Contents

Accession to the throne of the new 3D king, ATI RADEON 9700 Pro, isn't celebrated any more, such cards are widely available on the market and their prices have already dropped from $450 to $350. Now we are waiting for a new pretender for the throne - nVIDIA NV30. When ATI managed to snatch the leadership after several years of struggle (though only from the technological standpoint, not in terms of sales), it switched over to the mainstream market promising great profits and to the intermediate "demanding" sector of the price range from $180 to 250. Before that ATI had nothing in these sectors that could compete against NVIDIA GeForce4 Ti4400, Ti 4200 or MX 460. Now it has the RADEON 9000 Pro which is gradually rolling down to the sub$100 sector, the RADEON 9000 which is yet cheaper, the RADEON 8500(LE) priced at $110-140 which is living its last days, and the RADEON 7500 whose price fell down marginally. So, in the sub$100 category the Canadian company has a lot solutions that can be opposed to NVIDIA (GeForce4 MX 440 etc.). 

Of course, in the spheres where the GeForce4 Ti 4200 and Ti 4400 (and the cheapest samples of Ti 4600) are thriving ATI has nothing to offer. The flagship RADEON 9700 Pro is still overpriced, but are there many customers out there willing to buy a card at $350-370? The largest profits are pouring into the pockets of companies manufacturing Ti 4200. This processor is still beyond any competition in its price niche (see our 3Digest). ATI thus has a problem of making a solution based on modern technologies of the RADEON 9700, but less powerful and cheaper. That's good ATI didn't begin strengthening the RADEON 9000 and acted vice versa. Let's take a look at the GeForce Ti4200 - which has come down as well - as the junior version of the NV25. The new capabilities of the RADEON 9700, albeit in a slower mode, could be an important argument for promotion of the 9500 series. Moreover, the SmoothVision anti-aliasing inherited by the the 9000 series from the RADEON 8500 is too slow, and anisotropic quality is suffering a lot in spite of the fact that it doesn't affect performance much. The support of unnecessary pixel shaders 1.4 remained a vain marketing trick. Well, the choice is made - the 9500 will be based on the R300 core (later it will be possible to use one or another lite version of the chip but in the beginning the whole 9500 series will be built on the original R300). 

It's certainly attractive to develop a new product for the price niche of $130-150 with the complete support of DirectX 9.0 and all new architectural features of the RADEON 9700 Pro. As ATI didn't have much time, they decided to reduce the number of pixel pipelines and use a 128-bit memory bus. First of all, production of the lite chips will be cheaper, and the cost of the PCB will get noticeably smaller. Besides, mounting of the slower and cheaper memory will have a positive effect on the overall costs. And finally, it's possible to use culls of production of the RADEON 9700 Pro. 

ATI, thus, got a change to make the competitor for NVIDIA's solutions with minimal research expenses and as quickly as possible. And ATI did it by announcing a new line at the end of October. Here is how the RADEON 9500/9700 family looks like now: 

• RADEON 9700 PRO - 325 MHz chip (8 pipelines), 128 MB 310 MHz (DDR 620) 256 bit local memory;  
• RADEON 9700 - 275 MHz chip (8 pipelines), 128 MB 270 MHz (DDR 540) 256 bit local memory;  
• RADEON 9500 PRO - 275 MHz chip (8 pipelines), 64/128 MB 270 MHz (DDR 540) 128 bit local memory;  
• RADEON 9500 - 275 MHz chip (4 pipelines), 64 MB 270 MHz (DDR 540) 128 bit local memory.    

Before we jump into analyses of all the cards except the first one I must remind you that operation of the RADEON 9700 Pro was closely examined in the following reviews: 

Theoretical materials and reviews of video cards which concern functional properties of the VPU ATI RADEON 9700 (Pro)

Now let's turn back to the line we were speaking about above. The RADEON 9700 is just a little slower version of the RADEON 9700 Pro, but its price promises to be tasty - $250-270, which makes this card a direct competitor for the GeForce4 Ti 4600 (though their speeds are yet to be compared). The RADEON 9700 will be sold both by ATI and its partners. 

The RADEON 9500 Pro is based on the RADEON 9700 with a narrower memory bus: 128bit instead of 256bit. This solution will be selling both by ATI and its partners as well. The price is going to be around $200 (190-210). This is a competitor for senior cards on GeForce4 Ti 4200 and 4200-8x. 

The RADEON 9500 is the junior solution in the line of DirectX 9.0 accelerators whose chip has a twice lower fillrate compared to the RADEON 9700 (4 shading pipelines instead of 8) and whose memory has a 128bit bus. This solution is positioned as a competitor against junior models of GeForce4 Ti 4200, that is why its price will first make $120-150. It will hardly drop down as that niche belongs to the RADEON 9000 Pro and upcoming RADEON 9100. 

In spite of such an attractive Canadian policy of fighting against NVIDIA, it has one significant downside: Ti 4200/4600 models have been produced for a long time already, and manufacturers can start reducing their prices, while prices of RADEON 9500/9700 based cards will hardly creep down straight away. Besides, the Xmas sales are right around the corner, but the ATI's new-comers are not available yet. ATI and its partners are only sending preproduction samples. ATI should hurry or the pie will be eaten by others, in particular, NVIDIA with its Ti 4200 (just look at our reviews of the latest cards from Albatron or ASUS). 

Taking into account that all claimed functions of RADEON 9700 Pro are supported, and only quantitative parameters of the graphics processor are reduced, we will omit these functions (you can read about them in our first reviews from the list above). 

And today we will see how much the performance is affected by such stripping off and how much the speed of separate units controlling shaders has fallen down. 

Note that we haven't got yet RADEON 9500 Pro based cards; and beside the ATI's sample on the RADEON 9500 we managed to get an almost production card from Gigabyte based on the RADEON 9500. Later you will find out why we are focusing on it. 

Cards

ATI RADEON 9700 128MB
ATI RADEON 9500 64MB
Gigabyte MAYA II RADEON 9500 64 MB

ATI RADEON 9700 128MB
The card has AGP x4/x8 interface, 128 MB DDR SDRAM located in 8 chips on both PCB sides.  The memory chips are from Samsung marked as K4D26323RA-GC2A, BGA form-factor. The highest frequency is 350 (700) MHz, hence the access time of 2.8 ns. The memory runs at 270 (540) MHz by default and the chip works at 275 MHz. However, we couldn't verify these claimed data. Besides, it's not necessary to use such efficient memory for 270 MHz; it seems that the RADEON 9700 Pro card got a new BIOS version with such frequencies, because the PCB is the same.
ATI RADEON 9500 64MB, Gigabyte MAYA II RADEON 9500 64 MB
Both cards have AGP x4/x8 interface, 64 MB DDR SDRAM located in 4 chips on both PCB sides.  The memory chips are from Hynix and marked as HY5DU283222 F-36, BGA form-factor. The highest frequency is 275 (550) MHz. The memory runs at 270 (540) MHz by default and the chip works at 275 MHz. It can be proved by the PowerStrip utility.

 

Comparison with the reference design, front view
ATI RADEON 9700 128MB	Reference card ATI RADEON 9700 Pro 128MB
ATI RADEON 9500 64MB
Gigabyte MAYA II RADEON 9500 64MB

 

Comparison with the reference design, back view
ATI RADEON 9700 128MB	Reference card ATI RADEON 9700 Pro 128MB
ATI RADEON 9500 64MB
Gigabyte MAYA II RADEON 9500 64MB

All the cards are based on the reference design of the RADEON 9700 Pro, and the ATI RADEON 9700 is a pure copy of the RADEON 9700 Pro with just another BIOS version and another ID. 

The 128bit memory bus was obtained by removing half of the chips from the RADEON 9500. Each chip consists of 32 bits, 4 in all, i.e. it amounts to 128 bits. 

The graphics processor looks similar on all the cards: 

I've got two questions here: 

1. Why so expensive PCB which was developed for a 256bit bus is used for these relatively cheap cards ($130-150)? 
2. Why the cut-down RADEON 9500 is packed into the same FC-BGA case which is also designed for a 256bit bus? 

However, Gigabyte is not going to wait for the new design. They either found a cheaper way of production of such expensive PCBs or decided to speed up sales of the cards; anyway, the Gigabyte MAYA II RADEON 9500 is a production model. It is also possible that the card from ATI on the same chip will be put into the mass production, though not in large quantities. Well, the ATI's clan does its best to make RADEON 9500 based cards available on the retail market as soon as possible. 

In fact, it's not bad that the RADEON 9500 is going to use the PCB from a senior card. Moreover, it will be really pleasing for overclockers as the potential of the RADEON 9500 on such a smooth board must be superb. But whether it is rational to produce RADEON 9500 on such expensive PCBs depends on the price of such cards. 

As for the VPU, the RADEON 9500 is reportedly a redesigned chip based on the RADEON 9700, but it will be put into production a little later, and at the start they will use original R300 (RADEON 9700) with pipelines and probably some other blocks locked on the hardware level. But is it really hardware locking? And is it done on the chip's or BIOS's level? Is this operation reversible? We have no answers today, but the tested cards house exactly these R300 chips. They are possibly discarded samples which can't work as normal RADEON 9700 Pro or even RADEON 9700, or maybe they are just properly programmed samples that passed all the tests. 

Now look at the cooling system of the cards.   

ATI RADEON 9700 128MB, ATI RADEON 9500 64MB
This cooler is traditional for the RADEON 9700 Pro, which confirms again the origin of the RADEON 9700, as well as of the RADEON 9500.
Gigabyte MAYA II RADEON 9500 64MB
This is a more appealing cooler. The shape is the same, but it's golden and has a different fan.

All the cards have outputs to two monitors (one through DVI) and TV-out. 

Only the Gigabyte's card ships in the Retail box; so, let's pop into: 

Here is the box: 

Overclocking

The latest version of the PowerStrip (3.32) is able to work with the RADEON 9500, but there are problems with overclocking of the RADEON 9700. Earlier I mentioned that these cards have the PCB from RADEON 9700 Pro. And ATI is certainly not going to allow its partners to buy the RADEON 9700 (which is cheaper than the Pro version) and turn it into RADEON 9700 Pro. 

There is some tricky solution, like in the RADEON 9000, that makes impossible to increase the frequencies. But while frequencies of the RADEON 9000 are seen and even can be changed (though it won't affect the speed), none of the tweakers can define frequencies of the RADEON 9700. That is why we can't even verify that the card's working at 275/270 (540) MHz. The other cards have the following scores: 

What an excellent overclocking potential! Even at 368 MHz, which is the maximum frequency the PowerStrip supports the chip works stably even without additional cooling. The memory chips have reached their limit, but still, 325 MHz (from 270) is fairly good. The twice smaller number of pipelines could have a positive effect on overclocking. 

Note 

• in course of overclocking you must provide additional cooling, in particular, for the card (first of all, for its memory):
 



• overclocking depends on a definite sample, and you shouldn't generalize the results of one card to all video cards of this mark or series. The overclocking results are not the obligatory characteristics of a video card. 

Test system and drivers

Testbeds: 

• Pentium 4 based computer (Socket 478): 
• Intel Pentium 4 2530 MHz; 
• ASUS P4T533 (i850E) mainboard; 
• 512 MB 32bit RDRAM PC4200; 
• Seagate Barracuda IV 40GB; 
• Windows XP. 

In the tests we used ATI's drivers 6.200 (7.79). VSync was off in the drivers, texture compression was off in the applications. DirectX 8.1 was used. Texture detail level was set to High Quality. 

For comparison we used results of the following video cards: 

• Gacards: 
  • Gainward Powerpack Ultra/750 GS (GeForce4 Ti 4600, 300/325 (650) MHz, 128 MB, driver 40.72); 
  • Triplex Millennium Silver GeForce4 Ti 4200 (250/256 (512) MHz, 64 MB, driver 40.72); 
  • Albatron GeForce4 MX480 (MX 440 with AGP8x, 275/256 (512) MHz, 64 MB, driver 40.72); 
  • Hercules 3D Prophet 9700 Pro (RADEON 9700 Pro, 325/310 (620) MHz, 128 MB); 
  • ATI RADEON 8500 (275/275 (550) MHz, 64 MB); 
  • Hercules 3D Prophet 9000 Pro (RADEON 9000 Pro, 275/275 (550) MHz, 64 MB). 

  Drivers' settings

   

  
  
  
  
  
  
  
  
  
  Remember that only DirectX 8 drivers are available today! The DirectX 9.0 drivers are expected only after the DX9 itself. The control settings are standard for the whole new series of the CATALYST drivers except the settings of the RADEON 9700 in the AA (SmoothVision II) and anisotropy sections. For anisotropic filtering you can choose a mode of operation (expect the levels): performance/quality. In the above reviews we discussed how these modes differ from each other (in short, they differ in support of joint operation of trilinear filtering and anisotropic one). 

  The AA has three modes: 2x, 4x and 6x. 

  That's all about quirks of the drivers. 

  Test results

  2D graphics

  Let's start with 2D. The 2D quality is really perfect, like in case of the previous RADEON 9700 Pro based cards! 

  Remember that estimation of 2D quality can't be objective. It depends on quality of a sample, and a card/monitor tandem (you should pay special attention to quality of the monitor and the cable). 

  3D graphics, MS DirectX 8.1 SDK - extreme tests

  We used modified examples from the latest official final version of the DirectX SDK (8.1) for testing the extreme characteristics of the chips. 

  Optimized Mesh

  This test defines a real maximum throughput of an accelerator as far as triangles are concerned. For this purpose it uses several simultaneously displayed models each consisting of 50,000 triangles. No texturing. The dimensions are minimal - each triangle takes just one pixel. It must be noted that the results of this test are unachievable for real applications where triangles are much greater, and textures and lighting are used. The results are given only for 3 rendering methods - model optimized for the optimal output speed (with the size of the internal vertex cache on the chip accounted for) - Optimized, Unoptimized original model, and Strip - unoptimized model displayed in the form of one Triangle Strip: 

  
  
  
  This test depends mostly on the chip's geometrical performance. The number of pixel pipelines which is twice smaller compared to the RADEON 9700 at the same core's frequency of the RADEON 9500 doesn't affect the performance. Well, the geometrical subsystem of the chip wasn't stripped off and shows excellent efficiency outpacing (except the Unoptimized case) the GeForce Ti 4600. 

  Vertex shader unit performance

  This test allows determining the maximum performance of the vertex shader unit. It uses a complex shader which deals with both type-transformation and geometrical functions. The test is carried out in the minimal resolution in order to minimize the shading effect. The Z buffer is disabled and HSR can't affect the results: 

  
  
  
  The difference from the RADEON 9700 in this test is caused by the dependence on the fillrate. Well, this test needs a proper substitute. Fortunately, a new synthetic test included into the RightMark 3D packet which is due to appear soon doesn't depend on a fillrate. 

  Vert on a fillrate. 

  Vertex matrix blending

  This T&L's feature is used for verisimilar animation and model skinning. We tested blending using two matrices both in the "hardware" version and with a vertex shader that implements the same function. Besides, we obtained results in the software T&L emulation mode: 

  
  
  
  

  Here all ATI's solutions go behind the NVIDIA's cards irrespective of the RADEON's type. It seems that NVIDIA has realized better this function. But with spreading of vertex shaders and DirectX 9.0 the possibility of a hard hardware blending won't be important anymore. 

  EMBM 

  In this test we measure performance drop caused by Environment mapping and EMBM (Environment Bump). We also measure a fillrate in case of single texturing. We set 1280x1024 because exactly in this resolution the difference between cards and different texturing modes is the most discernible: 

  
  
  
  

  In this test the performance depends on a fillrate and, therefore, on the number of pixel pipelines. Well, in the simplest mode the RADEON 9500 is at the twice lower level compared to its senior brother. With EMBM the dependence is not so direct as there are some other influencing factors beside the fillrate. Here the RADEON 9500 loses to the NVIDIA's products, and this means that it won't be able to fight against the Ti 4200 in popular games that depend on a fillrate. 

  Performance of Pixel Shader 1.0 

  We used again a modified example of the MFCPixelShader having measured performance of the cards in high resolution in implementation of 5 shaders different in complexity, for bilinear-filtered textures: 

  
  
  
  

  The RADEON 9500 has a twice lower level than the RADEON 9700 in the "pixel" issues. As the shaders become lengthier and more complicated the gap gets greater as well. In case of very complex shaders it's impossible to stand against the GeForce Ti 4600 and even against the Ti 4200 - remember that the DX9 chips process shaders differently, they fulfill instructions in turn, and on DX8 shaders they will lose to the solutions of the DX8 generation equipped with pixel stages at the same number of pipelines and frequency. The NV25 belongs to them, which allows it to catch up with the RADEON 9700 coming with the twice more pipelines and twice wider bus on the long DX8 shaders. 

  So, the synthetic tests shows that the RADEON 9500 has half of the pixel pipelines disabled. But the geometrical subsystem and cache weren't changed, which often lets the RADEON 9500 keep up with the RADEON 9700 when processing geometry and has better than just the half performance in case of complex shading (EMBM). 

  Now we have the synthetic tests of the 3DMark2001: 

  3D graphics, 3DMark2001 SE - synthetic tests

  All measurements in all 3D tests were taken in 32-bit color. 

  Fillrate

  
  
  
   
  
  
  The test confirms our supposition about 4 pipelines and proves the test results of the DX8 SDK. 

  Scene with a large number of polygons

  In this test you should pay more attention to the minimal resolution where the fillrate makes almost no effect: 

  
  
   
  
  
  

  In the minimal resolution the RADEON 9500 is able to fight against the Ti 4200. But in higher resolutions, i.e. when the synthetic tests come closer to the real applications, it gets weaker. 

  Bump mapping

  Look at the result of the synthetic EMBM scene: 

  
  
  
  And now the DP3 relief: 
  
  
  The scores are identical to those obtained in the DX8 SDK tests. 

  Vertex shaders

  
  
  
  

  In a low resolution, where the fillrate isn't important much, a higher geometrical performance of the RADEON 9500 (compared to the RADEON 9700, according to the DX8 SDK tests) lets it easily outdo the NVIDIA's solutions. Well, in applications with rich geometry and numerous vertex shaders in low resolution the latest ATI products will be preferable compared to the previous generation of NVIDIA. But what will be when the next NVIDIA's generation appears? 

  Pixel shader

  
  
  
  

  The double defeat again, in accordance to the number of pipelines. But the scores correspond to the Ti 4200. Let's see what do we have in the Advanced Pixel Shader test. 

  
  
  
  

  The layout is the same. 

  Sprites

  
  
  
  

  This test was never a strong side of ATI. The GeForce Ti 4600 with its twice narrower memory bus is close on the heels of the today's king RADEON 9700 PRO. And the Ti 4200 beats the RADEON 9500 in all aspects. 

  So, the problems of the RADEON 9500 are caused primarily by the twice lower number of pipelines. It can stand against the Ti4200 but only in low resolutions. I wonder how it can be in real applications. Will the 8-pipeline brother of our hero (RADEON 9500 PRO) be able to eliminate all its weak points? The first question will be given an answer to today, and the second will be answered in the upcoming review of the RADEON 9500 PRO. 

  Before we turn to the games, I must say that we tested the RADEON 9500 not only at the default and extreme frequencies but also at 325/620 MHz (modest overclocking) to estimate how lighter the chip has become and how it can perform in contrast to its normal brother RADEON 9700 Pro. 

  3D graphics, 3DMark2001 - game tests

   

  3DMark2001, 3DMARKS

  
  
  
  

  
  

  3DMark2001, Game1 Low details

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	-6.1
  1280x1024 with AA4x	65.6
  1600x1200 with ANISO 8x/16xQ	21.2
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-30.4
  1024x768 with AA4x	-0.8
  1024x768 with ANISO 8x/16xQ	-31

  The RADEON 9700 outedges the Ti 4600 in the heavy modes but the RADEON 9500 falls far behind its rival. 
   

  3DMark2001, Game2 Low details

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	16.6
  1280x1024 with AA4x	141.7
  1600x1200 with ANISO 8x/16xQ	127.3
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-22.1
  1024x768 with AA4x	41.1
  1024x768 with ANISO 8x/16xQ	25.2

  The RADEON 9700 has brilliant scores and the RADEON 9500 has improved its results as well, - with AA or anisotropy it outshines its competitor. 
   
   

  3DMark2001, Game3 Low details

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	16.0
  1280x1024 with AA4x	153.6
  1600x1200 with ANISO 8x/16xQ	189.4
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-22.9
  1024x768 with AA4x	48.7
  1024x768 with ANISO 8x/16xQ	39.8

  All the same. 
   
   

  3DMark2001, Game4

  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	28.7
  1280x1024 with AA4x	86
  1600x1200 with ANISO 8x/16xQ	16.4
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-19.6
  1024x768 with AA4x	-1.3
  1024x768 with ANISO 8x/16xQ	-42

  While the RADEON 9700 stands firm, the RADEON 9500 lags behind again. 

  So, what is the outcome of the 3DMark2001? The RADEON 9700, being equal to the GeForce4 Ti 4600 in price, showed brilliant results. But the RADEON 9500 turned out to be weaker than expected. And even AA and anisotropy can't help it. The truncated bus badly affected the speed of AA and anisotropy. In addition, it has 4 texture units against 8 of the Ti 4200... Even the higher frequency (275 MHz) doesn't help it crash the 250MHz Ti 4200. 

  3D graphics, game tests

  For estimation of 3D performance in games we used the following tests: 

  • Return to Castle Wolfenstein (MultiPlayer) (id Software/Activision) - OpenGL, multitexturing, Checkpoint-demo, test settings - maximum, S3TC OFF, the configurations can be downloaded from here
  
   
  • Serious Sam: The Second Encounter v.1.05 (Croteam/GodGames) - OpenGL, multitexturing, Grand Cathedral demo, test settings: quality, S3TC OFF 
  
   
  • Quake3 Arena v.1.17 (id Software/Activision) - OpenGL, multitexturing, Quaver, all settings - maximum: detailing level - High, texture detailing level - #4, S3TC OFF, smoothness of curved surfaces is greatly increased with variables r_subdivisions "1" è r_lodCurveError "30000" (at default r_lodCurveError is 250!), the configurations can be downloaded from here
  
   
  • Comanche4 Benchmark Demo (NovaLogic) - Direct3D, Shaders, Hardware T&L, Dot3, cube texturing, highest quality
  
   
  • Unreal Tournament 2003 Demo v.927 (Digital Extreme/Epic Games) - Direct3D, Vertex Shaders, Hardware T&L, Dot3, cube texturing, default quality
  
   
  • AquaMark (Massive Development) is a game that demonstrates operation of a card in the DirectX 8.1, Shaders, HW T&L. 
  
   
  • RightMark Video Analyzer v.0.4 (Philip Gerasimov) - DirectX 8.1, Dot3, cube texturing, shadow buffers, vertex and pixel shaders (1.1, 1.4). 

  Quake3 Arena, Quaver

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	2.3
  1280x1024 with AA4x	80.1
  1600x1200 with ANISO 8x/16xQ	31.2
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-37.5
  1024x768 with AA4x	-3.2
  1024x768 with ANISO 8x/16xQ	-24.1

  The RADEON 9700 is still better than the GeForce4 Ti 4600, while the RADEON 9500 manages to beat only the GeForce4 MX. 

  Serious Sam: The Second Encounter, Grand Cathedral

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	17.2
  1280x1024 with AA4x	191.8
  1600x1200 with ANISO 8x/16xQ	27.7
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	3.3
  1024x768 with AA4x	79.7
  1024x768 with ANISO 8x/16xQ	11.6

  Both RADEONs of the new generation thrive in this test. 

  Return to Castle Wolfenstein (Multiplayer), Checkpoint

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	20.5
  1280x1024 with AA4x	119
  1600x1200 with ANISO 8x/16xQ	68.3
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-28.7
  1024x768 with AA4x	22.2
  1024x768 with ANISO 8x/16xQ	-5.5

  The RADEON 9700 is still unconquerable; the RADEON 9500 can grasp the palm only with the AA. 

  Comanche4 DEMO

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	7.3
  1280x1024 with AA4x	72
  1600x1200 with ANISO 8x/16xQ	18.5
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-7.9
  1024x768 with AA4x	-40
  1024x768 with ANISO 8x/16xQ	-32.1

  No comments. 

  Unreal Tournament 2003 DEMO

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	29.4
  1280x1024 with AA4x	151.6
  1600x1200 with ANISO 8x/16xQ	96.7
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-30.7
  1024x768 with AA4x	22.9
  1024x768 with ANISO 8x/16xQ	1

  AquaMark

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	3.7
  1280x1024 with AA4x	56.3
  1600x1200 with ANISO 8x/16xQ	104.2
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	-37.7
  1024x768 with AA4x	22.7
  1024x768 with ANISO 8x/16xQ	50.9

  RightMark 3D

  
  
  
  
  
  

  Test type	Gap between RADEON 9700 and GeForce4 Ti 4600, %
  1600x1200 without AA and ANISO	28.2
  1280x1024 with AA4x	63.6
  1600x1200 with ANISO 8x/16xQ	38.2
  Test type	Gap between RADEON 9500 and GeForce4 Ti 4200, %
  1280x1024 without AA and ANISO	8.3
  1024x768 with AA4x	27.4
  1024x768 with ANISO 8x/16xQ	-3

  In this synthetic application which tests operation with shaders the RADEON 9500 doesn't look that bad. Well, in future games where such technologies will prevail the RADEON 9500 has every chance to win. 

  Conclusion

  Note that we haven't estimated quality of the RADEON 9500/9700 today as it doesn't differ from the RADEON 9700 Pro (you can find it in the reviews listed above). 

  1. Today we have examined one version of the RADEON 9500. Note that some manufacturers are against production of such cards on the expensive PCB from RADEON 9700 Pro, and we can soon get other cards. It is also possible that a new chip, not a reprogrammed version of the R300, will be developed. Its results can be different from what we have obtained today! But remember that Gigabyte already supplies the RADEON 9500 based cards with the original R300. 
  2. As you can see, the RADEON 9500 falls behind its competitor GeForce4 Ti 4200 when AA and anisotropy are disabled. The new baby of ATI can boast of the DirectX 9.0 support (though it doesn't affect the performance today), an improved TV-out, more effective AA in most cases and anisotropy which sometimes brings good profits. ATI and its partners should consider price cuts for the RADEON 9500 to make it a strong competitor against the Ti 4200 considering the price/speed/features ratio. If the cards become available at $120, it will be very good. 
  3. The RADEON 9700, having the same price as the GeForce4 Ti 4600, shows a very high level of performance although its frequency is much lower compared to the Pro version. 
  4. RADEON 9500 Pro wasn't tested as there are no such cards yet; they will probably come onto the scene when the demand for the RADEON 9700 weakens and prices of the RADEON 9500 fall down. 
  5. The production card from Gigabyte fully uses the ATI's design. I hope such an expensive PCB won't tell on the card's price. As usual, Gigabyte provides a rich set of accessories, though the games are quite old. Moreover, this company is almost the first to bring such card to the retail market. 
  So, ATI stormed into the market with a strong competitor against the GeForce4 Ti 4600, though it has nothing to oppose to the GeForce4 Ti 4200 in its price niche; we still don't know how the RADEON 9500 Pro will perform, I can only assume that it will be a direct competitor for the Ti 4200. Besides, it's possible that ATI releases a crippled chip of the smaller square and prime cost if it considers that the candle is worth the game. I wonder if its performance will differ from the original R300 working in the mode of RADEON 9500 which was dealt with today. 

  The DirectX 9.0 support will be a real advantage for a card priced at $120. It will let game developers use new technologies, and DirectX 9.0 oriented games will arrive sooner. 
   
   
   

  Andrey Vorobiev (anvakams@ixbt.com)
  Alexander Medvedev (unclesam@ixbt.com) 
   

  Write a comment below. No registration needed!