Done for 2009: The Holiday MacBook Pro Roundup
by Anand Lal Shimpi on November 10, 2009 12:00 AM EST- Posted in
- Mac
Ugh, SSDs in the MacBook Pro
Last summer I posted my MacBook Pro review and I soon started hearing complaints about the SATA interface only running at 1.5Gbps speeds and not 3.0Gbps. The SATA 3Gbps standard has been around since 2004/2005 and the NVIDIA chipset Apple uses in the unibody MacBook Pro supports 3Gbps mode.
To understand why Apple limited the machines to 1.5Gbps I first went to NVIDIA to see if they made a custom version of the GeForce 9400M chipset without SATA 3Gbps support. Apple's partners are always nervous about talking, so NVIDIA's response was carefully worded:
"[The] GeForce 9400M offers complete support for SATA Gen2. You need to ask Apple if you have any questions specific to their systems."
In other words, the chipset supports it, Apple did something funny with its systems to break support for it.
Apple eventually released a firmware update to enable SATA 3Gbps support, but the update carried a strange warning from Apple:
About MacBook Pro EFI Firmware Update 1.7
MacBook Pro EFI Firmware Update 1.7 addresses an issue reported by a small number of customers using drives based on the SATA 3Gbps specification with the June 2009 MacBook Pro. While this update allows drives to use transfer rates greater than 1.5Gbps, Apple has not qualified or offered these drives for Mac notebooks and their use is unsupported
Apple's official statement is that 3Gbps SATA drives aren't supported? What? This is 2009 right?
Now the performance difference between 3Gbps and 1.5Gbps SATA operating modes isn't noticeable in real world usage. You need an SSD to consistently saturate 1.5Gbps SATA and even then, it is only in sequential reads. In fact, it's not the performance loss of going down to 1.5Gbps SATA that I care about. It's the next problem.
This is an OCZ Agility SSD, it's based on the Indilinx Barefoot controller. With the 1.5Gbps version of the unibody MacBook Pro firmware (EFI v1.6) this drive doesn't work properly. Given enough disk activity and the machine will freeze, requiring a hard reset (hold down the power switch for a few seconds). I know because I put one in my 15-inch MacBook Pro.
OCZ's Agility. A great drive, it just doesn't work in a 2009 unibody MacBook Pro under OS X.
Updating the MacBook Pro to EFI 1.7, thus enabling 3Gbps operation, makes it even worse. You can't even reliably install OS X half the time.
This is a SuperTalent UltraDrive GX, another Indilinx Barefoot based SSD:
It, too, doesn't work in the unibody MacBook Pro. Once more, I know because I tried it in my own personal system. In fact, the only non-Apple supplied SSD I could get to work in the system without constant crashes was Intel's 80GB X25-M G2 (I didn't try the G1 or the 160GB G2).
I asked Apple what was going on and the only response I got was that the drives I was using weren't supported. These drives work perfectly in my testbeds and in other notebooks and my Nehalem Mac Pro, but not in my unibody 15-inch MacBook Pro.
I've been a staunch advocate of Apple's hardware and software for years now, but this is a dangerous precedent that Apple is setting. The point of standard specifications is so that end users can buy any component adhering to the spec knowing that it will work in any system implementing the spec. This isn't an OS X incompatibility, this isn't even an Apple incompatibility, this is a specific issue between certain SSDs and the 2009 unibody MacBook Pros. Instead of admitting to that and committing the resources to fixing it, Apple is making the mistake of stating that non-Apple hardware isn't supported. These are storage devices, the only things that you absolutely didn't have to have an Apple logo on in order to guarantee operation.
Note that this wouldn't even be an issue in the first place if Apple offered halfway-decent SSD options on its systems.
If it were an Indilinx problem, why not work with Indilinx to fix it? There's a definite incompatibility but I'm willing to bet it is related to whatever kept Apple from enabling 3Gbps on these machines from the start. Some have speculated that the custom SATA cable Apple runs from the motherboard to the drive bay is to blame. It's the only explanation given that these problems only happened in the 2nd generation unibody MacBook Pro systems.
It's unacceptable and I'm guessing the next version of the MacBook Pro will magically fix all of the problems.
Update: I've received at least one report of the latest Indilinx firmware fixing the issues outlined above. I tested with the latest firmware in both of the examples I used, but there's been some theorizing that the problems are related to the chipset/motherboard which would explain the variance between systems. At the same time I've received additional reports of Indilinx drives not working from other users so unfortunately it appears that there's no complete solution to the problem at this time. The recent report of it working shows that there may at least be hope if you get the right combination of drive and MacBook Pro.
In case you're wondering, Apple's recently released OS X "Performance Update" that was designed to address drive stuttering issues does nothing for SSDs - it was simply intended to fix issues with mechanical drives. I confirmed this with Apple.
Unfortunately that leaves users with very few options. You could either buy the overpriced SSDs directly from Apple, or give the third party versions a try (presumably OCZ's Summit and Corsair's Performance series use a similar enough controller that they should work). I've been using the X25-M G2 in my 15-inch MBP for a little over a week now and haven't had any issues with the EFI 1.7 update. You could stick with a hard drive but anyone who has gone down the SSD route knows that's not a good option. Or just wait and hope Apple fixes it with the Nehalem update next year.
If you haven't heard, I'm a bit coo-coo for SSDs, because the performance improvement you can get from a good SSD is just awesome:
| 15-inch MacBook Pro | 250GB HDD | 80GB Intel X25-M G2 SSD |
| Time to Launch 15 Applications Sequentially | 40.7 seconds | 16.0 seconds |
Application launch time is reduced significantly, in general everything opens like it was just in memory. CPU intensive tasks don't see an improvement but day to day usage is where you'll notice it the most. And that's where it's most valuable honestly.
Facebook
Twitter
RSS
115 Comments
View All Comments
ltcommanderdata - Tuesday, November 10, 2009 - link
I believe Anand did comment on screen resolution on page 4 where he found the 15" resolution just right for the screen between space and not being an eyestrain. Personally, I agree that 1440x900 on a 15" screen is about right. I've found 1680x1050 on a 15" screen to be a bit too dense. Although that just brings up Apple's lack of progress on resolution independence which has been in development since Tiger.On the topic of screens, I wonder what are the chances that Apple will move to 16:9 screens as on the iMac and as other notebook manufacturers are doing. I hope not as 16:10 is wide enough and it's not like watching movies is the only thing people do on their laptops. Besides, for the creative industry users working on 16:9 content, 16:10 screens should make sense to leave room for toolbars and such for editing.
secretanchitman - Tuesday, November 10, 2009 - link
agreed. 1680x1050 on the 15" would be absolutely perfect. i would rather have 16:10 rather than 16:9.i would also love to see core i5/core i7 (doubtful on core i7 though), a much faster gpu (next gen nvidia/ati), and the expresscard slot to make a return as an option. also, most of all, PLEASE make sure the matte screen and silver bezel return!
darwinosx - Tuesday, November 10, 2009 - link
I like the current resolution. Anything higher would be too high for me.ltcommanderdata - Tuesday, November 10, 2009 - link
I don't know exactly how it works, but I've read before that 64-bit Safari is able to use a 32-bit Flash plugin through InterProcess Communications (IPC). Presumably, this causes more overhead and the 32-bit Flash plugin wasn't designed/optimized for this in mind, so that could explain why battery life is so poor under 64-bit Safari with Flash.In terms of Macs, I have to agree that qualities like battery life, weight, look, and feel are things that can be worth paying extra for even if they are harder to definitively measure and compare.
I wonder what Apple is going to do about the IGP situation when they move to Arrandale? It definitely seems strange that nVidia's CEO picks now of all times to advertise his love for Macs after ATI pushed nVidia out for discrete GPUs for the iMacs and on the eve of Arrandale's IGP and DMI link. Presumably Apple could consider using switchable graphics between Arrandale's IGP and discrete GPUs even for low-end models. The effort put into Intel IGP drivers in Snow Leopard bringing the GMA X3100 up from OpenGL 1.2 support in Leopard to OpenGL 2.0 support seems to indicate a readiness of Intel IGPs. It will of course force the introduction of another chip on the motherboard, which is a concern for the space constrained 13" MacBook Pro.
If Intel won't license DMI to nVidia, I wonder if Intel would license it to their good buddy Apple. Apple could then license nVidia graphics technology to serve as the IGP for their own custom chipset. Apple used to design their own chipsets in the PowerPC days and certainly have the resources to do so now with the P.A. Semi team. Besides a better IGP, a custom chipset would also allow Apple to integrate functionality currently done by external chips like Firewire controllers and the multi touch touchpad controller as well as adopt new technologies like USB 3.0 and Lightpeak without waiting on Intel. I believe Jobs did say in a previous conference call their intention to differentiate themselves by offering technology that no one else has and this is certainly one way to do it.
This article seems to mention Arrandale a lot as the future processor of the MacBook Pro, but hopefully Clarksfield will also be available on the 17" and high-end 15" models. Clarksfield clock speeds are disappointing though (1.6GHz - 2GHz) so it'll be interesting to see how Apple plays it.
JarredWalton - Tuesday, November 10, 2009 - link
Clarksfield is a power hog, and there's no getting around that. I've got an i7-720QM based laptop and if I use absolutely minimal performance settings (CPU at 0%, HDD turn off at 1 minute, DVD and webcam turned off when on battery) I barely get 90 minutes of battery life with a 55Wh battery.I can't imagine Apple will ever use Clarksfield in a laptop, because they are targeting mobile battery life far more than mobile performance. Frankly, even the top-end 17" MacBook Pro is pretty weak in many of the performance areas we'd look at on Windows. A 9600M GPU is nothing to write home about -- 32 SPs is what we had back in the days of the GeForce 8600. OS X seems more dependent on CPU, though.
I'm also curious about which CPUs Apple is using; previously they used the SL9000 and SP9000 I'm pretty sure, but now it looks like they're using standard P8000, P9000 and even T9000 parts (though only on the 15" and 17" for the T9000 I'm sure). Maybe Anand can say which specific CPUs are in the Macs now.
ltcommanderdata - Tuesday, November 10, 2009 - link
I didn't know Clarksfield battery life was so bad at idle/light usage. I thought that peak power consumption would be higher, but idle/light power would be comparable or better then high-end mobile Core 2 Duos due to the Power Control Unit. I guess Clarksfield could serve as the real test of Apple's power management abilities. Although with Anand finding that a MacBook Pro gets 78% better battery life in OS X than Windows 7, and a 17" MacBook Pro having a 95Wh battery compared to your 55Wh battery, if you get 90 minutes with the i7-720QM, Apple could get around 4.6hrs in OS X on a 17" MacBook Pro with that processor which is okay. Although Clarksfield would probably be limited to BTO only to avoid confusion.And I agree that Apple's GPU choices are generally questionable, especially seeing OS X's heavy reliance on the GPU starting with Quartz Extreme acceleration in Jaguar in 2002, Core Image in Tiger in 2005, and now OpenCL in Snow Leopard. I guess the only small consolation is that Apple always uses at least GDDR3 with their GPUs, even low end ones like the HD2400XT in the iMac, and doesn't resort to advertising high VRAM capacities and then sneaking in low clock speed DDR2. Hopefully, Apple's notebook refresh jumps directly to DX11 GPUs, presumably ATI since they seem to be first out of the gate. Preferably, they will go with a high mid-range GPU, although I guess heat and power and always Apple's concerns in a 1" case.
darwinosx - Tuesday, November 10, 2009 - link
OS X is not "dependent" on cpu but it can use the cpu for things other than standard vid card work.dagamer34 - Tuesday, November 10, 2009 - link
Clarksfield is a 45W CPU, which Apple will never use due to heat concerns.I'm really hoping that Apple changes the screen to a 16:9 ratio, which would be perfect for watching HD content. Something like: 1366x768 for the 13", 1680x950 for the 15" and 1920x1080 for the 17" would be a great improvement.
kapute - Wednesday, November 11, 2009 - link
16:9 screen ratio is terrible idea for laptops. Ok for watching movies but not when using a word-processor as all the tool bars etc diminish vertical space leaving a tiny letter box to type in. Better to have black bars top and bottom when watching a movie and more vertical space for everything else.ltcommanderdata - Tuesday, November 10, 2009 - link
Well, I'm pretty sure Apple's top end CPUs like the 2.8GHz Core 2 Duo and 3GHz BTO in the high-end 15" and 17" MacBook Pros are 35W TDP processors. The PM55 northbridge has a 7W TDP. It's hard to isolate a comparison to the 9400M chipset since it has both northbridge and southbridge integrated. Still, 35W CPU + ~7W for a northbridge compared to 45W for an integrated CPU + NB in Clarksfield doesn't seem unreasonable, from a heat dissipation perspective.