Page 1
Article Type: Review
Article Date: October 02, 2001
Hercules chose the KYRO II chip for their 3D Prophet 4500. KYRO II uses tile-based rendering to streamline graphics production. The hardware tests whether or not a polygon is rendered before it is actually drawn. This method saves the hardware from having to draw and output polys to the screen that will end up being hidden by other objects anyway. The increase in rendering efficiency is as high as 300%. While there is a similar ability in the GeForce3 boards under DirectX 8, PowerVR appears to have taken the technology to new heights in the KYRO II.
Admittedly, KYRO II needs that increased efficiency since the core clock only runs at 175MHz. This chip only sports 15 million transistors, compared to the roughly 24 million of GeForce2.
Early tests were promising, placing the Hercules 3D Prophet at least equivalent to the GeForce2 MX boards.
Where T&L is not a consideration, there is only a single further weakness in the KYRO II chip. It is DirectX 8 compatible, but not compliant, meaning that the 3D Prophet does not support all the features of the newer video hardware. But those features, as seen in GeForce3 and the soon to arrive Radeon 8500, come at a price premium anyway. As a competitor for GeForce2, the new chip looks to be an excellent solution.
On the positive side of the ledger, all internal rendering in KYRO II is done at a full 32-bits, giving superior image quality when compared to the GeForce2 boards. It is also claimed that there is less performance penalty for anti-aliasing than with GeForce2. To check out these claims is the aim of this article. Let’s pop the clutch and see how the 3D Prophet performs.
WARNING! The installation process is flawless if you FIRST close background applications that monitor memory. Either my firewall (Zone Alarm) was the issue or my IPE anti-virus software. After rebooting my hung machine, I closed all these background memory monitoring applications and reran the setup program. As with many video accelerators, the install notes recommend bypassing the Hardware Wizard and instead running the included Setup program.
After closing the background applications I was able to complete the installation. Once the board was properly installed I had to adjust the display positioning slightly. I found this odd, but I assume that the new installation uses a different refresh rate than my default setting. There was no flicker from the display which would have been noticeable at 72Hz or less.
Virtual pilots have two primary considerations when comparing video hardware for the PC: performance per dollar and compatibility. We want to know we can run today's games and tomorrow's games and we all want the most horsepower for our cash.
3DMark2001 is the latest iteration in MadOnion’s popular synthetic gaming test suite. It offers sixteen tests to determine gaming performance, and then averages the results of these tests to present a 3DMark score. It is well known that there is some bias toward hardware T&L in the benchmark, but actual in-game tests will balance our final results.
This first chart represents the result of those tests.
The 3DMark suite performs identical tests on an identical system. You can see from the individual scores that the diversity of tests is substantial. The “nature” test requires hardware T&L but failing the test does not affect the overall score.
One test that the 3D Prophet completed that the GF2 MX did not complete is the Environmental Bump Mapping. This feature, introduced by Matrox some years back, gives very lifelike effects to textured surfaces, like lakes and ponds. While GeForce3 supports this effect under DX8, GF2 does not.
While the 3D Prophet suffers from not supporting hardware T&L, it is reasonable to compare it to the GF2 MX since most simulations do support this feature and the lack of it may affect your gaming.
The overall 3DMark of the two boards is very close. As we would predict, it appears that whenever the complexity of a scene and its lighting increase, the 3D Prophet falls behind the GF2 MX.
This in-game measurement was taken with anti-aliasing ON, and shows the GF2 MX outperforming the 3D Prophet by roughly 40 percent.
The likely cause is DX8. WarBirds is written for DirectX 8, whereas CFS2 was written for DirectX 7. The additional reliance of the more modern games on hardware based T&L appears to be a challenge for the 3D Prophet, particularly when run with anti-aliasing on. The WarBirds test was run in free flight mode offline, with the frame rate noted in the outside view.
The difference was less pronounced in EF Typhoon, where the frame rate gain was roughly 20 percent, with the GeForce2 MX outperforming the 3D Prophet. I tested both internal and external view, and the difference was present in both places.
The next question is image quality. I found the general in-game screens to be very difficult to evaluate unless I took a screen capture to stop the action. In these instances I found color saturation to be excellent with board boards. There was one qualitative difference, however.
An internal rendering pipeline of 32-bits should make the image quality of the KYRO II on the 3D Prophet superior. In fact, it is the anti-aliasing implementation where the image is superior.
The first image is a screen capture taken running the GF2 MX in IL-2 Sturmovik under 2x FSAA using GF2 MX. Sample 2 is the same image running under the 3D Prophet. To be fair, running the 3x FSAA OpenGL setting under GF2 gives roughly the same image with the same framerate as the 3D Prophet.
Numerous reviewers have noted the impressive anti-aliasing ability of the KYRO II hardware. Others have noted that images in general are richer in texture. I found this latter claim more difficult to verify.
If, however, you can find a 3D Prophet for around $60, and you aren’t worried about hardware support for T&L, then it might be worth considering. Otherwise, I would wait for the new version of the board when the KYRO III becomes available.
Test System
Printer Friendly
Hercules 3D Prophet 4500 64 MB
by Len "Viking1" HjalmarsonArticle Type: Review
Article Date: October 02, 2001
My! How Things Have Changed
With the closure of 3dfx this year the 3D accelerator race appeared to resolve to a fine point, with serious competition between only ATI and NVIDIA. Matrox admits to no longer participating in the gaming hardware race, content to produce more for business machines and video editing. Who would have thought PowerVR would suddenly re-enter the picture?  |
| Hercules 3D Prophet 4500 - 64 MB |
Hercules chose the KYRO II chip for their 3D Prophet 4500. KYRO II uses tile-based rendering to streamline graphics production. The hardware tests whether or not a polygon is rendered before it is actually drawn. This method saves the hardware from having to draw and output polys to the screen that will end up being hidden by other objects anyway. The increase in rendering efficiency is as high as 300%. While there is a similar ability in the GeForce3 boards under DirectX 8, PowerVR appears to have taken the technology to new heights in the KYRO II.
Admittedly, KYRO II needs that increased efficiency since the core clock only runs at 175MHz. This chip only sports 15 million transistors, compared to the roughly 24 million of GeForce2.
Early tests were promising, placing the Hercules 3D Prophet at least equivalent to the GeForce2 MX boards.
180 Days Later
Here we are, roughly six months after release of the 3D Prophet, and drivers have matured considerably. In-game tests are revealing output equivalent to GeForce2 Pro in some instances, though generally between GeForce2 MX and the full GeForce2 boards. The battle lines appear to be drawn somewhere around the Transformation and Lighting (T&L) mark. Games which more completely support hardware T&L, or which use more light sources, benefit more by the hardware T&L abilities of GF2 and GF3, and tend to leave the 3D Prophet in the dust.Where T&L is not a consideration, there is only a single further weakness in the KYRO II chip. It is DirectX 8 compatible, but not compliant, meaning that the 3D Prophet does not support all the features of the newer video hardware. But those features, as seen in GeForce3 and the soon to arrive Radeon 8500, come at a price premium anyway. As a competitor for GeForce2, the new chip looks to be an excellent solution.
On the positive side of the ledger, all internal rendering in KYRO II is done at a full 32-bits, giving superior image quality when compared to the GeForce2 boards. It is also claimed that there is less performance penalty for anti-aliasing than with GeForce2. To check out these claims is the aim of this article. Let’s pop the clutch and see how the 3D Prophet performs.
Installation
Modern video boards rarely give much trouble during installation, unless the driver versions are very early. I was a bit surprised, therefore, when I immediately ran into installation troubles with the 3D Prophet.WARNING! The installation process is flawless if you FIRST close background applications that monitor memory. Either my firewall (Zone Alarm) was the issue or my IPE anti-virus software. After rebooting my hung machine, I closed all these background memory monitoring applications and reran the setup program. As with many video accelerators, the install notes recommend bypassing the Hardware Wizard and instead running the included Setup program.
 |
| Properties Tab in WIN ME |
After closing the background applications I was able to complete the installation. Once the board was properly installed I had to adjust the display positioning slightly. I found this odd, but I assume that the new installation uses a different refresh rate than my default setting. There was no flicker from the display which would have been noticeable at 72Hz or less.
Benchmarks and Game Testing
- GF2 MX (driver 12.41)
- Hercules 3D Prophet 4500 (driver 7.114)
Virtual pilots have two primary considerations when comparing video hardware for the PC: performance per dollar and compatibility. We want to know we can run today's games and tomorrow's games and we all want the most horsepower for our cash.
3DMark2001 is the latest iteration in MadOnion’s popular synthetic gaming test suite. It offers sixteen tests to determine gaming performance, and then averages the results of these tests to present a 3DMark score. It is well known that there is some bias toward hardware T&L in the benchmark, but actual in-game tests will balance our final results.
This first chart represents the result of those tests.
 |
| 3DMark Scores |
The 3DMark suite performs identical tests on an identical system. You can see from the individual scores that the diversity of tests is substantial. The “nature” test requires hardware T&L but failing the test does not affect the overall score.
 |
| 3dMark Details |
One test that the 3D Prophet completed that the GF2 MX did not complete is the Environmental Bump Mapping. This feature, introduced by Matrox some years back, gives very lifelike effects to textured surfaces, like lakes and ponds. While GeForce3 supports this effect under DX8, GF2 does not.
While the 3D Prophet suffers from not supporting hardware T&L, it is reasonable to compare it to the GF2 MX since most simulations do support this feature and the lack of it may affect your gaming.
The overall 3DMark of the two boards is very close. As we would predict, it appears that whenever the complexity of a scene and its lighting increase, the 3D Prophet falls behind the GF2 MX.
In-Game Tests
3DMark appears somewhat biased toward raw throughput, but the KYRO II doesn’t work that way. It could be that the test itself isn’t able to accurately represent the tile-based rendering of the KYRO II generation. Benchmarks in Microsoft's Combat Flight Simulator 2 (CFS2) gave the lead to the 3D Prophet 4500 by nearly 25 percent. On the other hand, when I ran benchmarks in WarBirds 3 and Eurofighter Typhoon, I saw the opposite result.  |
| WarBirds Benchmark |
This in-game measurement was taken with anti-aliasing ON, and shows the GF2 MX outperforming the 3D Prophet by roughly 40 percent.
The likely cause is DX8. WarBirds is written for DirectX 8, whereas CFS2 was written for DirectX 7. The additional reliance of the more modern games on hardware based T&L appears to be a challenge for the 3D Prophet, particularly when run with anti-aliasing on. The WarBirds test was run in free flight mode offline, with the frame rate noted in the outside view.
The difference was less pronounced in EF Typhoon, where the frame rate gain was roughly 20 percent, with the GeForce2 MX outperforming the 3D Prophet. I tested both internal and external view, and the difference was present in both places.
The next question is image quality. I found the general in-game screens to be very difficult to evaluate unless I took a screen capture to stop the action. In these instances I found color saturation to be excellent with board boards. There was one qualitative difference, however.
An internal rendering pipeline of 32-bits should make the image quality of the KYRO II on the 3D Prophet superior. In fact, it is the anti-aliasing implementation where the image is superior.
 |
| GF2 MX |
 |
| 3D Prophet |
The first image is a screen capture taken running the GF2 MX in IL-2 Sturmovik under 2x FSAA using GF2 MX. Sample 2 is the same image running under the 3D Prophet. To be fair, running the 3x FSAA OpenGL setting under GF2 gives roughly the same image with the same framerate as the 3D Prophet.
Numerous reviewers have noted the impressive anti-aliasing ability of the KYRO II hardware. Others have noted that images in general are richer in texture. I found this latter claim more difficult to verify.
Conclusion
At one time the 32 MB version of the 3D Prophet was an excellent value. With the release of the GF2 MX 400, however, you can now obtain a 64MB GF2 board for about the same price. Under those circumstances, it’s difficult to recommend the 3D Prophet.If, however, you can find a 3D Prophet for around $60, and you aren’t worried about hardware support for T&L, then it might be worth considering. Otherwise, I would wait for the new version of the board when the KYRO III becomes available.
Test System
- MoBo / CPU: Abit KT7 with 945MHz AMD Athlon
- RAM: 384MB SDRAM
- Video: OCZ Titan 3 GF 3
- Sound: Creative SB Live! X-Gamer 5.1
- CD-ROM: Toshiba 72x TrueX CD ROM
- OS: WIN ME
Printer Friendly