by Alexey Stepin , Yaroslav Lyssenko, Anton Shilov
10/13/2005 | 05:25 PM
RADEON X1300 PRO 256MB – the most powerful entry-level product from ATI’s new breed of graphics chips – demonstrated serious potential in synthetic tests when it came to pixel shaders (for details please see our article called ATI RADEON X1000: Brand-New Graphics Architecture from ATI Explored ), but ignited doubts about possibility for the part to compete against others in real games due to pretty moderate results in tests that rely on geometry and vertex shaders: while the part has 4 pixel processors, it only sported 2 vertex processors, the same amount as in previous generation entry-level family RADEON X300.
<%BANNER[article]%>The RADEON X1300-series of products will replace the X300 lineup in the entry-level market and will bring users a bit higher speed, Shader Model 3.0, Avivo video engine and some other improvements. However, the mission of the RADEON X1300 is not only to offer a decent alternate to the RADEON X300, but also solidify ATI’s positions in the segment, where NVIDIA pushes pretty hard. Let us examine whether the RADEON X1300, or code-named RV515, is capable of doing so.
But before we proceed, we’ll bring you architectural highlights concerning the new part from ATI:
We used the following test system for our today’s test session with the new graphics accelerator family from ATI:
ATI and NVIDIA drivers were configured as follows:
ATI CATALYST:
NVIDIA ForceWare 78.01:
The following games and applications were our benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
Simulators
Real-Time Strategies
Semi-Synthetic Benchmarks
Synthetic Benchmarks
Since RADEON X1300 PRO is a budget solution, we shouldn’t expect it to perform very fast with enabled full-screen anti-aliasing. Therefore, for the eye candy mode we will consider only the 1024x768 resolution.
Wow, the testing session has just started and the results we see are simply impressive! RADEON X1300 PRO managed to provide acceptable performance in Battlefield 2 even in 1280x1024 letting only RADEON X700 get ahead. And all this was possible with only 4 pixel processors! Of course, the clock frequencies did have their effect here: RADEON X1300 PRO works at 600MHz chip and 400 (800) MHz memory frequency, while by RADEON X700 these numbers are 400MHz and 350 (700) MHz respectively. The same picture can be observed in eye candy mode: RADEON X700 is the leader here and RADEON X1300 PRO follows close behind him, yielding only 4fps to the winner. You can play just fine in this mode, however the performance will definitely drop below 30fps in “heavy” scenes. Note that unlike RADEON X1600, our today’s hero, RADEON X1300 PRO, looks much better balanced: it boasts 4 TMUs per 4 pixel processors. So, the only bottleneck of this solution can be the 2 vertex processors. However, high working frequencies should actually make up a little bit for so few vertex processors onboard.
Chronicles of Riddick is one of NVIDIA’s territories, as this game works via OpenGL API and uses UltraShadow technology very actively. As a result, RADEON X1300 PRO falls very far behind GeForce 6600 and in high resolutions its performance drops down almost to the level of GeForce 6200. When we enable FSAA and anisotropic filtering, the newcomer rests at the same performance level with GeForce 6600, because the major limiting factor in this mode appears the memory subsystem performance and GeForce 6600 cannot boast faster memory subsystem than the new ATI baby. In fact, these results are still of pure theoretical interest to us: you can play comfortably only in pure gaming mode in resolutions not exceeding 1024x768. AT least with the maximum level of detail.
Doom III belongs to the same type of games as the just discussed Chronicles of Riddick : it uses OpenGL and a lot of stencil shadows. Of course, these are far not the best conditions for ATI graphics cards, which have always suffered from the inefficient OpenGL driver. In this case RADEON X1300 PRO suffers an even more devastating defeat than in the previous game: despite high working frequencies it is beaten even by GeForce 6200 in all resolutions. Only in eye candy mode this graphics accelerator manages to catch up with the NVIDIA rival. Just like in the Chronicles of Riddick the minimum acceptable speed is revealed in 1024x768 with disabled FSAA and anisotropic filtering.
The same can be observed in d3dm4 level, although all the testing participants show somewhat better results here.
RADEON X700 features 8 pixel pipelines that is why it outperforms RADEON X1300 PRO, however, the latter also demonstrates pretty good results in the first two resolutions. At least, you can have enjoyable gaming experience in both of them. It is pretty surprising but GeForce 6600 falls behind the new ATI solution in the lowest resolution only, and in all other cases goes neck and neck with it/ it is hard to tell what is the reason for that: this graphics card works at only 300MHz chip and 250 (500) MHz memory frequencies, which is a way lower than the working speeds of the RADEON X1300 PRO. Moreover, GeForce 6600 features 8 pixel processors but only four rasterizers. Maybe the reasons are again originating from the not quite finalized ATI drivers. In eye candy mode RADEON X1300 PRO and RADEON X700 perform equally fast and even get beyond the acceptable minimum of 30-35fps.
The situation repeats at the Research level. Despite the Shader Model 3.0 support, RADEON X1300 PRO outperforms RADEON X700 only with enabled FSAA and anisotropic filtering having demonstrated excellent performance of 42fps.
I have to say that despite the attractive graphics and interesting gaming experience, Half-Life 2 is not such a demanding game at all. So all the today’s testing participants, except GeForce 6200 and RADEON X300 showed very acceptable results in all resolutions.
The performance in the Canals level depends a lot on the pixel shader 2.0 performance, especially with Reflect All option enabled. These shaders do not seem to be complicated enough for the RADEON X1000 architecture to show its potential to the full extent. That is why RADEON X1300 PRO yields about 20-30% to RADEON X700 here depending on the resolution. Nevertheless, the newcomer appears a worthy competitor to GeForce 6600. By shifting to eye candy mode with enabled FSAA and anisotropic filtering we bring the RADEON X1300 PRO to the second position in the race after RADEON X700. In this case, the performance gap between the two ATI solutions gets limited to 8-10% due to a more efficient memory controller of the new RADEON X1300 PRO and higher memory frequency of 400 (800) MHz against 350 (700) MHz.
RADEON X1300 performed much better in the street battle scene recorded on d3_c17_02 game level: here our hero yielded only to RADEON X700 with 6 vertex processors onboard working at 400MHz frequency, which is evidently more than RADEON X1300 could do with geometry processing. In addition to that, RADEON X700 features 8 fully-fledged pixel pipelines. Unfortunately, the performance gap between these two cards increases together with the resolution: from 8% in 1024x768 up to almost 20% in 1600x1200. Also, the newcomer yielded about 8%-9% to the RADEON X700 in eye candy mode.
Painkiller is not a demanding game, and the performance of our RADEON X1300 PRO drops below the 100fps level only in 1600x1200. In fact, 91 fps framerate is more than enough for comfortable gaming. In eye candy mode the new ATI solution is even ahead of RADEON X700, most likely because of its higher working frequencies.
This gaming test is the most demanding one, none of the testing participants manages to reach even 30fps performance although RADEON X700 does get very close to that. Despite more advanced pixel processor architecture, RADEON X1300 PRO falls behind the RADEON X700 by 7-8% unfortunately. Enabled 4x anti-aliasing and anisotropic filtering cause the performance of all testing participants to drop below 10fps. So, if we disregard the fact that it is impossible to play at this framerate, then we can conclude that RADEON X1300 PRO competes quite successfully with RADEON X700.
Although the Pariah gaming engine uses different effects created with the help of pixel shaders, these shaders are not very complex and hence they do not give any advantages to the new RADEON X1300 PRO. In all resolutions RADEON X700 performs faster. Just like in the previous cases the performance of RADEON X1300 PRO drops faster with the increase of the screen resolution than that of GeForce 6600.
The shaders used in Project: Snowblind game are complex enough for the new generation RADEON solution to be able to demonstrate its advantages. RADEON X1300 PRO proves a worthy competitor to RADEON X700, although the performance of the latter doesn’t slow down as fast as that of the RADEON X1300 PRO when moving from lower resolutions to higher ones. Although in 1600x1200 none of our today’s testing participants can guarantee acceptable performance rate. In eye candy mode with enabled FSAA and anisotropic filtering the situation remains the same.
According to our theoretical testing results, the fillrate provided by RADEON X1300 PRO is not that bad at all, especially in case there are a lot of textures involved. As a result, this solution can compete successfully with RADEON X700 in a not very complex game like Unreal Tournament 2004 . Besides, RADEON X700 (RV410) is known to have huge issues with texturing because of smaller caches, so no wonder RADEON X1300 PRO won this round.
The same can be seen on the Metallurgy level. Here the advantage of RADEON X1300 PRO is even higher than in the previous case, because the action takes place mostly indoors. Of course, RADEON X700, RADEON X1300 and GeForce 6600 provide sufficient performance level for comfortable gameplay in Unreal Tournament 2004 even in 1600x1200 or in 1024x768 with enabled full-screen anti-aliasing and anisotropic filtering.
Even though RADEON X1300 PRO shows practically the same fillrate as RADEON X700 in a scene with a lot of textures (you shouldn’t forget that the latter also suffers from some texture caching issues), its performance in Prince of Persia is quite poor: it falls more than by 40% behind the competitor in 1600x1200. It is probably the fact that it features only two vertex processors against 3 vertex processors of GeForce 6600, even though the latter works at lower frequencies.
Splinter Cell: Chaos Theory supports Shader Model 3.0, and in this mode the image quality is much higher than in SM 1.1 and SM 2.0. Of course, you have to sacrifice something for the sake of great visual quality that is why RADEON X1300 PRO processing pixel shaders 3.0 yields to RADEON X700 working in a lighter mode. The performance difference as we can see is about 20%-25%. In eye candy mode the gap shrinks to less than 10% difference. Note that the performance of competing ATI solutions remains quite high for comfortable gaming even though the best performance margin still belongs to RADEON X700.
The number of pixel processors and their working frequency exert the most influence on the graphics cards performance in Colin McRae Rally 2005 . It doesn’t matter that much how efficient they are with complex shaders, because there are none in this game. RADEON X700 features 8 pixel processors, while RADEON X1300 PRO has only four of them. And even though higher working frequency makes up for it in a certain way, it is still not enough for the newcomer to win the first prize, so RADEON X1300 PRO has to put up with being the second fastest. Unfortunately, we haven’t answered the question about what prevents RADEON X1300 PRO from winning in high screen resolutions: in 1600x1200 its speeds almost drops down to the level of GeForce 6600. As for the eye candy mode with enabled full-screen anti-aliasing, the performance gap between RADEON X1300 PRO and RADEON X700 is minimal in this case, at least in low resolutions. You should also keep in mind that the new ATI solution offers smaller “safety factor” than the predecessor: just look at the numbers for minimal performance.
Even the acceptable minimum in Pacific Fighters is a way beyond the capabilities of our RADEON X1300 PRO: this game uses OpenGL. However, RADEON X700 cannot boast anything better either.
The same is true for Lock On, although this flight simulator doesn’t use OpenGL. The 20fps framerate with the minimum of 7fps can hardly be regarded as acceptable gaming conditions. So, RADEON X1300 PRO, as well as RADEON X700, will never become the choice of flight simulator lovers.
Since RADEON X1300 PRO copes pretty well with complex pixel shaders, it runs almost neck and neck with RADEON X700 in Perimeter game, which is very rich in shaders. The performance in this case is not high enough to ensure comfortable gaming however if we compare the results of our hero with those of the other testing participants, the newcomer will look just fine.
The scenes in the War of Dawn game are rich in high resolution textures and complex geometry: there can be more than 100 models in a single frame, each of which consists of about 2,000 polygons. RADEON X1300 PRO had no problems with the fillrate, however it yields to RADEON X700 in geometry processing speed. GeForce 6600 catches up real close with the ATI solutions in low resolutions, because GeForce 6600 supports UltraShadow technology (and this game uses stencil shadows). In 1600x1200 RADEON X1300 PRO again gives in, as we have already seen multiple times today, but as soon as we enabled FSAA and anisotropic filtering, it managed to get ahead of all the other testing participants. The game requires quite a lot of resources from the graphics accelerator, I should say, that is why you may encounter problems even in the 1024x768 resolution. Even here the minimum performance score doesn’t ever go beyond 20fps.
Aquamark3 uses a lot of Shader Model 1.1 shaders, while there are not that many other shaders there: for example, you can only come across 3 vertex and 4 pixel shaders from Shader Model 2.0. The major workload here falls upon the vertex processors and hidden surfaces removal algorithms. The performance of two vertex processors of our RADEON X1300 PRO is about as high as that of three vertex processors of GeForce 6600 during simple vertex shaders processing and much higher than that during complex vertex shaders processing.
The results obtained in this test are quite logical: RADEON X1300 PRO outperforms GeForce 6600 in low resolutions and runs equally fast with the rival in high resolutions. RADEON X700 is beyond any competition in pure speed mode, as well as in eye candy mode with enabled FSAA and anisotropic filtering. In the latter case GeForce 6600 slows down dramatically because of its slower memory.
This benchmark is very likely to require high fillrate and fast geometry processing. The only one that outpaces RADEON X1300 PRO in these parameters is RADEON X700, so the second prize won by our today’s main hero is well deserved.
Although 3DMark03 favors GeForce 6/7 graphics card families more than the ATI solutions, RADEON X1300 PRO manages to get ahead of all, at least in the total score chart. What determined this victory for a solution featuring only 4 pixel and 2 vertex processors?
Game 1 test requires high fillrate and efficient fixed TnL emulation. The entire RADEON X1000 family as well as RADEON X1300 PRO in particular have some issues with this, as we have already found out. Despite this fact, RADEON X1300 PRO retains its stable second position after GeForce 6600 in all resolutions up to 1600x1200. Unfortunately, in the highest resolution the newcomer surrenders even to GeForce 6200, which can only be explained by the raw drivers at this point. So far, raw drivers have been a real pain for the entire RADEON X1000 graphics card family. Enabled full-screen anti-aliasing and anisotropic filtering allow RADEON X1300 PRO to take the lead. This graphics card boasts the fastest memory subsystem of all the tested graphics solutions that is why it performs so well in eye candy mode.
High GPU frequency, efficient work with the Z-buffer, and relatively powerful memory subsystem help RADEON X1300 PRO to win in Game 2 test from the 3DMark03 suite. Only in 1600x1200 with disabled FSAA and anisotropic filtering the new ATI solutions slows down for some reason.
We observe a pretty similar picture in Game 3 test. RADEON X1300 PRO copes quite well with the increasing geometrical complexity of the scene despite the availability of only two vertex processors.
Game 4 3DMark03 test is much more complex than any of the previous three: it contains about 800,000 polygons and uses a lot of pixel and vertex shaders 2.0 as well as Overbright effect. The pure computational power of the graphics processor is very important here. From this standpoint, RADEON X700 seems to be the best solution that is why it wins the race. RADEON X1300 PRO falls about 15% behind the leader, but keeps the good second place throughout the entire test, and after switching to eye candy mode even catches up with the leader. Although Ring Bus architecture of the new RADEON X1300 is not involved here, its memory controller is still more efficient than that of the RADEON X700, and its memory bandwidth is higher: 12.8GB/s vs. 11.2GB/s.
So, the total score of the new ATI solution as we have just seen in the first diagram has been absolutely justified: RADEON X1300 PRO outperformed RADEON X700 in the first three tests and yielded only 10% of its performance to the closest rival in Game 4 test. The total scores could have been different in higher resolutions, because RADEON X1300 PRO was defeated by the previous generation solution in all four tests in 1600x1200. However, we all know that the default resolution for 3DMark03 is 1024x768, so, the laurels indisputably go to the today’s newcomer - RADEON X1300 PRO.
In 3DMark05 RADEON X1300 PRO and X700 run almost neck and neck with each other: the performance difference is only 15 points, which can be regarded as a measuring error. Let’s take a look at each of the three game tests included into this testing suite.
As we have just seen in our articles devoted to RADEON X1800 and RADEON X1600 graphics card series, the new ATI architecture proves very efficient in Game 1 test that uses dynamically generated shadows and 8 light sources. RADEON X1300 PRO proved this fact once again. Although the performance difference was not that high it still managed to outperform RADEON X700 in pure speed mode. In eye candy mode the advantage over the predecessor reached 40%, which is a great result.
Game 2 test loads the vertex processors much heavier as it generates a lot of vegetation. Two vertex processors of the RADEON X1300 PRO cannot compete with six RADEON X700 vertex processors, even despite the architectural advantages and higher working frequencies. Therefore, we are witnessing a completely different picture from what we have just discussed in Game 1 test. Full-screen anti-aliasing and anisotropic filtering allow RADEON X1300 PRO to win the first prize, however, forget about the 40% advantage: in the best case it will be 10% ahead of the closest competitor.
The last Game 3 test from the 3DMark05 suite loads the pixel processors as well as their TMUs to the maximum. Here we observe parity between RADEON X1300 PRO and RADEON X700 in pure speed mode. The latter has some problems with massive texturing, although in this case the number of pipelines compensates for it. Just like in the previous cases, FSAA and AF allow new ATI solution to dash forward. The performance advantage is not as impressive as in Game 1 test, of course, but it is also not as small as in Game 2 test: 18%.
Throughout the entire 3DMark05 session RADEON X1300 PRO demonstrated very good performance for the graphics accelerator of its class. It yielded to RADEON X700 only in one test that is why the victory in the total score chart is absolutely deserved. Unlike the situation in 3DMark03, we didn’t see any dramatic performance hits in 1600x1200, so we wouldn’t question the final verdict here.
Speaking in general, the RADEON X1300 PRO is a pretty fine performer for a casual user. Sometimes it is a bit faster than the RADEON X700, sometimes it is a bit slower, but in addition it brings Avivo video engine. For 3D gamer, the problem for the part is that at $149 price-point it will not compete against the RADEON X700, but against the RADEON X800 GT and the GeForce 6600 GT that both radically outperform even the RADEON X700 PRO in absolute majority of games.
Provided that the price is right for a 3D gaming enthusiast, e.g., the RADEON X1300 PRO costs the same, or a bit more, amount of money as the GeForce 6600 and the RADEON X700, the part is an incredible performer – it beats the predecessor and competitor in almost all benchmarks save for Doom III and Chronicles of Riddick games. But given that the RADEON X1300 PRO 256MB costs the same amount of money than a more powerful RADEON X800 GT, the part is unlikely to become popular among games shortly.
Video enthusiasts may find Avivo video engine an interesting option and get the RADEON X1300 PRO just because of improved video image quality and H.264 hardware decoding. Furthermore, some future proof in terms of video processing may appeal to people who acquire computers for educational purposes or for kids, who are definitely looking forward new generation movies on HD DVD or Blu-ray discs.
To sum it up, the RADEON X1300 PRO at this point and at $149 price-point is definitely not the best option for gamers. However, for casual users, who are more interested in, for example, video editing, the model X1300 PRO may be a fine choice.
Highs:
Lows: