Which is fastest, a "Closeout" G4/400 Yikes or a
G4/350 Sawtooth Rev B?
12/28/99
© 1999 Rob Art Morgan, editor of BARE FEATS
rob-art@barefeats.com

 

After reading the G4/400 Yikes versus the G4/400 Sawtooth comparison, many readers said, "But I'm trying to decide between the G4/350 Sawtooth selling for $1595 and the "closeout" G4/400 Yikes selling at $1499. How do they compare?" AND.... is the G4/400 Sawtooth worth $450 more? Here's your answer:

CPU (& FPU) crunching...

Reaction: If you do a lot of 3D rendering, get the fastest G4 Sawtooth you can afford.

 

Reaction: If you are a Photoshop jockey, any model of G4 with AltiVec enabled is wicked fast. A G3/400 took twice as long (12.8 seconds average) for the 4 Photoshop filters!

 

Hard Disk speed comparison

Reaction: All three machines were writing to the same drive,
a Seagate Barracuda 20 gig Ultra-ATA/66. And as you can see, CPU speed DOES determine disk writing speed. By the way, when I ran the ATTO Benchtest, all three machines scored an averaged sustained 27MB/sec read/write (8MB max file size).

  

2D scrolling

Reaction: I was shocked with how slow the RAGE Pro scrolled in Photoshop. I had to go back and retest because I didn't believe the stopwatch. I didn't make sense. It still doesn't.

 

3D Game Frame Rates

Reaction: If your main performance need is 3D gaming, the 350 is just as fast as either 400. Game speed is determined more by the type of graphics board when comparing machines this close in CPU speed.

 

Reaction: CPU speed was more of a factor with Unreal.

 

Reaction: Note the effect of CPU speed on minimum frame rate. The two 400MHz G4's never dropped below 19 fps when using Voodoo3 boards. (The Jerky threshold is 16 fps.)

 

Conclusion: If you are operating on a tight budget, I would get the G4/350 Sawtooth. It's not as easy to upgrade as the Yikes in the short term, but long term it will give you plenty of bandwidth to work with. (I know of at least one third party upgrade company working on a CPU daughtercard for it.)

And unless you are as obsessive about speed as I am, you would need a stop watch to tell the difference between the 350 and 400 MHz Sawtooth in every day life.

Finally, if you bought a G4/350 or G4/400 Yikes "closeout" already, don't feel too bad. For the time being, they are not being smoked by the Sawtooth models... with the exception of the RAGE 128 board. For a $100 or so, you can fix that--buy a 3Dfx Voodoo3 3000!

 

 

Test Configurations & Procedures

	Yikes!	Sawtooth
CPU	G4/400	G4/350
Bus Speed	100Mhz	100Mhz
Cache	1 MB at 2:1 (200MHz)	1MB at 2:1 (200MHz)
RAM	256 MB (8ns)	256 MB (8ns)
Graphics	66MHz PCI slot with either RAGE 128 or 3Dfx Voodoo3 3000	133MHz AGP 2X slot with either RAGE Pro 128 or 3Dfx Voodoo3 3000
Hard Drive	Seagate Barracuda 20 Gig 7200rpm ATA/66 drive connected to a ATA/33 interface	Seagate Barracuda 20 Gig 7200rpm ATA/66 drive connected to a ATA/66 interface
"Stripped Cost" (64MB RAM,10 Gig Drive, RAGE card)	$1499	$1595

(It's cheaper to buy the "stripped down" version of the Sawtooth and add the memory, bigger/faster hard drive, etc., from other online upgrade sources.)

 

• Bryce 3D 4.0 rendering was used to test a combination of CPU and FPU speed.
• Photoshop 5.5 was given 117MB of app size on each machine. The test file was 17MB. AltiVecCore was enabled. And average time was computed for 4 functions:
  • Rotate (30 degrees)
  • Gaussian Blur (30 pixels)
  • LightingEffects (Flashlight - default settings)
  • Motion Blur (30 degrees, 30 pixels)
• For the DUPLICATE THREE 27MB DOCS test, I copied onto each drive. Then I timed how long it took to duplicate three copies of a 27MB Adobe Photoshop document on the same drive. Finally I calculated megabytes per second by with the formula: folder size * 2 / seconds.
• For the DUPLICATE 91Mb FOLDER test, I took the 91MB Unreal Tournament DEMO game folder containing 202 documents and I timed how long it took to duplicate it on the same drive. Finally I calculated megabytes per second by with the formula: folder size * 2 / seconds.
• The 2D scrolling test was done with Photoshop 5.5. A 17MB photo file was zoomed in at 400% after various filters had been applied. Then, holding down the scroll arrow, I timed how long it took to scrool from top to bottom.
• Quake II test was run by pressing "~" after startup and typing "timedemo 1" (return) and "~" again. When the first person dies, I press "~" again to get the frame rate.
• Unreal 3D test used the Castle Flyby that comes up at startup. To get the average and minimum frames per second with version 2.2.x of Unreal, you press "~" and enter "timedemo 1." Then press "~" to open up the full screen again. It begins the fly sequence at the castle entrance. (Get the latest latest Mac version of UNREAL Beta from Westlake's Site.)