BARE FEATS - real world Mac speed tests

MAIN INDEX of latest speed tests

iMac G5 versus iMac G4 in 3D Gaming:
Three Times Faster? NOT!
(or Why Apple Should Correct Their Graph)

Originally posted September 21st, 2004, by rob-ART morgan, mad scientist
October 11th, 2004 -- Our latest findings refute Apple's "3 times faster" claim
December 27th, 2004 -- Comment on Apple's updated graphs

This article was a challenge to Apple's original claim that the iMac G5 was 3 times faster than the iMac G4 when running UT2004 and Halo. On the week of Christmas 2004, two readers alerted me that Apple had updated their graphs to show a smaller gap between the iMac G5 and iMac G4. They now show a 47% advantage for Halo and 99% advantage for UT2004... which very closely match our findings of 39% and 93% respectively in the original article found below.

THE ORIGINAL RESULTS USING APPLE'S SETTINGS

WHAT DOES "APPLE'S CLAIM" AND "ACTUAL" MEAN?
The results for the 17" iMac G4/1.25 labeled "Apple" and "Apple's claim" were extrapolated based on the graph on Apple's site and their testimony to me that their 17" iMac G4/1.25 only attained half the frame rates that our 20" iMac G4/1.25 attained -- even though both machines were configured EXACTLY the same and the Halo settings were exactly the same. "Actual" means as of 5AM Sunday morning, October 10th, our call for "independent" 17" iMac G4/1.25 results bore fruit. Our "remote scientist" tested his 17" iMac G4/1.25 with less memory than Apple's test unit (768MB) using Apple's settings and got 30.1 fps average for the two best runs -- DOUBLE the frame rates Apple claims for its 17" iMac G4/1.25 and virtually IDENTICAL to what we saw on our 20" iMac G4/1.25.

That means that Apple's graph showing HALO running 179% faster on the iMac G5 than it did on the iMac G4 must be A MISTAKE. Our results showing a 39% gain are MORE ACCURATE.

Once again, we have results from an independent source refuting Apple's claim that the 17" iMac G4/1.25 is only HALF as fast as the 20" iMac G4/1.25.

That means that Apple's original graph showing UT2004 running 212% faster on the iMac G5 than the iMac G4 must be A MISTAKE. Our results showing a 52% gain are MORE ACCURATE.

There's MORE! Apple ran the 3236 build of UT2004. We did, too, in the graph above. However, take a look at the graph below to see what happens when we used build 3323 of UT2004:

The iMac G5's frame rates increase significantly when build 3323 of UT2004 is used instead of build 3236. The advantage of the G5 iMac over the iMac G4 jumps from 52% to 93% -- though far from Apple's claimed 212% gain.


HERE ARE THE DETAILS EXPLAINING THE RESULTS

THE BIG FLAP
My posting on Sept 21st of HALO and UNREAL TOURNAMENT 2004 (UT2004) results created such a controversy that I decided to devote a special page to iMac G5 versus iMac G4 3D Gaming. I chose those two apps since they are among the newest and most popular of first-person-shooter 3D accelerated games.

Not having paid close attention to Apple's iMac G5 web pages, I didn't realize that Apple had compared the iMac G5 and iMac G4 using the SAME TWO 3D games that I chose. Their original graph showed the iMac G5 to be 179% faster than the iMac G4 running Halo and 212% faster than the iMac G4 running UT2004. My initial tests showed it to be 11 and 12% faster respectively -- but I was using different settings, different memory config, different botmatches, etc.

My report obviously got Apple's attention because they contacted me to see what I was doing differently. We compared notes. I decided to restest both of our iMacs exactly the way they did it -- hoping to see the same results.

THE PLOT THICKENS
When we used their settings, memory config, and app versions, we saw the iMac G5 advantage grow to 39% with Halo and 52% with UT2004 -- but still not close to Apple's "three times faster" numbers. After more head scratching early last week, Apple asked what size screen our iMac G4 had. We answered, "20 inch." Apple said, "Then that's the problem." Seems Apple used a 17" iMac G4. The two 1.25Ghz models have identical specs except for screen size, according to Apple's own documents. They are supposed to be identical in performance -- but Apple insisted that the 17" is slower -- for reasons they promised to explain a detailed technical email -- which I'm still waiting to receive a week later.

THE FINAL TWIST
My request for readers with an iMac G4/1.25GHz machine (or access to one) has paid off. As of 5AM Sunday, October 10th, I received HALO and UT2004 results for a 17" iMac G4/1.25. The results matched our 20" iMac G4/1.25, thereby refuting the claim by Apple that a 17" iMac G4 is inherently slower. Apple's test iMac G4 unit is either "uniquely slow" or something else is "haywire."

CONCLUSION
More than two weeks have passed since my first posting of results. It's been a wild ride. I've contended from the start that iMac G5 is a great machine at a great price. It has as much as a 93% advantage over the iMac G4 running iLife apps.

However, when it comes to two of the top 3D "run-and-shoot" games, rather than the 179% to 212% advantage proudly reported by Apple, we are now convinced that the real numbers are more like 39% to 93%. Those are still good numbers. We still think the iMac G5 is a great replacement for the iMac G4, but words like "unparalleled 2D and 3D graphics performance" are a bit over the top.

December 27th, 2004 -- Last week two readers alerted me that Apple had updated their graphs to show a smaller gap between the iMac G5 and iMac G4. They now claim a 47% advantage for Halo and 99% advantage for UT2004 -- which very closely matches our findings of 39% and 93% respectively.


MEMORY CONFIG DOESN'T MATTER -- As long as you have enough.
We've run tests using 9 applications in an iMac G5 with a single 1GB module versus dual 512MB modules. Average gain = 0%. More on this later.

IN CASE YOU WANT TO REPLICATE OUR RESULTS...

APPLE'S HARDWARE SETUP:
iMac G4/1.25GHz 17" SCREEN with 1GB of memory (two 512MB PC2700 modules).
iMac G5/1.8GHz 17" SCREEN with 1GB of memory ( two 512MB PC3200 memory modules as recommended in Apple iMac G5 tech docs.) Energy Saver set to "HIGHEST"

BARE FEAT's HARDWARE SETUP:
iMac G4/1.25GHz 17" SCREEN with 768MB of memory (one 512MB PC2700 module + one 256MB PC2700 module).
iMac G4/1.25GHz 20" SCREEN with 1GB of memory (two 512MB PC2700 modules).
iMac G5/1.8GHz 17" SCREEN with 1GB of memory ( two 512MB PC3200 memory modules as recommended in Apple iMac G5 tech docs.) Energy Saver set to "HIGHEST"
iMac G5/1.8GHz 20" SCREEN with 1GB of memory ( two 512MB PC3200 memory modules as recommended in Apple iMac G5 tech docs.) Energy Saver set to "HIGHEST"

APPLE'S HALO TEST SETUP
HALO 1.05 TimeDemo (Average of 3 runs)
HW Shaders = No Shaders, FSAA = none,
Lens Flare = medium, Model Quality = high
VIDEO: Resolution 1024x768, Refresh = 0, Framerate Throttle = no vsync
Specular = off, Shadows = off, Decals = off, Particles = off
Texture Quality = medium
SOUND: Sound = on, Sound Quality = low, Sound Variety = medium

APPLE'S UNREAL TOURNAMENT 2004 SETUP (Build 3226)
SantaDuck LCDBench was used to run this test (average of 3 runs):
Botmatch -- > Asbestos --> 1024x768 --> Maximum Settings which translate to:
12 bots, Colordepth 32bit, Texture Detail HIGHEST, Character Detail HIGHEST, World Detail HIGH, Physics Detail HIGH, Dynamic Mesh LOD HIGHEST, Decal Stay HIGH, Character Shadows FULL, Decals ON, Dynamic Lighting ON, Coronas ON, Trilinear Filtering ON, Projectors ON, Foliage ON, Weather Effects ON, Fog Distance Bar MAX, Sound Detail LOW.

...FOR 3D GAME RESULTS AT OUR "NORMAL REAL WORLD" SETTINGS compared to Power Macs and PowerBooks , SEE PAGE THREE.

...DOES MATCHING PAIR OF MEMORY MAKE THE iMAC G5 GO FASTER? SEE OUR ANSWER.

...FOR RESULTS ON THE IMAC G5 vs OTHER MACS running AppleWorks, iMovie, iTunes, FileMaker Pro, Photoshop and Motion, SEE PAGE ONE.

WHERE TO BUY MACS

When ordering products from Apple Store USA, please click THIS TEXT LINK or any Apple display ad as your "portal" to the online store. In so doing, you help to support Bare Feats. (DON'T LIVE in the USA? See links for Apple online stores in other countries.)

WHERE TO BUY REFURBISHED or CLOSEOUT iMACS (12 month warranty)
For refurbished, reconditioned, open box (as well as new) iMacs, check with Small Dog and Power Max. Also look for the "SPECIAL DEALS" link on Apple's online store.

Has Bare Feats helped you? How about helping Bare Feats?

2004 Rob Art Morgan
"BARE facts on Macintosh speed FEATS"
Email , the webmaster and mad scientist