Tyan Tachyon G9000 Pro Review October 30, 2002 Dave Barron

Introduction

Page:: ( 1 / 10 )

With ATI having taken a significant technology lead with their R300 core, leaving themselves exclusively to the high-end market was certainly not a part of their agenda. With lower margins and higher volume, the low-end and mid-range markets are prime territory for anyone, and certainly not segments that ATI intended to neglect. Developing a scaled back version of the R200’s core, RV250 was created. From that, ATI introduced RADEON 9000 PRO, a high performance, low-cost derivative of last year’s RADEON 8500.

Being aware of the detriment brought about by relying on one's boards sales exclusively, ATI has allowed other board manufactures to distribute products based on their architectures. Tyan, being one such board manufacturer, has produced their Tachyon 9000 Pro. Based entirely off of ATI's RADEON 9000 PRO reference design, both RADEON and Tachyon include 64 MB of 275 MHz DDR memory (though R250 supports up to 128MB of memory), with core clocks operating at 275 MHz.

Letting ones chip into the open market for a variety of board manufactures to produce, while primarily advantageous, can also be a detriment. While all based on the same chip that one produces, higher margins are gained through direct board sales. Competition is also created, which can drive down prices without some type of price regulation on the chip manufacturer’s part. In the case of ATI's partnership with Tyan, Tachyon sales allow for further market penetration, while potentially reducing ATI's own board sales.

Specifications

The feature set of RV250 is primarily based on that of R200. Running at a 275 MHz core clock allows for RV250 to provide a theoretical fill-rate of just over 1 Gigapixel/sec while supporting pixel shader version 1.4. Unlike R200 though, RV250 includes only a single texture unit, requiring all multi-texture rendering to be multi-cycle. This, however, does not require multi-pass rendering, as 6 textures can be written in a single pass via loopbacks, or reuse of texture coordinates. From this we can ascertain that applications that require multi-texturing will operate more efficiently on R200-based products.

-ATI® RADEON 9000 PRO 128-bit VPU; 275MHz
- 1100 MPPS Theoretical Fill-Rate
-64MB DDR memory; 275MHz DDR
-VGA, DVI-I, and S-Video connectors
-AGP 4x/2x/1x compatible
- Pixel Shader 1.4
- Vertex Shader 1.1

Shaders/SMOOTHVISION

Page:: ( 2 / 10 )

As with the pixel shader, the vertex shader onboard RV250 is nearly identical to that found on R200, supporting version 1.1. This is met with the support of 128 constant registers, allowing for the full number of operations to be performed. T&L performance with one light is in the 19-21 million vertices/sec range in 3Dmark 2001.

SMOOTHVISION

ATI's SMOOTHVISION technology is used in RV250 as well. It is in many ways similar to the anti-aliasing implementation found on Voodoo5, with some slight variations. Where Voodoo5 used a fixed jittered sampling position, SMOOTHVISION uses variable sampling, based on a select set of sampling patterns.

The human eye was designed in a way to be pattern sensitive. Patterns can easily be found, because they are typically very uniform, repeating identically over and over. Random noise is the opposite of a pattern, making it extremely difficult for the human eye to detect. The more random something is, the less likely it is to be noticed. This principle applies to anti-aliasing, with the more random a pattern, the less likely it is to be noticed. For example, one image can have a 6x ordered grid and another with a 4x random grid. Which will deliver a better result? Displaying noise, the 4x grid will have better results, going unnoticed by the eye. While the 6x grid is taking additional samples, these samples still remain in a uniform shape, making them easily noticed.

SMOOTHVISION makes a legitimate attempt at crossing the world of uniform patterns and noise. While definitely not noise, SMOOTHVISION provides several different sampling patterns to be chosen from. By doing this, a uniform pattern is no longer present; it becomes more difficult for the eye to detect. This provides a notable increase in anti-aliasing quality, with little to no variation in performance when compared to uniform sampling.

Anisotropic filtering

ATI's implementation of anisotropic filtering has certainly been a topic noteworthy of discussion. Anisotropic filtering is unique in that it filters in a non-uniform pattern, increasing overall quality and image accuracy. ATI's implementation is strictly bilinear in that it only filters within a single mip-map level. Other implementations, such as NVIDIA's, provide greater quality by using a trilinear anisotropic filter, filtering across mip levels.

ATI additionally provides a dynamic filtering implementation, meaning that at 64-taps (16x), implementation will provide a maximum of 64 samples, with some surfaces using a reduced number. This makes considerable sense as not all surfaces are of the same size and orientation, thus not necessitating the use of all 64 samples. With such an implementation, it is possible for the algorithm to miscalculate certain surfaces, not providing enough samples. This has been known to happen, but certainly it is not a common occurrence. With the performance advantage associated with this, it is seen to be a worthwhile implementation.



SIDEBAR: ATI recently dropped their official price on the RADEON 9000 PRO from $149 to $129.

HyperZ II/Card details

Page:: ( 3 / 10 )

Occlusion culling

HyperZ-II is found onboard Tachyon as well, as with all RV250 based boards. Besides a variable rate lossless Z-Buffer compression, a hierarchical Z-buffer is the primary feature of this. ATI's implementation of this uses 8x8 pixel blocks for hierarchical Z-buffering. The nearest vertex value from the 8x8 block is taken and compared to a stored reference value. This comparison tests to see whether any of the pixels within the block are visible. If the nearest pixel in the block is deeper than the reference value, the pixel block is discarded. If the vertex is nearer than the reference value, the block is rendered.

This implementation of occlusion culling presents both certain advantages and disadvantages. In its favor, hierarchical Z-buffering is highly advantageous in memory bandwidth consumption. Being that the entirety of the block is compared on chip to the stored reference value, there is little access to external memory. On the negative side, if only a single pixel is visible the entire block must be rendered.

The Tachyon G9000 card

[image]

Overclocking

Overclocking the Tachyon proved fairly painless. Loading up a third party overclocking utility such as Powerstrip, it was not an issue to pull a few more frames per-second out of the board. When overclocking the core, Tachyon did find itself with a potential advantage in that it provided an overall better active cooling system than Radeon 9000 Pro. The heatsink onboard Tachyon was roughly double the size of ATI’s RADEON 9000 PRO card, providing more efficient heat removal.



Running Tachyon at 290/290 did not produce any apparent artifacting or stability issues. Fill-rate levels were increased somewhat, as was overall performance.






SIDEBAR: Tyan bundles an S-Video to composite cable, S-Video cable, and Composite video cable with the Tachyon G9000 Pro.

System Setup

Page:: ( 4 / 10 )

AMD Athlon XP 1900+

Shuttle Spacewalker AK32A KT266A motherboard

512 MB DDR SDRAM

ATI Radeon 9000 Pro 64 MB
Tyan Tachyon 9000 Pro 64 MB
Catalyst 2.3 Drivers: v6.13.10.6166

Hercules Muse 5.1 DVD

Windows XP Professional

DirectX 8.1

Desktop Resolution 1024x768x32 at 75 Hz

All benchmarks done at 32-bit color

Benchmarks

3D Mark 2001 Second Edition Build 330 – 32-bit color
Quake III: Arena
Serious Sam: The Second Encounter




SIDEBAR: Tyan also ships the Tachyon G9000 Pro with a copy of WinDVD 4

3DMark 2001SE

Page:: ( 5 / 10 )

3D Mark 2001 SE

Notes

3DMark gives us an overall picture of the Tachyon's performance. Fill-rate was found to be slightly lower overall on Tachyon, which for the most part was consistent with other benchmarks. As seen by the recorded number, Tachyon and Radeon both come up more than 400 MPPS short of their theoretical peak. This shortage can be attributed to a variety of factors, yet specifically memory bandwidth.

Notes

The 3DMark lobby scene is certainly indicating a CPU and T&L limitation. Fill-rate remains available to spare for resolutions up to 1024x768. From this point we see a noted fall off, indicating that a fill limitation is taking place.




SIDEBAR: The Tachyon G9000 Pro ships with a 3-year warranty

Serious Sam/Quake 3

Page:: ( 6 / 10 )

Serious Sam 2

Notes

Serious Sam 2 showed consistent playability, even at 1600x1200. Granted, at this resolution, some level of slow down exists, 1280x1024 showed definite playability levels. What we also note is that Tachyon showed a slightly reduced level of performance throughout all resolutions when compared to Radeon.

Quake III Arena

Notes

Performance levels in Quake III were pretty much expected. Up to 1024x768, there are definite CPU limitations. From there, fill-rate becomes a factor and performance degrades. The variations between Tachyon and Radeon are clearly minor, with Tachyon taking the lower resolutions and Radeon taking the upper of those.







SIDEBAR: The Tachyon G9000 Pro has already been certified by ATI’s new certification program.

Anti-aliasing Performance

Page:: ( 7 / 10 )

Quake 3 – 2x AA

Quake 3 – 4x AA

Notes

Anti-aliasing performance was pretty low, as is typical with super-sampling implementations. Using a RGSS technique, quality, however, was extremely high. As seen, with 4x AA, anything over 1024x768 overflows the 64MB of available memory, making it impossible to store the necessary buffers. With 2x AA, this overflow occurs at 1280x1024.




SIDEBAR: Tyan is also working on a RADEON 9700 PRO-based card, but it isn’t out yet.

Anisotropic Filtering

Page:: ( 8 / 10 )

Quake 3 –16x Anisotropic filtering

Notes

Anisotropic performance was fairly consistent, showing little degradation. When compared to bilinear performance levels, the change is actually rather small. This can be attributed to the use of dynamic filtering levels, as well as ATI’s use of a bilinear anisotropic filter.

Quake 3 – 16x Anisotropic filtering/4xAA

Notes

With anisotropic filtering having such a minimal performance impact, as was expected, using anisotropic with anti-aliasing had minimal impact on performance when compared to simply using anti-aliasing alone.




SIDEBAR: Tachyon: A hypothetical subatomic particle that always travels faster than the speed of light.

Ballistics Report

Page:: ( 9 / 10 )

Pros

Performance: For a low-end board, as with Radeon 9000 Pro, performance is fairly impressive. Each successive generation of low-end products definitely raises the bar for what one considers low-end. Historically, low-end products had out-of-date feature sets and under-par performance. Yet, with RV250 based boards, such as Tachyon, the latest DirectX feature set is supported and performance is more than adequate.

Cooling: Tachyon 9000 Pro provides better overall cooling than ATI’s incarnation of the same board. While enthusiasts may replace the entire cooling system on their own, Tyan was considerate in providing a larger, more efficient cooling system. Doing so provides for greater potential overclocking directly out of the box.

Cons

Memory: While 64 MB of memory is adequate, when it comes to using high resolutions, and especially anti-aliasing, it simply doesn’t cut it. Anti-aliasing quickly consumes available memory, reducing the amount of potential texture memory.

Availability: Finding a Tyan Tachyon G9000 Pro online is rather difficult; many of the larger online retailers don’t seem to carry graphics products. Considering that this is the company’s first foray into graphics, this isn’t too surprising, but it is rather frustrating if you have your heart set on purchasing the Tachyon G9000 Pro over other cards based on the RADEON 9000 PRO.

Tyan has been in the motherboard business for over a decade, perhaps they can use some of these relationships to work their way into more retail channels.



SIDEBAR: Tyan was AMD’s launch partner for the 760MP chipset.

Final Verdict

Page:: ( 10 / 10 )

Speak!