 |
|
|
By Bryan William Jones Review date: 9/21/99 |
|
|
|
|
|
|
|
| Applications/General use comments | |
|
Application performance included tests in Adobe Photoshop, Bryce 2, JMP (a statistics program by SAS Institute), EASI/PACE 6.1 (this is a remote sensing/image processing package from PCIGeomatics), IDL (a data visualization program from Research Systems) and MacSpartan plus (a molecular modeling program from Wavefunction). |
|
| Adobe Photoshop Tests: | |
|
As in the previous review, all operations were performed on a 44.4 MB image in Adobe Photoshop 5.02 with 300 MB of memory allocated to the application to minimize any hard drive access for the test. Times reported are for a complete series of operations encompassing image transitions, RGB conversions, and a standardized sequence of filters and mask operations.
The tests in Photoshop reveal that for the 466 MHz G3 card the savings
in time for a 44.4 MB image was almost two minutes per image. Or
about a 18% increase in performance for Photoshop over the 604e@300Mhz.
People who make their living with Photoshop will certainly appreciate the
time savings here. It should be noted that even with with large allocation
of RAM, Photoshop still relies heavily on the hard disk, and as such a
fast hard drive is critical for good performance in opening and saving
images in Photoshop or while working with larger images that may exceed
the RAM allocation.
|
|
| Bryce 2 Tests: | |
|
I don't have the latest version of Bryce, however Bryce 2 still gives a good indication of the relative 3D rendering performance of the two CPU's. At any rate, Bryce 2 is a 3D design tool that allows one to easily perform rendering of 3D environments, however, like any 3D rendering application, you need lots of horsepower. In keeping with previous reviews with Bryce 2, sample scenes in the KAI folder "Magical Night I", and "Alexandria II" were used to gauge performance. Again Bryce 2 was left with the default memory allocation of 16980k, and anti-aliasing was on. All times are reported in seconds. 
With 3D rendering in Bryce 2, the Newer G3 running at 466 Mhz proved itself
to be from 40-50% faster than the 604e@300 Mhz. |
|
| JMP Tests: | |
|
JMP is a statistics program designed to make data exploration through statistics easier to visualize. Times reported are for a correlation matrix calculation of a large spreadsheet database. Times reported are in seconds.
|
|
| EASI/PACE & IDL Tests: | |
|
The specific test performed in this program was an image classification. This is a computationally intensive mathematical procedure originally developed by the Department of Defense (DOD) and later NASA to classify pixels in a multilayered or multispectral image into categories based on hyperdimensional statistical calculations for the purpose of interpreting satellite images from space. Originally, these techniques were used to help cull valuable defense information from the thousands of satellite photographs combined with the hyperspectral databases being generated by the first series of satellites with multispectral capabilities. Currently the CIA, NRO, and NIMA use these same statistical analysis to analyze multispectral information from the KH series of satellites. In looking for specific things, one can inform or seed a calculation to look for certain signatures in images that correspond to items of interest, like concrete disguised to look like natural rock for instance. NASA and other organizations and businesses use these techniques to prospect from space or map patterns of deforestation on earth in addition to determining the composition of other planets or objects.
These tests indicate an increase in performance of the G3@466 over the 604e@300 Mhz of approximately 37%. These are about the same results as the last review with the 400 Mhz G3 over the 300 Mhz 604e. I now believe that this performance plateau is due to the origins of this software. This software was originally written for UNIX workstations when memory was very expensive and storage on magnetic tape was much cheaper and easier to implement. For instance, with the EASI/PACE software, even when you allocate more memory to the application under the MacOS paradigm, the software only uses up to 32 MB of RAM. Operations performed with this software are disk intensive rather than memory or CPU intensive. While EASI/PACE is a very powerful and flexible application, in order to get an appreciation for the relative performance of these two CPU's with a more modern software implementation I ran tests with IDL from Research Systems Inc. IDL is a flexible 4th generation programming environment suited for the visualization of large volumes of data from a variety of disciplines. The tests implemented were again a classification of the same image with a program custom written in IDL. My primary occupation is not that of a programmer, and I am not as proficient as coding or porting in IDL as I could be so the code is probably not as optimized as it could be, but given that, the performance of the G3 CPU can more accurately be determined with an application that makes more efficient use of memory and CPU resources as can be seen in the following graph.
The performance in IDL of the G3@466 Mhz is approximately 46% faster
than the 604e@300 Mhz.
|
|
| Unreal TimeDemo Tests | |
|
Using the latest version of Unreal (The Westlake folks must be going nuts by now with all of the updates to Unreal they are porting) I tested the frame rate differences between the 604e@300 Mhz and the G3@466 Mhz. All tests were run at 800x600 with all optimizations set for maximal detail and sound quality. The latest beta drivers for the VoodooII card from 3Dfx were used giving much better results than the drivers developed by Microconversions.
As can be seen from this graph, the frame rates increased about 10%
from the 604e chip to the 466 MHz G3 card.
|
|
|
|
|
| The next page describes the software cache control supplied with the card. Or you may use the links below to jump to a specific page. | |
|
|
|
|
Intro | Benchmarks | Appl. Tests | Software/Controls | Documentation | Specifications | Summary - or -
|
|
|
Copyright © Michael Breeden, 1999. No part of this sites content or
images are to be reproduced or distributed in any form without written
permission.
Users of the web site must read and are bound by the terms and conditions of use. |
