by Alexey Stepin , Anton Shilov
02/26/2006 | 05:17 PM
Multi-GPU technologies, including the mature scalable link interface (SLI) from Nvidia Corp. and CrossFire technology from ATI that only begins to gain market acceptance, are intended for relatively small group of enthusiasts who desire tomorrow’s performance today. The enthusiasm towards high-peformance multi-GPU systems is akin to interest in luxurious sports cars: hundreads dream of Ferraris, but end up with Fiats.
Graphics chip companies present multi-GPU technologies as a way to get next-generation speed out of the currently existing hardware and as an option of upgrading later without need to get rid of the current graphics card. Both theories have a right to exist, but you should remember that future graphics cards will bring in a new feature-set; while in case you won’t sell your graphics board tomorrow, but install a similar one, you’ll need to sell already two boards the day after tomorrow.
Indeed, multi-GPU is generally needed for high-end computers, in fact, they do bring a lot of benefits for their owners: ability to play in high resolutions, ability to enable more advanced full-screen anti-aliasing methods compared to single-GPU setups, generally higher speed and some others. Keeping in mind both these advantages as well as somewhat debatable ritorics of graphics chip suppliers, it is clear that there is a need for ultra high end über-performance graphics solutions for those who demand it at any cost.
For about a year Nvidia’s SLI was the only option for those who would to have the highest speed: the Radeon X850 CrossFire arrived too late and was still slower compared to Nvidia’s offerings, the Radeon X1800 CrossFire arrived just weeks before the Radeon X1900 and also had some performance issues when compared to the highest-end dual-GeForce 7800-series setups. Today we try to find out whether two of the Radeon X1900 XT graphics cards deliver the unbeatable performance and really represent a too fast and too furious tandem.
Everybody was amazed with 16 pixel processors of the GeForce 6800 Ultra and the Radeon X800 XT a little bit less than two years ago: performance increases the new chips brought were just in time to enable smooth speed in the new breed of major titles: Far Cry, Doom III and Half-Life 2.
The new Radeon X1900-series has delivered an astonishing raise of the number of pixel shader processors: the new chips feature 48 of them. A question that has been discussed in our Radeon X1900 XTX review was whether we really need 48 pixel shader processors amid only 16 texture units. Even in synthetic benchmarks, for example, in pixel shader 2.0 test from the 3DMark03 the speed of the Radeon X1900 is limited by the speed of texture fetches; performance of Serious Sam 2 game was also limited by the same factor.
While the Radeon X1900-series does have some future-proof because of its astonishing arithmetic power, some titles that the gamers play today do not take much advantage of the newcomer. Those, who acquire a pair of the Radeon X1900 family products, should understand clearly that in some games performance increases over the previous generation CrossFire will not be high due to limitations in texture fetch performance, however, in games that utilize math1ematically intensive pixel shaders the new duo should offer much higher scaling.
There were quite some games where the Radeon X1900 XTX did not offer significantly improved performance over the predecessor – the Radeon X1800 XT. In those cases, the difference between the Radeon X1800 XT CrossFire and the Radeon X1900 XT CrossFire will be evem smaller. Moreover, when acquiring the Radeon X1900 XT CrossFire system you should make sure you have a processor fast enough to avoid limitations in performance in CPU-bound cases (future games will demand a rapid microprocessor anyway).
So, what should you generally expect from the Radeon X1900 XT CrossFire, keeping in mind that there is virtually one tangible improvement over the Radeon X1800 XT the new Radeon X1900 XT has: increased amount of pixel shader processors? We believe that the following:
A funny thing about 96 – or nearly 100, an unbelievable number a couple of years ago – pixel shader processors is that it took ATI two graphics chips to assemble such an array, whereas Nvidia Corp. will use “96 pixel processors” moniker to boost up interest to 4-way SLI systems coming out this spring.
The third incarnation of the ATI CrossFire multi-GPU platform is very similar to the second one, based on the Radeon X1800. Customers have to use either Radeon Xpress CrossFire Edition (RD480)-based mainboard, or Intel 955/975-based chipset platforms that have two PCI Express x16 slots to build a CrossFire system. Additionally, in case you are building a Radeon X1800 or X1900 dual-GPU system, you will need one appropriate CrossFire Edition graphics card with compositing engine.
ATI’s CrossFire platform allows to use graphics cards clocked differently and with various memory configuration. This means that users will be able to use Radeon X1900 XTX, Radeon X1900 XT or even lower-clocked Radeon X1900-series graphics card along with the Radeon X1900 XT CrossFire Edition board. Theoretically, graphics cards will continue to operate at their default frequencies, unless the differences are significant. It should be noted that it is not possible to assemble a multi-GPU array consisting of one Radeon X1800 and one Radeon X1900.
There are no differences between the working principles of the Radeon X1900 CrossFire platform compared to earlier Radeon X1800 incarnation: those who are not familiar with the matter should refer to the previous article called Caught in the Cross-Fire: ATI RADEON X1800 XT CrossFire Edition Review .
The multi-GPU modes also remained the same:
As previously, the CrossFire rendering mode is activated automatically if the Catalyst A.I. feature is turned on. Generally, SuperTiling was designed for Direct3D, while Scissor was developed for OpenGL application programming interface (API). AFR is used only when there’s an appropriate application profile in Catalyst. If Catalyst A.I. is disabled, SuperTiling is used for Direct3D games and Scissor in all other cases. The Super AA mode is automatically activated as soon as 8x/10x/12x/14x FSAA is turned on.
Those who would like to have a closer look at the Radeon X1900 CrossFire Edition board can do it in our Radeon X1900 XTX review and also find some more details in the Radeon X1800 XT CrossFire-related article . Keep in mind that the CrossFire Edition boards do not have TV-outs, hence, it does not make sense to acquire them for multimedia-oriented systems if you want to use conventional cables. Nevertheless, if you have DVI or HDMI supporting TV, you will be able to output onto a big screen.
We tested the performance of the ATI Radeon X1900 CrossFire series on this platform:
Since only nForce4 chipsets officially support Nvidia SLI, the results for two GeForce 7800 GTX SLI graphics cards were obtained on a different testbed:
We set up the ATI and Nvidia drivers in the same way as always:
ATI Catalyst:
Nvidia ForceWare:
We select the highest graphics quality settings in each game, identical for graphics cards from ATI and Nvidia, except for the flight sim Pacific Fighters that requires vertex texturing for its SM 3.0 rendering mode. We do not edit the configuration files of the games. If possible, we use the in-game benchmarking tools to record and reproduce a demo and then measure the reproduction speed in frames per second. Otherwise we measure the frame rate with the Fraps utility. If it is possible, we measure minimal as well as average fps rates to give you a fuller picture.
We turn on 4x full-screen antialiasing and 16x anisotropic filtering in the “eye candy” test mode from the game’s own menu if possible. Otherwise we force the necessary mode from the graphics card driver. If possible, we use the game’s integrated benchmarking tools to record and reproduce a demo and measure the reproduction speed in frames-per-second, and if not, we measure the frame rate with the Fraps utility. We measure minimal as well as average fps rates whenever possible to draw a fuller performance picture.
Unfortunately, we didn’t have a second sample of GeForce 7800 GTX 512 and had to use a SLI configuration with two ordinary GeForce 7800 GTX as an opponent to the RADEON X1900 XT CrossFire system.
We tested the following single cards in 4x FSAA + 16x AF mode for the sake of comparison:
These games and applications were used as benchmarks:
3D Shooters with First-Person View
3D Shooters with Third-Person View
Simulators
Strategies
Semi-synthetic benchmarks
Synthetic benchmarks
Like with the Radeon X1800 XT CrossFire, we couldn’t make the game run in extreme full-screen antialiasing modes, so there are no numbers for Super AA/SLI AA 8x/14x/16x.
The traditional “eye candy” mode is rather too easy here to become a serious trial for modern multi-GPU configurations.
In 1600x1200 the Radeon X1900 XT CrossFire platform is somewhat faster than the Radeon X1800 XT CrossFire, but the performance seems to be limited by the texture fetch speed. This small advantage, however, confirms the status of the Radeon X1900 XT CrossFire as of the highest-performing multi-GPU graphics subsystem among the tested ones.
The Radeon X1900 XT brings about a noticeable performance increase over the Radeon X1800 XT CrossFire, but it is also clear that the speed of this game doesn’t scale up too readily on the Radeon X1000 architecture.
Note that there’s a very small gain over the single graphics card within the Radeon X1900 series. This is expectable, though, since The Chronicles of Riddick uses the OpenGL API and stencil shadows rather than the shadow mapping technique.
Note also that the multi-GPU configurations as well as the single GeForce 7800 GTX 512 all ensure a comfortable frame rate in this game.
The new solution is on top in the Super AA 8x mode, almost making 1280x1024 resolution playable. The speed gain over the Radeon X1800 XT CrossFire isn’t too big here, again, since the clock rates of the two CrossFire platforms differ but very little.
Not too many changes as we switch over to the more resource-consuming Super AA 14x mode. Why? It is the memory subsystem load that is higher here, and the Radeon X1800 XT CrossFire and the Radeon X1900 XT CrossFire have near the same memory subsystem performance. The resolution of 1024x768 pixels is available for play, but would you want to use it on an ultra high-end gaming computer which is sure to be equipped with a 19-21” monitor?
Using 4x FSAA we again see both the CrossFire platforms delivering the same speeds in the first two resolutions; in 1600x1200 the Radeon X1900 XT CrossFire is slightly ahead of the older platform. Call of Duty 2 just can’t load all the 96 pixel processors of the new multi-GPU config as fully as to conceal the insufficient texture fetch speed.
Having much humbler technical characteristics, the Radeon X1800 XT CrossFire system is not, however, much slower than the new Radeon X1900 XT CrossFire. Both the multi-GPU configurations from ATI Technologies ensure a comfortable speed in resolutions of 1024x768 and lower.
What we’ve just said above can be applied to the results of the Super AA 14x mode, too. The newer CrossFire is a mere 10-11% faster than the older platform in the hardest test mode. This is not enough to yield a playable frame rate in 1600x1200 resolution; like in the previous case, you can play the game comfortably in 1024x768 only.
Note also the extremely low performance of the SLI configuration with two GeForce 7800 GTX 256MB cards. If you are going to assemble one, you should be aware that you will have to reduce the display resolution or disable some special effects to achieve an acceptable game speed when an extreme level of full-screen antialiasing is turned on.
The Radeon X1900 XT CrossFire proves its superiority in 1600x1200 resolution only, yet even then it evidently has a lower efficiency than the Radeon X1800 XT CrossFire. This is likely to be due to insufficient texture fetch speed and/or too low memory bandwidth.
It’s similar in the harder antialiasing modes: the Radeon X1900 XT CrossFire doesn’t have much advantage over the Radeon X1800 XT CrossFire in resolutions lower than 1600x1200, but the performance gain is less than 10fps even in the latter case. That’s not enough to make that resolution playable. It’s the same with SLI AA and the GeForce 7800 GTX SLI, though.
Can the 96 pixel processors make a difference in games that make a wider use of long and complex shaders?
The CrossFire platforms both reach easily the maximum possible speed on this level and have almost the same speed thereafter, indicating insufficient texture fetch speed or insufficiently high memory bandwidth.
This is not a big problem here, though. You can play in any of the standard resolutions with turned-on 14x Super AA on the Radeon X1800 or X1900 CrossFire. The GeForce 7800 GTX SLI doesn’t permit this.
The advantages of the newer graphics subsystem from ATI are more prominently outlined on the Research map, especially in high resolutions. Yet even here, with more complex shaders to be executed, the performance gain is too small to justify an upgrade of your current Radeon X1800 XT CrossFire configuration.
Note that if you don’t plan to use Super AA, it may be more profitable for you to purchase a single Radeon X1900 XTX than two Radeon X1800 XT for a multi-GPU config. These two solutions differ but little in their performance, but a single Radeon X1900 XTX costs much less and doesn’t require a CrossFire-compatible mainboard and also consumes less power.
ATI CrossFire technology is not as efficient here as Nvidia SLI. Moreover, the Radeon X1900 XT CrossFire is slower than the single Radeon X1900 XTX in the first two resolutions! We can’t find a good reason for this behavior since the Radeon X1800 XT CrossFire ensures a small but stable performance gain in comparison with the single Radeon X1800 XT. The two graphics cards may have got out of sync due to the difference in the clock rates of the Radeon X1900 XTX and Radeon X1900 XT CrossFire Edition.
It’s all different in the Super AA 8x mode. The Radeon X1900 XT CrossFire has a slightly worse minimum speed than the Radeon X1800 XT CrossFire, but has a much higher average frame rate, too. The speed gain varies from 20 to 60%. The resolution of 1280x1024 pixels seems to be playable, but the speed may bottom out to 21fps, so you may have some slowdowns even playing the game on a CrossFire pair of Radeon X1900 cards.
The Radeon X1900 XT CrossFire is over 40% faster than the previous multi-GPU platform, Radeon X1800 XT CrossFire, even in 1024x768 resolution. The average performance of about 60fps in 1024x768 allows playing F.E.A.R. comfortably with enabled 14x Super AA which is an achievement indeed.
On the other hand, the min frame rate is no higher than 23fps and, moreover, this low resolution will hardly be called for in a gaming computer of that class. Anyway, these results make us hope the 96 pixel processors may bring about a considerable performance gain in shader rich applications. In other words, the future perspectives of the Radeon X1900 XT CrossFire look bright indeed.
Half-Life 2 is not as hard a trial for modern graphics hardware as it used to be for previous-generation GPUs, so your using two Radeon X1900 in multi-GPU mode can hardly give you any advantage over the single Radeon X1900 XTX or over the Radeon X1800 XT CrossFire. The low texture fetch speed affects the results again.
In high resolutions, with extreme antialiasing modes enabled, the new multi-GPU system from ATI Technologies is faster than the older one, but the difference is not bigger than 10-11fps, the average frame rate being near 75-80fps.
The tech demo Half-Life 2: Lost Coast abounds in advanced SM3.0/HDR special effects and is hard enough even for the Radeon X1900 XT CrossFire, at least when you use extreme full-screen antialiasing modes.
Regrettably, this super-powerful graphics subsystem is still not powerful enough to make resolutions other than 1024x768 playable. But if you compare the Radeon X1900 XT CrossFire with its predecessors and competitors, you may note its excellent scalability which again confirms our supposition of the hefty performance reserve the new implementation of ATI’s multi-GPU technology has for upcoming applications.
The problem we described in our two previous reports on CrossFire still persists. This game still doesn’t like multi-GPU technologies and suffers a performance hit as soon as you try to use them.
The Super AA and SLI AA modes (when the additional GPU is only used to improve the antialiasing quality) work correctly, although we can’t see any speed growth over the Radeon X1800 XT CrossFire. The gap between the two versions of ATI CrossFire is less than 1fps at the biggest.
Like in many cases above, the Radeon X1900 XT CrossFire is only any better than the older platform in 1600x1200 resolution and when extreme antialiasing modes are turned on.
As you see, the balance between the TMU performance and pixel processor performance in the GeForce 7800 GTX is more optimal for today’s games than what Radeon X1900 XTX has to offer. Unfortunately, the low texture fetch speed limits the performance of the new solution greatly in a number of games, including Quake 4 .
Shaders in Serious Sam 2 contain a lot of texture fetches (up to 8) which is one of the main reasons why the game is predisposed to the GeForce 7 architecture. The computational pixel processors of Radeon X1800 and X1900 cards are just idle most of the time here. As a result, the Radeon X1900 CrossFire is only faster than the single GeForce 7800 GTX 512 in 1024x768 and 1280x1024 resolutions. You could bet safely a SLI configuration with two GeForce 7800 GTX 512 would leave no chance to the new solution from ATI in this test.
The gap between the Radeon X1900 XT CrossFire and the Radeon X1800 XT CrossFire doesn’t grow up in higher resolutions as in the previous tests, but on the contrary, gets narrower. With a min speed of 25fps and an average speed of 55fps, the Radeon X1900 XT CrossFire allows playing the game comfortably in 1024x768. This is quite a nice result, considering the complexity of the application.
The average performance of the Radeon X1900 XT CrossFire drops but slightly, but the min speed is below 25fps, making it uncomfortable to play with enabled 14x Super AA. The Radeon X1800 XT CrossFire is less than 10% slower than the newer platform.
There are still problems with the performance of Radeon X1900/X1800 in Unreal Tournament 2004 . This is especially conspicuous in resource-consuming FSAA modes, probably due to the low texture fetch speed again.
We also suspect the SLI AA 16x mode works incorrectly in this game as the results in this mode don’t almost differ from those with SLI AA 8x.
The central processor limits the speed of the whole system in low resolutions, but in 1600x1200 the Radeon X1900 XT CrossFire is only 10% ahead of the Radeon X1800 XT CrossFire although the game uses SM3.0 and HDR.
The new CrossFire pair enjoys the same or slightly bigger advantage in the Super AA 8x mode. The resolution of 1600x1200 seems to be playable now with more comfort – this is the only tangible effect of the 96 pixel processors here.
The same is true for the Super AA 14x mode: the threefold increase of the pixel processors yields a mere 10-12% of performance growth even in a game which is rich in advanced pixel shaders.
Note that the GeForce 7800 GTX SLI configuration ensures a good speed in the SLI AA mode in 1280x1024 resolution and an excellent speed in 1024x768.
The Radeon X1900 XT CrossFire is almost as fast as the Radeon X1800 XT CrossFire in nearly all the modes and resolutions even though this game is favorable towards ATI’s Radeons. Well, all the virtual racers should anyway rejoice as CrossFire technology ensures an excellent performance everywhere in this game. The GeForce 7800 GTX SLI looks well with 8x antialiasing in resolutions up to 1280x1024 and offers a playable speed in 1024x768 with 16x antialiasing.
The main performance-limiting factors for the Radeon X1900 CrossFire are its texture fetch speed and memory bandwidth which are the same as those of the Radeon X1800 CrossFire.
The performance of ATI Radeon based graphics cards is not the most optimal in this game because the engine of Pacific Fighters was originally developed to fit well the architectural features of GeForce 6 and 7 family cards. It uses the vertex texture sampling and the textures are sampled frequently, so the GeForce 7800 GTX and 7800 GTX 512 deliver the best performance here.
The GeForce 7800 GTX SLI is on top until the Super AA 14x/SLI AA 16x mode where it loses its ground due to the complexity of this antialiasing method and the low graphics memory frequency. The Radeon X1900 XT CrossFire limits comfortable play to 1024x768 with enabled 14x Super AA.
CrossFire technology couldn’t produce correct results in this game in the Super AA mode, so we don’t publish them here.
The efficiency of the Radeon X1900 XT CrossFire configuration grows up along with the display resolution. There’s almost no advantage over the single Radeon X1900 XTX in 1024x768, but then it amounts to as much as 50% in 1600x1200. The GeForce 7800 GTX SLI is thus left behind by 2fps.
A SLI configuration with two GeForce 7800 GTX 512 would probably be faster than the Radeon X1900 XT CrossFire, but we can’t yet check this out as we don’t have a second sample of that card.
The Radeon X1900 XT CrossFire is quite noticeably behind the Radeon X1800 XT CrossFire – by about 10%. We don’t have a clear explanation but we suspect it is due to desynchronization between the Radeon X1900 XT CrossFire Edition and the Radeon X1900 XTX. As for performance, the GeForce 7800 GTX SLI has better results than the Radeon X1800 XT CrossFire due to its UltraShadow technology and more efficient processing of textures.
The results of the Radeon X1800 XT CrossFire and the Radeon X1900 XT CrossFire almost coincide. Although the multi-GPU systems from ATI are slightly ahead of the GeForce 7800 GTX SLI in the extreme antialiasing modes, the latter ensures a higher minimum frame rate, so the player runs a lower risk of experiencing slowdowns during play.
There’s a somewhat bigger gap between the two CrossFire systems in the Super AA 14x mode. They are both ahead of the Nvidia solution, but the latter offers a higher minimum frame rate even though works in the more resource-consuming SLI AA 16x mode. The 29-44fps range provided by the two GeForce 7800 GTX is better to have in this case than the 17-49fps range from the Radeon X1900 XT CrossFire.
The scalability problems with Radeon X1800 CrossFire and Radeon X1900 CrossFire must be due to insufficiently fast texture fetches as they have only 32 texture-mapping units against 48 in the GeForce 7800 GTX.
Aquamark3 is not a representative test anymore when it comes to ultra-high-end graphics solutions, especially multi-GPU ones. That’s why it’s normal that both the CrossFire platforms and the NVIDIA SLI deliver almost the same performance, with a minor advantage on the part of the former.
The lower-performing graphics memory subsystem is the reason why the GeForce 7800 GTX SLI loses to the Radeon X1800 XT and 1900 XT CrossFire in the antialiasing modes equivalent to 8x FSAA. The results of the two CrossFire platforms are roughly the same due to the identical texture sampling speed.
The same can be observed with the extreme antialiasing modes, but the performance of the GeForce 7800 GTX SLI is even lower here due to the use of true super-sampling.
As we noted in our Radeon X1900 XTX review , 3DMark03 is not a favorable environment for the new graphics card series from ATI to show their best in. The two Radeon X1900 cards working in CrossFire mode perform much alike to the single Radeon X1900 XTX. They are faster, but not very much faster, than the Radeon X1800 XT CrossFire system and stop 1613 points short of the 30,000 points mark.
The Radeon X1800 XT CrossFire and the Radeon X1900 XT CrossFire are equally fast throughout almost the entire first test which lacks any pixels shaders altogether. The higher clock rates of the Radeon X1900 XTX that plays the role of the Slave in the new CrossFire configuration mean nothing here because the Master card works at the standard frequencies of Radeon X1900 XT. If there’s any performance gain, it is fully negated by the asynchronous frequencies of the Radeon X1900 XT CrossFire Edition and the Radeon X1900 XTX.
The second test produces a similar picture: Shader Models 1.1 and 1.4 aren’t anything serious for modern graphics cards, so the GPU and memory frequencies are still the main performance-influencing factors, especially when extreme antialiasing modes are in use.
It’s somewhat different in the third test where the GeForce 7800 GTX SLI gains the lead in the 4x FSAA + 16x AF mode. This graphics subsystem is thereafter not as far behind the CrossFire platforms as in the previous case, probably because of a higher geometry-processing efficiency (the geometry of the third test scene is more complex than of the second one).
The Radeon X1900 XT CrossFire proves its superiority over the Radeon X1800 XT CrossFire in the fourth test which provides some room to showcase the increased math1ematical performance, even though the employed shaders aren’t too complex by today’s standards.
As a result, the Radeon X1900 XT CrossFire scores more points overall than the Radeon X1800 XT CrossFire mostly due to the results of the fourth game test. In the other tests the two CrossFire platforms have near the same performance.
The efficiency of the Radeon X1900 XT CrossFire relative to the Radeon X1900 XTX is much lower than that of the Radeon X1800 XT CrossFire. Perhaps 3DMark05 depends but too heavily on the CPU performance or the CrossFire systems are limited in speed by their low efficiency at processing numerous complex textures. Let’s see.
The CPU obviously sets a limit to speed in the 4x FSAA mode – the ceiling goes at about 40-41fps. In higher resolutions and in the Super AA/SLI AA modes the CPU-imposed limitations slacken a little, allowing the Radeon X1900 XT CrossFire to show its best and outperform the Radeon X1800 CrossFire by 20-30%.
The new graphics solution behaves in the same way in the second test, too, but unlike in the first one, it has a smaller advantage over the Radeon X1800 XT CrossFire in the extreme antialiasing modes, no more than 20%.
In the third test, which uses the longest and most complex shaders in 3DMark05, the Radeon X1900 XT CrossFire is 10% ahead of the Radeon X1800 XT CrossFire from the very beginning. The gap is never bigger than 30% in this test, though.
So, the overall score of the new graphics subsystem seems well deserved considering that 3DMark05 runs in 1024x768 without FSAA by default and both the CrossFire versions are limited by the CPU performance in this case even in the third and hardest test.
3DMark06 uses a number of very complex shaders of versions 2.0 and 3.0, shadow mapping and HDR to create the special effects in its graphics tests. Having 96 pixel processors is more than welcome here and the Radeon X1900 XT CrossFire has the highest overall score as well as the highest results in the SM2.0 and SM3.0/HDR tests taken separately. Note also that the GeForce 7800 GTX SLI is faster than the Radeon X1800 XT CrossFire in the SM2.0 tests despite the lower clock rates. The math1ematical performance of the pixel processors is not as important there as it is for the SM3.0/HDR tests.
512 megabytes is an insufficient amount of graphics memory to turn on Super AA 14x or SLI AA 16x in 3DMark06, so we didn’t perform the tests in this mode. In the other cases the Radeon X1900 XT CrossFire is an unrivalled leader, delivering a 10% to 40% performance gain over the Radeon X1800 XT CrossFire depending on the resolution.
The gap between the two CrossFire configurations doesn’t grow up that rapidly in the second SM2.0 graphics test, but the newer config is obviously better here, too. The Fetch4 feature helps it cope well with the complex shadows the test scene abounds in.
GeForce 7800 series cards don’t allow using FSAA and HDR simultaneously, so they are expelled from the SM3.0/HDR tests. We use 4x FSAA even in the easiest test mode during this test session.
The first SM3.0/HDR test features very complex pixel shaders and dynamic shadows, so it is not a wonder the new CrossFire solution enjoys the biggest advantage over the older one here – up to 50%.
Take note that the two CrossFire systems don’t differ much in 1024x768 resolution. The frame rate of 36fps seems to be the ceiling for the platform with an AMD Athlon 64 4000+ processor. In higher resolutions and with enabled 8x Super AA the Radeon X1800 XT CrossFire loses its ground quickly as it can’t compete with its 32 pixel processors against the 96 pixel processors of the Radeon X1900 XT CrossFire.
So, the overall superiority of the Radeon X1900 XT CrossFire in 3DMark06 is well grounded. This graphics subsystem is the best in all the four graphics tests in the default 1280x1024 resolution.
It is not a news that the CrossFire multi-GPU technology has matured significantly since its first introduction in Septermber, 2005, along with the Radeon X800-series graphics cards. The Radeon X1800 CrossFire already offered enough stability, quality, reliability and robustness, so, the main question to the Radeon X1900 CrossFire is generally about performance.
A pair of two Radeon X1900 XT CrossFire is among the fastest graphics cards configurations money can buy these days: it features insane performance in modern games, it can enable extreme full-scene antialiasing and even has some proof for the future games. If you do want the highest performance, quality and can afford it, getting the Radeon X1900 XT/XTX along with the Radeon X1900 XT CrossFire Edition is definitely an idea to consider: it is really a fast and furious combination!
Unfortunately, the Radeon X1900 CrossFire tandem has some drawbacks: very often its speed is limited by texture fetching performance and/or memory bandwidth. As a consequence, the majority of games today run at very similar speeds on the Radeon X1800 and Radeon X1900 CrossFire configuration. The disadvantage of the relatively slow memory speed is so significant that in a lot of cases the new duet of graphics accelerators does not outperform the old champion tandem: the Radeon X1800 CrossFire.
The main advantage of the Radeon X1900 CrossFire duo is definitely its orientation on the future: it already demonstrates high framerates in games where pixel shader contain a lot of math1ematical instructions. The upcoming games should take full advantage of the 96 pixel shader processors for sure.
In case we are talking about the market perspectives of the Radeon X1900 CrossFire, we should also note that there are more Nvidia SLI-supporting mainboards for AMD processors, which are the most popular among gamers these days, on the market compared to CrossFire-supporting platforms. The availability of SLI ecosystem along with the higher customer awareness about the technology allows Nvidia not to release any direct competitors for the Radeon X1900 CrossFire configuration shortly and do not lose much of sales to the rival. Nevertheless, we should point out that Nvidia has reasons to worry: CrossFire supporting systems are already available from makers of high-performance computers, such as Alienware and VoodooPC.
Highs:
Lows: