System Benchmarks

How a system performs a task, or how fast a system is, is due to the sum of its parts, and the way the parts interact with the software in use.
In the other pages, we have looked at software performance as determined by the system, hard drive, and graphics card in use.
On this page, I hope to show examples of how a machine may benefit from increased RAM, and how a machine can benefit from a faster CPU.

A note about the benchmark utilities:
XBench is an often berated utility.
Myself included, find the results hard to categorize as “definitive” simply due to the fact that scores are difficult to replicate, with unacceptably high variations in scores.

Each test, with Xbench, was preceded by a system Restart, and run with no other application running (including as many background applications as possible, such as Spotlight).
In this manner, test scores can be replicated (usually) with only minor variation in results.

Geekbench provides for relatively consistent test results.
Geekbench, as well, was run after a Restart and run as the only application in process.
In Memory tests, Geekbench does not show sensitivity to the amount of memory installed, but does show the impact of a CPU and the architecture of such, in the performance of the memory bus in a machine.


A bit harder to show, but of utmost importance, is the value of increased memory in a system.
The following graph is to show the results of the impact on system scores in Xbench when upgrading Memory from 256MB through to 1.5MB.
The test machine is a Quicksilver 2002 800 MHz model running OS X 10.4.11, with a single PATA drive and Radeon 7500 graphics card.

 “Should I upgrade my memory?”

Quicksilver 2002 RAM Upgrade

We can see that increasing the memory has a significant impact on a machines performance.
For the test machine, the biggest boost was in moving from 1 512 MB DIMM to 2.
Though having the slots full (1.5 GB) doesn’t seem to provide much of a performance boost over 1 GB, real world use will show, undeniably, the benefit of having at least 1.5 GB RAM in a G4.

Should you upgrade your memory? 
For any G4 model, of any configuration, running any OS, the answer is “Of course”.


System Upgrades

In the next set of tests, I take a 800 MHz Quicksilver 2002 machine in shipped configuration (256 MB RAM, Radeon 7500 graphics), upgrade the RAM (1.5 GB) and graphics card (Radeon 9800 Pro 128 MB VRAM), and then upgrade that configuration with faster processors.
Finally, I add my favorite graphics card, the Geforce 6800 GT and a SATA 150GB Raptor boot drive (via Seritek PCI controller) to show the upgrades total performance benefits as measured with Xbench.
All tests use the same motherboard, PATA drive (except last) and the same 3x 512 GB DIMMs.

Quicksilver System Upgrade

The tests show that as faster parts are added and memory is increased, performance is dramatically increased.
The biggest single step is in the move from the 800 MHz processor to the 933 MHz processor.
The addition of the L3 cache of the 933 MHz processor along with increased frequency, combine to improve performance 212% over the 800 MHz QS shipping configuration.

In the final configuration, we see that performance gains of 280% can be achieved over a base Quicksilver configuration.

Though the test machine is a Quicksilver, the basics are the same for all G4 lines, and upgrades to these machines will follow similar trends.

In this following test, I use Halo to show the improved performance of the above 4 configurations that use 1.5 GB RAM and the Radeon 9800 Pro.

Processor vs. Halo Performance

As can be seen, the increased CPU power takes Halo from an unacceptable frame rate, to a very acceptable frame rate (30 fps and higher desired).
Of course, lower resolutions and lower quality settings can allow for good game play on slower machines, but the point is that there is no substitute for powerful components in a system.

See the Halo page for broader hardware test results.


Processor and Memory

Geekbench provides us with a way to quantitively measure a processors performance, and qualitatively compare the processor and memory performance of various machines in various OS versions.

Geekbench 2.1 - OS X 10.4.11

There is a measurable performance hit when running OS X 10.5x on a G4.

Geekbench 2.1 - OS X 10.5.6
The 800 MHz QS is NOT used in this test.

Not only memory performance takes a hit in Leopard. There is also a dramatic reduction in the Processor scores as compared to Tiger.

This performance hit has also been noted in other tests within these pages.