The Source for Mac Performance News and Reviews

Review: Powerlogix's 1GHz G4 Upgrade for the Cube Tests vs Original G4/450 and Dual G4/500 By Mike Published: 8/19/2002

Benchmark Test Results

Intro | Benchmarks  | Apps Tests | Game Tests | Installation | Specs/Design

Benchmark Tests

Note: This review used the 1MB L3 cache version. New models have 2MB of L3 cache in the 1GHz and 800MHz models. Benchmark tests were run with SpeedRun (OS X and OS 9), Altivec Fractal demo, CineBench 2000, Throughput 1.5 and memory bandwidth tests. Of course benchmarks are less important than real world applications and game tests, but many want to see them regardless. (The Apps tests page has results of tests with iMovie2, iTunes3 (OS X), Photoshop 7.0 and Appleworks. The game tests page includes results with popular 3D games like Quake3, Unreal Tournament, Medal of Honor Allied Assault and Return to Castle Wolfenstein.) SpeedRun 1.1.3 OS X/OS 9 Tests: Note: All tests run at the 22" Cinema display's native 1600x1024 resolution, but SpeedRun requires thousands color mode to run its graphics tests. (The graphics test isn't very extensive and is not a test of 3D performance. See the apps/game tests page for real-world apps/game test results.) The Above graphs show how the 1GHz upgrade compares to the original G4/450 CPU and a Dual G4/500 in the Cube compared. Here's a table (sorted by Overall Score) showing my results compared to other systems in the SpeedRun 1.1.3 database. (My results from this review are underlined in red.) SpeedRun 1.1.3 OS X Comparison I can't say as I'd agree with the ratings in some cases, but this is just a sort of the results from this benchmark. (For example the Cube is not a faster overall system than a Dual GHz Mac, assuming the 1000MHz G4 in the table above is a dual model. And note the overall score from someone's dual 800 G4 was lower than a G4/800 powerbook.) SpeedRun 1.1.3 OS 9 Comparison OS 9 and OS X versions of Speedrun are available here. (I used v1.1.3, latest as of this review date.) Fractal FPU Benchmark: I used the freeware Altivec Fractal (Carbon) demo program (available here) to compare FPU performance of the Cube system with the original G4/450 CPU, the Dual G4/500Mhz and the Powerlogix G4 upgrade. Default settings were used. (Note: MegaFlops/sec = Millions of Floating Point Operations per Second.) It's odd that both the dual 500 and single 1GHz had the exact same score on the inital run. "Refreshing" (repeating) the test showed a score of 3,351.4 for the DP500 and 3,365.3 for the single 1GHz. As a FYI, a Dual 1GHz Quicksilver 2002 scores 7,625. CineBench 2000 v1.0 Results: Since probably less than 1/2 of 1% of Mac users own Maxon's Cinema4D application, I use the freely available CineBench2000 benchmark. (Which does more than just a simple render of a scene - it also does shaded and wireframe model spins, shaded and wireframe "fly-throughs" of a scene to show graphics card and cpu performance, not just rendering times.) For more info see www.maxon.net/pages/download/benchmarks.html. For a full explaination of the benchmark and how to read the results, see this page. All tests were run at 1024x768, millions color mode. Longer bars are faster. The Dual 500 shows rendering benefited from both CPUs (Cinema's engine uses both for rendering), but the faster single CPU scored much better in the other tests which show graphics card performance (fly-through, manipulating shaded and wireframe objects, etc.). You can search for other Mac system/CPU/Video card CineBench 2000 results at my Mac Game/Video card performance database. If you download CineBench 2000, please enter your results. ThroughPut 1.5 Tests: Rene Trost's ThroughPut 1.5 benchmark was also used to see how the CPUs affected its scores. Here's how Rene describes this 'pure' benchmark: " ThroughPut is a little application, written in PPC assembler, that tests how much data your Mac can push through the PCI or AGP port to feed your graphics card with data." See his web page for more details, but basically his benchmark tests the amount of data the system can deliver to the graphics card using several modes - CPU (32bit stream), FPU (64bit stream), Altivec (128bit stream) and CopyBits (256bit busmaster). [Altivec requires a G4 CPU of course.] Memory Bandwidth Benchmark Tests (OS 9): I used a benchmark called MemoryWatch that I'd had for some time (I think Michiro Isobe may have sent this to me.) A later version of this benchmark generated some NaN (not a number) results with the 1GHz upgrade, so I used an older version. Note how the 7455 and 7400 Read/Write rates compare (and how rates compare with sizes larger than the 1MB cache). The 7400 based 450/500MHz CPUs have always scored higher on simple memory bandwidth tests (with older benchmarks at least - some may say this is compiler related perhaps). Although note that the much more recent SpeedRun Benchmark's Memory scores were much higher with the 1GHz upgrade, but I don't know the details of its memory tests. *Update 8/21.2002* Michiro Isobe wrote with comments on MemoryWatch: " Dear Mike, I have read your great review on PowerLogix G4 upgrade. I will comment on MemoryWatch results.. Briefly, current and older revisions of MemoryWatch cannot measure the performance of G4+ processors (MPC7450, MPC7451 and MPC7455). So, the write score of PLG4/1GHz is not accurate. The details is shown below. The purpose of the utility is the measurement of pure memory bandwidth bound by nothing but hardware. So, the measurement code of MemoryWatch was written in assembly language and optimized for each combinations of CPU and motherboard. As for the measurement of "read" operation, the optimization is not so difficult and the same code is used for almost all systems. But, the measurement of pure "write" operation is far more difficult. This is because when a CPU executes a write operation, several read and write operations are caused by the cache replacement mechanism, that is very different in each systems. About current revision, MemoryWatch is not optimized for extra load/store stage and L3 cache of G4+ processors. Old revisions cannot measure memory speed of 2MB L3 systems because they measure up to 2MB array. Best Regards, Michiro Isobe " (Note: remember Michiro's comments above, so these results are not really accurate/indicitive of the 7455's memory write performance.) Newer Tech's GaugePro has a simple memory bandwidth test included (reportedly emulating typical OS 9 Finder memory transfers). Since this benchmark is old, not updated to even properly recognize 745x CPUs (or their correct speeds), I really don't put a lot of weight in these results, but people always ask about it so I'm listing it here. I've seen a lot of variation with this utility in the past, so all my tests were run immediately after a clean reboot, with no other apps running. Again the more modern benchmark (SpeedRun) shows higher memory scores with the 7455 than GaugePro's dismal results and real-world performance is what matters most of course. (GaugePro 1.1 is available at OWC's Newer Tech software mirror page. FYI: As stated before, it does not properly recognize many later CPUs nor accurately report their clock speeds or L2/L3 caches.)

For comparisons to other CPU Card upgrades and systems see the site list of CPU Card Reviews and Systems page.

The next page has results from real world applications. Or you may use the links below to jump to a specific page.

Index of PL 1GHz Upgrade Review Pages Intro | Benchmarks  | Apps Tests | Game Tests | Installation | Specs/Design - or - Back to WWW.XLR8YOURMAC.COM

Copyright © 2002. No part of this sites content or images are to be reproduced or distributed in any form without written permission. All brand or product names mentioned here are properties of their respective companies. Users of the web site must read and are bound by the terms and conditions of use.