MacCetera LLC. Unique Business and Personal Macintosh Support for Southeastern Wisconsin, Creative Web Site Design and Hosting for the World

WelcomeMac StuffWeb Stuff Marc's NotesCool LinksContact Us

Marc's Notes: Comments, Ramblings, Rants & Tips

G4 Digital Audio CPU Benchmarks

This weekend (early June, 2006) I upgraded my G4 Digital Audio to it's ultimate performance level. For a shade under $600 Other World Computing shipped a Sonnet Duet 1.8 GHz G4 processor upgrade to me on a Saturday Morning - 9:30AM - And I'm creating this web page at 5:00PM running on this wicked dual processor G4 tower.

Part 1: Xbench CPU Comparisons

The transition from the OWC 1.33 GHz single G4 was not without a few glitches, and I had a graduation party to attend mid afternoon, so the applied time to figure this all out and work around the issues took somewhere around 4-1/2 hours. I'll get into the "gotchas" in a minute, but first let me wow you with some raw performance data.

Xbench Version 1.1.3
System Version 10.4.6 (8I127)
Physical RAM 1536 MB
Mac Model PowerMac3,4
Bus Frequency 134 MHz
  Apple 533 MHz G4 OWC 1.3 GHz G4 Sonnet 1.8 GHz Dual G4
Processor PowerPC G4 @ 534 MHz PowerPC G4 @ 1.33 GHz PowerPC G4x2 @ 1.80 GHz
Version 7410 (Nitro) v1.3 7455 (Apollo) v3.3 7447a v1.1
L1 cache 32K (instr), 32K (data) 32K (instr), 32K (data) 32K (instr), 32K (data) per CPU
L2 Cache 1024K @ 267 MHz 256K @ 667 MHz 512K @ 900 MHz per CPU
L3 Cache —NA— 2048K @ 4.13 GHz —NA—
CPU Test 44.95   157.77   217.24  
GCD Loop 42.47 1.66 Mops/sec 145.45 5.68 Mops/sec 223.01 8.71 Mops/sec
FP Basic 57.22 206.94 Mflop/sec 157.35 569.03 Mflop/sec 211.19 763.72 Mflop/sec
AltiVec Basic 25.04 727.51 Mflop/sec 163.49 4.75 Gflop/sec 209.32 6.08 Gflop/sec
vecLib FFT 63.79 990.32 Mflop/sec 166.74 2.59 Gflop/sec 219.05 3.40 Gflop/sec
FP Library 68.50 2.74 Mops/sec 157.53 6.31 Mops/sec 224.53 8.99 Mops/sec
Thread Test 46.09   112.34   274.31  
Computation 31.82 429.50 Kops/sec 81.17 1.10 Mops/sec 218.54 2.95 Mops/sec
Lock Contention 83.57 1.05 Mlocks/sec 182.37 2.29 Mlocks/sec 368.31 4.63 Mlocks/sec
Memory Test 66.47   114.75   94.30  
System 59.66   169.83   111.28  
Allocate 228.67 149.16 Kalloc/sec 690.87 450.65 Kalloc/sec 897.60 585.50 Kalloc/sec
Fill 91.64 729.44 MB/sec 253.80 2020.21 MB/sec 176.10 1401.74 MB/sec
Copy 91.64 142.86 MB/sec 81.45 407.26 MB/sec 49.59 247.95 MB/sec
Stream (AltiVec) 75.02   86.65   81.81  
Copy 79.62 582.03 MB/sec 85.97 628.45 MB/sec 81.26 594.00 MB/sec
Add 75.73 558.89 MB/sec 89.84 663.02 MB/sec 80.85 596.70 MB/sec
Scale 72.19 462.01 MB/sec 85.52 547.34 MB/sec 84.31 539.60 MB/sec
Triad 73.00 446.03 MB/sec 85.40 521.79 MB/sec 80.90 494.31 MB/sec
RESULTS 58.15 100% 125.24 215.37% 174.07 299.35%

Note— The dual G4 benchmark doesn't represent the true picture... the XBench application, for the most part, only beats on one of the processors. Something that measures the base CPU ability. In the real world this is a true measure of how a single application will cook on the processor, but it isn't being dragged down at all by the other tasks running under OS X - which also rocks on a dual CPU machine.

Part 2: Sanity Check

First question: Why, on the dawn of the day of Mactel machines, would an otherwise sane Mac consultant spend six bills to soup up his 2001 model year G4 tower?

Easy answer: Because I can, so there!

Hard answer: This is my primary machine. Having a stable box is critical, and I want to make sure that the "first edition" bugs are dead before leaping on the next technology band wagon. I will be buying a new MacPro in early 2007, probably when Leopard is released and my primary applications are Universal. Then again, my Pismo is getting a bit long in the tooth, and my foray with the Intel side of Macs may first go to the MacBook or MacBook Pro as I'll need intel capability for at least one of my clients. Stretching the G4 out for another year or two works.

Second question: Wasn't the 1.33 Ghz OWC processor fast enough to meet your needs.

"Yes" answer: For most of my needs, the single G4 OWC upgrade rocked.

"Maybe" answer: There were some stability issues that I never tracked down and I started getting the gut feeling that it might be a CPU bug. But then I also had a flaky RAM module I replaced at the same time as today's upgrade, and there were some other "phase of the moon" issues. It's a pain when Photoshop CS2 panics the machine 5 times in a row when I'm trying to get work done.

"No" answer: The dual processor solution has real applications in the heavy processing end of things, like video processing and the higher end applications. Since I am not buying a G5 tower to tide me over until the MacPro move, I wanted (probably as much as needed) the boost that TWO processors bring to the table—and if I'm going to slap down cash for an upgrade, it made sense to get the top end product. The G4 will move to very a very respectable server job down the road.

Part 3: The Ugly Side of Upgrades

Usually the gods are either smiling—or at least not paying attention—when I start messing with my own hardware. The Apple to OWC Mercury processor upgrade a couple years ago took literally 10 minutes, and yes, it's been up for most of the past few years without too many kernel panics and the like (OK PC guys, Mac's do crash, and we've been blowing smoke when we say they don't... happy now?)

Today's upgrade, turned into a 4-1/2 hour adventure. Here's the short process for the Sonnet processor installation:

  1. Boot into Mac OS 9.2 to install the Sonnet firmware patch — power down the machine.
  2. Install the processor — three screws out and in.
  3. Power up — be amazed.

Well, that step 1 was a bear. I hadn't booted 9 on this Mac in a while, and it simply refused to be coerced into starting up in Mac OS 9. So I booted the 9.2.1 CD and installed a fresh 9 on one of the OS-less drives. I "option-booted" to ... the OS X drive (???), installed the original Apple CPU (hey - I got original benchmark numbers while there) ... and up came OS X (???!!), and all sorts of other junk (time so far: 2-1/2 hours) ... and OS 9 was still not an option (#%$$?@).

The dawning of the root problem was in the hard drive controller card. I have a Sonnet Tempo Ultra ATA 133 card supporting four Maxtor internal beasts, and I had never tried booting Mac OS 9 after installing that card 6 months ago. Duh!

I grabbed a spare "expendable" drive, a short ATA cable, and plugged it into the Apple ATA connector after pulling the cables off the ATA card. Then I booted the Mac OS 9.2.1 CD and wiped the target, installed OS 9, and presto, the "Happy Mac" boot icon greeted me from the spare drive.

I popped the Sonnet CD in, applied the patch, did the interrupt-start to flash the firmware, and the next startup ended with a dialog telling me it was OK to install the processor card (time so far: 3-3/4 hours).

Off too Sarah's graduation party... argh.
Really Sarah, congrats! You're awesome!

Pulled the old CPU, buttoned in the Sonnet, restored my drive configuration, pressed the power button... Joy!

This processor smokes! Really fast, but the best news is that the kernel panics and random crashes are gone. As I complete this more than two weeks after the upgrade, the only time the box has been restarted was after a software update 13 days ago. Sweet!

(1) The opinions expressed in Marc's Notes: Comments, Ramblings, Rants & Tips are exclusively those of Marc Wolfgram.
(2) Only one Macintosh™ computer was permanently harmed - link. (3) Any references to real people may be intentional.
(4) Don't try certain things while driving or at home without proper adult supervision. (5) Microsoft Windows—Just say NO!

Copyright © 2003-2008 by MacCetera, a Wisconsin LLC owned by Marc & Tammy Wolfgram
Mac OS, Macintosh, and other like terms are all trademarks of Apple, Inc. - 25-Oct-2009

Valid HTMLValid CSS