by Alexey Stepin , Yaroslav Lyssenko
08/30/2006 | 09:36 AM
PC gamers often have to choose between speed and silence. Some of them are ready to put up with the roar of two Radeon X1900 XT in a CrossFire subsystem, but most people wouldn’t want to have a computer that is as loud as to require you wearing earphones to protect against noise. On the other hand, the graphics subsystem must be powerful enough to run modern games at a high speed. It won’t make a good gaming platform otherwise, however quiet it may be.
Graphics card manufacturers try to meet users’ demands and equip their solutions with coolers that have better noise characteristics than the reference ones, yet perfect silence could only be achieved with a passive cooling system. And here the manufacturers encounter the heat dissipation problem: powerful gaming solutions generate too much heat for a passive cooler to cope with. Particularly, our tests of graphics cards from ASUS with the SilentCool cooler – the 7800 GT Top Silent and the Extreme N6600 GT Silencer models – showed that despite the elaborate design of the cooler those graphics cards were prone to overheat without additional air cooling from a GPU fan or system fan on the side panel of the PC case. This might have been expected since Nvidia’s G70 and NV43 chips installed on the mentioned solutions dissipate quite a lot of heat whereas any passive cooler that relies only on inter-case ventilation is inferior in efficiency to even the simplest of active coolers.
Theoretically, it is possible to make a passive cooler acceptably efficient by increasing the area of its heat-dissipating elements, but this only leads to an increase in size and weight beyond all reasonable measures. The mentioned ASUS SilentCool is not small or light as it is, and would become downright monstrous if it were adapted to cool a Radeon X1900 XTX that dissipates about 120W of heat power. Until recently, the performance vs. silence dilemma was extremely hard to solve and only entry-level devices used to come out with passive coolers. Such solutions have a big market but are mostly intended for installation into multimedia centers for which noise level is a key characteristic: there are few people who would enjoy watching a movie or listening to music under an incessant humming of the cooler. Graphics cards of this kind had no worth in gamer’s eyes due to very low performance in latest games. But heat dissipation of GPUs declined after the leading manufacturers transitioned to 0.09-micron tech process, so it has now become possible to create a silent and relatively compact graphics card that would suit for playing games.
Of course, this doesn’t refer to top-end solutions. It’s the realm of high frequencies, of power consumption and heat dissipation of 80W and higher, even beyond 100W. It is obviously impossible to cool such a graphics card with a passive cooler. On the contrary, some devices like the Sapphire Toxic Radeon X1900 XTX come equipped with liquid cooling systems even!
We are now talking about mainstream-class devices, particularly about the Nvidia GeForce 7600 GT. The heart of this graphics card is the G73 chip, designed specially for mainstream products and manufactured on TSMC facilities using 0.09-micro tech process with low-k dielectrics. The GeForce 7600 GT is quite fast, although doesn’t allow using high resolutions along with FSAA in latest 3D games. The peak power consumption of this graphics card is a mere 35W which makes it plausible to create a passively cooled, i.e. absolutely noiseless, version of this device. Gigabyte tried that and has released its GV-NX76T256D-RH “Silent Pipe II” to the public. Let’s learn more about this device!
The package of the GV-NX76T256D-RH “Silent Pipe II” has vertical orientation which seems to have become an established tradition because Gigabyte’s GV-RX19X512VB and GV-NX79X512DB models we’ve reviewed on our site earlier had boxes of this same form-factor, too.
The package is designed in the Civilization 4 style, indicating which game is included with the graphics card. The numerous captions and logotypes tell you the technical characteristics and capabilities of the enclosed product. The design of new packages from Gigabyte is elegant: it doesn’t make use of gaudy colors, yet is capable of attracting a potential customer. The positioning of the card as if for playing Civilization 4 appeals to strategic minds that don’t need as powerful graphics cards as people who play first-person shooters.
Besides the graphics card, the package contains the following:
The user manual contains a detailed description of the graphics card and driver installation, so even inexperienced users shouldn’t have any difficulties at all.
Being not a very heavy application, but still a very exciting game, Civilization 4 makes a good companion to a mainstream graphics card.
So, the accessories to the Gigabyte GV-NX76T256D-RH aren’t gorgeous, but a full version of a popular game makes a very nice addition. After all, the key point of this graphics card is not in its accessories but in its design and ability for silent operation!
Nvidia doesn’t watch so intently that the manufacturers stick to the reference design in mainstream products as it does with regards to top-end GeForce 7900 series cards, so there are some unique designs available on the market. The Gigabyte GV-NX76T256D-RH “Silent Pipe II” is among them. The GeForce 7600 GPU series requires a very simple PCB thanks to a 128-bit memory bus and low power consumption and any advanced graphics card manufacturer, especially such a giant as Gigabyte, can take up the job of designing and releasing such a card. So, we won’t talk about differences from the reference GeForce 7600 GT because the Gigabyte card has a completely new PCB design developed from scratch.
There’s not the least similarity here, from the color of the lacquer the PCB is coated with to the power circuit design and the placement of the memory chips. The cards only have the DVI/S-Video outputs and the MIO connector (it connects two cards in a SLI tandem) positioned in exactly the same way.
The power circuit of the Gigabyte card is considerably simpler than the reference card’s. It uses fewer electrolytic capacitors, MOSFETs and case-less coils, yet its heart is still the same pair of Intersil ISL6594 controllers located on the reverse side of the PCB. With a peak consumption of 35-37W, the certain simplification of the power circuit shouldn’t affect the stability of the device. The elements are all neatly assembled, except for the 60T03GH MOSFET in the bottom right corner of the PCB which doesn’t stand upright for some reason.
In the top left corner of the card, near the MIO connector that is covered with a plastic cap, there is a seat for a 4-pin connector like those that are used on sound cards to attach TV tuners or CD/DVD drivers, but the connector itself is missing on our sample of the Gigabyte GV-NX76T256D-RH. The same connector used to be installed on ATI’s Radeon X800/X850 cards to attach a video input located on the front panel of the PC case. The PCB developed by Gigabyte engineers doesn’t have a place for a VIVO chip, so this connector must be intended to provide a video output in any place of the case as is convenient for the user. The lack of VIVO functionality is not regrettable at all in this graphics card, even though it is positioned as a solution for home multimedia entertainment centers. Such centers are always equipped with TV-tuners which, besides their main purpose of reproducing TV channels, allow connecting to analog video sources. The quality of video capturing is generally higher with TV-tuners than with graphics cards that offer VIVO functionality due to the use of better media processors. Besides that, some advanced TV-tuners provide the option of hardware compression of the captured video signal into MPEG-2 or even MPEG-4 format which is unavailable on simple and cheap VIVO chips that some graphics cards are equipped with.
The memory chips are rotated by 90 degrees in comparison with the reference GeForce 7600 GT. Their shorter side is now parallel to the GPU package sides. Unlike an overwhelming majority of graphics cards tested in our labs, this device employs graphics memory from Infineon rather than from Samsung. The four chips in 136-pin FBGA packaging are marked as HYB18H512321AF-14. This is GDDR3 memory. Each chip has a capacity of 512Mb, is designed as 16Mx32, and works at 2.0V voltage.
The “-14” suffix denotes an access time of 1.4 nanoseconds and, accordingly, a rated frequency of 700 (1400) MHz. The memory chips are indeed clocked at this frequency, in full compliance with the official GeForce 7600 GT specification. Organized as 16Mx32, four such chips provide 256 megabytes of memory accessed across a 128-bit memory bus. We’ll check out soon how good this memory is at overclocking.
Notwithstanding the passive cooler, the graphics processor of the GV-NX76T256D-RH card is clocked at 560MHz, exactly as in the official GeForce 7600 GT specification. This shouldn’t become a problem considering the low heat dissipation of the chip and the design peculiarities of the Silent Pipe II cooler. However, we wouldn’t recommend using this card in a poorly ventilated system case, which is in fact a general recommendation for any passively cooled graphics card.
So, the GV-NX76T256D-RH is equipped with an original passive cooler called Silent Pipe II which deserves a separate section in this review. There are no special innovations in this cooler; it uses the time-tested heat pipe technology, yet its design is very original, at least we haven’t yet met a solution like that in our practice.
The GPU die has contact with a small copper heatsink two heat pipes are pressed into. Each pipe is connected to its own heat-spreading section. The first section consists of 20 aluminum ribs and is located on the GPU side of the card. It is installed on a double-height mounting bracket and the rounded ends of the ribs go outside the bracket, i.e. outside the PC case. This is meant to improve the efficiency of the cooler by means of natural convection (the heatsink is passive, i.e. no fan is blowing at it), but the protruding part isn’t longer than 1-1.5cm and is unlikely to have a big effect on the cooler’s performance. Moreover, this placement of the front heatsink makes the card a dual-slot device, which may prove inconvenient for small system cases or barebone systems.
The second heat-dissipating section is installed on the reverse side of the PCB and consists of a copper base and thin aluminum plates. The bracket under the heatsink serves as a limiter that prevents the heat pipe the heatsink is fastened to from unbending and breaking. Since this is its only fastening, the heatsink actually hangs in the air and can move to and fro in the range defined by the limiters on the bracket. This section is the main heat-dissipating element of the Silent Pipe II cooler. It is placed at the reverse side of the PCB and, in a tower system case, will turn to be on top after installation which will improve heat transfer due to convection: the hot air from this heatsink will be rising up to be replaced with cooler one. An intake fan at the bottom of the system case would help the process. The heatsink ribs are positioned perpendicularly to the side panel of the PC case, so any system fans on this panel will help improve the cooling of the Gigabyte GV-NX76T256D-RH, too.
The Silent Pipe II cooler developed by Gigabyte looks a well-thought-out and effective solution that has only one significant drawback. It is rather big and may not fit into a small system case. This cooler should cope well with cooling a G73 chip, but you may want to provide additional air cooling to increase stability, especially at overclocking. The memory chips are not cooled at all, but GDDR3 chips in modern 136-pin packages are stable at frequencies up to 700 (1400) MHz without enforced cooling. Of course, you should take care about memory if you are overclocking your Gigabyte GV-NX76T256D-RH, not only to protect it from damage but also to reach higher frequencies.
The Gigabyte GV-NX76T256D-RH comes with a passive cooling system which is absolutely silent at work. A computer with that card installed would produce noise with an intensity of 40dBA, i.e. exactly with the same intensity as with any other passively cooled graphics card.
As usual, we tried to overclock the tested product. The card being equipped with a passive cooler, we put an additional 120mm fan to blow across the card’s PCB. We at first reached 680/800 (1600) MHz frequencies, but the system hung up after a couple of minutes in 3DMark06. The card was somewhat more stable at a memory frequency of 780 (1560) MHz, but again we had problems after a while. At 660/780 (1560) MHz frequencies the card was working normally for an hour, yet proved to be unstable in the end. So, the highest frequencies the graphics card was stable at, i.e. passed the full cycle of tests without ever showing image artifacts, were 660MHz GPU and 760 (1520) MHz memory.
When we were overclocking GeForce 7600 GT cards with the reference design we couldn’t notch even 600MHz, so the 100MHz GPU frequency gain we managed to get from the Gigabyte GV-NX76T256D-RH is an excellent result especially as we didn’t resort to extreme overclocking methods, modifications, etc. We’d like to warn overclockers against thinking that Gigabyte’s design suits better for overclocking than Nvidia’s reference design. The result we achieved doesn’t mean that the PCB developed by Gigabyte improves the overclockability of the GeForce 7600 GT because we tested only one sample of the card. More tests are necessary to collect statistical data and claim that one PCB design is better than another. We don’t have such statistics and don’t have time to gather it, so we are not sure that each sample of the Gigabyte GV-NX76T256D-RH is going to be as good at overclocking as ours. However, 660MHz is so far the best result we’ve achieved in our labs with a GeForce 7600 GT without extreme overclocking methods (which might have led to even better results).
The memory frequency growth is, on the contrary, moderate. 60MHz isn’t much (for comparison: Samsung K4J52324QC-BC14 chips with the same access time are pre-overclocked to 800 (1600) MHz by the manufacturer on the XFX GeForce 7600 GT XXX Edition graphics card and are absolutely stable at that). It must be the higher voltage of the Infineon chips (2.0V against the Samsung chips’ 1.8V) that resulted in higher heat dissipation, but we aren’t absolutely sure. Perhaps another sample of the Gigabyte GV-NX76T256D-RH would have provided a higher memory frequency growth.
We measured the performance of the Gigabyte GV-NX76T256D-RH on a testbed that was configured like follows:
ATI’s and Nvidia’s drivers were set up in our standard manner:
We select the highest graphics quality settings in each game, identical for graphics cards from ATI and Nvidia. We do not modify the games’ configuration files and do not use the drivers’ profiles optimized for specific games. The frame rate is measured with the game’s own tools or, if not available, with the Fraps utility. We also measure minimum frame rate where possible.
Since we gave up testing in the 1024x768 resolution, and in higher resolutions with enabled FSAA the performance of mainstream graphics cards is often below the acceptable level, we decided to stick to testing in two work modes: with maximum level of anisotropic filtering and with full-screen anti-aliasing (FSAA 4x) enabled at the same time. Besides the standard 1280x1024 and 1600x1200 resolutions we have also added the tests in 1920x1200 widescreen resolution for those games that support it.
We turned on FSAA and anisotropic filtering from the game menu. If such options were unavailable, we forced FSAA and AF through the appropriate options of the ATI Catalyst and Nvidia ForceWare driver.
Besides the Gigabyte GV-NX76T256D-RH, the following graphics cards took part in the tests:
The cards were tested in these games and benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
Despite the 128-bit memory bus, the GV-NX76T256D-RH is no worse than the Radeon X1800 GTO in the “pure speed” mode and even delivers slightly higher average and minimum frame rates. And you can also see that a comfortable speed is achieved in 1600x1200 resolution. The overclocking brings about the biggest gain in 1280x1024 since the memory subsystem affects the overall performance less in this mode (as you remember, the memory chips on the GV-NX76T256D-RH were not good at overclocking).
The Gigabyte is also good in the “eye candy” mode, at least it provides a good reserve of speed in 1280x1024. At an average speed of 70fps and with occasional slowdowns to 45fps you won’t feel any discomfort in the hardest scenes. The Radeon X1800 GTO slows down less in such scenes, though.
Call of Duty 2 is a more demanding game, so there’s no talking about playing it at an acceptable speed on a mainstream graphics card even in 1280x1024 (at the highest graphics quality settings, that is). Putting aside the practical value of the results, we can note that the Gigabyte GeForce 7600 GT is always slightly slower than the Radeon X1800 GTO in the “pure speed” mode, but is suddenly ahead at the “eye candy” setting, losing only to the much more powerful GeForce 7900 GT in high resolutions.
Still remaining one of the most beautiful 3D shooters, Far Cry is by far not the most demanding of them. The Gigabyte card easily delivers the required 60fps in 1920x1200. This is a little less than the Radeon X1800 GTO with its 256-bit memory bus can offer, but the GV-NX76T256D-RH closes the gap at the overclocked frequencies.
The “eye candy” settings put a bigger load on the memory subsystem and the 128-bit memory bus of the Gigabyte GeForce 7600 GT cannot cope with it in 1600x1200, although the Radeon X1800 GTO has an average speed of about 55fps in that resolution. The memory chips on our sample of the new card proved to be poor at overclocking, so the maximum playable resolution for the Gigabyte card is 1280x1024 with enabled FSAA.
The second clip we use in our tests was recorded in the underground caves of the Research map. The requirements to the graphics subsystem are quite different here: fast execution of version 3.0 pixel shaders is more important than high memory bandwidth. The Radeon X1000 family usually performs worse in this test in comparison with the GeForce 7 series, obviously due to the specifics of the implementation of SM 3.0. This is exactly what we see here: at the default frequencies the GV-NX76T256D-RH beats the Radeon X1800 GTO and is very close to the Radeon X1800 XL (but only overtakes it at the overclocked frequencies).
As we’ve said above, high memory bandwidth is not called for in the caves, yet it comes in handy after we enable full-screen antialiasing. That’s why the Gigabyte GeForce 7600 GT isn’t that brilliant in the “eye candy” mode. However, it is not as far behind the Radeon X1800 GTO as on the Pier map, and the average speed is generally higher. This means you can’t play the game in 1600x1200 because Far Cry is not all about closed environments and you have to go out into the open and experience an unavoidable performance hit. So, if you want to use FSAA along with anisotropic filtering, you have to switch to 1280x1024.
Although Far Cry ’s implementation of HDR (FP16) is optimized for GeForce 7 cards, solutions from the class the GV-NX76T256D-RH belongs to are not strong enough to enable the user to use this mode in 1280x1024, let alone more extreme modes. The overclocking helps achieve a certain performance growth, but it is not enough to make the resolution of 1280x1024 playable.
If you’ve got a Gigabyte GeForce 7600 GT in your system, you may only be interested in 1280x024 resolution with disabled FSAA. In that case this card delivers high enough performance and is 10% faster than the Radeon X1800 GTO and a mere 5% slower than the Radeon X1800 XL. The average frame rate is too low for comfortable play in the other modes.
The use of the so-called deferred rendering makes it impossible to turn FSAA on in this game, so only “pure speed” mode results are published. Besides that, the game automatically narrows the range of available resolutions and graphics quality settings for graphics cards with less than 512 megabytes of memory on board. The cards in this review all have 256MB of memory, so we couldn’t launch the game in 1600x1200 on them.
So, 1280x1024 is the highest resolution available on the participating cards, but the best result, shown by the GeForce 7900 GT, is only 38fps. Weaker products, including the GeForce 7600 GT, have lower speeds and do not allow playing this game with any degree of comfort.
Take note that we achieved a nearly 20% average speed boost by overclocking our GV-NX76T256D-RH, which is due to a low load on the memory subsystem. Besides the relatively low resolution and lack of FSAA, Ghost Recon Advanced Warfighter has not very high-quality textures as a consequence of its being a multi-platform project. That’s why the game is sensitive to the GPU frequency and reacts readily to any increase in it.
Half-Life 2: Episode One supports FSAA and resolutions up to 1920x1200, but the GeForce 7600 GT only allows playing in resolutions up to 1280x1024 with comfort. The GV-NX76T256D-RH can be overclocked to deliver enough speed to play in 1600x1200 without FSAA or in 1280x1024 with 4x FSAA. The Radeon X1800 XL delivers as much performance, but it consumes more power, needs an additional power connection, and produces more noise whereas the Gigabyte card is absolutely noiseless.
Although Prey runs on a modernized engine from Doom 3 and has inherited all of the latter’s key features, it offers better-quality and more detailed visuals, which makes it more difficult for the graphics subsystem. We could only achieve an acceptable speed in 1280x1024 by overclocking our Gigabyte GV-NX76T256D-RH. There’s no talking about the higher resolutions or full-screen antialiasing. You need a graphics card at least one class higher than the one the GeForce 7600 belongs to.
Yet another project on the Doom 3 engine from id Software, Quake 4 is not as resources-hungry as Prey . Working at its default frequencies, the Gigabyte GeForce 7600 GT is almost as fast as to provide a comfortable gaming experience even in 1920x1200, being a mere 3fps behind the Radeon X1900 GT.
Unfortunately, it’s not that well in the “eye candy” mode, yet the GV-NX76T256D-RH isn’t far worse than the Radeon X1800 GTO in resolutions up to 1920x1200 in which the memory bandwidth becomes the most crucial parameter. This doesn’t matter much when the average performance is only 20-30fps. The GeForce 7900 GT is the only card in this review that allows using FSAA in Quake 4 .
Besides high requirements to the amount of graphics memory, Serious Sam 2 also makes wide use of pixel shaders with multiple texture lookups. That’s why the GeForce 7600 GT with its 12 TMUs and 256 megabytes of memory cannot hope for high results here. As it is, the Gigabyte can’t notch 40fps even at the overclocked frequencies, not to mention 50fps which is achieved by the GeForce 7900 GT with its 24 TMUs. Take note of the min performance numbers: they are very low with all the participating devices, so you are unlikely to create comfortable playing conditions even by reducing the resolution and the level of detail.
You don’t need as high an average speed to play a third-person shooter as you do to play a first-person one, yet the fans of the Hitman series might like to purchase something better than a GeForce 7600 GT. You can hardly be satisfied with an average of 35fps in 1280x1024. On the other hand, Blood Money locations are mostly small, and good overclocking can boost the performance of that card almost to the level of the GeForce 7900 GT. You’ll need a GPU frequency growth of 100MHz to achieve that, but we are not sure if many other GeForce 7600 GT models, besides the Gigabyte GV-NX76T256D-RH, are capable of that. Higher resolutions and/or FSAA can only be used on high-end or better products.
Splinter Cell is older than Hitman: Blood Money and is much more moderate in its requirements. If you’ve got a Gigabyte GV-NX76T256D-RH, or any other version of the GeForce 7600 GT, you can play this game with comfort in 1280x1024 at the highest graphics quality settings. The same is true for the Radeon X1800 GTO – the two cards deliver comparable performance in this test, although the ATI solution has a better min speed. To play in 1600x1200 and/or use full-screen antialiasing, you should overclock your card or buy a more expensive solution, e.g. a Radeon X1900 GT or a GeForce 7900 GT.
It goes without argument that you can enjoy a game fully only by running it at the highest possible graphics quality settings. That’s why we benchmarked the cards in Tomb Raider: Legend in the Next Generation Content mode which uses Shader Model 3.0 capabilities that are currently supported by all modern graphics cards.
Alas, this mode proves too hard for mainstream and even performance-mainstream products. The GeForce 7900 GT copes with this task the best of all, but has moderate results anyway. Its speed is only 17fps at the minimum. The Gigabyte GV-NX76T256D-RH performs even worse, but has a much better “pure speed” in comparison with the Radeon X1800 GTO. The overclocking provides a 20% performance boost, but it means only an insignificant increase in the average frame rate from 30 to 36fps. Like in many other modern games, there’s no talking about enabling FSAA on mainstream graphics cards.
The Nvidia GeForce 7 architecture doesn’t allow using HDR (FP16) along with FSAA, so we benchmarked the cards in TES IV: Oblivion using anisotropic filtering only because the game’s visuals degenerate too much when you disable HDR. Speed was measured with the Fraps utility, so the numbers may be somewhat inaccurate.
Well, the speed of the Gigabyte GV-NX76T256D-RH isn’t too high, yet it is only second to the Radeon X1900 GT in average performance in the resolution of 1280x1024 pixels and even rivals it at the overclocked frequencies. Like all graphics cards with a relatively small number of pixel processors (less than 24), the tested graphics card has a very low minimum of speed, so slowdowns are unavoidable. We recommend using 1024x768 resolution on GeForce 7600 GT to have a bigger reserve of speed. You can also disable HDR or lower the level of detail, but this would greatly reduce the visual appeal of that beautiful game.
Contrary to the test in a closed environment, the overclocking has a small effect on performance of the Gigabyte GV-NX76T256D-RH in open scenes because the memory frequency is more important here than the GPU clock rate and we haven’t achieved a good memory frequency growth at overclocking. The game is generally slow, but don’t forget that the vast vegetation-rich open areas of the Cyrodiil province easily put even high-end graphics cards like Radeon X1900 XT or GeForce 7900 GTX to their knees, let alone less powerful solutions.
This game is sensitive to the graphics card’s fill rate and amount of memory, but it can be played with comfort at an average speed of 30-40fps. The Gigabyte GV-NX76T256D-RH gives you that speed, especially at the overclocked frequencies, in 1280x1024 without FSAA. This is good enough considering that the GeForce 7900 GT is the only card in this review that makes the resolution of 1600x1200 playable.
Graphics cards from the Radeon X1000 family do not support vertex texturing, so they can’t use SM3.0 to render the water surface with the highest possible quality. Only Nvidia GeForce 7 cards can do that.
Pacific Fighters is a rather old title and it is also optimized for Nvidia’s graphics architecture. That’s why the GeForce 7600 GT looks brilliant here, making all resolutions through 1600x1200 playable in the “pure speed” mode and up to 1280x1024 at the “eye candy” settings.
The same-class Radeon X1800 GTO is slower in each resolution, and the Radeon X1900 GT in resolutions higher than 1280x1024. What’s curious, the gap is even bigger in the “eye candy” mode, although the memory bus load is higher in it. The Radeon X1800 XL is the only worthy rival to the Gigabyte GV-NX76T256D-RH thanks to its 16 TMUs.
The engine of X3: Reunion prefers the ATI Radeon X1000 architecture. The Gigabyte manages to reach the level of the Radeon X1800 GTO only at the overclocked frequencies. Different sectors of the game world vary in difficulty from the graphics card’s standpoint, so the results of the GV-NX76T256D-RH aren’t very high and suggest that you switch to 1024x768 resolution where there is a bigger reserve of speed. Full-screen antialiasing is not to be used on the GeForce 7600 GT, at least in 1280x1024.
Either because of pixel shaders with multiple texture lookups or certain peculiarities of the implementation of the HDR (INT) mode, or perhaps due to architectural specifics of the R520 and R580 chips, but the game invariably shows its liking towards the Nvidia GeForce 7 family. As a result, the Gigabyte card easily beats all the opponents from its own product class in the “pure speed” mode, being only slower than the performance-mainstream solutions Radeon X1900 GT and GeForce 7900 GT. However, only the results in the lowest resolution we use, i.e. 1280x1024 pixels, have any practical value.
Like in many other cases, the “eye candy” mode is unplayable on mainstream graphics cards, but note that the Radeon X1800 XL closes the gap to the Gigabyte GeForce 7600 GT and even overtakes it in 1280x1024.
Although created by another developer and not using SM3.0 and HDR, this game behaves much alike to Age of Empires 3 . The average level of performance in Rise of Legends is lower than in Age of Empires 3 , even though it uses less advanced graphical technologies, and the best the Gigabyte GeForce 7600 GT can give you is 27-30fps. This is not bad in comparison with the Radeon X1800 GTO and XL, but is not comfortable as it is. So, you will have to lower the resolution or the level of detail if you want to play this game on a GeForce 7600 GT.
The Gigabyte GV-NX76T256D-RH is 864 points behind the Radeon X1800 GTO at the default frequencies. The gap is diminished t 319 points when the Gigabyte is overclocked, yet it still occupies the last place among the participating graphics cards.
The GV-NX76T256D-RH is as fast as the Radeon X1800 GTO in the “pure speed” mode, but slower at the “eye candy” settings due to the narrower memory bus.
There are different conditions in the second test: the scene is small, so the fill-rate parameter and memory bandwidth are not crucial factors, but there’s always a lot of work for pixel as well as vertex processors (the latter rendering the lush vegetation). Thanks to its 8 vertex processors, the Radeon X1800 GTO is ahead of the Gigabyte GV-NX76T256D-RH in every mode. The Gigabyte closes the gap by means of overclocking, but only in the “pure speed” mode.
The Gigabyte GV-NX76T256D-RH performs better in the third test, but only in the “pure speed” mode again. Surprisingly, the effect from our overclocking the Gigabyte card shows up stronger in this very large scene at the “eye candy” settings than in the second test. The performance gain is big enough to put the GV-NX76T256D-RH on the same level with the Radeon X1800 GTO, notwithstanding the latter’s 256-bit memory bus.
The low overall score of the GV-NX76T256D-RH is probably only due to its loss in the second test. In the first and third tests it is at least no worse than the Radeon X1800 GTO.
All versions of the GeForce 7600 GT, including the Gigabyte GV-NX76T256D-RH, look much better in 3DMark06 than in the earlier version of 3DMark. The Gigabyte reaches the level of the Radeon X1800 XL at the overclocked frequencies.
The first group of tests makes use of Shader Model 2.0. The Gigabyte GV-NX76T256D-RH is much faster than its opponent here. The results of the separate tests will explain why.
The Gigabyte is a little slower than the Radeon X1800 GTO in the SM3.0/HDR tests. It is also unable to compete with the Radeon X1800 XL at the overclocked frequencies as it did in the SM2.0 tests.
Mainstream graphics cards are too slow in 3DMark06 if used with enabled full-screen antialiasing, which increases the measurement error. That’s why we decided not to benchmark them in the “eye candy” mode, limiting ourselves to anisotropic filtering only. Moreover, the GeForce 7 family cannot use HDR (FP16) and FSAA simultaneously.
The results of the first SM2.0 test are good enough for a solution with only 12 pixel processors, 12 TMUs and 8 ROPs: the Gigabyte GV-NX76T256D-RH isn’t any worse than the Radeon X1800 XL. It is close to the Radeon X1900 GT in 1280x1024 and even outperforms it at the overclocked frequencies.
The second SM2.0 test is technically similar to the second game test from 3DMark05, but the scene has been made more complex, so it is only in 1920x1200 that the Gigabyte card falls behind the Radeon X1800 GTO.
The first SM3.0/HDR test produces opposite results: the Gigabyte GV-NX76T256D-RH is slower than the Radeon X1800 GTO in 1280x1024 only.
Demonstrating dynamic shadowing and HDR capabilities, the second SM3.0/HDR test is quite favorable towards the Gigabyte GeForce 7600 GT. This card is as fast as the Radeon X1800 XL in 1280x1024 without any overclocking. It slows down to the level of the Radeon X1800 GTO in the higher resolutions, but not a step lower. So, the small advantage the Gigabyte GV-NX76T256D-RH enjoys over the Radeon X1800 GTO in the total scores is indeed confirmed by the results of the separate tests from the 3DMark06 suite.
Now that we’ve tested Gigabyte GV-NX76T256D-RH “Silent Pipe II” we can finally answer the question about whether the company managed to create a noiseless gaming graphics card. And the answer is yes. Gigabyte used an original passive cooling system that dissipates some of the heat generated by the graphics card outside the system case and transfers the remaining heat to the heatsink located on the reverse side of the card PCB. At the same time the Silent Pipe II cooling system turned out much smaller than the solutions from ASUS we have reviewed in the past. The compact PCB size of the card and no need for any additional power supply make Gigabyte GV-NX76T256D-RH «Silent Pipe II» an excellent choice for those who do not want to sacrifice the graphics card’s performance for the sake of its quietness. The only drawback of Silent Pipe II is its height, that makes this card a dual-clot solution, which may cause some problems if the card has to go into a compact system case.
The card revealed excellent overclocking potential and allowed raising the GPU frequency by 100MHz without any special efforts. We don’t know if all Gigabyte GV-NX76T256D-RH «Silent Pipe II» graphics cards overclock that well, but if it is so, then you will definitely achieve even better results if you use some extreme overclocking tools. And Gigabyte definitely deserves all the credit for the successful graphics card design.
We won’t go into details about the performance of the new Gigabyte solution, because it works at Nvidia’s recommended clock speeds. However, I would like to point out that the tests proved GeForce 7600 GT to be one of the best mainstream solutions, which often outperforms Radeon X1800 GTO. Of course, Gigabyte GV-NX76T256D-RH «Silent Pipe II» may face some difficulties with its 128bit memory bus in case full-screen antialiasing is enabled. However, the realities of the gaming world suggest the graphics solutions of this category are anyway not powerful enough to ensure acceptable level of performance with enabled FSAA and/or in the highest resolutions.
As for the advantages that are not directly connected with the graphics card itself, it is pretty hard to say anything. The accessories bundle of the Gigabyte GV-NX76T256D-RH «Silent Pipe II» is not very rich, although it contains everything necessary and even offers a quality strategy game – Civilization IV. Anyway, all the major advantages of the new Gigabyte GV-NX76T256D-RH «Silent Pipe II» are different: from the very beginning this graphics adapter was designed as a quiet, economical and pretty high-performance gaming solution and it truly is like that. We have every right to call Gigabyte GV-NX76T256D-RH «Silent Pipe II» one of the today’s best solutions in this category.