November 13, 2006, 9:01 PM PST
The unofficial eight-core Apple Mac Pro
Posted by: Daniel A. Begun

CNET Labs might be ahead of Apple's product release cycle, and we likely violated our Mac Pro's warranty, but we just had to see what the Apple Mac Pro could do when populated with a pair of Intel's brand-new, quad-core Xeon 5355 processors.

Today marks Intel's first official day of the quad-core processor era with the release of quad-core processors for enthusiasts (the Intel Core 2 Extreme QX6700) and for servers and workstations (the Intel Xeon 5355)--and Intel was kind enough to supply CNET Labs with a pair of 2.66GHz Xeon 5355 processors. As the Xeon 5355 is pin-compatible with the Xeon 5160 processors that came installed in our Mac Pro, we proceeded to swap out the two dual-core processors with the new quad-core processors. (We highly advise you not to try this at home! The Mac Pro case is not designed to allow the end user to perform CPU surgery--and we've got the cuts and bruises to prove it.) With the pair of Xeon 5355 processors installed, we booted the system back up and were greeted with eight active processing cores in both the Mac OS and Windows XP via the Boot Camp Public Beta. With the transplant successful, it was time to run our benchmarks...

Even though dual-core processors have been around for a while now, you'd still be hard-pressed to find many mainstream applications that can efficiently take advantage of both processing cores at the same time (typically referred to as a multithreaded-application). Double that number to four processing cores, and the list of supported multithreaded applications gets even shorter. Double it again to eight...and you get the idea. Some professional multimedia and scientific applications, however, are designed to take advantage of as many processors as are present--and performance will scale accordingly, based on the number of processors available.

Both the Cinebench and PyMOL tests use all available processing cores and hit 100-percent total CPU utilization on every configuration we tested. We saw a 31-percent performance increase on the Mac OS X version of the Cinebench test from the two dual-core chips to the two quad-core chips. Although we doubled the number of cores, we didn't see twice the performance. This is for a few reasons: The quad-core chips are actually running at a slower speed (2.66GHz) than the dual-core chips (3.0GHz). Also, the extra cores introduce some additional computational overhead to the overall workload. Additionally, our "octo-core" rig is our own unsanctioned rig, and therefore isn't benefiting from any of Apple's special sauce, such as firmware and driver updates to better optimize the system for the additional cores.

Our multimedia multitasking test performs a QuickTime encode in the foreground while iTunes is simultaneously encoding in the background. On systems with two or fewer cores, this workload typically saturates the total CPU utilization at 100 percent. With four cores, the system hovered around 40-percent CPU utilization, but dropped to about 23-percent when using eight cores. Interestingly, the actual performance gain we saw between four and eight cores was less than 10 percent. To truly see a significant benefit from the additional cores while performing multiple tasks, you will have to perform a massively multitasking scenario--something we unfortunately did not have time to do for this story.

Our iTunes and Quake 4 tests are more representative, however, of what you are likely to see with most mainstream applications in a nonmultitasking scenario. The results for both of these tests (as well as with other apps, not shown here, such as Photoshop CS2) indicate that what influences the speed of these tasks is primarily CPU speed. Four cores running at 3.0GHz consistently outperform eight cores running at 2.66GHz. (Note that iTunes is better optimized for the Mac OS, and Quake 4 is better optimized for Windows XP.)

It will be interesting to see how long it is before Apple migrates the Mac Pro over to the new quad-core Xeon chip and makes an eight-core system publicly available. But unless you do work normally relegated to high-end workstations, perform massively multitasking workloads, or just want the bragging rights, eight cores is definitely overkill...at least for now. As more applications become available that support multithreading across multiple processing cores, the benefits of quad- and octo-cores will be realized.

System configurations:

Intel QX6700: 4 cores @ 2.66GHz, Windows XP
Windows XP Professional SP2; 2.66GHz Intel Core 2 Extreme QX6700; 2,048MB DDR2 SDRAM 800MHz; 256MB ATI Radeon X1900; 74GB Western Digital 10,000rpm SATA/150

Mac Pro: 4 cores @ 3.0GHz, Mac OS X
OS X 10.4.8; 2x 3.0GHz Intel Xeon 5160; 2,048MB DDR2 FB-SDRAM 667MHz; 512MB ATI Radeon X1900; 500GB Seagate 7,200rpm SATA/150

Mac Pro: 4 cores @ 3.0GHz, Windows XP
Windows XP Professional SP2; 2x 3.0GHz Intel Xeon 5160; 2.048MB DDR2 FB-SDRAM 667MHz; 512MB ATI Radeon X1900; 500GB Seagate 7,200rpm SATA/150

Mac Pro: 8 cores @ 2.66GHz, Mac OS X
OS X 10.4.8; 2x 2.66GHz Intel Xeon 5355; 2,048MB DDR2 FB-SDRAM 667MHz; 512MB ATI Radeon X1900; 500GB Seagate 7,200rpm SATA/150

Mac Pro: 8 cores @ 2.66GHz, Windows XP
Windows XP Professional SP2; 2x 2.66GHz Intel Xeon 5355; 2,048MB DDR2 FB-SDRAM 667MHz; 512MB ATI Radeon X1900; 500GB Seagate 7,200rpm SATA/150

Permalink | 29 comments

• Read also: Apple, Desktops
• Email to a friend | Print this post

August 23, 2006, 2:03 PM PDT
Mac Pro fails the Boot Camp test
Posted by: Daniel A. Begun

When CNET looked at the Mac Pro last week, we were very impressed with its performance. As more Universal applications become available, the Mac Pro's burgeoning potential will be even better realized. And if our Photoshop CS2 test is indicative of other non-Intel-native apps's performance, then even apps that run under Rosetta will still be relatively speedy.

One of the advantages that Intel-based Macs provide is that they are also capable of running Windows XP natively with Apple's Boot Camp Public Beta. Full disclosure: I purchased my very first Mac--a MacBook Pro--earlier this year so that I could have a single system running both Windows and the Mac OS. While I am likely in the minority for Mac owners, I am certainly not the only one enjoying the best of both worlds--as can be evidenced by the traffic on Apple's discussion boards. My MacBook Pro runs Windows just fine, as do apparently many other MacBooks and Mac Minis out there.

Needless to say, we had very high expectations when we installed Boot Camp and Windows XP onto our Mac Pro. The first indication that something was not quite right was the slow speed at which the OS installed. Once we had it all set up and the apps were loaded, we put it to the test. Windows performance was much slower than we anticipated. To figure out what might have been wrong, we turned to the true Mac experts: the Mac users who frequent Apple's discussion forums. The group's general consensus was that the current version of Boot Camp supports only Serial ATA hard drives in the slower PIO mode, and not the significantly faster DMA mode.

It turns out that the Mac community had it right: an Apple spokesperson confirmed to us that Boot Camp currently supports only PIO mode for Serial ATA hard drives. Keeping things in perspective, Boot Camp is still in beta, so you can't necessarily expect it to run flawlessly...yet. Unfortunately, this means that for the time being, Mac Pro users probably won't want to run Windows natively using Boot Camp. Another option would be to use Parallels Desktop for Mac to run Windows on the Mac OS as a virtual machine. Our own performance testing with a MacBook Pro, however, indicates that performance using Parallels is significantly slower than you'd get running Windows natively.

There is still a part of this mystery yet to be solved: Why didn't this show up as a significant issue with the Serial ATA drives in the MacBook Pro and the iMac we previously tested with Boot Camp? I've posed this question to Apple, and I hope to have an answer soon.

In the meantime, Apple is aware of the issue and is investigating it. There is no official word on when it will be resolved.

Permalink | 15 comments

• Read also: Apple, Computers, Desktops
• Email to a friend | Print this post

June 22, 2006, 12:57 PM PDT
Heresy: Windows XP performance on a Mac
Posted by: Daniel A. Begun

The Mac cult, er, I mean, community seems to be of two minds about the fact the Apple has switched to Intel processors for its computers: One group seems to believe that this is tantamount to sacrilege and Apple has forever sullied its good name. The other group thinks that this is just the right move to motivate all those poor Windows suckers (the meager other 89 percent of all computer users) to finally move over to the Mac platform--also known to some as "the light side of the force."

While my vastly oversimplified exaggerations might incur the wrath of die-hard Mac fanatics, the truth is that the rest of us can easily enjoy the best of both the Mac and Windows XP worlds on a single system--as long as that system is an Intel Mac. Apple's own Boot Camp Public Beta allows you to install Windows XP SP2 onto an Intel Mac, giving you a dual-boot system. Parallels takes a slightly different approach with its Parallels Desktop for Mac, a virtual machine application. Instead of Boot Camp's dual-boot approach, Parallels Desktop runs Windows XP directly on the Mac OS desktop (in what Parallels calls "near-native performance")--allowing you to run both OSs simultaneously and switch back and forth seamlessly.

The final version of Boot Camp is supposed to be integrated into the next major release of the Mac OS, Leopard--due out sometime when Steve Jobs probably thinks we'll least expect it. Expect to see a full review of Parallels Desktop for Mac on CNET very soon. To whet your appetite in the meantime, however, CNET Labs ran a few of its home-brewed benchmarks on a MacBook Pro with the Mac OS, Boot Camp, and Parallels Desktop to see how application performance stacks up between the three.

For our tests we used a 17-inch Apple MacBook Pro , running Mac OS X Version 10.4.6, with a 2.16GHz Intel Core Duo processor, 2GB of DDR2 SDRAM, a 100GB 7,200rpm hard drive, and an ATI Radeon X1600 graphics chip. Boot Camp was set up with a 10GB partition for Windows XP; while the Windows XP virtual machine in Parallels was set up with a 20GB virtual disk.

Strictly speaking, our Photoshop CS2 test isn't completely fair. The Mac OS version of the app does not run natively, but instead runs under Apple's Rosetta emulation layer. But the Mac version still outperforms the Windows XP version on Parallels. Our test is designed to be very CPU, memory, disk, and graphics intensive. Parallels just wasn't up to the challenge. Additionally, each Parallels virtual machine is represented only by a single processor thread. So despite the multithreaded nature of both Photoshop CS2 and our test workload, Windows XP on Parallels does not support multithreading. Score one for Boot Camp.

Our Microsoft Office 2003 test is a Windows-only test. There was no contest here: Boot Camp ran circles around Parallels, with Boot Camp running the test almost 2.4 times faster than Parallels. Score another one for Boot Camp.

iTunes is an Apple app after all, so it's no real surprise that the Mac OS was the speediest on our iTunes encoding test. Boot Camp wasn't too far behind, and Parallels once again brought up the rear.

Parallels doesn't officially support DirectX yet, but that didn't stop us from trying to run Quake 4 on it anyway. Alas, it was an exercise in futility--it did not work--and in fact, at one point, even forced a spontaneous system reboot! Surprisingly, with the game resolution set to 1,024x768, Quake 4 actually produced noticeably faster frame rates on our test than with Boot Camp. Perhaps the Mac has a future as a gaming platform after all?

The one test where Parallels was the speediest was on our boot time test. The Mac OS was right behind it, with Windows lagging behind. But, hey, it's Windows--what do you expect?

But this also just goes to show the danger of drawing too many conclusions from this type of testing. Making adjustments to system settings, files that load at boot time, disk cache sizes, and so on can all have significant impact on not just boot times, but even on overall performance. A few tweaks to Photoshop CS's settings, and we probably could have gotten it to run faster on the Mac. If we spent more time tweaking Parallels, we probably could have beefed up its overall performance.

Other than our Quake 4 glitch, Parallels was very stable. And we truly appreciated the flexibility of being able to switch back and forth from the Mac OS and Windows XP by merely moving the mouse. As to the mundane tasks that probably make up most of your computing time (such as word processing, e-mail, and browsing the Internet), you're not likely to notice Parallel's performance degradation. I'll admit that I didn't spend a ton of time with Parallels, but what I did see was a useful, stable platform that allows me to have my cake and eat it, too.

Permalink | 40 comments

• Read also: Apple, computers, desktops, laptops, notebooks, software
• Email to a friend | Print this post

June 19, 2006, 2:04 PM PDT
Windows Vista beta 2 application performance
Posted by: Daniel A. Begun

There's been plenty of lively chatter lately about Microsoft Vista's hardware requirements, software compatibility issues, bugs, and more. Hey, it's a Microsoft operating system. And it's in beta--an easy target! (But try publicly poking holes in the Mac OS, and the hordes will come after you with torches and pitchforks.)

We've already seen that the current Vista beta is not very battery-friendly for laptops. Hopefully, the final version will have more robust power-management features. Otherwise, future Vista users are in for some boring cross-country flights. But this begs the question, how does Vista's performance hold up with regular applications? We're glad you asked!

We loaded Windows Vista beta 2 (build 5384) and Windows XP Professional SP2 on a 3.2GHz Pentium 4, with 1GB of DDR2 memory running at 664MHz and an ATI Radeon X850 XT graphics card. We ran a few of CNET Labs' home-brewed tests on the two operating systems to see if we could discern any significant performance differences between XP and Vista. The results were surprising.

We expected to see "beta bloat" hamper the overall performance of the Vista beta, but that was not the case. Vista and XP both ran our iTunes encoding test in the same amount of time. Even more surprising, Vista was actually almost 6 percent faster than XP on our Adobe Photoshop CS2 image-processing test. The F.E.A.R. scores were close but not identical. The takeaway from the F.E.A.R. scores, however, is that at least for this one game, DirectX 10 is not only stable, but the current DX 10 beta version is showing comparable performance to DX 9.

This testing only scratches the surface and in no way is meant to be definitive. But it is promising. For more on Microsoft Vista, check out CNET's Windows Vista: Everything you need to know about Microsoft's next OS.

Permalink | 22 comments

• Read also: Software, computers, desktops, laptops
• Email to a friend | Print this post

June 09, 2006, 7:17 AM PDT
Intel Core 2 Duo preview benchmarks--a different approach
Posted by: Daniel A. Begun

Intel next-gen benchmark results. At yesterday's Intel Core Technology overview, Intel had two types of systems available for us to benchmark: one with an Intel Core 2 Duo E6700 and one with an Intel Core 2 Extreme X6800. I would have loved to have spent some quality time with the X6800 system, but Intel permitted us to run only a few select games on it. I was more interested in how the new Core Technology handled multitasking, so I focused on the E6700 system--on which Intel installed a few additional applications.

Intel was very strict about what we could and couldn't do with the systems. As Rich Brown pointed out in yesterday's blog post, this was the same set of tests that Intel has been allowing tech journalists to run on the Core 2 system for the last few weeks, and you can find plenty of results online from many of the journalists who took Intel up on its offer. I decided to take a different approach.

I stayed within the ground rules that Intel established: We could use only the apps and the files that were preinstalled, but we could change the program settings. And the company didn't say we couldn't run more than one app at the same time. So with only a limited time to test, access to someone else's tests and test files, and nothing relevant to actually compare the results to, I decided to spend my time seeing how the systems handled multitasking loads.

I mixed and matched a number of multimedia applications and was pleasantly surprised by the results. The combination of Adobe Photoshop Elements 4 processing multiple JPEG images while Adobe Premiere Pro 2 simultaneously rendered video effects on its own file was especially impressive. With both apps chugging away, the Core Duo E6700 took only 11 percent longer to complete all tasks than Premiere working on its own--and this was despite the fact that Windows was reporting 100 percent CPU utilization when both apps were working.

The other two multitasking combinations didn't have quite as impressive a showing; but nevertheless, the test system still performed the combined tasks in reasonable time, where other systems might have been brought to their knees. Of course, this is all conjecture without being able to run these tests on a comparison system. For that you will have to stay tuned for when we get these CPUs and systems in our Labs and can put them through our own torture tests, er, I mean benchmarks.

Permalink | 11 comments

• Read also: Desktops
• Email to a friend | Print this post