octocore macpro vs macbook NONpro - first pass

i just bought a new $899 macbook NONpro (because i just needed another machine and it was the most reasonably priced option)

just for @#$%& and giggles i did a race between it and my 16gig 2.8 octocore macpro. on rendering a little 10 second, 3 layer text animation to the desktop in after effects.

the macbook actually BEAT the macpro by about a second.

i didnt take time to check all the settings or eliminate all the variables. but i wouldnt think that a nonPRO macbook with 1/8 as much ram should even come NEAR an octocore macpro in any way shape or form. in ANY configuration...

stay tuned.

Have you set the MacPro to use all 8 cores with enough memory assigned to each?



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

very much so. yes.

Ten seconds, three layers? I'm just taking a wild guess here, but it sounds like that test is too short to reveal anything meaningful. It's not the same, but I've done a lot of little encoding jobs with Compressor, and the results are wildly variable on a fast eight-proc system like a Mac Pro or an Xserve. Sometimes, just by pure luck or whatever, the computer does things in such-n-such order to make them take five seconds, sometimes ten seconds. It seems pretty random. But on longer jobs ? five minutes, ten ? things settle down and get pretty consistent.

Something I've also noticed is that if you do the same encoding job with Compressor a few times in a row ? like you do when you screw up your settings and have to redo it ? the job goes faster the each subsequent time after the first. I'm not sure I'm right about this, but I think it's filesystem I/O caching at the OS level. Mac OS X will cache disk I/O in RAM, which is why (for example) it's so much faster to launch a big program like After Effects the second time after a reboot. The system has to read in a bunch of data just to load the program, and after the first time you do that, even if you quit the program, those pages of disk data are still cached in main memory, so loading them goes faster. I think the same underlying function affects everything the system does, including encoding jobs.

I don't think that would apply to After Effects, since it's mostly writing to disk when rendering, rather than reading from disk. But all I'm saying is that little effects like that can add up when you're doing short jobs.

I'd be curious to see how much faster the Mac Pro is on a big AE render. Take a giant 32-bit comp that takes half an hour to render out on the Mac Pro and see how long it takes on the Macbook. On a test that big, I bet all the little random things will average out.

i do plan to flush out the testing. just thought the initial bout was a bit surprising...

and yes again, i have noticed (more so on my intel box than any previous G_ machine) that FREQUENTLY on first attempt to do various tasks - things like transferring LARGE volumes, or big renders, the response and resulting transfer is crazy-ass slow. stop and redo it, and all is back to normal...

Do you have especially slow drives?

the test was internal drive to internal drive. so i should have even had the 7200 v 5400 rpm advantage...

not that it matters in this case (as far as i know) but all my typical work is on OWC or caldigit RAIDS on a caldigit fasta eSATA card

Well, the internal 5400 RPM drive on my MBP has always outperformed the internal 7200 RPM drive on the dual G5 at work, at least on the AJA disk whack test. I suppose the difference is that you have a much smaller platter on the 2.5" drives.



www.strypesinpost.com

Is the disk whack one the one that makes your drive seek all over the place instead of reading and writing contiguously? Yeah, I can see where a slower but smaller drive could win that test, just because of rotational inertia and head travel. It's kind of weird that such small things could make a difference, but I guess that's the world we're living in these days.

I remember hearing stories back in the 90s about how the Cray company had trouble making their supercomputers even faster than they were because they had to be so physically large that they ran into speed of light problems. Signals took such-and-such time to propagate through the system, and couldn't be sped up without making the wiring runs shorter, and the wiring runs couldn't be shorter without making the machines smaller, and they'd overheat. Or something like that. Kinda blew my mind. (Also may be apocryphal, for all I know.)

OMG- you're right. I'm dumping my Mac Pro for a MacBook right this instant due to the convincing and extensive scientific testing you've done. Wait for it.......... NOT.

-Noah

Final Cut Studio Training, featuring the HVX200, EX1, EX3, DVX100, DVDSP and Color at [www.callboxlive.com]!
Author, RED: The Ultimate Guide to Using the Revolutionary Camera available now at: [www.amazon.com].
Editors Store- Gifts and Gear for Editors: [www.editorsstore.com]

When life gives you dilemmas...make dilemmanade.

ok, no need to be snarky noah. just pointing out my surprise that it was even in the ballpark for aftereffects efforts.

stay tuned for more real world comparisons...

Not trying to be snarky at all- just pointing out this is an unscientific test when more reliable results already exist such as:

[www.barefeats.com]

-Noah

Final Cut Studio Training, featuring the HVX200, EX1, EX3, DVX100, DVDSP and Color at [www.callboxlive.com]!
Author, RED: The Ultimate Guide to Using the Revolutionary Camera available now at: [www.amazon.com].
Editors Store- Gifts and Gear for Editors: [www.editorsstore.com]

barefeats is the shizz-nizzle. I always check in there before buying any gear.

When life gives you dilemmas...make dilemmanade.

dont get me wrong, im never going to argue proven stats. but the application of a lot of that kind of data is esoteric and pretty subjective. its like the difference between a 2.8 and a 3.2 macpro.
is there a difference? sure.
is it a difference enough to justify an extra grand to anyone who isnt churning out deadline dependent product every minute of the day?
that's up for debate...

and thats the nature of my post.

im a "rubber meets the road" kinda guy. and before yesterday, id never have even given a non-pro macbook a second look. i just thought it was interesting that the thing even comes close to a usable production tool.

that makes more sense to me than a bunch of spec-head data.

Bear in mind that the computers we think of today as cheap and slow are somewhere on the order of one hundred skrillion times faster than the Onyx Reality Engine system that Inferno ran on back in the mid-1990s.

I might have my numbers slightly wrong, but I think I'm close. I did the math on the back of a cocktail napkin.

I'll take the "spec-head data" anyday rather than winging it - which is what you are doing. Rob @ barefeats knows how to set up the machines for best performance so tests are TRUE. If your plain old macbook beat out your macpro, you were not set up correctly for a proper benchmark testing.

When life gives you dilemmas...make dilemmanade.

I was mentioning about drive speed on read/write tests, not on renders. I'm not sure what the configurations are on your system, but there's no way a MB (or a MBP) can come close to beating an octocore when all 8 cores are utilized.



www.strypesinpost.com

Quote

Rob @ barefeats knows how to set up the machines for best performance so tests are TRUE.

I think you meant to say "so tests are often inapplicable to real-world situations."

I don't believe anybody has suggest that Wayne, y'know, did it wrong or anything. In that particular test, in that particular situation, those were his results. I floated my theory: The test was so short that random factors came into play. I'm not positive it's right, but I think it's a sound theory; anybody who's worked with parallel processing knows that scaling is never linear. It merely approaches linearity as the length of the job approaches infinity. A CPU-bound operation like an After Effects render ? one where the disk and memory bottlenecks aren't in play ? will be eight times faster on eight processors if you let it run all night, but if you only run the job for ten seconds, it probably won't be much faster than a single processor, and quite possibly a lot slower due to parallelization overhead.

Heck, you don't even have to know anything about parallel processing to get this. In After Effects CS3, it was often counterproductive to multiprocessor-render on small jobs because the mere act of the program having to fire off seven more render processes took longer than the whole job would on just one processor. They made some changes in CS4 to minimize that, though. I think now it fires off your render processes at application launch time rather than when you start a render, if I remember right.

Quote

i didnt take time to check all the settings or eliminate all the variables.

jeff,

I just gathered that this sentence meant the OP's machines were not using identical settings. THAT's what I meant. And Barefeats tests don't have to be real-world to show differential in performance specs.

When life gives you dilemmas...make dilemmanade.