Original Link: http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review
The next-gen MacBook Pro with Retina Display Reviewby Anand Lal Shimpi on June 23, 2012 4:14 AM EST
Last year when I wrote about the new MacBook Airs I offered two forward looking paragraphs:
What happens from here on out is what's really interesting. Intel has already committed to moving the TDP of its mainstream parts from 35W - 45W down to 10 - 20W. Since the Air is the new mainstream Mac notebook, Apple has already made that move. The performance in this 10 - 20W segment is going to get much better over the next two years, particularly once Haswell arrives.
The Thunderbolt Display is the first sign of what's to come. Moving IO controllers and expansion into the display, and potentially even moving discrete GPUs out of the notebook are all in store for us. Apple is really ahead of the curve here, but it's easy to imagine a future where laptops become a lot more like the new Air and shift to a couple high bandwidth ports instead of numerous lower bandwidth connections.
Perhaps I was being too aggressive in the prediction of a couple of high bandwidth ports. After all, the next-generation MacBook Pro with Retina Display features four such IO ports (2 x Thunderbolt and 2 x USB 3.0). But you get my point. Gigabit Ethernet and Firewire 800 are both gone. The discrete GPU is still present but I suspect even its days are numbered, at least inside the chassis. The personal computer as we knew it for so long, is changing.
The personal computer is getting thinner, lighter, more integrated and more appliance-like. The movement is no longer confined to just Apple either. The traditional PC OEMs are following suit. Even Microsoft has finally entered the PC hardware business, something it threatened to do for years but hadn't until now. Distribution models will change, the lines between different form factors will continue to blur. What was once a mature industry is going through a significant transformation. It’s exciting but at the same time it makes me uneasy. When I first got into this industry everyone had stories of companies with great ideas that just didn’t make it. As we go through this revolution in computing I’m beginning to see, first hand, the very same.
Apple makes the bulk of its revenue from devices that don’t look like traditional personal computers. For the past couple of years I’ve been worried that it would wake up and decide the traditional Mac is a burden, and it should instead be in the business of strictly selling consumer devices. With its announcements two weeks ago in San Francisco, I can happily say that my fears haven’t come true. At least not yet.
It’s been a while since Apple did a really exciting MacBook Pro launch. Much to my surprise, even the move to Sandy Bridge, the first quad-core in a MacBook Pro, was done without even whispers of a press conference. Apple threw up the new products on its online store, shipped inventory to its retail outlets, updated the website and called it a day. Every iPhone and iPad announcement however was accompanied with much fanfare. The MacBook Pro seemed almost forgotten.
With its WWDC unveil however Apple took something that it had resigned to unexciting, dare I say uncool status, and made a huge deal about it. Two weeks ago Apple did the expected and offered relatively modest upgrades to all of its portable Macs, all while introducing something bold.
Apple calls it the MacBook Pro with Retina Display. You’ll see me refer to it as the next-gen MacBook Pro, Retina MacBook Pro, rMBP or some other permutation of these words.
After using it for the past two weeks I can honestly say it’s the best Mac Apple has ever built. And there’s a lot more to it than hardware.
If you were hoping for a 15-inch MacBook Air, that’s not what the rMBP is. Instead it is a far more portable 15-inch MacBook Pro. I have to admit I was a bit let down the first time I laid eyes on the next-gen MacBook Pro, it looks good but it doesn’t look all that different. The disappointment quickly faded as I actually picked up the machine and started carrying it around. It’s not ultra light, but man does it make the previous chassis feel dated.
While I never really liked lugging around the old MBP (and it always made me feel like the old fogey at tradeshows where everyone else had something 13-inches or smaller), carrying the rMBP is a pleasure by comparison. Pictures really don’t do it justice. The impressively thin display assembly or overall chassis thickness look neat in a photo but it’s not until you actually live with the rMBP that you can appreciate what Apple has done here. I carry around a 15-inch MacBook Pro because it’s my desktop, and as such it’s incredibly useful to have with me when I travel. For my personal usage model, the Retina MacBook Pro is perfect.
If your workload demands that you need the performance of a MacBook Pro and your lifestyle requires you to carry it around a lot, the reduction in thickness and weight alone will be worth the upgrade to the rMBP. If you spend most of your time stationary however, you’ll have to be sold on the display and internal characteristics alone. The bad news is if the design doesn’t get you, everything else will.
The problem with being on the forefront of design is every iteration is expected to significantly outdo the one before it. The unibody MacBook Pro design took build quality to a new level for Apple. By constructing all parts of the machine that you generally interact with out of the same piece of aluminum, Apple significantly reduced the amount of flex and creaks you’d encounter during normal use.
The next-gen MacBook Pro chassis doesn’t revolutionize the design, but it does make some significant evolutionary improvements. The most tangible impact as I’ve already mentioned is the reduction in size and weight of the machine. At its thickest part, the 13-inch MacBook Air is actually a little thicker than the 15-inch Retina MacBook Pro. Unlike the MacBook Air however, the rMBP does not feature a tapered design. Instead you get a constant thickness which is definitely reminiscent of the previous design.
The backlit keyboard and glass covered trackpad remain, although the key travel has been reduced somewhat - likely to help thin down the chassis. It's not worse, just different in my opinion.
The reduction in thickness also comes at the expense of a missing optical drive and no mechanical storage. Once again Apple has opted to use its own custom form factor and custom SATA connector for the NAND based storage in the rMBP. You’ll hear no complaints from me on the move away from mechanical storage as I’ve been recommending SSDs as upgrades for the past few years. The battery continues to be integrated but it’s no longer easily user removable as the custom cells are now glued to the chassis. A few years down the road your rMBP will have to take a trip to the Apple store (or a clever third party service center) to get its battery replaced.
The MacBook Pro with Retina Display, Image Courtesy iFixit
This is the first Pro appliance that Apple has ever produced. The CPU, GPU, DRAM, battery, display and, for now, the SSD are either non-removable or at least not user-upgradeable. On a $499 iPad that’s one thing, but on a $2199 professional notebook that’s a completely different matter. I can even make an exception for the MacBook Air as it is more of a consumer device, where computing needs have largely slowed down over the past several years. But for a professional machine, to have such a fixed configuration seems very worrisome.
|MacBook Pro with Retina Display Comparison|
|15-inch Mid 2012 MacBook Pro||MacBook Pro with Retina Display|
|Dimensions||0.95 H x 14.35 W x 9.82" D||0.71 H x 14.13 W x 9.73" D|
|Weight||5.6 lbs (2.54 kg)||4.46 lbs (2.02 kg)|
|CPU||Core i7-3615QM||Core i7-3720QM||Core i7-3615QM|
|Base CPU Clock||2.3GHz||2.6GHz||2.3GHz|
|Max CPU Turbo||3.3GHz||3.6GHz||3.3GHz|
|GPU||Intel HD 4000 + NVIDIA GeForce GT 650M|
|GPU Memory||512MB GDDR5||1GB GDDR5|
|System Memory||4GB DDR3-1600||8GB DDR3-1600||8GB DDR3L-1600|
|Primary Storage||500GB 5400RPM HDD||750GB 5400RPM HDD||256GB SSD|
|Display Resolution||1440 x 900||2880 x 1800|
|USB Ports||2 x USB 3.0|
|Other Ports||1 x Firewire 800, 1 x Audio Line in, 1 x Audio Line out, SDXC reader, Kensington Lock slot||SDXC reader, HDMI out, headphone out|
|Battery Capacity||77.5 Wh||95 Wh|
Apple has definitely made accommodations to make this unupgradeable reality more palatable. Sure the primary silicon is fixed, but all Retina MacBook Pro configurations ship with a minimum of 8GB of DDR3L-1600 memory. The only available upgrade is a move to 16GB, which will surely suit most needs for at least a few years to come (if not more).
The SSD is physically removable although there isn’t presently a source of 3rd party upgrades. I suspect we will see some in the future although there are always concerns about any legal claims to Apple’s unique form factor and physical interface. Apple’s concession here is it offers as much NAND as is physically possible today: up to 768GB if you’re willing to pay the handsome upgrade fee.
We’ve long given up on upgrading mobile CPUs or GPUs, and more recently abandoning the removable battery in favor of increasing capacity and reducing form factors is a trade off we’ve accepted as well. Apple has tried to help on the memory and SSD sides but the whole package is still very...fixed.
Despite all of this my only real complaint about Apple’s fixed configuration is the $2199 spec comes with too little storage by default. If I want to carelessly use my machine and not worry about regularly deleting unused files I find that I need 512GB of storage. At 256GB it’s too easy to run out of space, particularly if I’m on the road and dealing with lots of photos and videos. If you’re like me then you’re shoehorned into getting the $2799 configuration as there’s no way to just upgrade the size of the SSD in the $2199 model. And if you’re spending $2799 you might as well get the 16GB memory upgrade, if you can convince yourself that you’re not going to buy the Haswell version next year.
Retina MacBook Pro Silicon: One Big Happy Family
Powering the Retina Display MacBook Pro is Intel’s latest and greatest quad-core Core i7. A part of the new 22nm Ivy Bridge family the Intel silicon in the system is well done as always. The default configuration ships with a 2.3GHz quad-core offering, while the upgraded option is clocked at 2.6GHz. Apple offers one more upgrade at 2.7GHz while upping the L3 cache to 8MB. Since these are i7s all of them have Hyper Threading enabled, making the major difference between parts frequency and cache size in the case of the most upgraded part.
|Apple MacBook Pro with Retina Display CPU Comparison|
|2.3GHz quad-core||2.6GHz quad-core||2.7GHz quad-core|
|Intel Model||Core i7-3615QM||Intel Core i7-3720QM||Intel Core i7-3820QM|
|Base Clock Speed||2.3GHz||2.6GHz||2.7GHz|
|Max SC Turbo||3.3GHz||3.6GHz||3.7GHz|
|Max DC Turbo||3.2GHz||3.5GHz||3.6GHz|
|Max QC Turbo||3.1GHz||3.4GHz||3.5GHz|
|Processor Graphics||Intel HD 4000||Intel HD 4000||Intel HD 4000|
|GPU Clock (Base/Max)||650/1250MHz||650/1250MHz||650/1250MHz|
Turbo Boost is supported and active on all options. As always I verified its support in OS X as well as its functional operation:
889A refers to the max number of speed bins supported by Turbo Boost for 4, 3, 2 and 1 active core, respectively, in hex. For example, the 2.6GHz base clock of the Core i7 in my test system can turbo up a maximum of 8 bins with three/four cores active (2.6GHz + 800MHz = 3.4GHz), or 9 bins with 2 cores active (3.5GHz) or 10 bins (A in hex) with 1 core active (3.6GHz). Just as with previous mobile Macs, CPU clocks remain unchanged regardless of whether the system is running on AC or battery power.
All of the CPU options feature Intel’s HD 4000 graphics, which handles the majority of graphics duties unless you fire up an application that triggers the discrete GPU. All of the rMBPs feature NVIDIA’s GeForce GT 650M equipped with 1GB of GDDR5. Apple went aggressive on the Kepler implementation and ships a full 384 core GK107 in the rMBP. The GPU clock is set at a very aggressive 900MHz with a 1254MHz memory clock. I do appreciate that there’s no variance in GPU/memory configuration across all of the Retina MacBook Pro options, it greatly simplifies the purchasing experience.
Ports & Expansion
Port layout has been greatly simplified on the next-gen MacBook Pro. Along the left side there’s now a MagSafe 2 connector, two Thunderbolt ports, one USB 3.0 and one 1/8” jack for mic/headphones. The right side has the other USB 3.0 port, a full sized HDMI port and an SD card reader.
MagSafe 2 is a thinner version of Apple’s MagSafe connector, and it’s used on both the Retina MacBook Pro and the 2012 MacBook Airs. The rMBP still ships with the same 85W power adapter as before, but now with an integrated MagSafe 2 connector. In order to deal with the change in connector, Apple offers a $10 converter that allows you to plug MagSafe 1 power supplies into MagSafe 2 systems. All new Thunderbolt Displays shipping from here on out will include the MagSafe 2 converter.
The absence of an integrated Gigabit Ethernet port will surely bother some, but Apple offers a Thunderbolt to GigE adapter for $30 to accommodate. Since Thunderbolt effectively offers an external PCIe interface, there’s no performance loss if you go this route vs. the old integrated GigE connector. I was able to sustain nearly 930Mbps between the rMBP with the Thunderbolt GigE adapter and last year’s MBP:
At a price of $30 Apple is most certainly using Intel’s Port Ridge Thunderbolt controller, a cost effective single-channel TB controller without any support for DisplayPort passthrough.
USB 3.0 is provided courtesy of Intel’s 7-series chipset. Apple supports the USB Attached SCSI protocol which should allow for even better performance than what I’m showing below (with all of my focus on Thunderbolt I actually don't have a 6Gbps UASP enabled USB 3.0 dock in house):
The King of All Notebook Displays
For years Apple has been shipping some of the best displays in consumer notebooks, but the MacBook Pro’s Retina Display is in a league of its own. While I never liked the phrase “painted on” in reference to the iPad and iPhone Retina Displays, that’s the best way I can describe the effect the MacBook Pro’s Retina Display has on me. Text really does look painted on. The effect is really the result of two things.
The first is Apple’s removal of its cover glass. LCD panels aren’t particularly attractive, they are ugly squares composed of two pieces of glass and a number of filters/polarizers. To hide the ugly edges, display makers wrap bezels around the display. Most people aren’t fond of bezels so next came a ton of effort to minimize bezel size. An alternative is to simply place a third piece of glass over the entire LCD assembly and make it look as if the bezel and LCD panel are integrated. This outermost layer is known as a cover glass and is what Apple uses on all of its glossy displays. If you’ve ever taken apart a Cinema/Thunderbolt Display or a newer iMac you’ll know that the cover glass is literally just a piece of glass that you have to remove with some suction cups.
The MacBook Pro’s Retina Display does away with the cover glass and instead uses a fairly unique LCD assembly. There are still two pieces of glass but the outermost glass is actually a different size and shape - it integrates a bezel. By integrating the bezel into the outermost glass in the LCD stack you get the same effect as a cover glass but without the added reflections it introduces.
You also limit the possibility of dust getting trapped between the cover glass and the LCD. The danger is that you no longer have a protective piece of glass in front of your expensive new LCD. If you scratch the display you're scratching the LCD itself. While this has been true for conventional matte displays for a while, it's worth mentioning if you're used to Apple's glossy displays where you did have that added security layer.
The Retina Display is also obviously an extremely high resolution panel at 2880 x 1800. Note that this is 44.6% more pixels than Apple’s 27-inch Thunderbolt Display, and 26.6% more pixels than the 30-inch panels that we’ve loved for so long - all in a 15.4-inch notebook display.
At 220 pixels per inch it’s easily the highest density consumer notebook panel shipping today. At normal viewing distances and even with my face closer than I’m comfortable putting it I simply cannot discern individual pixels.
It’s the combination of these two elements, the removal of the cover glass and the insanely high pixel density that makes everything from text to UI elements just look painted on the new Retina Display. And the effect is gorgeous. I’ve never seen a prettier panel and it’s actually ruined me for pretty much all other displays, notebook and desktop.
While I can appreciate the iPad’s Retina Display, the impact from the MacBook Pro’s display is even more significant. Perhaps it’s because I still spend so much time working on a standard, non-tablet display, but I’m far more excited about this display than anything else Apple has delivered under the Retina moniker.
It’s not just pixel density that Apple has to offer here. Similar to its Retina Displays in the iPhone and iPad, the MacBook Pro’s Retina panel ditches TN in favor of IPS technology. The result is an incredible improvement in viewing angles. On a notebook I don’t spend a lot of time viewing it from far left/right angles, although I see the benefit when I’ve got others huddled around my display. Here the panel performs admirably - you lose brightness at far left/right angles but there’s no perceivable color shift. In fact, the painted on effect is even more impressive at these far left/right viewing angles.
For a single user however the more impressive characteristic is just how good the display looks at vertically off-center angles. I wrote much of the initial parts of this review while on an airplane in coach, which with a 15-inch notebook on my lap means I’m going to be looking at the display at a weird angle to begin with. The thinner rMBP doesn’t do enough to make the airplane usage model any better if the person in front of you decides to recline, but the IPS panel does make the display perfectly usable at the off-center angle you’ll inevitably have to deal with.
The Retina Display in Numbers
I already published preliminary analysis of the Retina MacBook Pro’s display. In short brightness is down a bit, black levels are considerably improved and contrast as a result takes a huge step forward compared to previous models. My personal MacBook Pro used the anti-glare matte screen and the improvement in contrast ratio compared to that reference is over 50%.
Apple made no mention of impact to color accuracy or color gamut. It turns out that the omission was for good reason, the Retina Display offers no improvement along either vector. The numbers show a slight regression compared to last year’s panel but the difference is imperceivable.
Technically Apple’s use of the word Retina in reference to a display only refers to the inability for the human eye to resolve individual pixels at a specified distance (18-inches for the MacBook Pro). In practice however Apple has delivered tightly integrated IPS panels with wonderful performance characteristics as a part of the Retina brand. I do hope that for the years to come Apple does not compromise on these fronts.
The Software Side of Retina: Making it All Work
OS X, similar to iOS, uses points to represent display coordinates. Traditional OS/display combinations had a 1:1 mapping between points and pixels. Points in OS X are now floating point values, as a single point can be represented by multiple pixels on a high density display.
Images are the same size, but made of 4x the number of pixels on Screen 1 compared to Screen 0
How pixels map to points is determined by the backing scale factor. The backing scale factor can either be set to 1.0 or 2.0. In the case of the former you get 1:1 point to pixel mapping, while in the latter each point is backed by four pixels. The backing scale factor isn’t a global value, it can be set on a per element basis, allowing controls to to be legible while you get the benefits of a higher resolution for additional screen real estate. This aspect of OS X is key to enabling good behavior in applications as you’ll soon see.
Apple does a lot of the display handling for you so you don’t have to think about any of this. All vector based graphics and text using Apple’s APIs are automatically scaled up. Unmodified dialog boxes, toolbars, menus, etc.. should all look “normal” sized and just be ridiculously sharp on the Retina Display. Bitmapped images are scaled up using linear interpolation, but if higher resolution assets are provided OS X can simply swap and use those on a Retina Display.
Applications that render vector graphics, text and other elements to their own backing store will need hand tuning to look good on the Retina Display. These elements will receive the same linearly interpolated upscale I mentioned above.
It’s a bit complicated and confusing so let me try my best to explain what’s going on here in a practical sense. The 15.4-inch Retina Display has a native resolution of 2880 x 1800, that’s 2880 pixels across and 1800 pixels down for a total area of 5,184,000 pixels.
On the Retina MacBook Pro, Apple has done away with conventional resolution settings. Instead you get a horizontal list of scaling options (this applies to external displays as well):
In the default “best for Retina Display” setting, the desktop, menu bar, icons and Finder windows are drawn at 2880 x 1800, but they are drawn larger than they would normally be at 2880. Apple draws everything at 4x the size to make the desktop behave exactly as it would on a 15.4-inch 1440 x 900 display - this is the backing scale factor (2.0) at work. This approach provides the best image quality as there’s integer mapping from pixels on the panel to pixels on the desktop. No interpolation or filtering is necessary.
Third party applications without specific Retina Display support also operate in this same “looks like 1440 x 900” mode. If you fire up Chrome, Photoshop or Word you’ll see that everything looks identical to how it would look on a standard resolution 15-inch MacBook Pro. Again, the screen is drawn at 2880 x 1800 but everything is scaled up to be the same size it would be at 1440 x 900.
If third party applications use Apple’s standard methods of drawing text and windows, all of these windows will look super sharp. If they don’t however, whatever routines they use to display windows and text will need to be Retina aware otherwise they run the risk of not scaling text properly. The famous example at this point is Google’s Chrome which has its own offscreen text rendering buffer, even though it uses Apple’s text rendering APIs:
Chrome Canary fixes the text rendering issue but it has a similar problem displaying images, they simply look better in Safari:
Even though Adobe had a Retina-aware version of Photoshop running at Apple’s WWDC keynote, the publicly available version of CS6 doesn’t feature the same support. Here even open dialog boxes look bad:
Many of you asked about Office 2011. These apps just work like they would at at 1440 x 900, just with blurrier text unfortunately:
It’s not just third party applications that need updating however, even Apple’s own iWork suite has yet to be updated to take advantage of the Retina Display. As a result text in Pages is incredibly blurry. It has been roughly three years since Apple last updated the iWork suite, so the applications are definitely due for an overhaul. I am a bit surprised Apple didn’t update them at the launch of the rMBP to be honest. It’s quite possible that a major iWork update is imminent and Apple didn’t see the need to update 3 year old software in lieu of that.
Cocoa applications can be forced to open in magnified low resolution or high resolution modes by looking at the app's info window (cmd + i on a selected app in Finder):
Eventually most apps will by default open in high resolution, such as those that have been optimized for Retina Display operation (e.g. iMovie above). Those applications that are not yet Retina aware may default to opening in low resolution mode (e.g. Adium, Pages), in which case they'll look and behave like they would at 1440 x 900 but with all UI elements upscaled to fit the 2880 x 1800 panel. Non-Cocoa applications will have the resolution scaling option greyed out (e.g. MS Office apps).
Where things get really exciting is when you have an application that not only handles scaling properly, but also takes advantage of the added resolution. Take Aperture 3.3 for example. With OS X set to its “best for Retina display” mode, this is what Aperture looks like with a 2880 x 1800 image open and displayed at full size:
Here Apple is scaling the UI elements like the menus and widgets on the screen (backing scale factor = 2.0), but displaying the open image unscaled (backing scale factor = 1.0). As a result we can fit almost an entire 2880 x 1800 image on the screen without zooming out. Remember the backing scale factor isn’t global, individual elements on the screen can be scaled independently depending on their purpose.
The same thing happens when you look at applications like iMovie or Final Cut HD. The UI elements are scaled up but the video window is displayed unscaled, thus allowing us to display a full 1080p video alongside text and tools that are still legible.
It’s all handled amazingly well. It just works.
Oh but there’s more.
If you select the 1680 x 1050 or 1920 x 1200 scaling modes, Apple actually renders the desktop at 2x the selected resolution (3360 x 2100 or 3840 x 2400, respectively), scales up the text and UI elements accordingly so they aren’t super tiny (backing scale factor = 2.0), and downscales the final image to fit on the 2880 x 1800 panel. The end result is you get a 3360 x 2100 desktop, with text and UI elements the size they would be on a 1680 x 1050 desktop, all without sacrificing much sharpness/crispness thanks to the massive supersampling. The resulting image isn’t as perfect as it would be at the default setting because you have to perform a floating point filter down to 2880 x 1800, but it’s still incredibly good.
The same rules as above apply to Retina-aware applications. Take the Aperture example again, this time at “1680 x 1050”:
Note that we can fit the entire 2880 x 1800 image at 100% almost without having to scroll. This is possible because our screen is actually rendered at 3360 x 2100, with the text and UI elements scaled up so they aren’t super tiny, yet the image is left unscaled.
Here’s the same Aperture setup but at “1920 x 1200”:
The 2880 x 1800 image looks downright small since our desktop is rendered at 3840 x 2400. Despite the fact that we’re able to fit everything into a single screen, the text and other UI elements are totally usable at this setting. You get the benefits of additional application real estate without any of the downsides.
What happens if you decide to take Aperture full screen? The image is displayed almost completely at 2880 x 1800. You do lose a little vertical real estate but not much at all.
The 3D gaming experience is even simpler. Here you just choose the appropriate resolution and you get the same scaling you normally would in the game. I’ve already demonstrated support for 2880 x 1800 in titles like Diablo 3, although there is still a need for developer support as we see with the console window in Portal 2:
The flexibility offered by Apple’s handling of the Retina Display in OS X is unparalleled. What applications like Aperture, iPhoto, iMovie and Final Cut HD offer, is unbridled resolution independence. What Apple has done here is so much more difficult than what it pulled off in iOS with the Retina Display. It will take time for third party application developers to get on board, but with the power of the Mac app store and Apple’s growing install base of Mac users I suspect we will see incredibly quick adoption of support for the MacBook Pro’s Retina Display.
To make the MacBook Pro’s Retina Display a reality Apple had to work with panel vendors to build the panels it wanted at a reasonable cost, as well as deliver the software necessary to support insanely high resolutions. There was another problem Apple faced in making the rMBP a reality: the display pipeline of the GPUs Apple wanted to use didn't officially support scaling to the resolution Apple demanded of them. Let me explain.
All modern GPUs have fixed function scaling hardware that is used to efficiently scale between resolutions. A scaler either in your GPU or in your display panel is what lets you run non-native resolutions at full screen on your LCD (e.g. running 1680 x 1050 on a 1920 x 1080 panel). None of the GPUs used in the Retina Display MacBook Pro officially support fixed-function scaling of 3840 x 2400 or 3360 x 2100 to 2880 x 1800 however. Modern day GPUs are tested against 2560 x 1440 and 2560 x 1600, but not this particular 5MP resolution. Even 4K resolution support isn’t widespread among what’s available today. Rather than wait for updated hardware and/or validation, Apple took matters into its own hands and built its own GPU accelerated scaling routines for these higher resolutions. Fixed function hardware is almost always more efficient from a performance and power standpoint, which is why there’s some additional performance loss in these scaled resolution modes.
What’s even crazier is Apple wasn’t pleased with the difference in baseline filtering quality between the Intel HD 4000 and NVIDIA GeForce GT 650M GPUs. As the Retina Display MacBook Pro would have to regularly switch between GPUs, Apple wanted to ensure a consistently good experience regardless of which GPU was active. There are a lot of filtering operations at work when doing all of this resolution scaling, so rather than compromise user experience Apple simply wrote its own default filtering routines. Since you want your upscale and downscale quality to be identical, Apple had to roll its own implementation on both. Apple’s obsessive attention to detail really made it possible to pull all of this off. It’s just insane to think about.
Driving the Retina Display: A Performance Discussion
As I mentioned earlier, there are quality implications of choosing the higher-than-best resolution options in OS X. At 1680 x 1050 and 1920 x 1200 the screen is drawn with 4x the number of pixels, elements are scaled appropriately, and the result is downscaled to 2880 x 1800. The quality impact is negligible however, especially if you actually need the added real estate. As you’d expect, there is also a performance penalty.
At the default setting, either Intel’s HD 4000 or NVIDIA’s GeForce GT 650M already have to render and display far more pixels than either GPU was ever intended to. At the 1680 and 1920 settings however the GPUs are doing more work than even their high-end desktop counterparts are used to. In writing this article it finally dawned on me exactly what has been happening at Intel over the past few years.
Steve Jobs set a path to bringing high resolution displays to all of Apple’s products, likely beginning several years ago. There was a period of time when Apple kept hiring ex-ATI/AMD Graphics CTOs, first Bob Drebin and then Raja Koduri (although less public, Apple also hired chief CPU architects from AMD and ARM among other companies - but that’s another story for another time). You typically hire smart GPU guys if you’re building a GPU, the alternative is to hire them if you need to be able to work with existing GPU vendors to deliver the performance necessary to fulfill your dreams of GPU dominance.
In 2007 Intel promised to deliver a 10x improvement in integrated graphics performance by 2010:
In 2009 Apple hired Drebin and Koduri.
In 2010 Intel announced that the curve had shifted. Instead of 10x by 2010 the number was now 25x. Intel’s ramp was accelerated, and it stopped providing updates on just how aggressive it would be in the future. Paul Otellini’s keynote from IDF 2010 gave us all a hint of what’s to come (emphasis mine):
But there has been a fundamental shift since 2007. Great graphics performance is required, but it isn't sufficient anymore. If you look at what users are demanding, they are demanding an increasingly good experience, robust experience, across the spectrum of visual computing. Users care about everything they see on the screen, not just 3D graphics. And so delivering a great visual experience requires media performance of all types: in games, in video playback, in video transcoding, in media editing, in 3D graphics, and in display. And Intel is committed to delivering leadership platforms in visual computing, not just in PCs, but across the continuum.
Otellini’s keynote would set the tone for the next few years of Intel’s evolution as a company. Even after this keynote Intel made a lot of adjustments to its roadmap, heavily influenced by Apple. Mobile SoCs got more aggressive on the graphics front as did their desktop/notebook counterparts.
At each IDF I kept hearing about how Apple was the biggest motivator behind Intel’s move into the GPU space, but I never really understood the connection until now. The driving factor wasn’t just the demands of current applications, but rather a dramatic increase in display resolution across the lineup. It’s why Apple has been at the forefront of GPU adoption in its iDevices, and it’s why Apple has been pushing Intel so very hard on the integrated graphics revolution. If there’s any one OEM we can thank for having a significant impact on Intel’s roadmap, it’s Apple. And it’s just getting started.
Sandy Bridge and Ivy Bridge were both good steps for Intel, but Haswell and Broadwell are the designs that Apple truly wanted. As fond as Apple has been of using discrete GPUs in notebooks, it would rather get rid of them if at all possible. For many SKUs Apple has already done so. Haswell and Broadwell will allow Apple to bring integration to even some of the Pro-level notebooks.
To be quite honest, the hardware in the rMBP isn’t enough to deliver a consistently smooth experience across all applications. At 2880 x 1800 most interactions are smooth but things like zooming windows or scrolling on certain web pages is clearly sub-30fps. At the higher scaled resolutions, since the GPU has to render as much as 9.2MP, even UI performance can be sluggish. There’s simply nothing that can be done at this point - Apple is pushing the limits of the hardware we have available today, far beyond what any other OEM has done. Future iterations of the Retina Display MacBook Pro will have faster hardware with embedded DRAM that will help mitigate this problem. But there are other limitations: many elements of screen drawing are still done on the CPU, and as largely serial architectures their ability to scale performance with dramatically higher resolutions is limited.
Some elements of drawing in Safari for example aren’t handled by the GPU. Quickly scrolling up and down on the AnandTech home page will peg one of the four IVB cores in the rMBP at 100%:
The GPU has an easy time with its part of the process but the CPU’s workload is borderline too much for a single core to handle. Throw a more complex website at it and things get bad quickly. Facebook combines a lot of compressed images with text - every single image is decompressed on the CPU before being handed off to the GPU. Combine that with other elements that are processed on the CPU and you get a recipe for choppy scrolling.
To quantify exactly what I was seeing I measured frame rate while scrolling as quickly as possible through my Facebook news feed in Safari on the rMBP as well as my 2011 15-inch High Res MacBook Pro. While last year’s MBP delivered anywhere from 46 - 60 fps during this test, the rMBP hovered around 20 fps (18 - 24 fps was the typical range).
Remember at 2880 x 1800 there are simply more pixels to push and more work to be done by both the CPU and the GPU. It’s even worse in those applications that have higher quality assets: the CPU now has to decode images at 4x the resolution of what it’s used to. Future CPUs will take this added workload into account, but it’ll take time to get there.
The good news is Mountain Lion provides some relief. At WWDC Apple mentioned the next version of Safari is ridiculously fast, but it wasn’t specific about why. It turns out that Safari leverages Core Animation in Mountain Lion and more GPU accelerated as a result. Facebook is still a challenge because of the mixture of CPU decoded images and a standard web page, but the experience is a bit better. Repeating the same test as above I measured anywhere from 20 - 30 fps while scrolling through Facebook on ML’s Safari.
Whereas I would consider the rMBP experience under Lion to be borderline unacceptable, everything is significantly better under Mountain Lion. Don’t expect buttery smoothness across the board, you’re still asking a lot of the CPU and GPU, but it’s a lot better.
Boot Camp Behavior
When the MacBook Pro with Retina Display first started shipping Apple hadn’t even released Boot Camp drivers for the system. Since then Apple has uploaded a Windows Support package to its servers, and the Boot Camp Assistant will give you drivers for everything in the machine should you ask for it.
The driver bundle and Boot Camp Assistant work with both Windows 7 and the Windows 8 Release Preview, although Apple only explicitly offers support for the former. You don’t have to do anything to make the Windows 8 RP work with the Boot Camp Assistant, just supply the Win 8 RP ISO instead of the Windows 7 image during the Boot Camp setup and you’re all set.
Like all previous MacBook Pro Boot Camp installs, only the discrete GPU is “connected” under Windows. Apple relies on a lot of its own software to switch between processor and discrete graphics which obviously isn’t made available under Windows, thus you only get the dGPU. Apple uses NVIDIA’s 296.49 drivers for the GeForce GT 650M under Windows. Since the standard Windows desktop doesn’t support integer DPI scaling (see update below as you can force 2.0x scaling) Apple picked the next best option and configures Windows for 1.5x DPI scaling and 2880 x 1800 as the defaults. The result is a desktop that looks like this:
At the largest text scaling setting Windows is remarkably usable on the 15.4-inch display at 2880 x 1800. Unfortunately as I mentioned in the Zenbook Prime review, Windows 7 and third party handling of DPI scaling is hardly elegant. While applications that aren’t Retina-aware under OS X tend to simply have blurry text, those that don’t behave properly with DPI scaling under Windows just look odd. Some text elements will be huge and overflow outside of their normal borders, while others will ignore the scaling setting entirely and just be too small. It’s passable in a pinch but suboptimal for certain.
Windows 8 does a somewhat better job, but only under Metro. Metro supports integer DPI scaling at 1 and 2x factors, similar to OS X. With DPI scaling enabled under Windows 8, Metro looks like it would at 1440 x 900 - similar to the default setting for the Retina MacBook Pro under OS X.
200% DPI scaling under Windows 7
Update: As many of you have correctly pointed out, Windows allows you to specify custom DPI scaling modes including an integer 2.0 setting. The result is the same "1440 x 900"-like desktop you get under OS X. Application compatibility still appears to be an issue, check out the gallery below for some examples. Overall Windows is very usable on the rMBP, but just as under OS X the overall experience really depends on application support.
With the Retina MacBook Pro I get the distinct impression it was launched before the software was ready to support it. Apple did an amazing job enabling Retina support in all of the iLife applications, but iWork isn’t ready for it yet. The system technically launched without Boot Camp support although that was soon added. Then there are the UI performance issues in applications like Safari under the currently available version of Lion. There are also occasional graphical glitches under Lion. the occasional flashing of UI elements, nothing major but just not the polish we’re used to from Apple. Once again, Mountain Lion addresses a lot of this and is only a month out but you still have to deal with reality in the interim.
I also had a strange experience where the CPU clocks on the system magically decided to get clamped down to around 1.2GHz on battery power. A simple reset of the PRAM and SMC fixed the problem but it was odd behavior for sure. I can’t completely fault Apple for this one, as I know I subjected the rMBP to a bunch of strange tests over the course of the past week and a half.
If you’re buying today, just be aware that the best experience will come with Mountain Lion and even then you’ll have to wait an unspecified amount of time for Retina-aware app updates.
All Flash Storage
As expected, the next-gen MacBook Pro ditches mechanical storage in favor of a MacBook Air style NAND + SSD controller on a custom PCB. Apple refers to this solution as all-flash storage.
Apple’s distinction between Solid State Drives (SSDs) and all-flash storage boils down to what form the storage comes in. If it’s a standard form factor device in a chassis, it’s a solid state drive. If it’s just NAND + controller on a PCB? Then it’s all-flash storage. I suspect it’s a nicer way of saying proprietary SSD but either way they are technically the same thing, just in different forms.
The Samsung PM830 based rMBP NAND flash storage card, image courtesy iFixit
My Retina MacBook Pro was the upgraded model with a 512GB SSD, featuring Samsung’s PM830 controller. This is the same controller as in the Samsung SSD 830, which I’ve long felt was the best pair for Mac users who wanted an SSD upgrade. I’m not sure if other Retina MBPs may come with Toshiba’s SandForce based drive instead. I have one of these drives in house for a review but that’ll have to wait until next week.
Although both the Samsung and Toshiba/SandForce controllers support full disk encryption, neither hardware based encryption is supported by OS X’s FileVault 2. When OS X encrypts your boot volume not all areas are encrypted (such as the recovery partition). While I know SandForce allegedly offers multiple encryption levels across a volume I’ve never seen either controller claim support for partially unencrypted volumes. In this case it looks like for Apple to take advantage of SSD controller based encryption it would need more flexible encryption support on the controller level. If I were an SSD controller vendor I’d be paying close attention to this requirement.
Both the Samsung and Toshiba controllers support 6Gbps SATA - as a result performance is significantly better compared to previous Apple branded SSDs. I borrowed a friend’s 2010 MacBook Pro which happened to have a Toshiba based SSD installed and ran it through our standard Iometer four-corners test suite. This was a well used drive and thus the performance is even worse than last year's MacBook Airs. The improvement in performance is astounding:
The move to 6Gbps SATA is often associated with a huge bump in sequential transfer rates, but in this case Apple enjoys a significant increase in random speeds as well. Note that some of this improvement is going to be due to the fully populated configuration of the PM830 in the Retina MacBook Pro's SSD, but that shouldn't downplay the significance of the move to Samsung's latest controller. The previous generation controller used last year just wasn't very good, and the Toshiba alternative was even worse. This year, Apple finally has a good solid state story to tell.
How much of this are you going to be able to actually tell in day to day use of the system? The sequential transfer rates are most tangible when you are writing to or reading large files like movies to your drive. Obviously you need a source that's fast enough to hit these speeds. Although USB 3.0 can come close you're unlikely to have a USB 3.0 SSD that's as fast as the internal drive. Moving large files between your internal SSD and Promise's Pegasus R4/R6 is where you'll really appreciate this performance.
The random access improvements are likely overkill for most normal uses. Things like program launches, compiling, web browsing, and any other normal application IO will depend on a mixture of random and sequential IO. The key is to have good enough random IO performance to avoid becoming a bottleneck. I can safely say that the numbers we see here are more than enough.
While previous Apple SSDs were nice only from a convenience standpoint, at least the Samsung option in the Retina MacBook Pro is what I’d recommend even if Apple didn’t bundle it with the machine.
Apple’s 2011 Macs were the first to enjoy Thunderbolt, an interface co-developed with Intel that carries PCIe and DisplayPort over a single cable. As it derives most of its revenue from mobile, Apple wasted no time in bringing its Thunderbolt Display to market. A single Thunderbolt cable could bring Gigabit Ethernet, Firewire 800, high-speed mass storage, external audio and display to an otherwise IO-deprived MacBook Air.
At a high level, Thunderbolt is pretty easy to explain. The current implementation of Thunderbolt pairs four PCIe 2.0 lanes with DisplayPort, offering a maximum bandwidth of 2GB/s in either direction in addition to DP bandwidth. The Thunderbolt interface itself can deliver 10Gbps of bandwidth in each direction, per channel. The physical Thunderbolt port is compatible with mini DisplayPort to allow for the use of mini-DP displays as well as Thunderbolt chains. Each Thunderbolt port can carry up to two Thunderbolt channels, although one channel is typically reserved for DisplayPort duties.
In the past we measured a maximum of 1GB/s of unidirectional bandwidth for a single Thunderbolt channel in addition to video bandwidth over DisplayPort. There’s no shipping device that will deliver this sort of performance, I needed to outfit a Promise Pegasus with a handful of SSDs to truly saturate the bus.
In the 2012 Macs Apple, like the rest of the PC industry, has switched to using Intel’s 2nd generation Thunderbolt controllers codenamed Cactus Ridge.
The Retina MacBook Pro uses a four-channel Cactus Ridge controller and drives two Thunderbolt ports with it. Each port can drive a mini-DP display or a Thunderbolt chain with a mini-DP/Thunderbolt Display at the end of or in it. The rMBP can actually drive a fourth panel (counting the integrated Retina Display) via the integrated HDMI port although that’s not an officially supported configuration.
Unlike most other implementations, Apple hangs the Cactus Ridge controller off of the Ivy Bridge CPU rather than the PCH. The GeForce GT 650M in the system only gets the use of 8 PCIe 3.0 lanes instead of the full 16, but with PCIe 3.0 this is not an issue (it wouldn’t be an issue with PCIe 2.0 either to be honest).
I performed the same test as before to test if maximum bandwidth has gone up since switching to Cactus Ridge. Initial results remained unchanged, I was able to get north of 900MB/s to an array of SSDs in the Pegasus connected to a single Thunderbolt port. Now with two Thunderbolt on the rMBP however I was able to create a second chain of devices. I only have a single Pegasus so I resorted to chaining a LaCie Little Big Disk (SSD) and Elgato Thunderbolt drive. The combination of the two isn’t anywhere near as fast as the SSD array in the Pegasus but it allowed me to push the limits of the controller even more:
1380MB/s, over copper, to the rMBP. I suspect if I had another Pegasus SSD array I’d be able to approach 1800MB/s, all while driving video over the ports. Apple may limit the internal storage expansion of the rMBP but you still have a path to expansion for storage of large media files and other archives. And it’s very fast.
Unfortunately Thunderbolt behavior is still not perfect, although it is improved compared to previous Macs. If you write to Promise’s Pegasus for long enough while playing audio through Apple’s Thunderbolt Display you will still drop audio frames. Subjectively it seems to take longer to trigger this phenomenon but it does still happen. On my early 2011 MacBook Pro the problem has gotten so bad that I’ll even drop other USB packets for devices connected to the Thunderbolt Display. If I’m writing to the Pegasus I’ll miss keystrokes and the mouse will jump around until the high-speed write is complete. So far I haven’t had anything this bad happen on the Retina MBP but it took a while for this behavior to manifest on my early 2011 model so we’ll see what happens. I’m not sure what the fix will be for these types of issues as it seems there’s no good quality of service assurance for PCIe devices residing on Thunderbolt. As Thunderbolt was supposed to be as transparent as possible, it’s not surprising that even QoS overhead is nonexistent but it’s something that is clearly necessary. I’m not sure this is Apple’s fault as I’ve seen similar behavior under Windows. I suspect it’s something that Intel is going to have to figure out a way to address.
Vastly Improved Thermals
When the 2011 MacBook Pro hit I was so excited because it finally gave me a fairly portable chassis with my definition of a “fast-enough” CPU: a quad-core Sandy Bridge. As always, there’s a downside to being an early adopter. In this case the downside to buying Apple’s first quad-core notebook was that it got incredibly hot and was fairly loud at that.
It had good reason for being a space heater. Intel crammed nearly a billion transistors on its 32nm process into the quad-core Sandy Bridge. AMD contributed over 700 million transistors on TSMC’s 40nm process with the Radeon HD 6750M, and when you used the two in tandem things got bad.
A bunch of ridiculously smart fab engineers in pursuit of maintaining Moore’s Law gave us improvements on both fronts. Ivy Bridge uses Intel’s brand new 22nm 3D transistors, while NVIDIA’s GeForce GT 650M (codename: Kepler) uses TSMC’s 28nm process. Both architectures are very focused on power reduction and it shows.
To understand the improvement that the Retina MacBook Pro offers over its predecessor we need to really understand where last year’s model suffered. To put things in perspective I created three experiments, two focused on the CPU’s thermal behavior and one on the CPU and GPU.
To start I took Cinebench and ran its single threaded test over and over again, plotting performance vs. number of runs. I ran the test for about 20 minutes before calling it quits. As you can see both the 2011 MacBook Pro and 2012 rMBP delivered consistent performance between runs:
Subjectively, fan speed wasn’t an issue on either system during this test. The fans never ramped up to full speed on either machine, which tells me that despite the CPUs supporting Turbo Boost neither chip is actually dissipating all 45W allocated during this test. It’s possible the chips could be clocked higher, or just as likely that thermal hotspotting (dense areas of transistors that get too hot) is preventing them from running at higher frequencies - not the overall power budget of the chip.
The next test uses Cinebench but instead focuses on the multithreaded benchmark. Here all four cores (and eight threads) are engaged across both systems. I ran these tests, back to back, for around 20 minutes straight. Despite the increased load, neither system shows any throttling of performance:
Fan speeds definitely ramped up here, stabilizing at their maximum spinning speeds of just over 6000 RPM. The Retina MacBook Pro is subjectively quieter but I’ll discuss that in greater detail shortly.
The final experiment focuses on a hardly stressful game by today’s standards: Half Life 2 Episode Two. Once again, for approximately 20 minutes I ran our standard macbench timedemo test in HL2. I ran both systems at 1680 x 1050, without AA, but with all other quality settings maxed out. This is a bit more stressful than 20 minutes of actual gameplay since the timedemo runs renders all frames as quickly as possible rather than playing back the demo in real time. In the end it worked out to be nearly 40 consecutive runs of our benchmark.
The graph below gives you an idea of what happens to performance over time:
I’ve adjusted the y-axis on the chart to exaggerate the impact here a bit, but you get a clear idea of just how much heat both of these chips were putting out in the 2011 MBP. Either the CPU or GPU (or both) have to be throttled back over time in order to stay within their thermal and power budgets. As a result, in the 2011 15-inch MacBook Pro, performance drops by over 20% over the course of 20 minutes of this test.
The MacBook Pro with Retina Display, on the other hand, remains relatively stable across all runs. While its performance definitely dips, the impact is around 5% off of peak.
Intel and NVIDIA are really to thank here as both Ivy Bridge and Kepler are really mobile focused. Kepler in particular was a dramatic leap forward in power efficiency as we saw from our deep-dive on the architecture. Intel’s 22nm process, while relatively unimpressive on the desktop (a 10W savings isn’t anything to write home about in a tower) makes its first dramatic impact in the next-gen MacBook Pro.
Apple does play a role in the improved thermals of the Retina Display equipped MacBook Pro. Although very similar to the unibody chassis that Apple has been shipping for the past few years, the cooling system is significantly revamped.
2011 15-inch MacBook Pro heatsink, iFixit
The previous generation 15-inch MacBook Pro featured two fans, each cooling a heatpipe that ran across the CPU and GPU. The two were treated as thermal equals with neither residing in the heat shadow of the other. By running both heatpipes over both chips Apple guaranteed equal cooling when both chips were active, or quieter operation when only one was doing significant work. With the traditional MacBook Pro design, air was exchanged in a single location: along the display hinge.
The next-gen MacBook Pro still retains the two fan design, but there’s now a single large heatpipe that runs across both the CPU and GPU. Heatsink thickness appears to have been reduced as well, a concession made in order to decrease the overall z-height of the rMBP. The entire construction looks far more MacBook Air-like, which makes sense given the design of the notebook.
The rMBP heatsink, iFixit
Air is now moved through three sides of the notebook. Small slits on the left and right sides of the MacBook Pro act as intakes, while warm air is still exhausted out of the rear.
Apple also went to an new fan design with asymmetrically spaced blades. The idea is that by changing blade spacing you produce a sound that is spread out over more frequencies, and thus appears quieter.
The combination of silicon changes, side intakes and fan design results in three things:
1) Under light use the rMBP’s fans are virtually inaudible compared to the previous generation. In a quiet room you can still hear them spinning but it’s a significant improvement over last year’s model. Originally I thought the fans simply spun slower but running a light workload both last year's MBP and the rMBP spin their fans at 2000 RPM. The rMBP's fans simply sound quieter.
2) Under moderate use the fans in the rMBP are generally quieter than in the previous generation. Again this is likely due to the 22/28nm silicon in combination with the improved cooling system.
3) Under heavy use the rMBP fans are definitely audible, but the sound is different than what you get with the previous model. I was disappointed at first because I was expecting a dramatic reduction in noise, but instead I got the sort of improvement that you really notice once you go back to the older system.
To give you a sense of how the two sound I’ve embedded two audio clips of the 2011 15-inch MacBook Pro and MacBook Pro with Retina Display running a multithreaded Cinebench 11.5 test. In both clips you can hear me open the lid on the sleeping machine, quit open apps, use Spotlight to run Cinebench and click to start the benchmark (and click to repeat the test throughout the recording).
Neither system can make it through a single run of this test without ramping up the fans but note how quiet the Retina MBP remains in the early part of the test. Although the rMBP fans sound louder towards the end of the recordings, in person they actually sound quieter. Unfortunately I didn't have a good setup to really convey that.
Many have you have asked for specifics on thermal and acoustic performance. Here you go.
I ran both the upgraded 2011 MacBook Pro and Retina MacBook Pro through the same Half Life 2 torture test I described above, but at even more stressful GPU settings (1680 x 1050 with 4X AA enabled). I measured surface temperatures at the top and bottom of the notebook, as well as noise 3 inches above the trackpad on both systems. The results are in the table below:
|2011 MacBook Pro (Core i7 2.3GHz, 6750M)||Retina MacBook Pro (Core i7 2.6GHz, GT 650M)|
|Max CPU Temp||75C||63C|
|Max GPU Temp||69C||72C|
|Max Surface Temperature (Top)||47.3C||49.8C|
|Max Surface Temperature (Bottom)||49.8C||41.8C|
|Max Noise (3" Above Trackpad)||53.4dB||51.3dB|
In this particular test we see the two machines trade blows. The Retina MBP actually has a higher surface temperature on the top of the machine, but the underside remains significantly cooler. I believe the tradeoff is honestly worth it as the new machine has a tangible impact on how hot your lap gets. The Retina MBP is also noticeably quieter during this run, although it does technically let the GPU get a little hotter than last year's system. CPU temperatures are down considerably as well but we're also comparing 32nm vs. 22nm silicon at potentially different leakage levels.
Overall I’d consider the acoustic and thermal improvements here to be significant. Apple’s Phil Schiller got some of his strongest applause from the audience when he mentioned the fan design at the rMBP’s WWDC introduction. It’s attention to detail that Apple earns a lot of bonus points for; the fan design in the next-gen MacBook Pro isn’t revolutionary but it’s a tangible improvement that I can appreciate.
The next-gen MacBook Pro is equipped with a decidedly this-gen wireless stack. In other words it uses the same 3x3:3 WiFi solution that was present in the 2011 MacBook Pro and is present in the non-Retina 2012 MacBook Pro as well: Broadcom’s BCM4331. The wireless behavior characteristics are a bit different since this is a physically different chassis, but we’re still dealing with a 3-stream 802.11n solution - not 802.11ac. All three antennas are located in the Retina Display’s housing.
We have seen Apple be conservative with component choices in the past. Deciding to stick with Samsung’s 45nm LP process for the A5X instead of embracing 32nm LP sooner with the 3rd gen iPad is one example that comes to mind. Like a good silicon company Apple appears to mitigate risk in design by sticking with known-good components wherever possible. Major changes to the industrial design are typically paired with comparatively minor silicon changes, and other components are kept as static as possible so long as they don’t overly compromise experience. While 802.11ac dongles and routers are just arriving today, Apple likely froze the Retina MBP’s wireless configuration quite a while ago. Rather than be caught shipping potentially unratified hardware, Apple went the safe route and stuck with 802.11n.
That’s not to say Apple’s wireless implementation is bad. The 15-inch MacBook Pro has been one of the best behaved notebooks on wireless that I’ve had the pleasure of using. The MacBook Pro with Retina Display is no different. Just like before, the best case negotiated physical rate is 450Mbps when paired with a 5GHz 3x3 access point. Unobstructed, within a couple of feet of the AP, I measured as much as 230Mbps to the Retina MacBook Pro. I tested at three different distances from the AP, through walls and on both 2.4GHz and 5GHz bands. Overall performance seemed comparable to the standard 15-inch MacBook Pro, although it’s definitely faster in some areas and slower in others.
|Location 1||Location 2||Location 3|
|2011 MacBook Pro (2.4GHz)||124.0 Mbps||12.6 Mbps||61.6 Mbps|
|Retina MacBook Pro (2.4GHz)||117.9 Mbps||87.6 Mbps||44.0 Mbps|
|2011 MacBook Pro (5GHz)||186.8 Mbps||154.6 Mbps||24.7 Mbps|
|Retina MacBook Pro (5GHz)||227.7 Mbps||156.8 Mbps||33.7 Mbps|
The second test location consistently performed poorly on the 2011 MBP, only on 2.4GHz however. For the most part there were no real surprises otherwise.
The SD Card Reader
It was our own Brian Klug who clued me into the horrible behavior of the 2011 MacBook Pro’s SD card reader. Depending on the SD card used, the integrated SD card reader either performed admirably or was the most frustrating part of the Mac experience. Out of the three SD cards I frequently use: a Patriot LX series card, a Transcend and a new UHS-I Patriot EP Pro, only the Transcend card actually works remotely well with the 2011 chassis. Even then, it’s not perfect. I usually have to insert and remove the card at least once before the reader will recognize it. The LX and EP Pro on the other hand are measurably worse. To get the EP Pro to work in the 2011 MBP’s reader I usually have to push the card in then apply upward or downward force to the exposed edge of the card to get it to read properly. Even then it’ll usually disappear from OS X or be present but read at bytes per second. I doubt this is the fault of the card itself but rather the latest example of incompatibility with the horrible SD card reader in last year’s MacBook Pro.
At least with the cards I’ve tested, the Retina MacBook Pro exhibits none of these issues. Over dozens of insertions I had no issues reading from or writing to all three of these cards, including the problematic ones. I ran a Quick Bench test on the EP Pro as it’s the fastest of the lot and came away with reasonable performance as well. Roughly 80MB/s reads and 40MB/s writes. The numbers are shy of Patriot’s 90/50 spec but quite good.
One of the times I was able to get the EP Pro working in the 2011 MacBook Pro I managed to squeeze in a single Quick Bench run. Read performance was almost identical at 80MB/s, but write performance was far lower at only 10MB/s:
Shortly after the test completed I could no longer write to the drive in the 2011 MBP so I suspect the card reader was acting up again. Needless to say, if you like using SD cards with your MacBook Pro the Retina Display model appears to be much better. That’s not to say there couldn’t be other incompatibilities, but in everything I tested it looks like this problem is finally fixed.
Better Speakers and Dual Mics
Apple is proud of its new speaker design in the Retina MacBook Pro. There’s not a whole lot you can do for tiny laptop speakers but despite shrinking the overall volume of the chassis, Apple has managed to deliver much better sound out of the new speakers in the rMBP. Like most of the upgrades to the next-gen MacBook Pro, you really need to do an A/B comparison to appreciate the difference. And keep your expectations in line with reality, a good set of external speakers are always going to sound better. With that said, the new speakers definitely deliver a fuller, more rich sound than their predecessor. You can still tell you’re listening to some form of integrated speakers, but now they sound distinctly less like they’re coming from a inside a notebook.
In preparation for Mountain Lion's arrival with dictation support, Apple outfitted the next-gen MacBook Pro with dual microphones in order to better focus on your spoken voice and not on background noise. In practice the new mics work reasonably well, rejecting moderate volume background noise. Loud music nearby will still cause interference and as always, accurate dictation requires more than just good quality source audio to get right.
The Retina MacBook Pro can complete a full boot from power off to usable desktop in just over 17 seconds. It’s a hair faster than last year’s MacBook Airs, a bit quicker than the old SSD equipped MacBook Pro, and night and day compared to any Mac with a hard drive. Four years ago I said that Solid State Drives were the single biggest upgrade you could do for your computer, and it couldn’t be any more true today.
The Retina MBP behaves more like the 3rd gen MacBook Air in how it goes to sleep and wakes up. Both happen virtually instantaneously, and if your battery dies while asleep you don’t get the greyed out screen with progress bar as your environment is restored from disk - it just appears, taking a couple of seconds for the clock to update and everything else to come to life. It’s a small but subtle change that tells you the rMBP is in a distinctly different class. Apple’s tight control over firmware and storage interface help it deliver up to 30 days of standby power, a number I’ll really need to verify one of these days.
Raw CPU performance is up handsomely over the 2011 MacBook Pros, at least in their standard configurations. The 2.6GHz chip in the $2799 rMBP can turbo up to 3.6GHz when only a single core is active, delivering a 13% increase in performance over the previous generation 2.2GHz part. Compared to the upgraded 2.4GHz Sandy Bridge Core i7 from last year (not pictured) however, I would expect a sub-10% advantage.
With all four cores active the 2.6GHz chip can run at up to 3.4GHz, which it regularly hits as long as the Kepler GPU stays asleep. The 2.2GHz Sandy Bridge based 2011s we’re comparing to on the other hand can only turbo up to 2.8GHz. Here the advantage is a more tangible 19%, although once again if you are comparing to one of the 2.4GHz parts from last year I would expect notably smaller gains (mid to high single digit percentages). The improved thermal characteristics may allow mobile Ivy Bridge to operate in turbo modes for longer than Sandy Bridge, however I don’t have any data to actually support that claim. That doesn’t mean it can’t happen, it’s just complex to test and model.
While the Cinebench tests are largely CPU bound, many of the following tests are largely influenced by the SSD in the Retina MacBook Pro. Here we see some huge gains, especially compared to older HDD based Macs.
iMovie import time is heavily influenced by disk as well as CPU performance. As a result there are big improvements over both the HDD and SSD equipped 2011 MBPs.
Export time is more heavily CPU bound and here the advantage over the previous generation notebooks is pretty much nonexistent.
Our iPhoto import test stresses both disk and CPU, giving the rMBP a tangible advantage compared to the SSD equipped 2011 MBP. None of the dual-core or HDD based Macs stand a chance here.
Our Lightroom test continues the storage/CPU dependencies as only the SSD equipped 2011 MBP is able to come close to the rMBP’s performance. There’s not much of a performance advantage here when you compare similarly equipped systems though. Ivy Bridge may have been a good upgrade from a power standpoint, but it doesn’t tell a significantly different performance story in all cases.
The rMBP cranks through our Photoshop workload fairly quickly. The performance advantages here are likely due to increased memory, a much faster SSD and obvious CPU speed improvements as well.
The same holds true for the advantages in our Final Cut Pro X test. As a default configuration the $2799 MacBook Pro with Retina Display is easily the fastest notebook Apple has ever shipped. It’s only if you had an upgraded 2011 model (perhaps with an aftermarket SSD?) that you’ll be unimpressed by the move.
I can’t stress enough how much the new SSD improves the overall experience. It’s just so much faster than what Apple used to ship.
We’ve already established that NVIDIA’s Kepler architecture is fast, but the GeForce GT 650M used in the rMBP is hardly the best NVIDIA has to offer. The result however is a significant improvement in performance over the Radeon HD 6750M used in the previous generation model.
|15-inch MacBook Pro Model||Mid 2010||Upgraded Early 2011||Upgraded Late 2011||Retina|
|GPU||GeForce GT 330M||Radeon HD 6750M||Radeon HD 6770M||GeForce GT 650M|
|Memory Bus||128-bit GDDR3||128-bit GDDR5||128-bit GDDR5||128-bit GDDR5|
|Memory Data Rate||1580MHz||3200MHz||3200MHz||5016MHz|
The GT 650M offers fewer “cores” compared to the 6750M and 6770M used in previous MacBook Pros, but likely better utilization of the available hardware. NVIDIA also clocks the cores much higher in the 650M, the result is a ~20% increase in theoretical raw compute power.
The memory bandwidth story is also better on Kepler. While both the GT 650M and the 67xxM feature a 128-bit GDDR5 interface, Apple clocked AMD’s memory interface at 800MHz compared to 1254MHz on Kepler. The resulting difference is 80.3GB/s of memory bandwidth vs. 51.2GB/s.
The real world impact is most noticeable at higher resolutions, thanks to the tremendous amount of memory bandwidth now available. The other benefit from the new GPU is obviously things run a lot cooler, which as I’ve already shown to considerably reduce thermal throttling under load.
At 1440 x 900 we actually see a regression compared to the 2011 models, but differences in the AMD and NVIDIA GPU drivers alone can account for the difference at this hardly GPU bound setting. Look at what happens once we crank up the resolution:
At 1680 x 1050 with 4X AA enabled we see a modest 11% increase in performance over last year's MacBook Pro. As I established earlier however, the rMBP will be able to more consistently deliver this performance over an extended period of time.
What's even more impressive is the 42.4 fps the GT 650M is able to deliver at the rMBP's native 2880 x 1800 resolution. Even though I ran the test with AA enabled I'm pretty sure AA was automatically disabled. At 2880 x 1800 the rMBP is able to outperform the two year old MacBook Pro running at 1680 x 1050. How's that for progress?
While the gains we've shown thus far have been modest at best, Starcraft 2 is a completely different story. Here for whatever reason the IVB + Kepler combination can be up to 2x the speed of last year’s models. I reran the tests both on the older and rMBP hardware to confirm, but the results were repeatable. The best explanation I have is Starcraft 2 is very stressful on both the CPU and GPU, so we could be seeing some thermal throttling on the older SNB + Turks hardware here.
Once again we see playable, although not entirely smooth frame rates at 2880 x 1800. I've also included a screenshot of SC2 at 2880 x 1800 below:
Although gaming options continue to be limited under OS X, Diablo 3 is available and finally performs well on the platform thanks to the latest patches. Diablo 3 performance is appreciably better on the GT 650M compared to last year’s 6750M. There’s no FRAPS equivalent under OS X (free advertising to the first eager dev to correct that) so I have to rely on general discussion of performance here. The GT 650M is fast enough to drive the rMBP’s 2880 x 1800 panel at native resolution at playable frame rates, around 18 fps on average. Connected to an external 2560 x 1440 display however the GT 650M is fast enough to deliver around 30 fps in Diablo 3. For what it’s worth, performance under Diablo 3 is far more consistent with the rMBP than with last year’s MacBook Pro. I suspect once again we’re seeing the effects of thermal throttling under heavy CPU/GPU load that has been well mitigated by the move to more power efficient silicon.
For much of the past year I haven’t been pleased with just how good Apple’s caching has become both on OS X and iOS. Aggressively caching our test web pages produces artificially inflated battery life numbers and that’s no fun for anyone. I’m happy to say that I’ve fixed that problem with our OS X battery life tests.
The suite is completely redone although conceptually it’s quite similar to what I’ve run in the past. I have three separate workloads: light, medium and heavy, each one representing a different stress level on the machine and all three giving you a decent idea of the dynamic range of battery life you can expect from one of these notebooks. All three tests are run with the displays set to 100 nits (a little above the halfway brightness point on most MacBook Pros).
The light and medium suites are inherently related - they use the same workload and simply vary the aggressiveness of that workload. The light test hits four different websites every minute, pausing for nearly the entire time to simulate reading time. Flash is enabled and present on three of the sites. The long pause time between page loads is what really makes this a light test. Web browsing may be the medium for the test but if all you’re doing is typing, watching Twitter update and maybe lazily doing some other content consumption this is a good representation of the battery life you’ll see. It’s a great way of estimating battery life if you’re going to be using your notebook as a glorified typewriter (likely a conservative estimate for that usage model).
The medium test hits the same webpages (Flash and all) but far more aggressively. Here there’s less than 10 seconds of reading time before going onto the next page. It sounds like a small change but the impact on battery life is tremendous.
Both the light and medium tests are run in their default state with processor graphics enabled, as well as with the discrete GPU forced on. I run with the dGPU on as well because it’s far too often that a single application open in the background will fire up the dGPU and contribute to draining your battery. The goal here is to deliver useful numbers after all.
The final test is very similar to our old heavy multitasking battery life tests, but with some updates. Here I’m downloading large files at a constant 1MB/s from a dedicated server, while playing back a looped 1080p H.264 movie (the Skyfall trailer) all while running the medium battery life test. The end result is a workload that gives you a good idea of what a heavy multitasking usage model will do in terms of battery life. I’ve found that OS X tends to fire up the dGPU anyway while running this workload so I saw no reason to run a separate set of numbers for processor and discrete graphics.
Overall the rMBP pretty much behaves as expected. Apple claims up to 7 hours of battery life and using our light workload we see a bit over that. Fire up the dGPU and even a light workload will get cut down to around 5.5 hours. Moderate usage will drop battery life to around 5 hours, and if you fire up the dGPU you’ll see that cut down to 3.5. The heavy multitaskers in the audience will see a bit above 2 hours out of a single charge. Note that all of these numbers are at 100 nits, drive the 2880 x 1800 panel at its full brightness and you can expect a tangible reduction in battery life.
The rMBP’s integrated 95Wh battery is ginormous by today’s standards, but it’s really necessary to drive both the silicon and that impressive panel. Subjectively I did find the rMBP lasted longer than last year’s MacBook Pro, despite the similar max battery life ratings. My experiences were echoed by the results in our tests.
I suspect most users will see around 5 hours of battery life out of the system compared to a bit under 4 hours out of last year’s machine. At minimum brightness, typing a long document (similar to what I’m doing right now) you can significantly exceed Apple’s 7 hour estimate. As always it really depends on usage model. Professional users doing a lot of photo and video editing aren’t going to see anywhere near the max battery life, while the writers and general users will be quite happy.
One trick to maximizing battery life on light or moderate workloads is to keep an eye on what the discrete GPU is doing. I still find that OS X will wake up the discrete GPU far too frequently, even when in my opinion its services aren’t needed. As always I turn to Cody Krieger’s excellent gfxCardStatus app for keeping an eye on which GPU is driving the panel. The app has been updated and is now fully compatible with the rMBP.
What to Buy
I mentioned earlier that Apple mitigates risk in its designs much like a CPU maker. It’s always considered good practice to decouple major architecture changes from process node jumps (Intel’s famous tick-tock cadence embodies this). Apple similarly tries to alternate major changes to the industrial design from significant internal component changes. Although Ivy Bridge and Kepler are all new, the former is quite similar to Sandy Bridge while the latter is really no different than integrating any other discrete GPU. The more dramatic silicon departure comes with Haswell next year, and I suspect that’s why we got the rMBP this year.
In our performance investigations I mentioned that compared to an upgraded Sandy Bridge MacBook Pro (high clocks with SSD), you won’t see tremendous performance gains from the rMBP. A quick look around Apple’s website actually shows not even a single CPU bound performance comparison between the rMBP and last year’s MacBook Pro.
The logical thing to do, if you’re the owner of a recent (2010/2011) MacBook Pro, is to wait until next year at the minimum. Haswell should bring a significant performance increase (particularly on the processor graphics front) and you’ll get it in the same chassis as what you see today.
Most users however don’t upgrade annually. If you have an older MacBook Pro, the rMBP offers all of the benefits of last year’s Sandy Bridge upgrade but in a much better package, and with vastly improved thermal characteristics. If you fall into this group, the upgrade is a no-brainer. I won’t lie, the next two years are going to be tough. Haswell is looking very good, and if Intel can pull off 14nm on time, Broadwell will be even more impressive from a graphics standpoint. You can always make the argument to hold off on an upgrade as there’s almost always something better around the corner. In my opinion you really can’t go wrong picking any of the next three years to upgrade.
Should you decide to buy today, which model should you get?
As I mentioned before, the $2199 configuration is near-perfect in my opinion, save for the 256GB of NAND flash. Apple unfortunately won’t let you upgrade storage capacity on the base MacBook Pro with Retina Display so you’re left with two options: 1) live with the 256GB and hope someone will build an aftermarket SSD in the not too distant future, or 2) buy the $2799 model. While it’s quite likely that we’ll see third party SSDs for the rMBP, I seriously doubt you’ll find one with Samsung’s PM830 controller.
I do think 256GB is livable, it’s just that 512GB is so much more comfortable.
Apple has simplified things by not allowing multiple GPU options, and the CPU options are pretty cut and dry.
If you can live with 256GB of storage, the $2199 configuration is fine. Otherwise I’d go with the upgraded $2799 model.
The question of whether or not you should opt for the 16GB memory upgrade really depends on what you do with the system and how long you expect to use it. Without any form of socketed memory expansion, you’re stuck with the amount of memory you order on the system. Thankfully 8GB is healthy by today’s standards and likely will continue to be so for the next couple of years. If your present day workloads require 8GB of memory, then the 16GB option is a must have. If you’re looking at 16GB purely as future-proofing, chances are you’ll run into processor (or storage) limitations before you feel held back by memory. That being said, if you want to be kind to the next owner, ticking the 16GB box won’t hurt.
Apple has done an incredible job with the next-gen MacBook Pro. It brings a level of portability to the 15-inch chassis that we’ve never seen before from Apple, all while getting a good handle on some of the thermal and noise issues from last year’s model. If you’re like me and have to lug your 15-inch MBP around, the improvements in portability alone are worth the upgrade. But a lighter chassis is hardly all Apple is relying on to sell this system.
The internals are easily the best collection of parts Apple has ever assembled. Ivy Bridge and Kepler are natural fits, but shipping the machine with 8GB of memory by default is a much appreciated gesture especially considering its un-upgradeable nature. For the first time in Apple’s history of shipping NAND flash based storage in Macs, I actually have no complaints about the controller choice in the rMBP. Samsung’s PM830 (or the consumer, SSD 830, version) is what I’ve been recommending to Mac users for much of the past year. It’s still possible that you’ll end up with a non-Samsung controller, and I don’t yet know whether or not that’s a bad thing, but this is at least progress.
The connectivity story on the rMBP is near perfect. The pair of Thunderbolt ports allows extra flexibility as well as the ability to drive more bandwidth to external IO than any prior portable Mac. The Thunderbolt teething issues still remain unfortunately, but it looks like that’s going to require at least a partial act of Intel to rectify. USB 3.0 is a welcome addition to the Mac family. It took both Apple and Intel far too long to get to this point, but I’m glad it’s here.
All of this is really just wrapping however, as the real gift is the MacBook Pro’s first Retina Display. It’s easily the most beautiful display I’ve had the opportunity of using. Even more impressive to me than the iPad’s Retina Display, and enough to make me actually want to use the Mac as a portable when at home rather than tethered to an external panel. The added portability of the chassis likely contributes to that fact though.
The credit Apple deserves for the display extends beyond simply pushing LG to get a panel out on time and in large enough quantities. There’s a tremendous amount of software work that Apple put into making the Retina experience work under OS X. The OS and several key applications have been updated to properly support the MacBook Pro’s Retina Display, and things can only get better from here. Mountain Lion will improve performance and I would expect at least a few key app updates over the next year to bring increased Retina awareness.
There’s also the behind the scenes work Apple put in to make all of this happen. The pressure on the GPU vendors, as well as taking matters into its own hands with writing scaling and filtering routines to deliver a good experience are all noteworthy.
It’s because all of this that I’m doing something I’ve never done before in an Apple review. We rarely give out Editor’s Choice awards at AnandTech, and I’m quite possibly the stingiest purveyor of them. I feel that being overly generous with awards diminishes their value. In this case, all of the effort Apple has put into bringing a Retina Display to the MacBook Pro is deserving of one.
I’m giving the MacBook Pro with Retina Display our bronze Editor’s Choice award. Making it the first Mac to ever receive one. It would have been a silver had the software story been even stronger (iWork, Mountain Lion, Office and Photoshop being ready at launch would have been a feat worth rewarding). And it would have been a gold had Apple been able to deliver all of that but without sacrificing end-user upgradability. Which brings me to my final point.
I accept the fact that current mobile memory and storage form factors preclude the creation of the thinnest and lightest form factors. But I would like to see Apple push for the creation of industry standard storage and memory form factors that wouldn’t hinder the design of notebooks like the Retina Display equipped Macbook Pro. As Apple has already demonstrated that it has significant pull with component vendors, this should be possible. The motivation behind doing so is no different from the motivation driving the use of Retina Displays: for the betterment of the end user experience.
Sidebar: Impacting the Rest of the Industry
Apple’s impact on the industry has already been felt. The threat of Apple bringing Retina Displays to its entire lineup forced ASUS’ hand and gave us 1080p IPS panels in the new Zenbook Primes. This will undoubtedly continue. In the early days Apple simply raised the bar for a focus on industrial design. Apple’s influence quickly expanded to touch everything from packaging to trackpads. We’re now seeing PC OEMs focus far more on experience than they ever have before. Apple isn’t the only one to thank for this, but the company is a significant factor.
The fact of the matter is the days of blaming a lack of innovation on cost or the inflexibility of one’s suppliers are over. In fact, those days are long gone. Today the MacBook Pro with Retina Display exists at a very high starting price, but make no mistake, it won’t remain there indefinitely. Apple introduced this model as the next-generation MacBook Pro because it truly is a preview of what’s to come. Maybe next year’s model won’t be any cheaper, but the one after that definitely will be. Apple has a healthy obsession with high quality displays and it will put its might behind panel suppliers until it can put forth a lineup of top to bottom Retina Displays. There’s no doubt in my mind that within the next 12 - 24 months Apple will introduce an external 4K Retina Display. Whether you love, hate or are indifferent about Apple and its products, its impact on the industry is tangible. PC OEMs now care about display quality and keyboard feel. They care about trackpads and design. There’s only one motivator in this industry stronger than Moore’s Law: experience, and the PC OEMs finally care about that too.
Apple’s success hasn’t been because it is a vertically integrated company. On the contrary, everything Apple has done Acer, Dell, HP, Intel, NVIDIA and Microsoft could have done together. Apple is successful because its competitors have all been selfishly focused on themselves rather than all coming together to build better computers. Based on my conversations with Intel and some of the OEMs at Computex earlier this month, the wake up call has been heard. Intel seems quite motivated to help its OEM partners do better. It is a bit troubling for the ecosystem that Microsoft is throwing its hat into the ring as a competitor - especially as it was Microsoft's inaction on the software side that really hurt the PC OEMs over the past several years.
For years we’ve been pushing OEMs to focus on better displays, and for years we were given cost and customers-don’t-care as excuses for why we don’t get them. That’s all starting to change.