by Alexey Stepin
06/12/2006 | 11:04 AM
As every informed computer enthusiast knows, Gigabyte Technology is in the top five of the world’s major mainboard makers. It is a large corporation manufacturing a wide range of high-tech products: PC cases, cooling systems, notebooks, multimedia home centers, server solutions and even mobile phones. Of course, there are graphics cards in this list, too. Gigabyte is among the few companies that turn out products on graphics processors from both the leading developers, ATI Technologies and Nvidia.
Gigabyte’s model range includes entry-level graphics cards like Radeon X300 SE or GeForce 6200 as well as high-performance flagship solutions featuring the most advanced GPUs available at the moment. Last time we reviewed low-end and mainstream cards, but today we are going to review the fastest graphics cards Gigabyte offers: GV-RX19X512VB (Radeon X1900 XTX) and GV-NX79X512DB (GeForce 7900 GTX).
As you know from our reviews, all flagship products on GPUs from ATI Technologies and Nvidia are manufactured by companies like Sapphire Technology or Foxconn Electronics and then distributed ready-made among the GPU developers’ partners. ATI and Nvidia insist on keeping the reference design of such cards intact and allow no modifications. This ensures high quality and reliability, yet limits the graphics card manufacturers’ opportunities to distinguish their products on the market and make them more appealing for the customer. ATI’s and Nvidia’s partners have to experiment with the package design, accessories or cooling system to be different.
Let’s see how the new cards from Gigabyte differ from analogous cards other companies offer.
The package with a Gigabyte GV-RX19X512VB (hereafter referred to as Gigabyte Radeon X1900 XTX) is not intricately shaped like packages of XFX’s products, but still looks somewhat unusual due to its vertical orientation:
Fantasy-styled design is widespread among graphics card manufacturers. Step into any PC shop and there’ll be all manner of mages, dragons, elves and other mythology looking down on you from the shelves. Gigabyte’s designers couldn’t resist the temptation, but thankfully they got along without hackneyed monsters: a cute girl in a horny helmet and fanciful armor is depicted on the front of the box. The box is painted mostly red and pink which looks quite pretty and doesn’t irritate the eye. Moreover, red is the traditional color of ATI Technologies, so the box indicates what graphics processor is installed on the card.
The text at the bottom of the box tells you the basic technical specs of the product. There is also a sticker indicating that there’s additional software included within. According to the sticker, the Gigabyte Radeon X1900 XTX brings you a software pack costing as much as $190, which looks like a real bargain. On the other hand, it all depends whether you in particular are going to use this software. Besides the graphics card, the box contains:
The user manual is written in English and Chinese and is all right from the technical standpoint. It gives you a thorough description of the installation procedure as well as of the ATI Catalyst Control Center settings. The manual also explains how to use the video-editing program CyberLink PowerDirector and the media player CyberLink PowerDVD 6 that you receive with the card. The BIOS update procedure is also described. We don’t know if localized versions of the manual are supplied into the European regions.
The choice of games in the software bundle is somewhat surprising because they were both released in 2004, and while the racing sim Xpand Rally features good enough visuals even for today, Counter-Strike: Condition Zero uses the game engine from the first Half-Life. Yes, the game is highly popular, but it is nothing much in terms of graphics and doesn’t suit for demonstrating the capabilities of a Radeon X1900 processor.
Overall, the accessories to the Gigabyte Radeon X1900 XTX aren’t exactly gorgeous, especially for a flagship solution, yet you can’t call them paltry, either. You get some useful software and two popular games, even though not masterpieces of 3D graphics.
The Gigabyte Radeon X1900 XTX is an ordinary Radeon X1900 XTX manufactured for ATI and doesn’t differ at all from the reference card:
Only three things indicate that it is a Gigabyte version of Radeon X1900 XTX: the sticker with the model name near the PCI Express x16 slot, the picture on the cooler’s casing, and the Gigabyte logo at the bottom of the blower. The reference cooler is quite efficient, so Gigabyte didn’t try to experiment with it. There’s nothing particularly interesting on the reverse side of the PCB, just various stickers with serial numbers. The cooler is fastened reliably and safely for the PCB with 8 screws. There’s a metal back-plate with dielectric pads to ensure that the heatsink’s sole is uniformly pressed down to the GPU die.
The Gigabyte Radeon X1900 XTX has eight chips of GDDR3 SDRAM on board (Samsung’s K4J52324QC-BJ11 chips in 136-pin packaging). The total amount of memory is 512MB and it is clocked at 775 (1550) MHz, exactly as described in the Radeon X1900 XTX specification. This memory, however, is rated for work at 900 (1800) MHz and we are going to try to overclock it to this frequency.
The graphics processor is clocked at 650MHz, as described by the Radeon X1900 XTX specification, too. Thick dark-gray thermal paste with low thermal resistance is used as a thermal interface between the GPU die and the heatsink’s base; the memory chips contact the heatsink via elastic thermal pads. This should be enough considering that the power consumption (and, accordingly, heat dissipation) of K4J52324QC-BJ11 chips is very low – 2.6W per chip at maximum when clocked at the nominal frequency.
There’s no sense in describing the cooler installed on the Gigabyte Radeon X1900 XTX in details as we did it a number of times in our earlier reviews. In brief, the cooler consists of a massive copper well-ribbed heatsink equipped with heat pipes and covered with a plastic casing. The blower set up behind the heatsink drives air from within the system case through to heatsink and then exhausts it to the outside. So, this cooler help keep the system case temperature low, but there are some imperfections in ATI Technologies’ implementation. Particularly, the plastic casing acts as a resonator, adding an irritating plastic note to the sound of the blower when the latter works at an increased speed. Besides that, the reference cooler from ATI is very heavy due to the copper heatsink with a thick sole.
Like every other Radeon X1800/X1900 card, the Gigabyte Radeon X1900 XTX supports VIVO through a Rage Theater chip. This functionality isn’t in great demand in our era of digital video, but sometimes you may need to connect an analog video source to your PC.
Since two graphics cards from Gigabyte are covered in this review, we are going to show you the next one, Gigabyte GV-NX79X512DB, before proceeding to overclocking and gaming tests. Gigabyte’s products having not very clear model names, we will refer to the next card as to Gigabyte GeForce 7900 GTX.
The Gigabyte GeForce 7900 GTX graphics card comes in a box of the same shape as the Gigabyte Radeon X1900 XTX comes in – a rectangular vertically oriented box:
The package is generally designed alike to the Gigabyte Radeon X1900 XTX’s, but uses a black-green color scheme, green being the color of Nvidia. A young warrior is pictured on the box. A sticker informs you there’s additional software and games included. This package looks good overall, and here’s what we found in it, besides the graphics card:
The user manual is as exhaustive as the one included with the Gigabyte Radeon X1900 XTX. Moreover, it additionally describes how to build a SLI configuration and make it running. Connection of HDTV devices with YPbPr input is also covered in the manual. Everything is accompanied with illustrations so even inexperienced users shouldn’t have any troubles.
You get a copy of Serious Sam 2 with the card. That’s a good choice since this 3D shooter is not only popular, but also features up-to-date visuals with numerous shader-based effects and support for HDR. The game is very demanding to the computer’s graphics subsystem and thus suits well to showcase the capabilities of the Gigabyte GeForce 7900 GTX. What’s more, the game is an ideal match to this very card since its pixel shaders with multiple texture lookups are handled most efficiently by the GeForce 7 architecture.
There are no S-Video and RCA cables among the accessories, yet we think the included set is quite sufficient, especially with the popular 3D shooter. On the other hand, the company might have put in more and better accessories into the box with its top-end product.
Like the first card in this review, the Gigabyte GeForce 7900 GTX is an exact copy of the reference card.
There are even fewer discrepancies here. The only things that indicate this is a Gigabyte product are a sticker with the company logo on the fan and an elastic cap on the MIO connector. There’s also a sticker with the card model name near the PCI Express x16 slot, but it is not easily seen due to the massive cooler. The cooler’s casing bears two halves of the Nvidia logotype although Gigabyte could have placed an original picture there like they did with their Gigabyte Radeon X1900 XTX.
Besides the ordinary stickers with serial numbers, there’s one with the version of the BIOS employed by the card. This information may be useful in some situations. The Nvidia GeForce 7900 GTX’s reference cooler is very light, especially in comparison with the Radeon X1900 XTX’s, and is fastened to the PCB with 9 screws without a back-plate.
The Gigabyte GeForce 7900 GTX uses the same memory as the Gigabyte Radeon X1900 XTX: Samsung’s K4J52324QC-BJ11 chips in 136-pin FBGA packaging. The memory is clocked at 800 (1600) MHz, exactly as described in the GeForce 7900 GTX specification. The chips having a capacity of 512Mb each, the total of graphics memory is 512 megabytes. These memory chips have a very low access time, so there is some reserve for overclocking. We are going to check how big it is in one of the following sections.
The GPU frequency complies with Nvidia’s specs, equaling 650MHz. This is the frequency of the pixel processors and raster operators (ROPs), while the vertex processors are clocked at 700MHz. The card uses traditional thermal interfaces: dense dark-gray thermal paste between the GPU die and the cooler’s sole and non-organic fiber pads soaked in thermal paste between the memory chips and the cooler’s base.
The Gigabyte GeForce 7900 GTX uses a copy of the reference cooler which was first employed in Nvidia Quadro FX 5400 and won our praises for its near-silent operation. The secret of this cooler is that its 80mm works at a very low speed, about 1000rpm. Here’s how this cooler works: the copper base takes the heat from the GPU and transfers it to the two heatsinks on the right and left of the fan via four heat pipes. The heatsinks are covered under an air-directing casing.
Some of the hot air is exhausted out of the case through the slits in the graphics card’s mounting bracket. Some air makes its way back into the system case.
This cooler does its job well, but is not free from drawbacks. The air stream created by the low-speed fan is very weak. As a result, the heatsinks and heat pipes become very hot when the card is in 3D mode, despite the low heat dissipation of the G71 chip. You may get your fingers scorched if you touch the heat pipes when the card is working, so be careful. This shouldn’t be a problem in a closed and well-ventilated system case.
There is a seat for a video-encoding Philips SAA7115 chip on the card, but the chip itself is missing, depriving the Gigabyte GeForce 7900 GTX of VIVO functionality. But as we’ve said in our earlier reviews, this feature isn’t very important for a today’s graphics card. Digitals cameras communicate with the PC via FireWire. And if you really need to capture analog video, you can buy a cheap internal or external TV-tuner that provides a number of additional features, too.
This time we’ve dedicated a separate section to the noise characteristics of tested graphics cards because we are now going to evaluate them not subjectively but with a digital sound-level meter Velleman DVM1326. This instrument has a resolution of 0.1dB and allows measuring in a range up to 130dB with A or C weighting.
What do these A and C mean? The fact is the human hearing is a complex non-linear psycho-physiological process. Sounds that have the same sound intensity but different frequency are perceived by the human ear as differing in loudness. Particularly, medium-frequency sounds are perceived as being louder than low- or high-frequency ones. As a result of scientific research in the 1930s, the equal loudness contours were produced (the so-called Fletcher-Munson curves). These curves show the sound intensity of different-frequency tones that are perceived by us as having the same loudness.
Three such contours, for sound intensity levels of 40, 70 and 100dB, were used to create A-, B- and C-weighting curves which help describe loudness with a single value that accounts for the way the human ear perceives sound. Currently, the A curve with a decline in 20-1000Hz range is an industry standard; the C curve with an almost linear frequency characteristic is used but seldom. The B curve is not used at all and is not even supported by our sound-level meter. Of course, we used the A curve for our measurements, so the results are shown in dBA.
The decibel is a logarithmic rather than linear measurement unit. It means that an increase in sound intensity by 3dB is the same as a doubling of intensity, but due to the non-linear nature of human hearing, it is commonly assumed that the perception of a twofold increase of sound loudness corresponds to an increase of sound intensity by 10dB.
We don’t have a special acoustic chamber, so the numbers we get in our test will differ from what the manufacturer specifies, yet they will be closer to real conditions the PC is used under. The level of background noise in our test room was 36dBA. At a distance of 1 meter from the working testbed with a passively cooled graphics card our sound-level meter showed 40dBA. We’ll base our judgments upon these two reference numbers.
We measured the level of noise produced by the cards in three modes: 2D, 3D typical and 3D maximum (the maximum level of noise with the automatic fan-speed adjustment). The noise is measured at a distance of 1m and 5cm from the working testbed assembled in a Chieftec LBX-01 system case with the side panel removed. Here are the results:
The diagram shows that Gigabyte’s cards produce about the same amount of noise in 2D and 3D modes and are equally quiet. It’s only when the fan of the Gigabyte Radeon X1900 XTX begins to rotate at a higher speed, reacting to an increase in the temperature of the GPU, that the difference becomes noticeable – 2.4dBA. This is not little considering the decibel is a logarithmic measurement unit. These results also comply with what you can hear with your own ears: when working at an increased speed, the cooler of the Radeon stands out among the other components of the system with its irritating “plastic-colored” noise due to resonance in the air-directing casing. The Gigabyte GeForce 7900 GTX is almost imperceptible by the ear under the same conditions, and you won’t hear it at all if your system case is closed.
At a distance of 5cm the Gigabyte GeForce 7900 GTX is quieter than the Gigabyte Radeon X1900 XTX, too. This is expectable, considering its 80mm fan rotates at only 1000rpm. Subjectively, a soft whispering of the air can only be heard. The Gigabyte Radeon X1900 XTX again produces that irritating “plastic” rumble with its blower and resonating casing. You may find it uncomfortable to be near this card when the fan speed management system decides to increase its speed. Fortunately, this is a hypothetic situation as no one is going to remain so close to the system case for long. This test however shows that the acoustic characteristics of the GeForce 7900 GTX cooler are better than those of the Radeon X1900 XT/XTX cooler.
So, in the most typical situation, when the PC is about 1 meter away from the user, the level of noise produced by Gigabyte GeForce 7900 GTX is 4.6dBA higher than that produced by the same system but with a passively cooled graphics card inside. This value remains constant in all the work modes of the graphics card and, thanks to its spectrum, is almost imperceptible against the noise from other system components (CPU cooler, hard drive, etc).
The Gigabyte Radeon X1900 XTX has worse results when running 3D applications for long. Its noise is 7dBA louder than the noise from the same system but with a passively cooled graphics card – that was perceptible by the ear. Fortunately, the fan of the Gigabyte Radeon X1900 XTX is not always working at its highest speed. Moreover, most games have internal sounds that override the graphics card’s noise.
Gigabyte’s Radeon X1900 XTX and GeForce 7900 GTX are both top-end graphics solutions which means they are already working at the limit. Still, we made an attempt to overclock them. The maximum stable frequencies of the Gigabyte Radeon X1900 XTX were 670/810 (1620) MHz for the GPU and memory, respectively. The Gigabyte GeForce 7900 GXT was stable at 690/880 (1760) MHz.
The achieved GPU frequencies are quite high considering that we used only the standard overclocking tools. The memory only did well on the Gigabyte GeForce 7900 GTX. Considering that 1.1ns chips are capable of working at 900 (1800) MHZ frequency, the memory frequency growth on the Gigabyte Radeon X1900 XTX is disappointing. It’s possible the thermal pads employed by ATI Technologies are not very good.
Both the cards being in fact reference samples of the respective products, the 2D image quality was immaculate on either of them. We couldn’t spot any defects in the picture the cards produced via our Dell P1110 and P1130 monitors in any display mode, including 1800x1440@75Hz.
We tested Gigabyte Radeon X1900 XTX and Gigabyte GeForce 7900 GTX in the following testbed:
We set up the ATI and Nvidia drivers in the same way as always:
We selected the highest graphics quality settings in each game, identical for graphics cards from ATI and Nvidia, except for the Pacific Fighters flight simulator that requires vertex texturing for its Shader Model 3.0 rendering mode. Radeon X1000 doesn’t support this feature therefore we ran the game in Shader Model 2.0 in this case. We did not edit the configuration files of the games. We also didn’t use the driver profiles optimized for given games. To measure the performance we either used the integrated tools of the games we tested in, or if there were none available, resorted to FRAPS utility. If it was possible, we measured minimal performance as well.
To load the video subsystem to the full extent and to minimize the influence of the CPU speed on the performance results we didn’t test the systems in the “pure speed” mode. We only ran the tests in “eye candy” mode with full-screen anti-aliasing and anisotropic filtering. The graphics subsystem potential is used in a more optimal manner in this case. Besides, the image quality is considerably higher than in those cases when no full-screen anti-aliasing and/or anisotropic filtering is enabled.
We enabled FSAA and AF from the game if possible. Otherwise we forced the necessary mode from the ATI Catalyst and Nvidia ForceWare graphics card driver. We didn’t include any results of the overclocked graphics cards, because we didn’t manage to raise the clock speeds high enough.
These games and applications were used as benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
Both the Gigabyte cards deliver the same performance of over 120fps at the beginning, but the Radeon-based one comes out on top in 1600x1200, obviously due to its more advanced memory subsystem.
Here, all the cards provide a high enough average frame rate when we turn on 4x FSAA. The game was tested with the Fraps utility, so the numbers may be somewhat inaccurate.
The Gigabyte GeForce 7900 GTX beats the Gigabyte Radeon X1900 XTX in this test, but it is greatly aided by the design of the game engine which uses OpenGL and stencil shadows. Anyway, this advantage matters little here because both the cards deliver an excellent average performance in all the standard resolutions even with enabled full-screen antialiasing.
The Gigabyte GeForce 7900 GTX enjoys a certain advantage over its opponent in Call of Duty, too. But again, this doesn’t matter much. Both the cards are sufficiently fast in 1280x1024, but cannot maintain a playable average frame rate in 1600x1200. The gap between the two cards from Gigabyte is expectably getting smaller in higher resolutions. They differ by only 2fps in the hardest display mode.
Doom 3 prefers Nvidia’s GPUs, too, and the reasons are the same as with The Chronicles of Riddick: this is an OpenGL application and it uses rendering methods optimized for the GeForce 7 architecture. Yet even despite this handicap, the Gigabyte Radeon X1900 XTX is less than 10% behind the Gigabyte GeForce 7900 GTX in the standard resolutions with enabled full-screen antialiasing. The average performance of both the tested products isn’t lower than 57-62fps.
It’s just the opposite on Far Cry’s Pier map where the high pixel shader performance allows the Gigabyte Radeon X1900 XTX to take revenge. This is only visible in 1600x1200, though, because in the lower resolutions the top-end graphics cards are limited by the CPU performance.
The scene recorded on the Research map helps differentiate between the cards better. The action goes in catacombs and the CPU influences the overall performance less. Graphics cards on ATI’s Radeon X1000 GPUs are often slower than same-class GeForce 7 products in this test due to the way SM3.0 is implemented in Far Cry. However, the Gigabyte Radeon X1900 XTX carries 48 pixel processors on board and features a very efficient memory subsystem. This helps it outperform the Gigabyte GeForce 7900 GTX in high resolutions.
Far Cry is not a hard game by today’s standards, so both the cards from Gigabyte handle it well when full-screen antialiasing is enabled. Their average performance is always higher than 100fps even in 1600x1200.
Although HDR (F16) support for Radeon X1000 is not implemented well in this game, the Gigabyte Radeon X1900 XTX provides a good average performance in all the tested resolutions.
It’s the same when we use HDR on the Research map. The Gigabyte Radeon X1900 XTX’s 60fps in 1600x1200 is quite sufficient for normal play. The average speed of the Gigabyte GeForce 7900 GTX is higher by 10fps or about 15%.
Despite the huge number of complex shader-based special effects in this game, Gigabyte’s cards yields similar results: the Radeon X1900 XTX is a little better in high resolutions, but the GeForce 7900 GTX is one step ahead in lower ones. Note also that the Gigabyte GeForce 7900 GTX provides a higher minimum speed than its opponent. It’s not clear why the 48 pixel processors of the Radeon X1900 XTX don’t make it a winner in this test. It’s possible that this card’s performance is limited by its 16 TMUs (the GeForce 7900 GTX has 24 TMUs).
This shooter is quite a resource-consuming application, so you don’t have a playable frame rate in 1600x1200 even on single top-end graphics cards, not if you turn on full-screen antialiasing.
For reasons hidden in the driver or in the design of the Source engine Half-Life 2 runs on, graphics cards on GPUs from ATI Technologies have better results than Nvidia’s solutions in this test.
This is not a big problem, though. Both the Gigabyte cards deliver enough performance in all the resolutions with full-screen antialiasing. You can even increase the level of FSAA from 4x to 6x on a Gigabyte Radeon X1900 XTX to improve the visuals quality without losing anything in your playing comfort.
We suspect Half-Life 2: Lost Coast uses shaders with multiple texture lookups, but there may be other improvements in the new Source engine that lead to the defeat of the Gigabyte Radeon X1900 XTX. Note that the results of the Radeon X1900 XTX do not improve much at the increased frequencies which may be indicative of its having too few TMUs. Currently, it is only the GeForce 7900 GTX that can provide a comfortable average performance in 1280x1024 with enabled FSAA in this tech demo.
It’s logical that the Gigabyte GeForce 7900 GTX wins this round: the game is simple and doesn’t use math1ematics-heavy shaders, so the Gigabyte Radeon X1900 XTX just has nothing to prove its muscle on. Anyway, the average performance of the R580-based card is 72fps in 1600x1200 whereas its min speed is a mere 6fps lower than that of the Gigabyte GeForce 7900 GTX. The cards are never slower than 60fps even in the hardest scenes of this game which means comfort for the gamer in every resolution.
Project: Snowblind was tested with the Fraps utility, but we used a cut-scene available in the game and are quite sure about the accuracy and repeatability of the results.
Despite full-screen antialiasing and OpenGL, the Gigabyte Radeon X1900 XTX and Gigabyte GeForce 7900 GTX have the same performance in a normal gaming situation with lots of shots and action. Their average performance is over 100fps in all the resolutions. The minimum of speed is quite high, so you won’t be disappointed when playing Quake 4 on a top-end graphics card.
The game employs a large number of pixel shaders with multiple texture lookups and prefers graphics cards with more TMUs on board. In other words, GeForce 7900-based solutions have a certain advantage in this test over Radeon X1900-based ones.
Although the average performance of the Gigabyte GeForce 7900 GTX is quite high in 1280x1024, its min speed is as low as 13fps, making impossible to play the game’s most complex scenes normally. So, you have to limit yourself to 1024x768 in Serious Sam 2 if you insist on using FSAA. Note also that the Gigabyte Radeon X1900 XTX has a better minimum of speed than its opponent, but a worse average performance. The bigger reserve of speed must be due to the card’s having two times more pixel processors, even though they may not be getting as many textures as they could process.
We didn’t test this game in FSAA modes because it disables HDR on Nvidia’s GeForce cards in this case, making the image much less beautiful. Moreover, HDR is incompatible with the Bloom effect. TES IV: Oblivion doesn’t offer integrated benchmark options, so we had to test it manually with the Fraps utility; the numbers may be a little inaccurate as a consequence.
In dungeons and closed environments in general it is important that the graphics card be able to effectively process pixel shaders as well as lighting and shadowing. The advantages of the Radeon X1900 architecture show up in a rather peculiar way here: the average performance of the Gigabyte Radeon X1900 XTX differs but slightly from that of the Gigabyte GeForce 7900 GTX, but the Radeon has a considerably higher minimum of speed! It means the game is going to run smoother on the ATI-based card, giving you more playing comfort.
Considering Oblivion is not all about closed environments, we recommend you using a resolution lower than 1600x1200 even on flagships products from Nvidia and ATI to safeguard yourself against slowdowns.
The Gigabyte GeForce 7900 GTX is a little faster than the Gigabyte Radeon X1900 XTX on average, but has a much worse min speed (less than 20fps even in 1024x768). In other words, you can’t avoid the game’s slowing down in complex scenes on this graphics card. Thus, the Radeon X1900 XTX remains the best graphics card available to play TES IV.
The majority of special effects in Splinter Cell: Chaos Theory are created with math1ematics-heavy shaders, so the Gigabyte Radeon X1900 XTX feels at ease here, providing a min speed of 45fps in 1600x1200 with enabled 4x FSAA. The average performance of the two cards from Gigabyte is roughly the same. You are not going to see any difference until you come across a complex game scene where the Gigabyte Radeon X1900 XTX is going to ensure a smoother game-play.
The engine of this game being oriented at the GeForce 6/7 architecture, it’s no wonder the Gigabyte Radeon X1900 XTX is on the losing side. However, this graphics card is capable of giving you a playable frame rate in resolutions up to 1280x1024 with a min speed similar to that of the Gigabyte GeForce 7900 GTX. Unfortunately, the Radeon provides a worse-quality picture than the GeForce because the game requires support for vertex texturing, and Radeon X1000 series chips lack this capability.
The space simulator X3: Reunion favors graphics cards on ATI Technologies’ chips and the Gigabyte Radeon X1900 XTX gains a confident victory over the Nvidia G71-based solution. Both these cards provide a comfortable frame rate in 1600x1200 with enabled 4x FSAA, their speed never sinking below 40fps.
Age of Empires 3 using a lot of SM3.0/HDR special effects, graphics cards with more pixel processors, i.e. based on ATI’s Radeon X1900 processor (R580), have the best results in this test. The Gigabyte Radeon X1900 XTX beats the Gigabyte GeForce 7900 GTX in high resolutions as you can see. The Radeon looks preferable, yet the other card is also sufficient for playing this strategy in any resolution with enabled full-screen antialiasing.
It’s different in Dawn of War: the Gigabyte Radeon X1900 XTX has a min speed of less than 25fps even in 1024x768 whereas the Gigabyte GeForce 7900 GTX easily yields over 35fps in a higher resolution. Why? Because the game makes use of stencil shadows the Radeon X1900 lacks any means to accelerate the rendering of. When there are many units in the screen, each casting such a shadow, the GeForce 7900 with its UltraShadow II technology is obviously better than its rival.
In the total score chart Gigabyte X1900 XTX outperforms its brother based on G71 GPU by 527 points. Let’s see what results we will get from the individual tests. However, we should keep in mind that the individual test results were obtained with enabled anti-aliasing unlike the total score.
Game 1 test is the largest one and contains a lot of light sources, however the Gigabyte solution based on ATI R580 shows the best result here than the Nvidia G71 card with 24 TMUs. It should be the ringbus memory controller of Gigabyte Radeon X1900 XTX that played the most important part here.
Game 2 test doesn’t require high fillrate, but demands efficient work with lightning and shadows, as well as performance of vertex processors that are responsible for creating dynamic vegetation. Thanks to higher GPU speed, Gigabyte GeForce 7900 GTX manages to show very good result in 1024x768, but then it start falling behind Gigabyte Radeon X1900 XTX again. In 1600x1200 the gap reaches 10%.
Game 3 test requires high complex shader processing speed, and here Gigabyte X1900 XTX is beyond all competition right from the start, because it features 48 pixel processors onboard, which is twice as much as Gigabyte GeForce 7900 GTX has. The triumph of Radeon X1900 is slightly held back by the 16 TMUs it has, because the scene has very high math1ematical as well as texturing workload. All in all, the total score demonstrated by the Gigabyte graphics cards in 3DMark05 test is proven by the results of each individual benchmark.
In 3DMark06 the leadership belongs to Gigabyte 7900 GTX although it is only 142 points ahead of Gigabyte Radeon X1900 XTX. It is probably the first SM2.0 graphics test that determined these results, so let’s take a closer look at it now.
It is true, Gigabyte GeForce 7900 GTX has the biggest advantage in SM2.0 tests, which equals 257 points while the overall score lies around 2000 points. It is most likely to be the first test, a heavier version of Game 1 3DMark05 test, that imposes the biggest influence on the result.
In SM3.0/HDR tests the winner is Gigabyte Radeon X1900 XTX. This solution feels at home in these benchmarks with a lot of complex shader effects and HDR. Although I have to admit that this advantage is not as significant as the advantage of Gigabyte GeForce 7900 GTX in the previous group of benchmarks we have just discussed.
The supposition we made that Gigabyte GeForce 7900 GTX owes its victory to the results of the first SM2.0 graphics test have been proven: the card is 10%-15% ahead of the competitor in the first two resolutions. In 1600x1200 the gap almost vanishes and the reason evidently lies in the more efficient work with the memory subsystem of Gigabyte Radeon X1900 XTX. The nearly identical results of both, Gigabyte Radeon X1900 XTX and GeForce 7900 GTX, indicate that the TMUs limit the performance of our testing participants in 1024x768: we see that none of them can overcome the 17fps performance barrier.
The second SM2.0 test shows a totally different picture: the leadership here belongs to Gigabyte Radeon X1900 XTX and as the resolution grows, its advantage increases going beyond 20% in 1600x1200. Unlike the first test, this benchmark requires high texturing speed and Radeon X1900 XTX has always been great with CSM dynamic shadows (thanks to Fetch4 support).
So, Gigabyte Radeon X1900 XTX it is solely the first test that determined the failure of Gigabyte Radeon X1900 XTX in SM2.0 total score, because the first test is the most sensitive to the number of TMUs (and hence to the fillrate).
Gigabyte Radeon X1900 XTX and Gigabyte GeForce 7900 GTX graphics cards made a very decent impression although they couldn’t boast any extraordinary features, like Sapphire Toxic Radeon X1900 XTX, for instance (for details see our article called Radeon without Antidote: Sapphire Toxic Radeon X1900 XTX Review). These two solutions are the common representatives of the Radeon x1900 XTX and GeForce 7900 GTX families with all the typical highs and lows.
In particular, Gigabyte Radeon x1900 XTX demonstrates excellent performance in applications that require high shader processing speed and deal with complex math1ematical calculations, but gives in in tasks with high fillrate requirements. Among its major advantages we should certainly mention 48 pixel processors and extremely efficient memory subsystem with ring bus memory controller as well as simultaneous support of FSAA and HDR (FP16).
Gigabyte GeForce 7900 GTX feels great in games using OpenGL and rendering techniques benefiting from GeForce 6/7 architecture. Namely, these techniques include stencil shadows. Also, the 24 TMUs it has guarantee high performance in scenes with high-resolution textures and/or shaders containing a lot of texture samples. A great example of a game like that is Serious Sam 2.
To cut the long story short, Gigabyte Radeon X1900 XTX and Gigabyte GeForce 7900 GTX have their advantages and drawbacks and it all depends on the game. The performance of Gigabyte’s both flagship solutions is quite high for gaming with enabled full-screen antialiasing, however if we look at them from the long-term prospective, then the ATI R580 model will be preferable, because the shader effects with complex math1ematical calculations is going to grow bigger with the time.
As for the peculiarities of these two particular graphics solutions, there are not too many of them worth mentioning here. I would certainly like to point out great packaging design and free games included with the cards. And although Gigabyte GeForce 7900 GTX comes bundled with only one free game, the quality of the game makes up for the single title included. I have the feeling that they could have made a bit richer bundle for the high-end cards like that, but if you do not really care about the availability of free software, then either of these cards may become a great choice for you.