ASUS Radeon 9600XT/TVD

Everyone knows that the 9600 XT is one of the top mid-level graphics cards currently available, but what makes Asus’s offering different? How does it perform compared to other mid-level cards? Read this review to find out.

Introduction

It was about that time… games were coming out fast, and with each new game, my old Ti4400 seemed to be dropping a few FPS, and we hadn’t even really seen the introduction of any fully DirectX 9 compliant games yet. So, after seeing the “XT” launch featuring the Asus cards and then reading later reviews on both of the new XT cards, I decided I would go for a 9600XT. When I saw the ASUS 9600XT at my favorite online retailer at just about my “spending limit,” I did a little more research and then decided to order it.. This particular card comes at the stock speeds of 500MHz core and 600MHz RAM, but that’s where the similarity to the reference ATI design ends. Built on a nice orange PCB and adorned with an impressive shiny aluminium heatsink, this card definitely stands out and is definitely not a vanilla ATI graphics card

Now I’d like to reiterate that I personally purchased this card, so I will therefore be giving it a very thorough set of tests to find out if I made a wise decision buying it. I’m looking to find out just how fast this card is in comparison to my Ti4400 and FX5600 non-ultra. I am expecting to find that the Ti4400 outperforms both of the DX9-based cards in DX8 games and benchmarks when no AA (anti-aliasing) is enabled, but I expect it to fall behind both DX9 cards once AA is enabled (this is why I bought the card and for playing future and present day DX9 based games). I also expect the DX9-based cards – especially the 9600XT – to perform better than the Ti DX8 based card in more modern titles and tests.

When I received the card, I was not shocked at the size of the box – actually it is about the same size as the Ti4400 box. What really surprised me was when I opened the nice packaging and got the white card box out and opened that. What did I find? Well, what I found was a box that, although similar in size to most graphics cards packages (which generally only have the card and few sparse accessories and take up only 1/4 of the box), was full and needed every square inch to fit all the goodies in.

The packaging is very tasteful. The front lets you know the features you’ll be getting along with info telling you that you get Half Life 2 with the purchase of this card, while the back shows you what the card looks like and goes into a little detail on the different technology, software and hardware you get with it.

Features

• 500 Mhz VPU Core
• 600MHz (300MHz DDR) Memory
• 400 Mhz RAMdac
• Microsoft® DirectX 9 and OpenGL 2.0 support
• Shader 2.0
• DirectX 9 compliant
• AGP 8X/4X/2X support
• S-Video/RCA in/out
• DVI and VGA connectors
• Max Resolution 2048 x1536 32bit

ASUS Radeon9600XT/TVD equips with newest ASUS-specific award-winning features and ATI®’s Radeon 9600XT Visual Processing Unit(VPU), which owns both ultimate excellent cinematic resolution and fierce graphics computing power. Incorporating various state-of-the-art technologies such as ATI’s SmartShader 2.1, SmoothVision 2.1, Hyper Z III+, VideoShader, and unique ASUS innovations such as “ASUS Smart Cooling” ,”ASUS GameFace”, “ASUS VideoSecurity”, and “ASUS Digital VCR”make ASUS Radeon 9600XT /TVD a complete solution for the 3D market.

• SMARTSHADER 2 expands advanced abilities of vertex shader 2.0 and pixel shader 2.0
• SMOOTHVISION 2.1 generates the sharpest and clearest images by up to 6x Anti-Aliasing and 16x anisotropic filtering
• FULLSTREAM removes blocky artifacts from streaming Internet video
• VIDEOSHADER integrates shader features to provide unprecedented support for digital and high definition video.
• HYDRAVISION provides multi-monitoring management
• 128-bit, 64-bit & 32-bit per pixel floating point colour formats

You can see the real differences between the 9600XT and the 9800XT are in the clock speeds but more importantly the 9800XT has 8 pixel pipelines to the 9600XT’s 4. The difference between the 9600XT and 9600Pro is strictly in the clock speeds. Also, there is no need for an external power connection to the 9600XT/Pro; the AGP bus provides all the power the card needs.

For more detailed specs, please check out the ASUS 9600XT/TVD web page.

Package Contents

Once again an excellent package from Asus — three full games with a value of around $70-$80; software to use the card to its fullest potential, from editing and correcting your photos to creating and editing your own movies and burning them to DVD; a very nice orange plastic CD case that can hold up to twelve CDs (and the orange ties into the PCB colour theme); break-out VIVO cable of a good length (56″ or just over 4.5 feet); and last but not least, a DVI-to-VGA converter.

Construction and First Impressions

I had seen a few photos on the Asus web site and around the web, but they did not prepare me for the card in person. In looking at original pics of the 9800XT and the 9600XT, I thought that the cooling solution for the 9800XT looked much better, but once I got my 9600XT/TVD in my hands, I found that in person it is a very good looking design that suits the card. The cards construction is top notch as are all the components on it. I did however find it strange that there are no heatsinks on the RAM on the back of the card, but to my surprise after feeling it while running benchmarks, it felt very cool so there was no real need for them.

Another determining factor for my purchase of the XT was its VIVO capability. Below you can see the ATI Rage Theater chip — this is the same chip as you get on the AIW cards, but there is no TV functionality incorporated into this card 🙁 . The memory is 3.3ns Samsung. Doing the math, we see that it’s ….1000 / 3.3 = 303, which is equivalent to 606MHz DDR, so technically it is running right at its “correct” speed of 600MHz. Unfortunately, this doesn’t give it much headroom for overclocking.

Hardware Installation

As with all cards I have tried to install into my SFF (small form factor) system, it isn’t so easy. This is no fault of the card manufacturer, but rather just a limitation due to the very small confines of the SFF system and all the hardware crammed into it. This is a “one slot” cooling solution, but you still (in my humble opinion) can’t put a PCI card in the adjoining PCI slot as it would starve the VPU’s fan of air. Installing into my full tower PC-Toys CaseMAX 610 case was a breeze; none of the problems that faced me in the SFF case were present, and with the added room in the case, temps went down by a couple of degrees. You still can’t really put a PCI card in the first PCI slot, though.

You can see how nice the 9600XT looks installed in the SFF system (a Biostar iDEQ 200N with added windows), and it matches the red PCB of my MSI K7N2 Delta-L in my full tower case. It certainly looks better than a boring green “standard” PCB as evidenced by my modem below it…and it looks even better when I turn on my UV cold cathodes. 🙂

VPU Heatsink

Definitely not your reference ATI cooler… Asus has most assuredly made its card stand out from the rest with this change. Made of aluminium, it covers the entire VPU core and the BGA memory. The fan is controllable through the included Asus Smart Doctor software and has a max RPM of about 6,600, which produces a slightly annoying “whine.” I have found the VPU runs very cool and have therefore set the fan to a manual-controlled setting of 70%. You would expect that this corresponds to a RPM of 4620 (6600 * 0.7 = 4620) but strangely according to ASUS this means about 6100 RPM. Conveniently, this also takes care of the “whining” problem. Since I am using this card in my Biostar iDEQ 200N SFF system, which runs very quietly, noise is a factor for me (something I’ve only recently become annoyed by, thanks to the iDEQ 200N). No matter what, it’s still nice to have the ability to either manually or automatically adjust your video card’s fan speed.

Size DOES Matter

Size does matter, but in this case smaller is better. Below is a photo showing the size difference between the PNY Ti4400 and the Asus 9600XT. The 9600XT is over 1 1/2″ shorter. Also you will note that there’s no external power supply connection required for the 9600XT, a welcome fact since I’ll be using the card a lot in my SFF system, which only has a 200W PSU.

Software Installation

Installation of the ATI Catalyst drivers was a snap. I inserted the provided CD and installed the VGA, control panel and the WDM drivers and rebooted as prompted. That was about it. I then decided that it would be best to download the newest catalyst 3.9 drivers as the CD had only come with version 3.7, and just before I started this review, I downloaded the 4.10 drivers and used them for all benchmarking and tests.

You can click on the image of the ATI control panel below to see an animated GIF that steps you through all the tabs. If you’ve never seen the ATI panel, then it might be of interest. After being accustomed to an nVidia panel, this is a bit overwhelming at first, but you soon grow accustomed to it. This is where all adjustments are made for your video card: color, anti-aliasing, anistropic filtering, etc.

The Asus Smart Doctor 2 Utility was also included in the package, and the install went without a hitch. In no time at all, I was viewing my VPU temps, RAM temps, fan RPM and board voltages. I also had the ability to set warning levels for voltage, temperature and overheat shutdown features. Below, you can see an animated GIF of the Smart Doctor panel and setup panel.

Benchmarking

The test suite I’m going to be using consists of the following titles:

• SYNTHETIC Benchmarks – 3DMark2001 SE, Aquamark3, Codecreatures and ChameleonMark.
• GAME Benchmarks – Halo-Combat Evolved, GunMetal Benchmark 2, X2: The Threat – Rolling Demo, Unreal Tournament 2003, Call of Duty and Quake 3 Arena.

Due to all of the controversy surrounding the 3DMark2003 benchmark, this site has decided that for now we will drop it from our benchmark suite.

Test System:

• AMD XP2500+ Barton 333FSB ( 11 x 166 )
• MSI K7N2 Delta-L – 400FSB ready
• ASUS 9600XT/TVD -128MB – 500Mhz core / 600Mhz RAM
• PNY Ti4400 -128MB – 275Mhz core / 550Mhz RAM
• XFX FX5600 – 128MB – 325Mhz core / 550Mhz RAM
• Kingston HyperX PC3200/400 DDR – 2 x 256MB (512MB) run synchronously w/ CPU
• Memory Timings – 5,2,2,2
• Maxtor DiamondMax Plus9 80GB – 133 ATA – 7200 RPM
• Antec TrueBlue 480W
• PC-Toys CaseMAX 610 Aluminium case
• nForce Drivers – 3.13 WHQL
• ATI Catalyst – 4.10 WHQL
• nVidia Detonator – 53.03 WHQL
• Win XP Pro – SP1and most recent updates
• DirectX – 9.0b

I did not perform a clean install of the OS as that is not what most people would do, and with a total of about 25GB of programs and games installed, I wasn’t about to re-install all of them and re-configure them while trying to keep save games, configurations, etc. This way you’ll get real world performance results.

I used the new Benchmark batch program we have started using here at Bjorn3D.com called Bench’emAll, a very useful program that can be used to execute batch runs of numerous favorite game and benchmark utilities, including 3DMark2001 SE, Codecreatures, ChameleonMark, Unreal Tournament 2003, Halo, Quake 3 Arena, and Call of Duty.

AA (anti-aliasing) and AF (anistropic filtering) were set manually in all card’s control panels. I threw in the FX5600 non-ultra purely as a comparison, since it is a mid-low end but DirectX 9 based card. I was, however, quite shocked at the FX5600 results. I will say that out of the 5600 non-ultra’s I have seen and read about, this one puts up some pretty decent scores and frame rates.

Also please note that unfortunately due to circumstances beyond my control, I was unable to complete all the benchmarks with the Biostar SFF system and so decided to re-run all tests in my full tower setup and discard the couple I had gotten from the Biostar runs. I would however like to point out that the Biostar iDEQ 200N performed on par with its ATX form factor brethren. I included the photos of the 9600XT installed in the Biostar as this is where it will finally stay as a gaming-only rig and because this card suits a SFF system (low temps, no external power required, and good performance).

I thought this might be something some readers might like — a quick listing of all the 9600XT’s benchmark results. Full comparisons are listed below in their own individual tables, showing results from all three cards. So, if you’re in a rush and can’t go through all the benchmark tables, here’s a “summary” for a quick reference.

The 3DMark2001 SE benchmark is an older test but one that everyone is familiar with. I’m sure everyone has at one time or another benched one of their cards with it and is therefore familiar with the scores. The XT didn’t do as well as I expected compared to my Ti4400 when there was no AA/AF, but once AA/AF was turned on, the XT really shines and shows its newer technology. This is why I bought the card; I knew that the actual performance in most current games wouldn’t be much different without AA or AF enabled but quite a bit different once these were turned on. The FX5600 non-ultra loses by about 2,000 points at each resolution, which is still very good for a card that costs less than 2/3’s the price of the 9600XT.

nVidia’s drivers have definitely improved for the better — image quality-wise that is — Without too much of a loss of speed. I can see this in my Ti4400, which I haven’t used for some months now. I last used it with the 45.23 Detonator drivers when reviewing the Biostar iDEQ 200N SFF System. My 3DMark2001 SE score has dropped by around 600 points or so, and I noticed a few little differences in image rendering during the tests.

AquaMark 3 is one of the latest benchmarks to arrive on the scene. It uses Massive Development’s Aquanox game engine (used to power Aquanox and Aquanox 2), so we’re actually seeing results using a real life game engine instead of a synthetic test. We should be able to therefore take these results as a bit more reliable when it comes to actual game performance of these cards. There is a lot of eye candy in this test, and each scene tests a particular feature that you can clearly see. Here the XT really shows its stuff, taking about a 40% lead over both nVidia cards.

The Codecreatures benchmark is a very stressful DirectX 8 benchmark program and generally produces some pretty low figures even with high end cards. I’m testing mid-range cards, so I didn’t expect anything super from them, but I also didn’t expect the old Ti4400 to be on par and even beat the 9600XT. The 9600XT redeems itself, though, when AA and AF are enabled.

Although this is a benchmark designed by nVidia to bench its cards and show what they can do, it also works for ATI-based cards and shows some good comparison. Built using DirectX 8 technology, it can really stress a card more than you might think. I don’t know what happened here, but the XT certainly has “smoked” the Ti and FX at all resolutions and settings; this could have something to do with this being an nVidia-specific test, and the card maybe does not conform to it somehow…but I cannot say for sure.

Halo, the game of 2003 that brought even the highest end PC systems to their knees, definitely earned itself a reputation for being very hard on hardware. It was originally made for the Xbox, which by the way if you didn’t already know is a PIII 733MHz, 128MB SDRAM interface shared with on-board GeForce3 tech graphics (this fact is what really had people stumped when they tried to run it on their XP2500+ or 2.4GHz CPUs and FX5600 or 9600 GPUs and found themselves staring at a basic slide show at times). They did some tweaks (added shaders) to the game to improve the visuals. However, no one expected this big of a performance hit on much more powerful systems than an Xbox. Definitely a good test for a graphics card benchmark suite.

GunMetal Benchmark 2 has tons of eye candy to really stress your video card. You can see by the results at 1024 x 768 that this is very stressful indeed and wouldn’t be very playable at any resolution higher than 1024 x 768 on the 9600XT.

Recently released, this is “A first person space experience designed for today’s game players with sound and graphics to match.” This game has some really nice graphics that do stress the 9600XT. It’s a recent game using modern advanced graphics techniques, so it should really differentiate between the DX9 and DX8 cards, and clearly it does. I had problems when running the benchmark using the FX5600 at the 1280 x 960 resolution; when it was finished running the benchmark, I was greeted with a black screen that I could not change unless I did a shutdown and restart. The computer was not locking up as I could still see and move the mouse on top of the black screen. Using 1280 x 1024, though, I had no problems, so I’m not quite sure what happened.

The UT2K3 benchmarks were run using two fly-by’s and two bot matches. I added the fly-by scores and the bot match scores together and gave an average. The maps that the fly-bys and bot matches were run on were DM-Antalus and DM-Asbestos.

UT2K3 has playable frame rates on all three cards without any AA or AF turned on, but once you turn on the AA, this is where you really start to see the technological improvements the 9600XT has over the other two cards. In my experience playing UT2K3, I found the gameplay to be very smooth. I play at 1280 x 960 without any AA/AF since it seems to help me sight opponents faster. At this resolution and at my user settings, I have an FPS that never drops below 60, which I find more than acceptable.

Call of Duty uses an updated Quake engine that has been modified to enhance the graphics and overall experience when playing the game. This makes it easy to use as a good OpenGL benchmark. I ran the benchmark using the provided demobta00.dm_1 on MP_Caretan map.

As you can see by the results, Call of Duty has done some serious work to the Quake engine, and it now stresses the 9600XT much more compared to the original Quake engine. The 9600XT is dishing out the wins here, but just not by the margin you would have expected from a card that is two years ahead of the nVidia Ti card. Although the engine has had a serious rework and frame rates have dropped, we still see the Ti4400 nipping at the 9600XT’s heels by only a few frames, showing nVidia’s true power in OpenGL apps.

Quake 3 Arena, the original OpenGL game, was responsible for the Quake / Quack fiasco that didn’t go too well by ATI. This has all changed, and it looks like everyone is playing by the books now, as far as Quake is concerned. ATI’s offering looks pretty good. nVidia of course is known for its cards stellar performance in OpenGL-based games and especially in older titles. It definetly shows here; the aged Ti4400 (275/550) only falls short by a few mere percentage points. The core/memory speeds are running much slower when compared to the 9600XT (500/600), and the 9600XT clearly takes the lead here as the difference shows up when AA and AF are enabled.

Overclocking

Well, considering that this card does not have the reference design heatsink, I thought that it might get a bit better of an overclock than the reference ATI designs. I did manage to get it up to a clock of 320MHz core and 650MHz RAM; not much of an OC, but it still netted some good performance gains. This 9600XT does not support ATI’s overclock utility; instead you will have to use Asus’s Smart Doctor utility to achieve your OC and adjust anything else to do with the card in this respect. If you’re wondering if I tried any higher, the answer is yes. I took the core to 525, but the system reset while running 3DMark2001SE, and I decided to go back to 520, which was stable. If you want me to try for any higher, send me another card, and I’ll surely try for a 600MHz core. 🙂 I’m quite happy running it at stock speeds right now as I have no games that give it trouble.

Image Quality / Anistropic Filtering Test

To tell you the truth, I am not one who goes for all the image quality comparisons using a “magnifying glass” to discern the differences between two cards. I am, however, curious as to how each card manufacturer’s cards implement stuff like AF (Anistropic Filtering), so I used the Texture FilterTestApp Version 1.2 by Xmasto see just how the two brands of cards do this.

ASUS 9600XT Using the Tinted setting,  and object set to tunnel with 200 faces

ASUS 9600XT Using the Tinted setting,  and object set to tunnel with 3 faces

XFX FX5600 Using the Tinted setting,   and object set to tunnel with 200 faces

XFX FX5600 Using the Tinted setting,  and object set to tunnel with 3 faces

VIVO Functionality

I finally got around to testing the VIVO functionality of this card, and I would have to say I’m quite pleased with the results. The provided programs are quite sufficient for the average user (and that’s what I am when it comes to video editing). There are enough features and settings using the provided ASUS Digital VCR program for importing movies onto your computer. The video/TV out feature worked fine when I tried it using an S-video cable to a very nice Samsung flat screen 26″ TV. I did, however, experience that phenomenon that a lot of ATI users say is reserved for nVidia only based cards — the black border around the frame. No setting I could find in the ATI control panel helped me recover any screen real estate unfortunately.

Conclusion

Well color me pleased as punch! I’m definitely impressed with the performance and quality of my new card. The switch to an ATI-based card was a hard one for me in that I have only ever owned nVidia-based solutions and really like the “control panel” interface and controls. I quickly learned my way around the ATI control panel, but I still prefer nVidia’s panel. I guess old habits die hard.

When running the card without AA or AF on or at lower resolutions (1024 x 768 or below), you don’t get to see what this card can really do or any real performance difference between it and the Ti4400 (in any DirectX 8 or lower games/benchmarks), but enable AA and AF and go above 1024 x 768 resolution, and then you’ll start to see it pull away from the Ti4400; or play a game using DirectX 9 functions and see the beauty.

I’m not one to go after super duper visual quality for the games I play; I mean what’s the point when you’re flying around at mach-1 in games like UT2K3 and Q3A. You don’t have time to notice all those nice textures, and if you do, it probably means you’re dead. This being said, I have enjoyed being able to apply AA to a few of the slower paced driving games I play and still get acceptable frame rates. I’ve finally gotten Colin McRae Rally 2 to run nice and smooth and looking excellent using 4xAA and 8xAF with everything maxed out; it really does look and play so very nice now. On the “difference” issue between ATI and nVidia, all I can is that I can’t see the difference when playing games or on my desktop (maybe I’ve been lucky in my manufacturers that I choose and the quality of the cards they make), and I really couldn’t care less if you grab a screenshot and show me that this or that one renders better or more true. If I can’t see it when playing games, then I really don’t care; show me the FPS, that is what matters to me. For visual 2D quality, I can say the card does very well, as do my nVidia-based solutions. I don’t think this is much of an issue anymore, especially with a card made by a company like Asus, which is known as a tier-one manufacturer.

The VIVO functionality and bundled games, utilities and accessories are all very good and include all that you would need to use the features of this card. It is easy to see how it performs in games that can really show the eye candy and pull down slower, older cards. I have enjoyed relaxing while playing Battle Engine Aquila; I also played a little GunMetal, but I’d have to say I prefer the Battle Engine over GunMetal for its POV and controls. Eye candy and effects-wise, they both offer a lot.

Pros

• Very Good Performance
• Very Good Bundle
• Good Features
• Nice, Different Orange PCB colour
• Unique & Efficient HeatSink Design
• Cool Running GPU & RAM
• Good Price – Now
• 3 Year Warranty
• Half Life 2 Certificate STILL included (at least at time of my purchase)

Cons

• Was a little expensive when I bought it
• Not guaranteed which RAM modules you’ll get. (This might be dependent on the PCB revision)
  

Final Score = 9 and the Bjorn3D Golden Bear Award …… I have had the card for over 3 months now and have experienced no problems with it. It has played all the games I have tried and the VIVO functionality is something I really like and have started using to import and edit card racing videos for my uncle. I would definitely recommend this card to anyone looking for a very good performance mid-range card with a very good bundle and more features than the average 9600XT. I can truly say I’m happy with my purchase, the only thing I would wish is that I had managed to get a board using the 2.8ns RAM, so later on down the line when games are a bit stressful on the card, I could overclock it a bit to help out.

I’d like to make a quick note on the matter of the memory modules fitted to this board … I have contacted Asus asking why the board I bought had 3.3ns BGA Samsung memory modules (part# K4D263238E-GC33 rated at 300Mhz Max Frequency) on it and other sites that have been sent review samples of the same revision board have received them with 2.8ns BGA Samsung memory modules (part# K4D263238E-GC2A rated at 350Mhz Max Freq.) I carefully checked the accompanying review photos from these other sites to discern the memory speed and board revision. I have not received a response from Asus as yet and will therefore have to draw conclusions that maybe reviewers who are sent cards to review get them with faster memory than you or I who actually have to pay for them and most of the time don’t see them until the box gets to our house. Samsung’s site

Update: Asus finally replied, and this was their response

Email solution
———–
The standard RAM Spec. is 3.3ns for A9600XT, but depend on the material status, if the original material are in shorh supply, we will use the substitute to replace it, it always the same or higher speed of RAM Spec. But the A9600XT still run in the same speed (3.3ns) except overclocking.

I still find it strange though that all the review sites that receive these cards manage to get the ones using the 2.8ns memory modules. Don’t you?