Saturday, September 23, 2006

Sad, sad iPod mini

This little trooper has worked tirelessly for me for the last 1.5 years, but alas, my 6 GB iPod mini took one too many falls. (Yes, he's hit the ground more than once.)

I couldn't help but like Apple's sad iPod icon though:

However, not all their icons tug at the heart as much:

As far as I can tell, the drive is the culprit. I tried substituting in a CompactFlash card, but as expected it didn't work.

Apparently, the Apple MicroDrives have a custom firmware so that Apple MicroDrives won't work in third party devices, and third party drives/flash cards generally won't work in iPods. It's too bad the flash cards don't work, because it wiould be great to have a completely solid state mini. I could buy a replacement Seagate or IBM MicroDrive (even up to 8 GB), but it looks like it's time to upgrade.

Now, what to buy... I would have bought a new widescreen iPod, but they don't exist. The 8 GB nano is a great little unit, but unfortunately, it only comes in black. The shuffle is a nice convenient little item given its size and built-in clip. However, I think I'll use my Sony Ericsson K790a phone for now, while I contemplate my options. Where is that Apple iPhone when you need it?

Thursday, September 21, 2006

Parallels 1898 eliminates 64-bit Mac kernel panics

Parallels Desktop for Mac is an interesting program that allows one to run a Windows virtual machine from within Mac OS X on Intel Macs. Unlike Boot Camp, there is no need to shut down Mac OS X and reboot into Windows.

I had tested it out on my MacBook, and it worked just fine. It's not as fast running Windows natively under Boot Camp, but because it's not running under emulation, it's much faster than running Virtual PC on PowerPC Macs. In fact, Parallels is almost fast enough to playback 720p HD H.264 video files on an Intel iMac. This is a huge performance increase over Virtual PC. Although Virtual PC is a remarkable program, speed is not one of its fortes.

Unfortunately, on my 24" iMac with 7600 GT, starting a Parallels virtual machine with the latest release version of Parallels immediately causes OS X to kernel panic. Fortunately, there is a release candidate 1898 build which solves the problem for 64-bit Intel Macs, including the new Core 2 Duo iMacs. However, it (reversibly) limits memory usage of the machine to 2 GB. That for now is not a significant issue for me, as I only have 2 GB anyway, but it is a major limitation for Mac Pro users who have more than 2 GB RAM. There is light at the end of the tunnel for Mac Pro users though. Build 1908 is an internal build that solves the memory problem. It may not be completely stable for some people however.

Another glitch on my 24" iMac is that networking in Parallels through wired Ethernet does not work in Windows XP. Fortunately, wireless works, so I can use that for software updates, etc. This is very important on Windows XP for keeping my virus checker up to date of course.

[Update 2006-09-22]

It seems that the 2 GB limitation of Parallels 1898 isn't really 2 GB. It's slightly less than that:

I suppose I can live with the 0.05 GB loss for now. ;)

[Update 2006-09-25]

I installed build 1910 yesterday, which removes the 2 GB memory limitation. I haven't tested it extensively, but it seems to work fine on my Core 2 Duo iMac.

Apple underclocking 7600 GT in 24 inch iMac?

According to this MacRumors forum post, the 7600 GT in the 24" iMac runs at only 500 MHz core and 575 MHz memory. There could be a couple of reasons for this:

1) Apple cheaped out and is using cheaper components
2) Apple felt the heat generated from the GPU at higher clocks could compromise the iMac's reliability.
3) Apple felt the heat generated from the GPU at higher clocks would cause the fan to come on too often and make the iMac louder.

Whatever the reason, at least under Windows XP, one can clock it back to usual PC 7600 GT speeds. Whether that is safe or not is unknown at this time. It would be nice if someone could write such a utility for Mac OS X, to compare performance benchmarks with say Cinebench, Aperture, and 3D games.

Monday, September 18, 2006

iMac Core 2 Duo Benchmarks: It's Fast!

I received my 24" iMac a few days ago, and had a chance over the weekend to work with it a bit and run some benchmarks.

I purchased a nearly fully tricked out machine:

24" 1920x1200 widescreen LCD
2 GB 667 MHz PC-5300 DDR2 SDRAM
Core 2 Duo Merom 2.33 GHz, with 4 MB L2 cache
nVidia GeForce 7600 GT with 256 MB GDDR3
500 GB hard drive
8X DVD±R/W dual-layer burner
Wireless keyboard and wireless Mighty Mouse

The first thing one notices upon opening the box is just how large is the screen. It's a beautiful sight to see that LCD screen in all its 24" glory. In fact, the screen is so large that once I got it on my desk, I found myself having to look up to see the top of it. The 'chin' at the bottom of the iMac made the viewable screen area a little too high for my tastes on my table. I had to lower my table and adjust my chair to compensate for the height of the iMac, since its height cannot be adjusted. It's still not ideal, but I think I'll get used to it. If not, I could always purchase the VESA mount and an arm, so that I can bring the Mac down an inch or so. It's good that Apple brought back VESA compatibility to this design.

The screen quality is impressive, especially for a consumer desktop. The colours are good for an LCD, and the contrast appears better than my previous machine. The brightness is much higher, and actually, it may be too bright for the room its in. I have it on the lowest setting and it's still very bright.

One nice thing is that in Firewire target mode, this machine remains very quiet. This is in stark contrast to my G5 iMac, which does a very good impression of a vacuum cleaner when in target mode. It sounded like a server room in my den when I was copying files off my G5 iMac using this feature.

Apple's account data transfer application works quite well. It's remarkable that apps built for PowerPC work just fine on this Intel iMac with just a direct file copy, with no tweaking necessary. Well, all was not perfect. My install of Photoshop didn't survive the transfer. It ran on the Core 2 Duo iMac but had a couple of weird bugs such as not being able to read .png files. However, I just recopied the application over from the G5 iMac, and all was fine again. No re-install required.

The keyboard and mouse pair up with the machine quite nicely during bootup, much improved compared to the G5 iMac. It's much faster and more reliable. The design of the wireless Mighty Mouse is also an improvement. While I don't like the pseudo secondary button or the side buttons as implemented, the scroll ball is quite welcome. It can also be configured to function as an extra button, which is nice. On the downside, the new keyboard is a step in the wrong direction. The keys have less of a positive 'click' to them, which means that the keyboard feels a bit mushy compared to the older version.

Bootup is quite quick, as might be expected with the newer hard drive and the much faster dual-core CPU. The overall feel of the OS is very smooth, even when multitasking. The benefits of having two cores is obvious, and yet, it's extremely quiet too, even under full load. The rest of the machine is quite similar, aside from having a bigger screen and a louder speaker system. The bass is more full than my 20" iMac's system, but it doesn't sound that much better. While the bass is loud, it's not of stellar quality. Overall the speaker system is an improvement, but not a huge one.

So, just how fast is the new iMac? I'm glad you asked... I've run a series of benchmarks to answer this question.

The first test was Handbrake. In my previous Handbrake test the MacBook Core Duo (2.0 GHz with 2 GB RAM) dominated the G5 iMac (2.0 GHz 970FX, 1.5 GB RAM, Radeon 9600 128MB). The results with the iMac Core 2 Duo were similar. The test involved exporting a single 149 second DVD chapter (off the hard drive) to an MPEG4 file, using the default settings of the software.

iMac 2.33 Core 2 Duo: 52.2 seconds, 70.4 fps
MacBook 2.0 Core Duo: 54.9 seconds, 67.2 fps
iMac 2.0 G5 970FX: 266.4 seconds, 13.6 fps

In this test the iMac Core 2 Duo was only about 5% faster than the MacBook, despite having a 16.7% clockspeed advantage, as well as twice as much L2 cache. That was a little disappointing, as I was expecting more out of the iMac Core 2 Duo. The iMac G5 for some reason is extremely slow in this application.

The second test run was unzipping a file. To my surprise the Core 2 Duo just flew here. It was 2.6X as fast as the G5 iMac, and 1.6X as fast as the MacBook.

iMac 2.33 Core 2 Duo: 5.2 s
MacBook 2.0 Core Duo: 8.6 s
iMac 2.0 G5 970FX: 13.5 s

Cinebench was the third test. Again, the iMac Core 2 Duo shone, both in the CPU tests as well as the GPU tests.

Note that the MacBook's hardware shading score is worse than its software shading score, due to the fact that much of the MacBook's GMA 950 3D functionality is emulated on the CPU. The Radeon 9600 in the iMac G5 handily beats it, but is of course no match for the GeForce 7600 GT in the iMac Core 2 Duo.

The fourth test was another video encoding test, this time using the DivX encoder that ships with Roxio Toast 7. This is an interesting test because it only recently went universal, and I as able to test both the PowerPC version as well as the universal binary that supports Intel CPUs. The test involved encoding the 720p Macaulay Library clip to a DivX file at the same resolution. As expected, the Intel-based machines did very poorly with the PowerPC version of the program, since they needed to use Rosetta translation. They did much better with the Intel native version of the software, with the iMac Core 2 Duo about 30% faster than the MacBook Core Duo. However, the iMac G5 still was the fastest in this test by about 16%, despite having only one core to work with.

(Toast 7.0.2 is the PowerPC version, and runs under Rosetta on Intel Macs. Toast 7.1 is a universal binary and runs natively on both PowerPC and Intel Macs.)

iMac 2.0 G5 970FX (Toast 7.0.2): 11' 03"
iMac 2.33 Core 2 Duo (Toast 7.1): 12' 49"
MacBook 2.0 Core Duo (Toast 7.1): 16' 43"
iMac 2.33 Core 2 Duo (Toast 7.0.2): 19' 05"
MacBook 2.0 Core Duo (Toast 7.0.2): 29' 32"

One new test I tried was Geekbench 2006, a command line benchmarking application which is cross platform, supporting Mac OS X, Linux, and Windows. On Mac OS X it's useful not only to compare different machines, but also to compare native applications to applications running under Rosetta, as there are both versions of this program. The application consists of a series of independent tests, which are then combined to create a final score:

iMac 2.33 Core 2 Duo (native): 227.9
iMac 2.33 Core 2 Duo (Rosetta): 156.8
MacBook 2.0 Core Duo (native): 179.0
iMac 2.0 G5 970FX (native): 106.5
MacBook 2.0 Core Duo (Rosetta): 105.5

The iMac G5 did quite poorly, but a fair number of tests in this bench are of memory performance. The iMac G5 has 1.5 GB, unmatched 512 MB + 1024 MB, which may affect its performance in the synthetic memory tests.

The last test was with Aperture. It was amusing to see this screen when I installed Aperture 1.0 and tried to run it:

I guess the compatibility check routine didn't like Intel chips too much, which isn't surprising given that Intel Macs didn't exist when Aperture 1.0 was released. However, I continued with the installation of the updates and was greeted with a nice smooth running application in Aperture 1.1.2. In fact, it felt faster than even a dual G5 2.5 with Radeon 6800 Ultra I had tried previously (although that was with an older version of the software). To give you and idea of how fast it is with edits, I've captured a small video of Aperture in action (5.3 MB) with my camcorder. Edits were near instantaneous, although at times there was a split second delay before the controls responded. This is a huge improvement in feel over the G5 iMac, which often had a several second delay before any changes were applied. To better quantify the speeds, I exported a project of edited 8 Megapixel Canon RAW files to JPEG files. As expected, the iMac Core 2 Duo was the fastest by a large margin.

iMac 2.33 Core 2 Duo: 50.2 s
MacBook 2.0 Core Duo: 79.6 s
iMac 2.0 G5 970FX: 141.4 s

Overall, we knew the new 24" iMac was going to be fast, but it's also a nice design, at a good price. It's a clear winner, and has become Apple's desktop Mac of choice even for some wishing to run pro applications on a budget.

[Update 2006-09-19]

It seems the most significant issue with Aperture is memory usage. I've been playing around in Aperture some more, watching CPU and memory usage using Activity Monitor. While CPU usage can sometimes jump to around 180% (90% per core), it's the memory usage that can really slow things down. In my usage I often need over 1 GB for Aperture alone, which can cause my 2 GB machine to page out memory to disk.

When 2 GB SO-DIMMs become available at more realistic prices, I will be buying one for this machine, to take the memory to a total of 3 GB. (There are only two slots, so I will have to remove 1 x 1 GB.) This will decrease memory performance somewhat because the memory will be unmatched and the system will no longer be able to make use of memory interleaving. Because of that the performance will decrease in some parts of certain applications, including Aperture, but overall performance will improve because paging to disk will be reduced significantly.

It seems the ideal system for the iMac will be based on the upcoming Santa Rosa chipset. Not only will it support an increased memory bus speed (800 MHz), it will also support more memory. The current Napa chipset supports up to 3 GB only, hence the memory limitation of the current 24". Santa Rosa won't have this limitation, which means that 2 x 2 GB with full memory interleaving should be available in the the new iMac in spring 2007. Unfortunately, that's probably around 8 or 9 months from now, and while by that time 667 MHz 2 GB SO-DIMMs will likely have dropped significantly in price, 800 MHz 2 GB SO-DIMMs will probably be much more expensive.