In terms of serviceability, Apple dramatically improved access to the system CPUs, though you”ll recall that you can”t add a CPU to a Nehalem-based system. If you buy the system with one CPU, Apple ships a CPU that”s not dual-processor-capable. So the CPUs are easier to replace, but you can”t upgrade from single- to dual-processor configuration.
Figure 1. Open these to latches to access the CPU/RAM compartment.
RAM has been very easy to access for a while now, and it”s even easier on the new Mac Pro. Specifically, as you can see in Figure 1, at the bottom of the case is a compartment that you open by pulling two latches.
Then you pull the tray out, and you can swap CPUs or memory with ease (Figure 2). Guides built into the compartment and tray ensure that you don”t insert the tray improperly. I slid the tray in and out several times during my tests, and has long as you”re moderately aware of what you”re doing, it”s near impossible to mess up.
Figure 2. There”s the aforementioned CPU/RAM tray, where all components are very easy to access.
The only bummer was that I couldn”t convince Apple to send me more RAM to test performance at full capacity, so I was all dressed up with no RAM to insert. Sigh. So if you”re looking for performance statistics for varying levels of RAM, you”ll just have to find them somewhere else (and let me know if you do, because I”d love to see them).
I have in my hands an Intel Nehalem-based Apple Mac Pro, specifically a 2.93GHz dual-processor, quad-core unit running Mac OS 10.5.7 with 12GB of RAM and an ATI Radeon HD 4870 driving a beautiful Apple LED Cinema Display…
Apple was the first out of the gate with an Intel Nehalem-based workstation, in the form of the Mac Pro announced in early 2009…
Apple did send me two graphics cards: the stock Nvidia GeForce GT 120 and the ATI Radeon HD 4870, advising me that 3D preview would be much faster on the latter in programs such as Motion and Shake. Unfortunately, while I”m generally proficient in Final Cut Pro, Motion isn”t my thing.
After creating a preliminary project or two, and some reflection, I decided that I could spend half a day creating Motion projects and still not have information as relevant as the Motion 3.0.2 tests presented by Apple on its own website, which show the Radeon to be twice as fast as the GT 120. So no independent confirmation of those performance statistics here. And no, mom, I”m not being passive aggressive because Apple refused to send me that RAM. Really!
Figure 3. To install the drive, simply push it into the slot.
Apple also sent its new RAID card ($700), which I was eager to test and had a wonderful scenario for doing so. Specifically, I had just finished up another review that involved Blackmagic Design”s Decklink HD Extreme card, so I had a legitimate use for the performance offered by the RAID. Recall that my Intel Nehalem-based Apple Mac Pro is a 2.93GHz dual-processor, quad-core unit running Mac OS 10.5.7 with 12GB of RAM.
Apple provided the RAID card with a set of four disks, including one system disk that connected via SATA and three other 1TB drives configured in RAID 5, which offers a blend of protection from data loss and performance. To test performance with and without the RAID card, I tested first using the original system that Apple sent, which was equipped solely with SATA drives, then installed the RAID card and new system drive and tested again.
Figure 4. Blackmagic Design”s Disk Speed Test on the SATA drive.
Now would be a good time to mention that Apple also redesigned how you access and install drives in the new case, which simplified the disk-swapping process into about a 1-minute operation. Basically, it”s all toolless—you pull the drive out of the slot, then push the new drive back in. The RAID card also works without cables, simplifying the hardware installation even further.
Figure 5. Blackmagic”s Disk Speed Test on the RAID drive.
I started by running Blackmagic”s Disk Speed Test in both cases. Using the SATA drive, the test measured write performance at 100.8MBps.
As you can see in Figure 5, the RAID drive boosted write performance to 191.4 MB/second, almost double.
I then captured some test video clips from my Panasonic AG-HMC150, a full-resolution, full-data-rate AVCHD camcorder, using the Decklink”s HDMI connection. Between a recent piano recital of my darling daughters and some street shots from a recent music festival here in town, I had a nice selection of 720p and 1080i footage.
Figure 6. My irresistible Whatley, capturing fine in 720p in 10-bit Blackmagic format, but only to the RAID drive.
I tried capturing the two types of footage in three formats; Apple ProRes 422 HQ and 8-bit and 10-bit Blackmagic format, using the SATA drives in the first setup and the RAID drives in the second. The SATA drive did fine with ProRes 422 HQ in both resolutions, but it failed in all other attempts, even though Blackmagic”s Disk Speed Test seemed to indicate that the SATA drive could handle 720p in both Blackmagic formats. Go figure. As a retest, I tried capturing to the SATA system drive in the RAID setup and got the same result.
Scratching my head, I decided to check the data rates of the captured files using Final Cut Pro”s Item Properties window, and the reason for the failures quickly became evident. Even at 720p, Blackmagic”s native formats are enormous, and at 1080i, they”re simply overwhelming. You can capture either format with a SATA drive if you use ProRes, but otherwise, you need the RAID drives.
Just for fun, I wanted to see how much more headroom the RAID drives delivered during editing, and created two projects using multiple 720p files in ProRes 422 HQ format, with one in the background and the other streams inserted as 30-percent-sized picture-in-pictures. I set preview at Unlimited RT, with quality set to high and frame rate to full.
Table 1. Performance trials for SATA and RAID drives.
Retrieving from the SATA drive, I experienced the first dropped frames at six streams, while the RAID drive started hiccupping at 10 streams. While this test is fairly uninteresting from an editing perspective, since few projects involve realtime preview of in this or a similar configuration, the results bode well for performance in a video server environment, where the RAID cards should come close to doubling performance over SATA drives. As shown in the table, CPU use was actually fairly low in both scenarios, leaving plenty of headroom for other activities.
Overall, from all angles—performance, expandability, and serviceability—the new Mac Pro is an impressive piece of hardware.