GeCube Radeon 9550 Platinum Edition
Benchmarking Results
Synthetic benchmarks
Following the tradition, we start with our substantially amended benchmarking package.
ShaderMark v2.0
As is well seen from the diagram, we observe a traditional lag of the NVIDIA board at this test. Version 2.0 HLSL shaders have never been a strong point of the NVIDIA GeForce FX chip architecture. This time, the situation simply confirmed once again. Also note the results for GeCube Radeon 9550 Platinum Edition and the Gigabyte's board on the base of the ATI Radeon 9600 Pro chip.
They are practically always identical. All is natural enough, in view of the same graphic core in the first and second cases and the completely identical operating clock speed of the graphic processors in both these boards - the memory operating frequencies play a less significant part: all depends on the shader processing speed of the graphic processors.
D3D RightMark
Geometry Processing Speed
This test allows assessing the speed at which the geometry is processed by the accelerator. The test allows to choose the following illumination models (computed at the vertex level):
- Ambient Lighting - simplest constant illumination
- 1 Diffuse Light
- 2 Diffuse Lights
- 3 Diffuse Lights
- 1 Diffuse + Specular Light
- 2 Diffuse + Specular Lights
- 3 Diffuse + Specular Lights
We used the most advanced mode with three diffuse-specular light sources in combination with three different operating modes: the traditional TCL (Fixed-Function Pipeline), version 1.1 vertex shaders and version 1.1 pixel shaders, version 2.0 vertex shaders and pixel shaders, as well as the simplest mode of ambient illumination is also in combination with the three operating modes: the traditional TCL (Fixed-Function Pipeline), vertex shaders 1.1 and pixel shaders 1.1, vertex shaders 2.0 and pixel shaders 2.0. In the case of Ambient illumination (the simplest constant illumination) and transformation we reach the practical limit for the card's bandwidth at processing the triangles.
 For the case of the complex mode, we used the most advanced mode with three diffuse-specular light sources in combination with three different operating modes: with the traditional TCL (Fixed-Function Pipeline), version 1.1 vertex shaders and version 1.1 pixel shaders, version 2.0 vertex shaders and version 2.0 pixel shaders 2.0 we get quite predictable and familiar results as per the previous materials. In the Fixed-Function Pipeline mode, the MSI board built on the NVIDIA GeForce FX 5700 chip is beyond competition. This is quite a natural victory of NVIDIA chips in that ATI chips simply have no hardware support for T&L emulation, but NVIDIA chips offer the implementation of hardware units responsible for light source computation. Boards on the base of ATI chips offer results that are the same and stable in practically all the operating modes (the results for both ATI Radeon 9550, and ATI Radeon 9600 Pro are absolutely identical because of the same clock speeds of the GPUs - 400 MHz for each of the boards). The board built on the NVIDIA GeForce FX 5700 chip takes a noticeable lead in the Fixed-Function Pipeline mode, but gradually loses victory points as the version number and complexity of the shader go up (as is well seen in the VS 2.0 static flow control (HLSL) mode, the NVIDIA chip already lags well behind).
 In the case of the Ambient illumination (the simplest constant illumination) and transformation mode where we get the practical limit for the card's bandwidth at handling the triangles, the results for both ATI cards and NVIDIA cards are more uniform, although the NVIDIA GeForce FX 5700 is more biased to the Fixed-Function Pipeline mode, as can be easily seen =). Again, ATI boards are leaders at these tests - they offer a higher limit of triangle bandwidth (albeit not so much greater than in NVIDIA boards).
Pixel FillingThis test performs a number of various tasks, but we were mostly interested in the possibility of measuring the performance of frame buffer filling. We used two schemes in this test: both with a 256x256 texture and without such.
For the case of measuring the performance during filling the buffer frame without a texture, we can ascertain a leadership for ATI boards, and the difference between ATI Radeon 9550 and ATI Radeon 9600 Pro can already be seen to the naked eye - the memory bandwidth which is different in these two boards because of the different memory operating speed came into play (it is higher in ATI Radeon 9600 Pro). In this test, the 4x1 architecture of NVIDIA GeForce FX 5700 evidently shows results that fully match the given configuration in the number of pixel processors and texture units, together with the chip clock speed. The peak values differ from the theoretically admissible by merely 40-50 mln pixels per second, which in view of the produced absolute values can be regarded as a minor error. However, traditionally in this test, ATI chips do not reveal their theoretically declared pixel fillrate bandwidth.
 For the case of texturing (1 texture of 256x256 in size), we get entirely different results rather than in the case without a texture. Now, MSI's board based on the NVIDIA GeForce FX 5700 chip takes a lead, while ATI boards still demonstrate varied results (ATI Radeon 9600 Pro is a true leader), but they are close to the results produced for NVIDIA GeForce FX 5700.
Pixel Shading
This test in the D3D RightMark benchmarking package allows to estimate the performance of executing various pixel shaders of the second version. In this test, the geometry has undergone maximum simplification in order to minimize the dependence of the test results on the geometric performance of the chip and verify the operation of purely pixel pipelines. We've brought in the operational modes for both 16-bit and 32-bit floating-point precision (switching between the precision modes is topical for NVIDIA chips only).
In the test for calculating pixel shaders, all the three boards for the case of simple 1.1 pixel shaders do a more or less the same good job, although ATI boards anyway offer some advantage.
NVIDIA chips proved shattered at performing version 2.0 shaders under 32-bit floating-point precision. We can make reservation for that the ATI chip was running that time at its invariable 24-bit precision, but let's look at the fps values produced with the 16-bit floating-point precision. In this case, the NVIDIA chip is in a vantage point, since it uses less precision rate than the ATI chip (whose absolute values, as expected, do not change - it runs at its invariable 24-bit).
Yes, the fps values in NVIDIA chip went up as expected, but in the end it proved a fiasco again. By the results of the test, NVIDIA chip executes pixel programs twice as slower than ATI chips do.
Point Sprites
This test is aimed at revealing the accelerator speed at displaying point sprites. Let's enumerate the adjustable test parameters:
- Resolution
- Windowed and full-screen mode
- Testing times (statistics accumulation) in seconds
- Vertex shader software emulation mode
- Operation mode:
- Vertex Shaders 1.1 and Fixed Function Blend Stages
- Vertex Shaders 2.0 and Fixed Function Blend Stages
- Animation mode:
- No
- Enabled
- Lightning mode:
- No
- Enabled
In the test settings, we used 2 diffuse light sources and enabled animation. We also investigated the dependence of execution speed on displaying point sprites on the version of vertex shader used.
At point-sprites handling, no one-one conclusions can be made of course. As we see, the NVIDIA chip is inclined to handling smaller-size sprites, whereas ATI chips are more comfortable at handling mid-sized sprites and perform a bit faster (ATI Radeon 9600 Pro) with large sprites.
For the case of version 2.0 vertex shaders, MSI's board built on NVIDIA GeForce FX 5700 chip is consistently among the outsiders, but the gap is not so great like that in using, say, version 2.0 pixel shaders in the previous test (tasks that are done here are quite different).
Hidden Surface Removal
This test allows to estimate the efficiency of removal of hidden points and primitives by the accelerator. A randomly generated scene will then be displayed in one of the three selected modes:
- sorted, in the front to back order
- sorted, in the back to front order
- unsorted
Also, we can investigate the dependence of the efficiency of hidden points and primitives removal efficiency on the version of vertex shader used (1.1 or 2.0).
Albeit not in all modes, the HSR block runs definitely faster in the NVIDIA chip than in ATI chips, and at that our settings in the tests suit very well to NVIDIA chips - we perform texture sampling which may not give us 100% correct results in estimating the efficiency of the HSR block of the chip. In the nearest future, this part of our analysis of accelerator operation will be complemented with one more HSR efficiency estimation mode in order to produce a clearer idea in this regard.
In the case of migration to using version 2.0 shaders, the general situation with the alignment of forces between ATI and NVIDIA chips hasn't changed.
Codecreatures
The well-known Codecreatures benchmark demonstrates a leadership of MicroStar's board on the base of the NVIDIA GeForce FX 5700 chip. NVIDIA's optimizations in combination with the good characteristics of the NVIDIA GeForce FX 5700 chip itself give a good result - NVIDIA GeForce FX 5700 is a leader. How do you like the results for GeCube Radeon 9550 Platinum Edition? Results in a real application that is built on a real gaming engine confirm the results produced in specialized synthetic applications - GeCube Radeon 9550 Platinum Edition is practically on par with ATI's Radeon 9600 Pro and at $100!
3DMark 2003 v340
Now let's tamper a bit with 3DMark =). There were many times when we got reasons to doubt the evidence of this benchmark, so we'd better not judge once again who was right or wrong at that benchmark, but will simply present the results for this comprehensive package just for general information only =). Note that GeCube Radeon 9550 Platinum Edition again keeps a stiff upper lip =).
Real-world gaming applications
From synthetic applications, we are now moving on to analyzing the performance of the graphic boards in real gaming applications.
Unreal Tournament 2003
This test is already outdated, so we'd better rule it out of the test package and replace it with the fresher Unreal Tournament 2004 which is part of our today's testing tools. Despite that the game is well optimized for NVIDIA cards, the leadership of NVIDIA GeForce FX 5700 proved to be only relative to GeCube Radeon 9550 Platinum Edition. In the gaming application it lags well behind the opponent by Gigabyte built on the ATI Radeon 9600 Pro chip.
Unreal II: The Awakening
In the case of using the most complex lighting model, the geometry handling performance was higher than the similar parameter in NVIDIA chip (provided the complex modes with version 2.0 shaders missing in Unreal II: The Awakening are disabled). This is partly confirmed by the results of tests in Unreal II: The Awakening. The NVIDIA GeForce FX 5700 chip in MSI's board beats GeCube Radeon 9550 Platinum Edition, and at 1600õ1200 it shatters even the the good old ATI Radeon 9600 Pro in Gigabyte's make.
Call Of Duty
In the "pure speed" mode (without using any image quality improvement techniques), GeCube's board lags behind MSI's solution on the base NVIDIA GeForce FX 5700. In its turn, it competes well with Gigabyte's solution on the base ATI Radeon 9600 Pro. At 1600õ1200, the board on NVIDIA GeForce FX 5700 appears to be a bit faster than ATI Radeon 9600 Pro.
With the AA and AF (8 samples), there is a chance for MSI's board on NVIDIA GeForce FX 5700 to rush forward to absolute leadership at all the resolutions. Not too bad really.. GeCube's board also keeps a stiff upper lip among its competitors, albeit takes the last place at this test, but the price is attractive... =).
X2: The Threat
The "Ultra Shadow" technology implemented in NVIDIA chips should allow them easily bypassing ATI boards at this test. Due to this technology, stencil shades whose rendering technology is widely used in the game are built faster. And we can see leadership of MSI's boards on NVIDIA GeForce FX 5700 chip in the low resolution. At higher resolutions, the results get more or less equal because as the resolution goes up other factors limiting the performance of NVIDIA GeForce FX 5700 chips and giving the chance to ATI boards come into play - ATI boards handle stencil shades in greater number of passes and demonstrate results on par with NVIDIA boards. However, let's not forget that the board on the base of NVIDIA GeForce FX 5700 completely beats its direct competitor - GeCube board made on ATI Radeon 9550.
Similar effects (stencil shades) will be used in the forthcoming DooM III (in new NVIDIA chips, the UltraShadow II technology that allows better handling stencil shades has already been implemented). However, the first experiences of people who have already tried DooM III in action suggest that boards on the base of NVIDIA GeForce 6800 Ultra are not much faster than ATI Radeon X800 XT (for details, read http://cyberfight.ru/site/news/25473/?PHPSESSID=16c764d5efafdbf3330693763bbb7e1e).
That video cards on the base of NV40 (GeForce 6800), as was stated above are merely a little bit faster than cards on the base of ATI R420 (X800) is simply an outrageous statement. Originally, the use of stencil shades was planned as a factor of significant advantage of NVIDIA cards for both the previous generation (GeForce FX) and the next generation (NV40) - this game offers such shades in abundance. Now these statements albeit not furnished with proof, do not bode any good for NVIDIA. We can expect a flow of new drivers from NVIDIA which will give a "drop-by-drop" performance rise due to optimization the architecture of NVIDIA boards, but that is definitely not what Carmack and NVIDIA anticipated. The matter of performance on video cards of the previous generation - ATI Radeon 9800XT and NVIDIA GeForce FX 5950 - remains open. In any case, we can tell something certain once we look at the engine and feel it live =).
Unreal Tournament 2004
This is a new test in our set of benchmarks. To be more precise, not the most new - we have already used it in our review on Gigabyte GeForce FX 5950 Ultra, but that time we used an integrated benchmark which gave rise to unfavorable criticism. Now we recorded our own demo "3DNews003 Demo" which reflects a real gaming process in one of the death-match maps of Unreal Tournament 2004 (we had to relive a bit of the past and shoot at bots =)). As the testing tool-set to run the demo, we used BenchemALL, a software that is gaining increasing popularity among testers.
Let's see the results. In fact, the engine of Unreal Tournament 2004 does not bring nothing radically new at all as compared to Unreal Tournament 2003, so it would be logical to expect a repetition of the picture observed in Unreal Tournament 2003. However, that didn't come true - at that, NVIDIA GeForce FX 5700 loses to its direct competitor, GeCube Radeon 9550 Platinum Edition, and Gigabyte ATI Radeon 9600 Pro proves to be an absolute leader. In our view, that has nothing to do with the engine of Unreal Tournament 2004, but the cause is that optimizations running in Unreal Tournament 2003 have not yet been introduced into the latest drivers for Unreal Tournament 2004. With the release of new NVIDIA drivers (whose number is continuously increasing lately), we believe the performance at Unreal Tournament 2004 should become equal to that at Unreal Tournament 2003 (hopefully, not to the prejudice of quality).
 Enabling the AA allowed NVIDIA GeForce FX 5700 to rush forward relative to GeCube Radeon 9550 Platinum Edition, and even beat both boards on ATI chips at 1600x1200.
Anisotropic filtering, for which NVIDIA is currently doing optimizations for this game, allows producing results for NVIDIA GeForce FX 5700 very similar to those for GeCube Radeon 9550 Platinum Edition, and at 1600õ1200 the competitor from NVIDIA anyway rushed forward.
HALO: Combat Evolved
The game uses version 2.0 pixel and vertex shader programs. Besides, the game offers wonderful features for testers by allowing forced enablement of version 1.1, 1.4 and version 2.0 pixel and vertex programs. Nevertheless, we long ago gave up using shaders of version other than 2.0 for this test, because we are mostly interested in the board's DirectX 9.0 performance and also because we have repeatedly shown performance difference for video cards on the base of ATI and NVIDIA chips between shader versions 1.1 and 2.0.
And again we get practically close and similar results for all the three boards. Gigabyte's board, as it should be, rushes a bit forward, whereas its direct competitors - GeCube Radeon 9550 Platinum Edition and MSI NVIDIA GeForce FX 5700 perform simply close to one another at all the resolutions. It's surprising how the results for these boards could merge to the tenths of fps =).
Tomb Raider: Angel of Darkness
Tomb Raider engine: Angel of Darkness is a technically advanced DirectX 9.0 solution. The game uses version 2.0 pixel and vertex shaders, which, considering the results of synthetic tests, is not going to bring anything good for NVIDIA boards. We should also make essential reservations for the results of tests in this application before the readers can make themselves aware of the test results. NVIDIA's boards at this test have serious problems with rendering. In particular, after applying patch 49 the rendering of shades by the main character simply vanished with all NVIDIA boards without exception. In the further patch (version 52) all has been fixed already, but the option of game benchmarking has been disabled. Therefore, the results of boards built on NVIDIA chips should be perceived with the reservation that rendering results can't be regarded canonically ideal (while this is OK with video cards built on ATI chips).
 The first scene that makes the most intense use of version 2.0 pixel shaders (which is confirmed by the analysis of the game's code) leaves no chances for NVIDIA's board based on the GeForce FX 5700 chip - the gap is disastrous and something prompts us that no new drivers will help fixing that =). ATI's boards cope quite well with that, the gap between Gigabyte's board on ATI Radeon 9600 Pro chips and GeCube's board on Radeon 9550 Platinum Edition is not that noticeable. Also note the absolute fps values. Both boards easily overcome the 30 fps "psychological barrier" at 1024x768, which allows us to claim these cards are suitable for active gaming in this application. Unfortunately, that can't be said about the NVIDIA cards...
 The second scene is abundant with shaders, but they are not so great in number than in the previous test scene, which allows MSI's board on the NVIDIA GeForce FX 5700 chip slightly reducing the gap between its direct competitor - GeCube's Radeon 9550 Platinum Edition. Change of leadership is out of the question here.
Shaders are almost not used at all, but anyway NVIDIA board failed to get the leadership. The same sad gap versus its direct competitor. Boards on the base of ATI Radeon 9600 Pro and ATI Radeon 9550 in GeCube's make offer the same performance difference.
Half-life 2 Leaked Beta
It is premature to make final conclusions on the performance of an accelerator basing on a raw and leaked version, since in the final release many things might change (we are more than confident that it will be the way). Nevertheless, I think I won't be wrong if I say that many readers would be curious to know the alignment of forces among graphic chips just at that benchmark. The game uses version 2.0 pixel and vertex shaders even more intensely than the previous game, which is found out through the analysis of the game's shader code. At the same time, the complexity of shaders is an order of magnitude higher than in Tomb Rider: Angel of Darkness.
So, here goes the first demo that we recorded. And again we see a disastrous gap for the board based on NVIDIA chip. ATI boards again show the same performance difference at all the resolutions. Again the 30 fps "psychological barrier" at 1024x768 was overcome.
The second demo appeared to be a too heave burden to cope for NVIDIA GeForce FX 5700 =). The lag between its direct competitor sometimes amounts to 2.5 times!
AquaMark 3
"Very nice!" - that's what we can say looking at the results in that complex semi-synthetic DirectX 9.0 benchmark for MSI GeForce FX 5700. The board demonstrates results close to those for ATI Radeon 9600 Pro and leaves its direct competitor, GeCube Radeon 9550 Platinum Edition, behind. Indeed, programmers from the Californian company are miracle-workers =).
But with the AF enabled (also activated in this mode) we get a quite different alignment of forces. Although the difference is not great, NVIDIA GeForce FX 5700 is among outsiders at this mode.
FireStarter
The graphic engine does not use DX9 functionalities. So, I'd rather relate it to the DX8 generation This bodes NVIDIA's board more or less good chances to not lose face.
As is seen from the results, that has proved the way. MSI's board on NVIDIA GeForce FX 5700 chip shows better results than GeCube Radeon 9550 Platinum Edition, however they have a too long way to go to reach the results for Gigabyte's board on ATI Radeon 9600 Pro chip - perhaps the higher memory bandwidth in Gigabyte's board makes itself felt.
In the gaming benchmark, the alignment of forces is similar to that in the previous one. The absolute values are also pleasing to see. Even at 1280x1024, the gameplay is on par for a game of this type.
FarCry
FarCry is a bright example of what we can expect from Hi-End boards with support for DirectX 9.0. The beauty demonstrated by the game gives cause to think of the futility of existence and that Middle-End boards might not be enough for future. =) The game uses version 2.0 pixel and vertex shaders (the forthcoming patch 1.2 of the game offers support for version 3.0 pixel shaders, which should stir up the interest to board on the base of NV40 with hardware support for version 3.0 shaders).
MSI's board on NVIDIA GeForce FX 5700 chip suffered from serious problems with rendering with this driver version for the current (1.1) version of the game. With other drivers, no problems like that were found. So we bring in the results for only two boards on ATI chips.
Painkiller
Looking at the game engine of Painkiller,you can't say it offers something special. Version 1.1 pixel and vertex shaders, standard effects and the like, but such impression stays until you look at the game alive. In dynamics, the engine looks not bad at all. Here is a brief list of technologies used in the game:
- dynamic shades using shadowmaps;
- 3D fog;
- reflexive and refracting surfaces;
- Fong's pixel-wise shading in static objects with dynamic lights;
- pre-pixel lighting in using normal maps on dynamic objects.
For tests, we used the most recent version 1.3 of the game. Note that in the new version of the patch the mode of integrated benchmark has changed - now there are player (more precisely, monsters) models on the map, which adds more load to the accelerator (and the CPU).
MSI's board on NVIDIA GeForce FX 5700 lags well behind. The game is definitely a hard nut to crack for the NVIDIA chips, despite that the engine should have been optimized for the NVIDIA GeForce FX architecture, since the game is part of NVIDIA's initiative called "The way it`s meant to be played".
Prince of Persia: Sands of Time
The game is very well made with the finely tuned gameplay, and, most importantly, the nice graphic component. The game engine does not bring anything revolutionary (hardly anything more can be expected from PS2 port). The game imposes rigid requirements upon the video card - it must support version 1.1 pixel and vertex shaders. The game simply won't start with cards that don't offer support for pixel and vertex programs. Unfortunately, we were unable to find more detailed information on the game engine (dubbed "JADE engine") used for the PC version of the game. However, we already know something =).
Now let's look at the results in the simplest mode, without any improvement techniques. It seems like NVIDIA's label "The way it`s meant to be played" is gaining revenge for the defeat in Painkiller game. The results were taken using the Fraps utility, which may add some 1 fps error, however, as we can see from the diagrams, the gap from both (!) video cards is much greater than 1 fps.
Breed
The game's engine dubbed Mercury can't boast any extraordinary abilities - like FarÑry's, the graphics boggles your mind. Nevertheless, the game uses version 1.1 pixel and vertex programs and stencil shades implemented on NVIDIA cards with the UltraShadow technology; on video cards built on ATI chips, because of the lack of support for this NVIDIA's proprietary technology, shades are rendered in greater number of passes, which results in drops of speed especially in applications where shades are in abundance. In other words, we can preliminarily claim that NVIDIA cards will have some advantage at that. However, let's not run ahead - tests will tell all.
 The engine of the game answered our expectations. The card on the base of NVIDIA GeForce FX 5700 chip offers an advantage over both (!) boards on the base of ATI chips. Stencil shades implemented on NVIDIA video cards with the UltraShadow technology indeed is a great help to NVIDIA boards. Well, let's wait for DooM III to find that out alive (or to disappoint ourselves in view of the above facts).
 |
Top Stories: |
 |
 |
|
MoBo:
| |
|
|
VGA Card:
|
|
|
|
CPU & Memory:
|
|
