ASUS Radeon 9600XT

ASUS Radeon 9600XT
VPU chip	ATI RV360
Memory	DDR 128 Mb; 2.8 ns
Frequencies:	500/600 MHz
Category:	Middle level `2004
Price:	200$

Asus' move to partnership with ATI turned out to be out of the blue. Albeit not earth-shattering, but at least unexpected enough. We've witnessed how one of the long-standing and reliable partners of NVIDIA's suddenly moved to ATI's camp. Of course, here we should make reservation for that Asus hasn't laid off the production of video cards built on NVIDIA chips. It simply released a line of boards based on ATI chips while preserving a complete line of video cards on the base of NVIDIA chips.

To date, Asus' line of video cards based on ATI chips looks like this:

As we see, Asus didn't start producing video cards on outdated chips, but simply overlapped all the market sectors (starting with the cheapest Low-end up to the most costly solutions like Radeon 9800XT ), awaiting what ATI might offer today.

Also, we can't help noting that Asus having joined ATI gave sort of an impetus to other competing manufacturing companies who produce graphic accelerators. Following Asus, a great many manufacturers announced their transition to production of video cards on ATI chips. So, why have the priorities changed? The answer is evident. The leader on the market of graphic chips has been changed. This is rooted in the evidently failing NV30, problems with first suppliers of chips, the GeForce FX 5900 made as a hastily released patch to NVIDIA GeForce FX 5800, general blunders in the whole FX family as to the operation of DirectX 9.0 generation pixel programs, and of course a series of scandalous stories of NVIDIA's "cheating" for specific applications. The latter is just a consequence of the very blunders of the company in the chip architecture (for details, read "ForceWare 52.16: NVIDIA's retaliation ", so it makes no sense repeating). On the contrary, ATI made much fewer mistakes than NVIDIA. As a result, we see the bright orange colored produce by Asus in our today's review.

Quite soon we'll be able to observe a new coil of rivalry between the giants of graphic chips production. The battle "R420 vs NV40" is going to be quite exciting and, unless NVIDIA with its new chip is able to leave ATI well behind, then we believe the "inertia of consumers" will go down and users will no longer buy NVIDIA cards because it is NVIDIA.

But meanwhile, we'll be reviewing ATI's Middle-end representative - ASUS Radeon 9600XT video card based in the ATI Radeon 9600XT chip - in great details.

The ATI RV360 Chip

In our comparative table on ATI Radeon 9600XT, we included the specifications of ATI Radeon 9600 Pro, a predecessor of ATI Radeon 9600XT in the ATI line, and those by NVIDIA. To date, ATI Radeon 9600XT's rival is NVIDIA GeForce FX 5700 Ultra (a quite detailed review was presented in "NVIDIA GeForce FX 5700 Ultra"), as well as NVIDIA GeForce FX 5600 Ultra - its predecessor in the line of NVIDIA chips. Therefore, we can compare the characteristics of video cards of both new and older generation, ATI's and NVIDIA's video cards of the same "weight".

First, let's start with the distinguishing features of 9600XT which proved to be not so many as compared to ATI Radeon 9600 Pro. The question is only about the increased frequencies of the graphic chip: from 400 MHz in ATI Radeon 9600 Pro to 500 MHz in ATI Radeon 9600 XT.

But NVIDIA's evolution in terms of a new chip for the Middle-End has gone even further. First, the operational frequencies of both the graphic chip and the memory have been increased. For the graphic chip: from 400 MHz to 475 MHz; and for the memory: from 400 MHz (800 MHz) to 450 MHz (900 MHz). Besides, the geometry handling speed in NVIDIA GeForce FX 5700 Ultra has been increased due to the raised number of vertex pipelines to 3. Of especial note is that NVIDIA GeForce FX 5700 Ultra uses the DDR II memory. To these, add the overall complexity of boards built on the NVIDIA GeForce FX 5700 Ultra chip. We have to admit, such solution by NVIDIA seems to be rather strange. For the Middle-End sector of graphic accelerator market, this happens not so often. Anyway, the costs for the manufacture of boards built on GeForce FX 5700 Ultra may be quite substantial, as we see it. What we have least doubts of is that the cost of NVIDIA's partners' finished products may exceed that of boards based on ATI Radeon 9600XT. Therefore, we can make an interim conclusion saying that NVIDIA manufactures and sells GeForce FX 5700 Ultra at the breaking point of the prime cost only for the sake of winning the crown in the Middle-End niche for today.

*1200x981; 234Kb

Package bundle:

We found the following contents in the very smart box bearing a logo and an image designed especially for the new series of ASUS video cards built on ATI chips:

• The graphics board itself;
• a VIVO splitter adapter;
• a DVI-to-D-Sub adapter;
• 5 CDs with software and drivers;
• 3 CDs with games and demo versions of new games.

Note the smart way the CDs are "packaged". They are shipped not in paper pouches, and even in not in small boxes, but in a special orange container accommodating as many as 12 CDs. Quite an interesting and most importantly a smart trick. Asus has always been notable among the mass of retail suppliers of boards built on NVIDIA chips and now those on ATI chips for smart ideas, which the vast majority of partnering manufacturers have lacked so much.

Design and layout:

The first what catches your eye is the bright orange color of the PCB instead of the already customary red PCB typical of ATI. Clearly, Asus couldn't help missing the chance of rising above at that =).

*1200x856; 238 Kb	*1200x856; 364 Kb

The board offers 128 Mb of DDR memory, 2x/4x/8x AGP, and a standard set of output connectors: one analogous, one digital, and one TV-OUT. To implement VIVO potentials, the chip offers a proprietary ATI's Rage Theater chip traditional for all ATI boards.

Evidently, Asus decided not to come up with innovations, so didn't install third-party VIVO chips. It's hard to assume if the company proceeded from the chip cost considerations or if it was caused by the functioning of the chips themselves. But anyway in the end we get an ATI Rage Theater chip onboard. The review of VIVO functionalities goes beyond the scope of this article, so we leave out the issue. The Anpec's two-phase SC1175CSW chip is used as a voltage regulator.

The video card is equipped with 128 MB DDR memory packaged in 8 chips (4 chips on each of the sides - front and rear) within the advanced BGA packaging, with the 128-bit memory bus. The memory is produced by Samsung (K4D263238E-GC2A), offers a 2.8 ns access time, which is equivalent to approximately 350 MHz of memory operation (700 MHz), but the memory runs at its intended frequency 300 MHz (600 MHz). That is, there is a small overclocking margin for the memory.

Also note that Asus is faithful to its traditions and installs Samsung memory both on video cards built on ATI chips and on those built on NVIDIA chips. The graphic chip also runs at 500 MHz as per the specifications.

The cooling is implemented in an interesting enough way. The average-size fan is installed on a solid radiator that covers both the chip itself and memory chips positioned on the front side of the PCB. Cooling of the memory chips located on the reverse side of the PCB is not provided. Therefore, the efficiency of the whole big radiator covering the chips of the PCB front side is reduced to zero. However, it can give you aesthetic pleasure - the engraved Asus emblem covering 2 last memory chips on the front side of the PCB looks presentable enough. The efficiency of the fan blowing about the chip arouses no doubts; but for the successful overclocking the best option would be use of additional cooling that blows primarily about the memory chips of the radiator.

Tests

We remove all the decorative "niceties" and set the operating system to the maximum performance.

Disable the Vsync forcedly via the drivers both in OpenGL and in Direct3D applications. The S3TC texture compression was also disabled.

Test software:

• Future Mark 3DMark2003; v340;
• ToMMTi-Systems ShaderMark v2.0 (a synthetic DirectX 9 HLSL benchmark for pixel shaders);
• D3D RightMark 1.0.2.7. (Public Beta 1) (comprehensive DirectX 9.0 synthetic benchmark);
• Unreal Tournament 2003 (Direct3D, Hardware T&L, vertex shaders, Dot3, cube texturing. "Antalus Flyby" demo);
• Codecreatures v1.0.0 (a DirectX 8.1 application, shaders on, Hardware T&L);
• Call of Duty 1.1 (OpenGL, multi-texturing. The image quality was set to the maximum. Used was one of our own recorded demo reels "3Dnews002" (flying on a rocket in the mp_rocket level)).
• Gun Metal Benchmark 2 v1.20s (a DirectX 9.0 benchmark, Vertex Shaders 2.0, Pixel Shaders 1.1, Hardware T&L);
• X2: The Threat Demo (Direct3D, multitexturing, Dot3, running in the benchmark mode embedded in the demo version);
• Final Fantasy XI Official Benchmark 2 (a benchmark for assessing the performance in the future game Final Fantasy XI. Unfortunately, the developers haven't presented any data on the gaming engine);
• HALO: Combat Evolved 1.2 (DirectX 9.0, Vertex Shaders 1.1/1.4/2.0, Pixel Shaders 1.1/1.4/2.0, Hardware T&L, quality set to the maximum possible);
• AquaMark 3 (DirectX 9.0, Vertex Shaders 1.1/1.4/2.0, Pixel Shaders 1.1/1.4/2.0, Hardware T&L, AquaMark3 Triscore mode);
• Half-life 2 leaked beta (DirectX 9.0, Vertex Shaders 2.0, Pixel Shaders 2.0, default quality);
• Unreal II: The Awakening (Direct3D, vertex shaders, Hardware T&L, Dot3, cube texturing, quality offered by the BenchemAll suite).
• FireStarter (DirectX 8.1/DirectX 9.0, pixel and vertex shaders, particles system, dynamic lights, projected textures - read more below).

For running the tests we took 2 boards at a time built on the base of ATI Radeon 9600XT's predecessor - ATI Radeon 9600 Pro:

• GeXcube Radeon 9600 Pro
• Sapphire Radeon 9600 Pro

Why have we included 2 ATI Radeon 9600 Pro boards at a time? The thing is that GeXcube Radeon 9600 Pro turns out to be an inferior Radeon 9600 Pro. It's memory clocking is downstated by 100 MHz, although no information regarding that can be found anywhere. Such "tricks" of video board manufacturers cannot by regarded good at all. We were interested to find out how the board in question would rank among other competitors, and we believe it's our duty to inform the users of emergence of such "hacks" bearing a label of full-featured cards.

We also included a board based on the NVIDIA GeForce FX 5600 Ultra chip -

• ASUS V9560 Ultra/TVD

And of course the hero of today's review - ASUS Radeon 9600XT =).

You can assess the performance difference between NVIDIA GeForce FX 5700 Ultra and ATI Radeon 9600 XT in our review NVIDIA GeForce FX 5700 Ultra. Anyway, we found it more interesting to compare the performance difference between ATI Radeon 9600XT's predecessor - ATI Radeon 9600 Pro, as well as find out how the performance boost matches the "swollen" prices for the cards =). Besides, we also wanted to compare the results of tests for ATI boards with the tests for video cards based on NVIDIA GeForce FX 5600 Ultra which are there on the market in greater quantities and appear to be more competitive solution than NVIDIA GeForce FX 5700 Ultra.

Benchmarking Results: Synthetic benchmarks

Following the tradition, we start with the radically amended benchmarking package.

ShaderMark v2.0

As we see, this test shows the traditionally grievous trend for all video cards of NVIDIA family: ASUS V8560 Ultra/TVD on the NVIDIA GeForce FX 5600 Ultra chip loses to its direct competitor - Sapphire's ATI Radeon 9600 Pro at handling all (!) the shaders suggested by the program. And if we look at the absolute fps values produced by the benchmark, you unwittingly think of the sense in buying an NVIDIA GeForce FX 5600 Ultra in particular, and buying a video card on the base of a chip of the FX family on the whole. We have used this benchmark several times, and the failure of NVIDIA chips at it is quite evident.

But it's still premature to make a conclusion. The benchmark is a pure HLSV application which, as we have repeatedly noted in our materials, produces a tragic effect on all cards built on NVIDIA chips. In real-world applications, things might get to the better (and, running ahead, we say that they really are getting better =)) due to the optimizations introduced into the game/benchmark from the part of the manufacturer/developer for the architecture of boards built on the base of NVIDIA GeForce FX chips.

As regards the alignment of forces in ATI, there is an evident ordinance in the frequencies of graphic chips and memory. Note that in most cases ASUS Radeon 9600XT leaves Sapphire on ATI Radeon 9600 Pro much further behind than the board Sapphire ATI Radeon 9600 Pro by GeXcube does relative to Radeon 9600 Pro with reduced memory clocking. This points to the fact that ATI Radeon 9600 Pro definitely doesn't suffer from shortage of memory bandwidth, which is a sign of well-balanced performance of the chip (the memory bandwidth more or less matches the needs of the graphic processor), but the increased frequencies of graphic chip in ATI Radeon 9600XT naturally yield a higher performance boost in handling shader operations.

3DMark 2003 v340

However, coming back to today's "probationers", we can see that the shader code compiler in ASUS V9560/TVD doesn't help much. The board anyway does lag behind its direct competitor - Sapphire Radeon 9600 Pro. Because of the reduced memory clocking, the board by GeXcube shows a much lower performance level than it should for the Radeon 9600 Pro label on. We'll review the prices in what follows, but such behavior of the manufacturer is simply outrageous. On the other hand, ASUS Radeon 9600XT here is an absolute leader.

Codecreatures

While NVIDIA is not superior in the pure HLSV, in a real-world gaming application (in this case, it's a synthetic application based on a game that has still been in development for quite a long while) things might radically change. As we see, the alignment of forces is already in favor of Asus' video card on the NVIDIA GeForce FX 5600 Ultra chip. The increased frequencies of graphic chip in ASUS Radeon 9600XT have produced a favorable effect on the board's operation speed in that particular application (for which there already were prerequisites from synthetic tests).

Benchmarking Results: Real gaming applications

From synthetic applications, we are now moving on to analyzing the performance of the graphic boards in real gaming applications.

Unreal Tournament 2003

The game Unreal Tournament 2003 has traditionally been more loyal to video cards based on NVIDIA chips. So, what do we see now? We see that the NVIDIA card based on a past-generation Middle-end chip takes a lead (albeit at a small margin) over today's ATI's Middle-end representative, ASUS Radeon 9600XT. And ATI cards without any surprises ranked in ascending order of clock speeds of graphic chips and memory.

In the 1024x768 resolution with 4x antialiasing enabled, the alignment of forces is not changing, but as the resolution goes up Radeon 9600XT is coming up as a leader.

With 8x anisotropic filtering enabled, ASUS Radeon 9600XT finally overtakes at both resolutions. However, the direct competitor of ASUS V9560/TVD - Sapphire Radeon 9600 Pro - still lags well behind its rival, so we can't claim there is an on-par struggle of ATI and NVIDIA among the cards of the same class in this gaming benchmark.

Finally, the most demanding mode combining antialiasing and anisotropic filtering. We see a sure leadership of NVIDIA card over a the ATI board of the same class, and a substantial gap from ASUS Radeon 9600XT in high resolution. The greater operating frequencies of the latter make themselves felt.

Unreal II: The Awakening

Situation in Unreal 2 is traditionally reverse to what we observed in Unreal Tournament 2003. A complete fiasco of the NVIDIA board. At the absolute fps reading, ATI Radeon 9600XT shows quite good frame-rate - it seems to be a good match for those into "shooting at good fps rate" at a reasonable "price/performance" ratio =).

Call Of Duty

In our test run with "the replacement to Return to Castle Wolfenstein", the ASUS V9560 Ultra/TVD takes the last place even failing to catch up with GeXcube Radeon 9600 Pro with reduced memory clocking. Isn't that a strange result in view of the ever high results for NVIDIA boards in OpenGL tests?

The general picture with 4x antialiasing enabled does not change except that NVIDIA GeForce FX 5600 Ultra does succeed in bypassing GeXcube Radeon 9600 Pro in both resolutions.

But to overcome anisotropic filtering was too high a price ASUS V9560/TVD had to pay...

The most demanding modes combining anisotropy and antialiasing make ASUS V9560/TVD a leader in low resolutions versus its direct competitor - Sapphire Radeon 9600 Pro. But at 1280x1024 we already see a complete parity. In all the modes, ASUS Radeon 9600XT is an undisputable leader.

Basing on the results produced with Call of Duty, it's highly difficult to make any distinct conclusions regarding the choice of a one-one winner. Anyway, the results in various operation modes strongly differ, which does not allow us making any one-one conclusions.

X2: The Threat Demo

The "Ultra Shadows" technology implemented in the NVIDIA GeForce FX 5600 Ultra chip allows the ASUS V9560/TVD board competing with ASUS Radeon 9600XT (!) on par due to that stencil shades in NVIDIA chips are handled faster with this technology than in ATI chips.

Final Fantasy XI Official Benchmark 2

Basing on our experience of this benchmark, we say that all is restricted by the raw fill-rate speed. At that, ASUS Radeon 9600XT leads with pretty good results. On the other hand, NVIDIA GeForce FX 5600 Ultra shows inexcusably low result as compared to Sapphire Radeon 9600 Pro.

AquaMark 3

Gun Metal Benchmark 2

Version 2.0 shader programs have always been a challenge for NVIDIA chips. The benchmark based on the Gun Metal gaming engine is in support of that. Although it doesn't use version 2.0 pixel shaders, but only 1.1, the load applied to the accelerators because of the engine specifics is tremendous (in support of which the absolute fps values tell). As we see, ASUS Radeon 9600XT leaves its competitors well behind, and the NVIDIA board lags behind the leader substantially.

HALO: Combat Evolved

The results here are similar to those in the previous test.

Half-life 2 Leaked Beta

In shader applications, NVIDIA GeForce FX 5600 Ultra (like all the boards of the FX family) loses significantly. And the pure HLSV in the form of Half-Life 2 is a vivid confirmation of that.

"FireStarter"

It's been the first time we run tests in FireStarter game created by Ukrainian developers from GSC GameWorld. That's why we're reviewing this benchmark in detail. As per the developers themselves, the FireStarter is capable of doing the following:

Background:

• Originally based on the BSP technology
• Originally aimed at API DirectX 8-9
• The fundamental feature is in building games with plug-ins available for all the game components.
• Easy to import to XBox
• The engine is designed for maximum speed and most dynamic gameplay

• The levels are partially opened and built with the BSP technology
• The editor is user-friendly and easy to master, does not require special training.
• 98 pre-defined entity classes like doors, lifts, platforms, jump-pads, triggers, bonuses etc.
• Standard protocol for plugging in any new entities
• Use of skyboxes to simulate skies/space etc.
• Illumination around light maps with varying details
• Use of the high polygonal model in creating games

• Standard protocol for implementing the character physics
• Specific implementations of animation systems (matrix animation, morphing, skeletal animation) are embedded in plug-ins and can be selected proceeding from particular requirements/design

• Particle systems, explosions, projectile flight effects are built with primitives and can be whatever complex
• All the effects are embedded in the plug-ins
• Standard protocol for programming effects
• 11 primitive effects like spray, debris, flashes, relay etc.
• A simple mechanism for plugging in effects to levels and association with entity levels
• Programmable recurrence of effects inclusion

• Vapor, smoke, fire, geysers etc. effects based on particle systems.
• Realistic localized 3D fog with the camera placeable both outside and inside the fog
• Global 3D fog throughout the whole level without restrictions for level geometry and shaders used
• Illumination with point-wise, directed 3D light sources, illumination with skybox
• Dynamic light sources
• Projectable textures
• Detail textures
• Auras, radiance around light sources

• Standard scripting language for creating multi-pass/multi-texture shaders
• Standard procedural deformations of vertex coordinates
• Standard procedural deformations of texture coordinates
• Standard protocol for plugging in shader effects (renderers) using the most recent functionalities of video accelerators like programmable vertex/pixel pipelines
• Plug-ins to do with diffuse/specular bump-mapping effects, EMBM, anisotropic illumination etc. supporting several rendering techniques enabling operation with any hardware starting with Riva TNT2
• A smart algorithm for rendering huge volumes of static and dynamic geometry
• Procedural textures such as water, fire, plasma etc.

Interview:

To get an in-depth idea of the technical details of the game engine, we contacted Vyacheslav Klimov, the chief programmer responsible for the gaming engine of FireStarter. Vyacheslav kindly agreed to answer our questions.

3DNews: First, please introduce yourself, tell about the company you work with, as well as the projects underway.

Vyacheslav Klimov: The company is called GSC Game World. I think it doesn't need much of introduction =), and I am more involved in the technical work rather than PR. The gaming projects I have worked at in the company are Hoverace and Firestarter.

• Hoverace (1999-2002), is a futuristic racing shooter.
• Firestarter (2001-2003) is an arcade RPG-like shooter.

Not telling you what I am working at now - it's a military secret :)).

3DNews: Since we decided to include tests of FireStarter game into our series of benchmarks which we use for testing video cards in 3DNews reviews, we would like to find out more about the technical specifications of the gaming engine. What DirectX generation does the engine, the versions of pixel and vertex programs and other specific features relate to?

Vyacheslav Klimov: The graphic engine does not use DX9 functionalities. So, I'd rather relate it to the DX8 generation On the whole, the engine runs as a mix of DX8 and DX9 to date. Regarding the graphics - switching between versions 8 and 9 is done at the compilation stage. The problem is that we came up against a serious performance drop (as much as 30%) in the engine compiled for DX9. Once the cause is found, the engine will be fully migrated to DirectX 9. The shader programs are not linked to the engine. The design philosophy of the engine implies embedding complex effects in dynamic libraries.

3DNews: Good. But we wanted to dwell on some specific issues. In changing the API generation, the performance drop is observed in video cards built on both ATI chips and in NVIDIA cards, isn't it? Or the drop of performance in video cards of one of the manufacturers is more dramatic?

Vyacheslav Klimov: I think it's about NVIDIA's cards. We haven't yet experimented with ATI cards. We were involved in transition to the 9th version immediately after the first drivers supporting DX9 were released. That time, it was no way spending much time experimenting, so we had to discontinue the work. Now more time is left - in the 3rd patch the engine might run solely under DX9. By the way, there will be fixed the "blue screen" bug that appears with recent NVIDIA drivers. Probably the problem was in the raw drivers, or maybe some fundamental changes have been introduced into DX9...

3DNews: You said "shader programs are not linked to the engine". Does that mean the engine supports all possible versions of pixel programs (1.1, 1.4 and 2.0) and switching between them is done "on the fly"? Does it also mean that ATI and NVIDIA cards can handle the same scene using different versions of pixel and shader programs?

Vyacheslav Klimov: The graphic engine operates geometry separated into threads - for coordinates, normal, color, texture coordinates, guides etc. In the material declaration you can specify execution of rendering not by the engine, but by a dynamic-link library (dll). In this case, the library accepts all the necessary data flows via a standard protocol and is thus responsible for the rendering. For pixel programs, the library is solely responsible for selection of the right version, and upon initialization it should choose a rendering algorithm that best suits the specific configuration. Such an arrangement allows building up graphic effects without any modifications to the engine. Examples of using the technology:

- bump-mapping;
- anisotropic illumination;
- procedure-driven textures;
- procedure-driven deformation of vertices/normals;
- video textures;
- rendering into the texture.

I must admit this mechanism is enabled in the game merely in part - for instance, we had to give up active use of bump-mapping and some other features. We might prepare a separate benchmark built on the FS engine that includes gaming and synthetic tests which use all the features of most recent video cards and DX9, - the engine has it all it takes. The problem is only with time and resources.

3DNews: Recently, they talk a lot about the opposition between NVIDIA and ATI on the market of graphic chips. Many ideas are spoken out regarding that etc. As a DX 8.1 and DX 9.0 developer, do you believe video cards built on ATI chips are better suited for that or is it the notorious code optimization for the architecture of different chips that determines it? Have any specific optimizations been made to the engine of FireStarter for ATI or NVIDIA video cards?

Vyacheslav Klimov: For now, I'd better refrain from statements regarding the opposition. We didn't do any optimizations for each architecture. During the operation, the engine was debugged in parallel on both ATI and NVIDIA cards. We were making sure everything was running properly without any essential performance difference. In my view, the problem is not much about the architectural differences but more about the differences in drivers, sometimes found in the cards of the same manufacturer. For example, it was a big surprise for us to find that a completely written and debugged game simply wouldn't start up with NVIDIA's new drivers. Some ATI cards showed jerking with older versions of drivers. Unfortunately, such things take a lot of time which could have been used more opportunely.

3DNews: OK. Your views are clear =). And how are things with the float-point precision for NVIDIA cards? It's all clear with ATI cards - the minimum admissible 24 bits. What could you tell about NVIDIA cards in FireStarter which can switch the floating-point precisions between 32 bit and 16 bit?

Vyacheslav Klimov: I've never gone so deep in that. No comments.

3DNews: The demo version of the game offers a quite impressive benchmark which gives testers good opportunities for testing video cards. Whose idea was to include such feature into the demo version and do you plan to further promote this direction?

Vyacheslav Klimov: That was a consolidated idea. In our view, it's a good tradition to equip games with benchmarks. We'll be following it further on. As I already said, we might prepare a separate benchmark based on the engine - it is a very interesting topic. By the way, in the demo version and the release we included an option requesting to send the report file to us into the benchmark . So far, many gamers have responded (and responses still keep coming is) - now we can gather a precise enough statistics on PC configurations and speeds.

3DNews: That's a really good idea, but will there be options of writing the user's own demos in the game in order to run tests using them?

Vyacheslav Klimov: Yes, we plan to include such option into one of the patches.

3DNews: Finishing the interview about FireStarter, I can't help asking about one more project of yours: S.T.A.L.K.E.R. =). What can you tell about the engine of this game? What is the fundamental difference from FireStarter?

Vyacheslav Klimov: A separate team is engaged in Stalker. I think you'd better ask about the engine the programmers directly :) I think the main difference is that the engine of Stalker was originally targeted at "tomorrow" of graphic technologies, while FireStarter is aimed at the maximum wide range of today's hardware.

3DNews: In conclusion, I think our readers are anxious to know when the game will see the light in its final make.

Vyacheslav Klimov: Regarding S.T.A.L.K.E.R - no comments. But Firestarter has been on sale for over a month by now, with two patches fixing the most critical bugs have been released. Currently, we are working at the third patch - it will fix the issue of compatibility to some drivers, and some gamers' suggestions will be implemented. The patch is to be released in March 2004.

3DNews: Well, I'd better stop my tortures at that =). Thank you very much for the interview.

Vyacheslav Klimov: Thank you for the questions. Good luck!

We now start running the tests of video cards in question in this very interesting game. Below, we are bringing in a screenshot with settings for the benchmark mode of the game:

We traditionally included maximum possible image quality (which is enabled through the game menu), we also ran our tests with purely 32-bit color.

As was already stated in the interview, no specific optimizations for NVIDIA cards were done, and thus the cards from ATI and NVIDIA are originally in more or less testing conditions. In the "Fast Benchmark" mode (camera flights around one of the game levels) we can see an essential lag of NVIDIA cards based on the NVIDIA GeForce FX 5600 Ultra chip from its competitor - Sapphire Radeon 9600 Pro. The results for ASUS Radeon 9600XT are excellent.

In the "Game BenchMark" mode (a demo recording of the gaming process) the picture doesn't change. These results (in the "Game BenchMark" mode) are important primarily for gamers and potential buyers of the game because they display true figure which the gamer should "feel" while playing. Also note the high absolute fps values in the game. Tests were run with a completely closed card, which together with the features of the engine did not apply much load on accelerators. As a result, it gives more than acceptable fps in any resolution for any of the cards tested. Nevertheless, NVIDIA's lag is substantial.

The AA and AF modes (including AA+AF) do not change the general alignment of forces. By the way, even in the maximum quality mode (AA6x + Aniso16x for ATI chips) all the ATI cards give an acceptable frame-rate, which can't be said about the NVIDIA card.

Final Words

Summing it up, we can only praise Asus for a successful debut with ATI chips. We already did that in the review on ASUS Radeon 9800XT. Now we see that in the Middle-End segment as well ASUS has succeeded from the first and made a really good and quality board based on the ATI chip. The decorative cooling system onboard was a bit annoying - radiator on the reverse side of the PCB would be of good use. Among the advantages of the board is the smart exterior design and an excellent package bundle. As usual, at first we can observe the prices for Asus produce a bit overstated. But with time the prices might go down to reasonable levels. Anyway, it's up to the buyer to decide if ASUS produce is worth the money.

Read more on this topic: