by Alexey Stepin , Yaroslav Lyssenko
04/10/2008 | 12:50 PM
It’s a kind of a tradition in the world of consumer 3D graphics hardware that one year since the release of a flagship product there appears a new card capable of delivering similar performance for only half the money. A striking example of this is the recently announced GeForce 8800 GTS 512MB. This graphics card, based on the G92 core in its full configuration with 128 streamed processors and 32 TMUs, proved to be competitive against the GeForce 8800 GTX whose performance level had seemed unattainable. The new card came out at a recommended price of $299-349 whereas the GeForce 8800 GTX was available for $549-649.
There are no traditions without exceptions, though. As you know from our reports, the Nvidia GeForce 8800 GTX carries 768 megabytes of graphics memory accessed across a 384-bit memory bus. The GeForce 8800 GTS 512MB has only 512 megabytes of memory with 256-bit access. As a consequence, the new card is often slower than the older one in the most resource-consuming modes (1920x1200 with 4x FSAA) and in games that have high memory requirements (e.g. Hellgate: London) due to lower memory bandwidth.
Can this gap be closed by increasing the GPU and memory clock rates of the GeForce 8800 GTS 512MB so that it would be the fastest card in every application? Factory overclocking is a widespread practice and we often review pre-overclocked graphics cards, usually from ASUS, Leadtek or Gainward. XFX is also known to have a passion for overclocking. We can recall the XFX GeForce 8800 Ultra Extreme as an example.
Of course, the company couldn’t pass by the GeForce 8800 GTS 512MB. So, here it is, the XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition.
The box with the card has medium width and height but it’s thicker than usual. It is designed in the company’s traditional restrained black-and-green style.
It’s easy to identify the product model: a special sticker tells you it’s an XXX series product. The GPU frequency is mentioned specifically whereas such important parameters as the amount and type of graphics memory and the version of the PCI Express bus are given in small print. Considering that some of Nvidia’s G92-based cards have compatibility problems with PCI Express 1.0/1.0a mainboard, we guess the user should be informed about PCI Express 2.0 support beforehand.
As we noted in an earlier review, the graphics card box from XFX is so thick because the manufacturer wants to protect the card as much as possible against possible damage during transportation and storage. There is a thick layer of foam rubber with cutouts for the graphics card and accessories in the box. Another sheet of foam rubber covers the contents of the box from above. The following accessories are included with the card:
People at XFX must think cables and adapters a redundant feature for a gaming graphics card. There is no power adapter even. On the other hand, that’s not a big problem today when every modern PSU offers at least one 6-pin PCI Express 1.0 connector. The gaming positioning of the product is indicated by the funky bonus: an I am Gaming, Do Not Disturb sign for your door handle.
The user manual covers the installation procedure, offers recommendations on how to cool your system case properly, and contains information about the compatibility of mini-DIN connectors for analog video output.
As a special bonus, the box contains a copy of Lost Planet: Extreme Condition. The game is on the list of benchmarks we use and boasts good visuals. The extremely high system requirements of this application can be relaxed considerably if you lower some settings to the Medium level. This increases the frame rate greatly while having but a small effect on the quality and detailedness of the picture.
So, the XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition is packaged well, its box being appealing to the eye and protecting the card against possible damage. The accessories might be more numerous, though. We guess a standard adapter from two Molex plugs into one 6-pin PCI Express connector should have been put into the box.
GeForce 8800 GTS 512MB cards are shipped to Nvidia’s partners in ready-made form, being copies of the reference card. The XFX card is no exception, the sticker on the cooler’s casing being the single difference from the Leadtek WinFast PX8800 GTS 512MB.
The GeForce 8800 GTS 512MB uses a unified PCB design originally developed for the GeForce 8800 GT. The elements of the fourth phase of the power circuit, which were missing on the GeForce 8800 GT, are installed now. The power circuit is based on two controllers: the Primarion PX3544 is responsible for the graphics core and the Intersil ISL6549CBZ is responsible for memory chips. The card carries one 6-pin PCI Express 1.0 connector with a load capacity of 75W. It’s more than enough for the G92 chip that is manufactured on 65nm tech process and has modest power requirements.
The card is equipped with GDDR3 memory (Qimonda HYB18H512321BF-10). The chips are rated for a voltage of 2.0V and a frequency of 1000 (2000) MHz. The card maker didn’t overclock the memory much. It is clocked at 986 (1972) MHz even on the XXX version of the card. Thus, the described graphics card has almost the same memory bandwidth as Nvidia’s reference card. It’s 63.1GB/s to be exact. The card is thus inferior to the GeForce 8800 GTX with its 86.4GB/s in this respect, and memory bandwidth may become a limiting factor for its performance at high resolutions, especially with enabled full-screen antialiasing.
The graphics core is overclocked more: the main domain frequency is 678MHz, and the shader domain frequency is 1700MHz. The core doesn’t have disabled subunits and incorporates 128 unified shader processors (grouped into 8 large modules), 32 TMUs and 16 ROPs. With such parameters the XFX is going to outperform the Nvidia GeForce 8800 GTX across most applications except when you use a high resolution together with 4x FSAA.
The card has two dual-link DVI-I ports and a mini-DIN connector for analog video output. Unlike certain models of GeForce 8800 GT, the card cannot send S/PDIF audio into HDMI. The MIO connector is for SLI technology, but SLI mode is not yet supported well under Windows Vista.
The XFX card carries a new reference cooler developed for the GeForce 8800 GTS. It is based on the most efficient design that used to be installed on Nvidia’s GeForce 8800 GTS 640/320MB. The cooler consists of a large heatsink connected to a GPU-contacting copper base with three heat pipes. The heatsink is cooled with a Delta BFB1012L fan (5.8W). The low heat dissipation of the G92 chip combined with the highly effective heatsink allow to keep the fan speed within reasonable limits. The cooler is silent. The hot air is exhausted out of the system case through the slits in the mounting bracket.
The card has the same thermal interfaces as Nvidia’s reference samples: dark-gray thermal grease for the GPU and fabric pads socked in white thermal grease for the memory chips and power circuit elements. The cooler is fastened securely; no misalignment is possible. We guess the cooler is virtually free from drawbacks. The dual-slot form-factor may be considered as such, but it has long become standard for top-performance graphics cards.
As usual, the card is loaded in 3D mode by the first SM3.0/HDR test from 3DMark06 launched in a loop at 1600x1200 with 4x FSAA and 16x AF. In the Peak 2D mode the card performed the 2D Transparent Windows test from PCMark05.
Click to enlarge
The peak power consumption of the GeForce 8800 GTS 512MB doesn’t reach 100W, which is impressive in comparison with the GeForce 8800 GTX. It is the external power connectors that is the most loaded in this case.
The card is not very economical in idle mode, being inferior to the ATI Radeon HD 3870 with its more advanced power management techniques. No wonder Nvidia is currently working on a technology to make up for the higher power draw of its GPUs. We discussed that technology, called HybridPower, in our GeForce 9800 GX2 review. It should also be noted that XFX’ factory overclocking hasn’t affected the results much. The peak power draw is still below 110W even.
We measured the level of noise produced by the card with a digital sound-level meter Velleman DVM1326 using A-curve weighing. The level of ambient noise in our lab was 36dBA and the level of noise at a distance of 1 meter from the working testbed with a passively cooled graphics card inside was 43dBA.
There is almost no difference from the reference card, though. The XFX card is silent in every mode, the fan speed being no higher than 800rpm even under load. The GPU temperature varies from 47-58°C in idle mode to 85-86°C under heavy load in 3D applications. These showings are acceptable for the G92. The card has no effect on the temperature of other system components because the hot air is driven out of the system case.
We tried to overclock the card without additional cooling and managed to increase the GPU frequency from 678MHz to 760MHz. The shader domain frequency grew from 1700 to 1906MHz. The memory chips were stable at 1100 (2200) MHz, which was quite good for the Quimonda chips rated for a frequency of 1000 (2000) MHz. Our sample of XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition worked normally at such frequencies, yet we don’t recommend you to use it in such mode. Setting the frequencies higher than the manufacturer’s 687/1700/1972MHz may render the card unstable. It’s no good having your system hang up due to a failure of the overclocked graphics card, especially at a crucial moment of the game.
We didn’t have any compatibility problems with the XFX card. It started up successfully on every PCI Express 1.0a mainboard we tried it with.
To test the performance of XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition in games we assembled the following standard test platform:
According to our testing methodology, the drivers were set up to provide the highest possible quality of texture filtering and to minimize the effect of software optimizations used by default by both: AMD/ATI and Nvidia. Also, to ensure maximum image quality, we enabled transparent texture filtering - Adaptive Anti-Aliasing/Multi-sampling for ATI Catalyst and Antialiasing – Transparency: Multisampling for Nvidia ForceWare. As a result, our ATI and Nvidia driver settings looked as follows:
For our tests we used the following games and benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
We selected the highest possible level of detail in each game using standard tools provided by the game itself from the gaming menu. The games configuration files weren’t modified in any way. The only exception was Enemy Territory: Quake Wars game where we disabled the built-in fps rate limitation locked at 30fps.
Games supporting DirectX 10 were tested in this particular mode. With a few exceptions, the tests were performed in the following most widely spread resolutions: 1280x1024/960, 1600x1200 and 1920x1200. If the game didn’t support 16:10 display format, we set the last resolution to 1920x1440. We used “eye candy” mode everywhere, where it was possible without disabling the HDR/Shader Model 3.0/Shader Model 4.0. Namely, we ran the tests with enabled anisotropic filtering 16x as well as MSAA 4x antialiasing. We enabled them from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia ForceWare drivers
Performance was measured with the games’ own tools and the original demos were recorded if possible. Otherwise, the performance was measured manually with Fraps utility version 2.9.1. We measured not only the average speed, but also the minimum speed of the cards where possible.
For the sake of more illustrative analysis we have also included the following graphics cards to participate in this test session:
This game doesn’t support display resolutions of 16:10 format, so we use a resolution of 1920x1440 pixels (4:3 format) instead of 1920x1200 for it.
The XFX card is but slightly ahead of the GeForce 8800 GTX, yet it’s a good start nonetheless. The reference card’s performance also suggests that G92-based solutions in full configuration can be competitive against flagship G80-based products.
BioShock doesn’t support FSAA when running in Windows Vista’s DirectX 10 environment. That’s why we benchmarked the cards without FSAA.
The GeForce 8800 GTS 512MB enjoys a bigger advantage over the GeForce 8800 GTX than in the previous case, especially at low resolutions where it amounts to 20%. The gap is getting closer as the display resolution grows up. At 1920x1200 it is 8-9% with the XFX card and less than 5% with the reference card. As you can see, overclocking the GeForce 8800 GTS 512MB has little practical value for BioShock.
The XFX card is competitive to the GeForce 8800 GTX at 1280x1024, but the high memory requirements of Call of Juarez DX10 Enhancement Pack show up at 1600x1200 already. The GeForce 8800 GTX is still strong enough to deliver a high frame rate, but the resources of both versions of GeForce 8800 GTS 512MB are exhausted and they fall behind the ATI Radeon HD 3870.
Overclocking provides a good increase in speed for the GeForce 8800 GTS 512MB in this game, but it has little practical value due to the very low frame rate. Even the most advanced of single graphics cards cannot cope with Call of Juarez DX10 Enhancement Pack at the highest level of detail and enabled FSAA.
The XFX is not even 1fps faster than the reference card in some applications. This difference is too negligible for practical gaming, of course. So, the XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition, Nvidia GeForce 8800 GTS 512MB and Nvidia GeForce 8800 GTX are peers in this test ensuring full comfort even if you play at a resolution of 1920x1200.
The game being too hard at its Very High level of detail, we benchmarked the cards without FSAA to get a more playable speed.
FSAA being disabled, the game’s memory requirements are only really high at 1920x1200 which makes the GeForce 8800 GTX the leader. At the lower resolutions the ex-flagship is no better than the reference GeForce 8800 GTS 512MB and is slightly slower than the pre-overclocked version from XFX. There is no playing comfort as the average frame rate is as low as 18-20fps.
The frame rate is fixed at 30fps in this game as this is the rate at which the physical model is being updated at the server. Thus, this 30fps speed is the required minimum for playing the game.
The increased clock rates of the XFX card give it no special advantage in this test. The performance growth is insignificant by itself, about 4-6%. This is enough for the card to become the winner, though. On the other hand, the game is just as enjoyable on Nvidia’s GeForce 8800 GTX and GeForce 8800 Ultra cards.
Like Battlefield 2142, this game doesn’t support resolutions of 16:10 format and we use 1920x1440 (4:3 format) instead of 1920x1200.
The XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition goes neck and neck with the GeForce 8800 GTX. It’s only at 1920x1200 that the XFX card is somewhat ahead, outperforming the opponent by 9-10%. The latter has a better minimum of speed, however. Overclocking gives no particular advantage to the GeForce 8800 GTS 512 because even the reference version runs Perseus Mandate at a comfortable speed in every display mode including 1920x1440.
Both versions of GeForce 8800 GTS 512MB, the reference and the pre-overclocked ones, reach the speed ceiling set by the configuration of our testbed at 1280x1024. The GeForce 8800 GTX does the same, though. The XFX card is superior at the higher resolutions, being slightly ahead of the reference card which in its turn is closely followed by the GeForce 8800 GTX. Every card delivers comfortable performance across all the display modes. The gameplay seems comfortable even at 1920x1200.
The game doesn’t support FSAA when you enable the dynamic lighting model, but loses much of its visual appeal with the static model. This is the reason why we benchmarked the cards in S.T.A.L.K.E.R. using anisotropic filtering only.
The Nvidia GeForce 8800 GTX wins low resolutions and is just as fast as the XFX card in high display modes. The reference GeForce 8800 GTS 512MB is close behind the two leaders, all of them delivering superb performance at every resolution. The frame rate is always above 25fps even at 1920x1200, making it easy to take your aim accurately.
Forcing FSAA from the graphics card’s driver doesn’t produce any effect as yet. That’s why the game is tested with anisotropic filtering only.
There is no definite leader among the top-performance graphics cards. Both G92-based products and the G80-based ex-flagship reach the speed ceiling at resolutions up to 1920x1200. The XFX card and the GeForce 8800 GTX have nearly identical results at 1920x1200. You are guaranteed to have comfortable playing conditions, the worst result being as high as 70fps.
The XFX card and the GeForce 8800 GTX reach the 25fps mark at 1280x1024 yet cannot overcome it. It means the game is still unplayable at the highest level of detail. The GeForce 8800 GTX is ahead at the higher resolutions thanks to its more advanced memory subsystem.
The XFX card is 12 to 14% faster than the Nvidia GeForce 8800 GTX at high resolutions in terms of average performance but the minimum speed is the same, which means the same level of playing comfort. The frame rate is high enough for this game genre. Like in many other tests, the GeForce 8800 GTS 512MB doesn’t speed up much when overclocked.
The XFX card is faster the GeForce 8800 GTX at low resolutions but the latter goes ahead starting from 1600x1200 thanks to its larger amount of graphics memory. As you may know from our reviews, Hellgate is most sensitive to this parameter. Anyway, the gameplay is perfectly comfortable at 1280x1024 but you have to watch for the barely comfortable minimum of speed at 1600x1200.
The current version of the game doesn’t support FSAA, so we performed the test with anisotropic filtering only.
Both versions of GeForce 8800 GTS 512MB are ahead in every resolution but they are no different from the GeForce 8800 GTX when it comes to practical gaming. In other words, you don’t have to replace your existing graphics card, but if you are buying a new one, you should choose the G92 core.
The game loses much of its visual appeal without HDR. Although some gamers argue that point, we think TES IV looks best with enabled FP HDR and test it in this mode.
The GeForce 8800 GTX beats both versions of GeForce 8800 GTS 512MB in one case only: at 1920x1200 in open scenes. This must be due to the 384-bit memory bus. In the other test modes the ex-flagship of the GeForce 8800 series is slower, even though maintains a high frame rate still.
The new add-on to Company of Heroes is tested in DirectX 10 mode only since it provides the highest quality of the visuals.
The overclocked frequencies help the XFX card keep up with the GeForce 8800 GTX at resolutions higher than 1280x1024, but both have a very low minimum of speed. We recommend you to stick to 1280x1024 if you don’t want to disable FSAA.
The game having a frame rate limiter, you should consider the minimum speed of the cards in the first place.
The frame rate limiter doesn’t allow this test to show a winner. All the cards are equally good for running Command & Conquer 3.
The increased clock rates give no advantage to the XFX card over the reference card from Nvidia. The G92-based solutions are both inferior to the GeForce 8800 GTX in terms of average performance but have a higher minimum of speed, at least at 1280x1024. You can try to play the game at that resolution with FSAA enabled, but you may find your gaming experience not very enjoyable.
There is almost no effect from the overclocked frequencies of the XFX card. It scores a mere 51 points more than the Radeon HD 3870.
You can see there is but a small performance gain from the pre-overclocked frequencies of the XFX card. The biggest gain is observed in the second and third tests at 1280x1024, which is due to the modest memory frequency growth relative to the reference GeForce 8800 GTS 512MB. The individual tests agree with the overall scores quite well.
The XFX card scores 376 points more than the reference sample. Not much, but let’s check out the detailed picture.
It is in the SM2.0 tests that we see the biggest performance growth, about 6%. It must be due to the first test which is very sensitive to the main domain frequency of Nvidia’s G8/G9 series processors. The performance is increased less in the SM3.0/HDR tests, by 2.5-3% only.
The results of the individual SM2.0 tests confirm our supposition that there is a performance growth in the first test. There are almost no performance benefits in the second test.
The XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition enjoys about the same advantage over the reference card in both of the SM3.0/HDR tests. The Nvidia GeForce 8800 GTX remains the leader of the second test, probably due to its more advanced memory subsystem.
The XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition left a positive impression overall, but that would be true for any other sample of GeForce 8800 GTS 512MB because there is virtually no difference between XFX’ product and Nvidia’s reference card except for the sticker on the cooler and the increased clock rates.
The pre-overclocked frequencies do not lead to a serious performance growth because the memory frequency is not overclocked much above the reference card’s. The performance gain varies from 5 to 8% and brings no practical benefits. On the other hand, the XFX card is indeed the leader in most of our tests, outperforming or at least matching the GeForce 8800 GTX. The ex-flagship is still strong in those applications that require a large amount of graphics memory and/or high memory bandwidth. These are mostly DirectX 10 applications such as Call of Juarez, Crysis, Lost Planet, Hellgate: London and World in Conflict but they are too heavy for the GeForce 8800 GTX to provide tangible benefits in them.
The XFX GeForce 8800 GTS 512MB XXX Alpha Dog Edition is unlikely to be found selling for Nvidia’s recommended price of $349 or lower because it belongs to the premium XXX series. Anyway, a new Radeon HD 3870 X2 will cost you considerably more money but will not deliver maximum performance in every game due to its dual-chip architecture. The new GeForce 9800 GTX cannot be much better, either, because its memory frequency is a mere 200MHz higher and its GPU frequencies are the same as those of the XFX card.
So if you are looking for a graphics card capable of delivering the performance of a GeForce 8800 GTX, you can choose the described product from XFX. It won’t disappoint you. But if you’ve already got a GeForce 8800 GTX, you shouldn’t worry at all: the pre-overclocked GeForce 8800 GTS 512MB may be faster in some cases, but not faster in general. As for the ATI Radeon HD 3870 X2, it cannot yet provide its maximum performance in all games.