Chaintech Geforce FX 5900XT SA5900X Review

VGA Reviews by jmke @ 2004-06-15

Chaintech has a very competitively priced video card in its products range which offers a great price/performance ratio. Just how great? We ran it through our benchmark mill so we could let you know!

In the box

Chaintech has been working hard on their imago, with a goal set to please the hardware enthusiast. It comes to no surprise that they included nVidia’s best price/performance ratio GPU in their video card line-up, the FX 5900XT.

Madshrimps (c)


The aptly named SA5900X is their entry in the budget video card segment of the market, since the GPU and PCB layout is pretty much the same on all 3rd party boards, it mostly comes down to features, cooling, overclocking potential and included goodies, when trying to find a winner.

Madshrimps (c) Madshrimps (c) Madshrimps (c)


The Chaintech takes a great start in the goodies department, you will find all the cables included to hook it up to your D-sub monitor or DVI TFT, further more you’ll find 2 different cables for TV-out and a Y-splitter for giving the card its needed juice.

Madshrimps (c) Madshrimps (c)


The card is an exact copy of the nVidia reference card, except for the cooling, Chaintech’s solution is very good looking and if you have a side windows in your case it will surely stand out! The little cooler pushed quite a bit of air, but unfortunately makes a lot of noise too.

www.HardwareZoom.com (c)


The memory chips used are from Hynix and according this table from OC.com.tw translate into 2.8ns or 350Mhz.

Madshrimps (c)


Which brings us to the card's specifications:

Clocks speeds Standard (2d mode):
  • Core clock: 300 Mhz
  • Memory clock: 350 Mhz (700 Mhz)

    Clock speeds Performance (3d mode):
  • Core clock: 390 Mhz
  • Memory clock: 350 Mhz (700 Mhz)

    Features:

  • 256-bit chip technology
  • 0.13 Micron Process
  • 256-bit DDr Memory Bus
  • AGP Bus 2*/4*/8*
  • 128 Mb Memory
  • Nvidia© Ultrashadow™ technology
  • Nvidia© CineFX™ 2.0 Engine
  • Nvidia© Intellisample technology
  • Nvidia© Nview™ multi-display technology
  • Nvidia© Digital Vibrance Control 3.0
  • Supports Microsoft® DirectX® 9.0
  • 128-bit, studio-precision colour
  • Vertex Shader 2.0+
  • Pixel Shader 2.0+

    Dual 400Mhz RAMDACs allow you to hook up 2 monitors and the card also supports high resolution HDTV decoding.

    On the software side of things you get 3 applications, 1 full game (MDK2) and 5 game demos, a more detailed line-up on the next page ->
  • Included Software

    Madshrimps (c)


    Onto the test setup ->

    Test Setup and Quality Setting

    Test Setup

    Our test system consists of a P4 system which is handicapped due to its CPU being a 2.0 “A”, this one is limited to 400Mhz Quad Bus memory speed, and even when overclocked to 3Ghz it still is being bottlenecked by this, although performance comes close to an out-of-the-box P4 3Ghz “C” (800Mhz).

    JMke's Setup
    CPU Intel P4 2ghz "A" @ 3 Ghz
    Mainboard Asus P4C800
    Cooling * Zalman CNPS-7000Cu
    Memory 2 * Corsair PC3200 Pro 512Mb
    Video * Chaintech nVidia Geforce FX 5900XT
    * PNY nVidia Geforce 4 Ti4600


    When the GF4 Ti4600 was launched it was the top performer, for a short period, until the Radeon 9700 Pro appeared on the scene. Nevertheless, today, it can still offer the enthusiast an enjoyable gaming experience.

    “Low” end FX 5900 series vs “High” end GF4 series


    Windows XP SP1 and the latest nVidia drivers (56.72) from the official website were installed on the tests system. The video card’s driver settings allows you to change Full Scene Antialiasing (FSAA), Anisotropic Filtering (AF) and overall mipmap quality settings. The slider gives you a choice between: Quality, Performance and High performance.

    I disabled FSAA & AF and loaded up a game of MotoGP2 (1280x1024 @ max detail) to see how much visual and performance difference it would be when choosing the different image quality settings.

    Average FPS recorded using FRAPS 2.0 during a replay of a spin around the “Le Mans” track gave the following results:

    Quality: 43.38fps
    Performance: 47.07fps
    High Performance: 48.89fps

    Well by moving a slider you can get up to 5 FPS extra, how much difference is there visually?

    Quality
    Performance
    High Performance


    The difference between Performance and Quality lies in very small details, High Performance versus the other shows a bit more difference, but all in all nothing earth shattering and certainly not what you could call “high image quality degradation”

    All the benchmarks were run with the “Image Quality” slider set to Performance as it offers a very good balance between speed and visually stunning graphics.

    On with the show ->

    Benchmarks UT/Q3A/FarCry/More

    Overclocking

    Out of the box the Chaintech could be clocked up to 450/750 (core/mem) without breaking a sweat. By adding extra cooling and exchanging thermal paste between Cooler and GPU even higher could be achieved

    Madshrimps (c)


    The added cooling power put the maximum stable core speed between 460-470Mhz, an extra 80mm fan blowing over the RAM heatsinks made them run stable at 800Mhz, quite a performer this Chaintech card.

    For some of the tests below I set the cards to their “out-of-the-box” OC speed of 450/750, while the PNY was clocked from 300/660 to 330/720 to see if the statement “overclocking your video card can postpone the need for an upgrade” is true.

    Madshrimps (c)


    It seems the 5900XT core offers quite a bit of headroom, while the memory is lagging behind the older Ti4600’s OC capability.


    UT2003 / UT2004

    First up is a popular Direct3D game from Epic, their latest creation uses an enhanced version of the UT2003 engine, aptly named “UT2004”. The Botmatch benchmark stresses the system more then it does the video card, this gives results which are close to each other, depending on the map.

    Madshrimps (c)


    At a resolution of 1024x768 the difference between the 2 cards is negligible except for the DM Map, increasing the resolution to 1280x1024 makes it clear that the 5900XT is the superior card, almost maintaining the same frame rate as it did with at a lower resolution.

    The UT2003 benchmark includes a flyby mode which really pushes each video card to its limits

    Madshrimps (c)


    Both cards are able to offer fluent gameplay but the 5900XT sticks out when a higher resolution is used.


    Chameleon Mark

    This is a pretty old benchmark, but it can quickly show you the performance of the older generation of shaders:

    Madshrimps (c)


    ChameleonMark is a performance tool for measuring pixel shader performance for a variety of shaders. It is based upon the popular "Chameleon" demo released by NVIDIA with the introduction of the GeForce3 line of GPUs.

    Madshrimps (c)


    Seeing as this is a BenchMark which could be used with GF3 generation of cards it comes to no surprise that both cards can play it fluently at 60+fps @ 1024x768. The 5900XT outperforms the GF4.. and not just “a bit”. 100+% at 1024x768 and even more at the higher resolutions!

    Quake III Arena

    An old classic, the Q3A engine is used in a lot of popular games now, although these engines are heavily modified to allow for greater detail. If these cards can run Q3A at ridiculously high frame rates then they stand a fair chance of running the other Q3A based games quite decently.

    Madshrimps (c)


  • Check! The 3Ghz system is holding back both cards from performing at their optimal speeds it seems :)


    3DMark2001 Se & 3DMark03

    When 3DMark2001 was fairly new companies used it to base their games playability on it, a PC which scored 2500 3Dmarks could run Max Payne fluently at higher detail. Today this benchmark offers the end-users a quick check to see if their system can handle DirectX 8 games without problem.

    Madshrimps (c)


    Although more system-stressing DirectX 8 games have hit the market since Max Payne, it seems both cards will give you a good shot at running the latest DX8 games at full or slightly reduced detail when keeping the resolution at 1024x768.

    The newer brother, 3DMark03, set out with the same goals as the 2001 version, unfortunately it has been used and abused that it no longer offers comparable results when used with cards from different manufactures (ATI/nVidia). But it does still shine a good light on the capabilities of a video card when used with a game which supports or even requires DX9 features to run.

    Madshrimps (c)


    As the GF4 is lacking in Pixel Shader 2.0 support, we see a rather large gap between the 2 cards.

    Far Cry

    A next gen game which can be used as an example is the extremely good looking Far Cry

    Madshrimps (c)


    While the 5900XT can give you fluent gameplay at “High” Quality settings, you will have to reduce details to get the same performance from the Ti4600.

    Speaking of details, how do these cards perform when we start moving the FSAA and AF sliders around? Find out on the next page ->
  • Benchmarks FSAA/AF UT/Q3A

    To make it easier for me to type and you to read, from this point forward: Full Scene Antialiasing = FSAA and Anisotropic Filtering = AF.

    First up is the UT2003 Benchmark, changing the quality settings to 2*FSAA/4*AF removes quite a bit of edges while making the surroundings look sharper.

    Madshrimps (c)


    I think we found the real power of 5900XT right here, when enabling FSAA and AF the performance surely drops, but the average FPS remains quite high and the game remains playable! The Ti4600 is completely crushed; at 1280x1024 any chance of fluent gameplay is thrown out of the window.

    This is pushing the limits of any card, even more when we are using a “budget” card, FSAA up to 4* and AF up to 8*.

    Madshrimps (c)


    The 5900XT’s performance is halved, and playing games at this setting is quite impossible as the lower dips in FPS will make you slide of ledges and plummet to your death due to the stut-t-tering. The GF4 makes playing UT2003 at these settings a slide-show: good looking, but nauseating.

    The visual difference on the FX 5900XT between vanilla and 2*FSAA / 4*AF can be observed in this screenshot:

    Madshrimps (c)


    Will the results in Q3A remain ridiculously high when we crank up the detail settings?

    Madshrimps (c)


    Both cards give playable FPS, although I didn’t see the 5900XT break a sweat.. at all!

    Madshrimps (c)


    The Ti4600 is out of the game, while the impressive XT is able to keep up with the pace. But don’t go thinking you can run those latest OpenGL games at these settings, although an older game like this will gladly “play along”, running 4*FSAA on this “budget” card is not yet an option.

    Visually the difference is less remarkable in this particular scene then with the UT2003 shot, it does give the 'ol Q3A a nice polished look:

    Madshrimps (c)


    With all the testing done, it’s time for a conclusion ->

    Conclusion

    Conclusion:

    Madshrimps (c)


    Having played more then a dozen games at different resolutions and quality settings I’m quite confident in saying that this video card can offer you a more then acceptable gaming experience with the latest games out there.

    Far Cry at its "high" settings was quite enjoyable, and older games ran fluently with graphic details cranked to the maximum.

    If you have an older video card and you don’t want to spend $300+ on a new one, then you’ll find the 5900XT to be a very appealing solution, as it comes with good performance at a great price. Chaintech’s edition of this budget power card is priced very competitively around ~$170 if you shop around a bit.

    The included extra connectors are a nice addition, performance at default and overclocked speeds is more then acceptable and it simply looks great. The only possible downside would be the rather noise cooler which sits on top of the GPU. Depending on your current setup your experience will differ, since it isn’t louder then your average 80mm CPU fan.

    PRO
    Good performance with and without FSAA/AF enabled in all games tested
    Very complete package
    Competitively priced


    CON
    Noisy GPU cooler


    I would like to thank Frédérique @ Chaintech.nl for helping us out with the review sample.

    Madshrimps (c)


    Questions/Comments: Forum Thread
      翻译: