The Performance Edge: How Much
Difference Does Having a Faster "Backside" Cache Make
- A Performance Review
A couple of years ago when the first G3's came out we took a
limited look at the effects of backside cache speed. At that
time there were upgrade cards whose backside cache speed ran
at full processor speed. These cards were, of course, very expensive.
Now most G3/G4 upgrade cards run at half the processor speed.
Some, like the XLR8 card, allow you to over-clock the backside
cache speed, if possible, from the card's recommended speed.
How far you are able to over-clock really depends on the quality
of each individual chip.
We decided to take a second, in depth look
at cache speed to get a fuller perspective on the benefits
of backside cache, and exactly what processes it speeds up.
To do this we used a XLR8
MAChSpeed ZIF G4/500/250 upgrade card installed in a Beige
G3/266. We clocked the processor to 400 MHz and then used
the control panel to clock-up and clock-down the speed of
the backside cache. For one series of tests we turned the
cache off completely.
The test results are interesting. While having
an operational backside cache as part of the processing system
results in a tremendous boost in performance (up to 300% in
one test), the speed of the cache seems less important, in
most tests. Tests affected by cache speed the most were our
AppleWorks scroll and search & replace tests, which showed
a 14% and 30% improvement respectively over the default setup
. Most other processor intensive tasks showed only between
a 0 - 6 percent improvement when cache speed was doubled.
This would seem to indicate that most light
tasks, such as scrolling, word processing and finder functions,
will see an increased responsiveness, with a high-clocked
backside cache. But for most heavy duty processing, a faster
cache makes minimal difference.
It will be interesting to see if these numbers
hold up when the cache is running at full processor speed.
G3/G4 Machine Cache Bus Scheme
For Great Prices On Upgrades Check The Vendors Below
By far the biggest boost to performance that the PowerPC
G3 offers can be credited to its incorporation of an approach
to level 2 cache memory known as backside cache. This approach
effectively bypasses limitations on the speed at which transactions
between the processor and the level 2 cache can occur. Earlier
PowerPC processors used the system bus to access both the
level 2 cache memory and the main memory, which could result
in conflicts. For example, under the previous approach,
at processor clock speeds above 200 megahertz, the CPU would
often stall as it waited for data to arrive from the level
2 cache. To prevent such slowdowns, the PowerPC G3 processor
features a new dedicated bus that handles only the CPU/cache
transactions. This bus can operate at higher speeds than
the system bus -- speeds that relate incrementally to the
clock speed of the processor. This enables the more effective
use of level 2 cache, because even the relatively large
amounts of data they can store can be accessed by the processor
rapidly and efficiently. In fact, as clock speeds increase,
so does the performance value offered by the backside cache
- Apple Computer
MacBench 5.0 Results
The scores below are from MacBench 5.0. MacBench 5.0 is
a subsystem-level benchmark that measures the performance
of a Mac's processor, disk, and graphics subsystems to name
a few. MacBench normalizes all scores relative to the base
machine, a Power Macintosh G3/300. The base machine receives
a score of 1000. For all MacBench tests, higher numbers indicate
"Real World" Tests
The tests below are from our newly revised suite of real
world application tests. These tests feature a diverse selection
of applications commonly used by the Mac community. The test
suite was designed to render an accurate and well rounded
picture of an upgrade's performance. Click
here for detailed information on each test and our machine's
configuration. All of the tests below (with the exception
of the Quake III tests) were timed with a stopwatch. The times
are then converted to percentages with the base configuration
set to 100%. Lower numbers indicate a better score.
Copyright 1996-2007 by Cider Press Publishing LLC all rights reserved.
MacSpeedZone is not authorized, sponsored, or otherwise approved by Apple
Computer. Apple, the Apple logo, Macintosh, iPod, iBook, iMac, eMac, and
PowerBook are registered trademarks of Apple Computer, Inc. Additional
and product names may be trademarks or registered trademarks and are hereby