Adobe Creative Suite) can leverage NVIDIA's CUDA API to improve performance.

Unsurprisingly, the video card you use will dramatically affect gaming performance. Although it still lags far behind Windows, OS X is slowly gaining some decent games, mainly via Valve's Steam platform. And of course if you build a dual-boot system that sees some Windows use, a more powerful video card might make sense even if it can't be fully leveraged under OS X.

As for cost...well, look at the components cost if we're building a system with no video card and the Core i3-3225 CPU:

This compares well with the $599 base configuration of the Mac Mini, which has less memory (4G), and a 2.5GHz Core i5 dual-core CPU (the exact model of CPU isn't specified, and Apple has been known to change CPUs during a model run). For about $20 more we get twice the memory, a Blu-ray drive (the Mini has no optical drive at all!) a (probably) faster CPU, a faster hybrid hard drive, and much more expandability. Sounds like a good deal to me.

COMMENT QUESTION: What would your Hackintosh performance goals be?

Related Articles:

• ASUS GeForce GTX-465 Video Card
• Olevia/Syntax 242T FHD 42-Inch LCD 1080p HDTV
• SilverStone SUGO SG09 Micro ATX Build Notes
• Biostar TA890GXB-HD mATX AMD Motherboard
• Best PC Video Games 2010 Holiday Guide
• Cooler Master Storm Enforcer SGC-1000-KWN1
• Xigmatek HDT-S1283 Exposed Copper Heatpipe CPU Cooler
• OCZ RevoDrive 3 X2 PCI-E Solid State Drive
• Zotac H55ITX-C-E Mini-ITX WiFi Motherboard
• HIS Radeon HD 7870 IceQ Turbo 2GB