by Alexey Stepin , Yaroslav Lyssenko, Anton Shilov
10/06/2005 | 08:23 PM
As we revealed in our ATI RADEON X1000 architecture overview (for details see our article called ATI RADEON X1000: Brand-New Graphics Architecture from ATI Explored ), the new RADEON X1800-series visual processing units (VPUs) feature a lot of promising technologies and manage to demonstrate enough potential in synthetic tests. The time has come reveal performance of ATI RADEON X1800 XT and ATI RADEON X1800 XL graphics products in real-world games and benchmark applications in an attempt to find out whether the GeForce 7800-series is endangered by the new lineup from ATI.
<%BANNER[article]%>But before we proceed, allow us to remind some peculiarities about the RADEON X1800-series of graphics processors.
All-in-all, the RADEON X1800 features everything what it takes to be called a modern graphics accelerator. Let’s see how the new graphics cards from ATI – the RADEON X1800 XT and the RADEON X1800 XL – show up in real life.
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
Battlefield 2 uses a new DirectX 9.0 graphics engine as well as advanced physics engine, which makes it a computational resources-demanding game in general. While the graphics quality the game provides is really impressive, all the graphics cards deliver sufficient performance in Battlefield 2.
Due to framerate limiter, Battlefield 2 equally well on both RADEON X1800 XT and the GeForce 7800 GTX, but the GeForce 7800 GT seems to be a little bit faster compared to the RADEON X1800 XL in high resolutions.
Once full-scene antialiasing as well as anisotropic filtering are enabled, the RADEON X1800 XT and the X1800 XL start to outperform the rivals from NVIDIA Corp’s camp modestly. In 1600x1200 resolution with all the eye-candy enabled, the top-of-the-range ATI product offers 22% more performance in Battlefield 2 compared to NVIDIA’s GeForce 7800 GTX. The RADEON X1800 XL cannot boast with that great advantage over the rival and is only moderately faster than the GeForce 7800 GT.
Performance advantage of the RADEON X1800-series in the Battlefield 2 may be conditioned by faster memory speed (in case of the model 1800 XT) as well as more efficient memory controller (possibly, the case of the model X1800 XL).
The Chronicles of Riddick game is not that shader intensive, but uses OpenGL and takes advantage of NVIDIA’s UltraShadow technology, it better performs on the GeForce 7800 rather than on the RADEON X1800.
While the speed provided by the RADEON X1800 XT should be enough to play the game, it is outperformed by the GeForce 7800 GTX by 35%. Due to lower core and memory speeds, the RADEON X1800 XL is even slower compared to the RADEON X850 XT Platinum Edition (PE).
The gap between competitors does not shrink even in the eye candy mode. The RADEON X1800-series is constantly slower compared to the competing GeForce 7800 lineup in the Chronicles of Riddick game.
Doom III has always been a strong point for NVIDIA GeForce 6 and 7 hardware and it is no surprise that the RADEON X1800 hardware is behind the GeForce 7800 in this test. But take a look at the considerably narrower gap between performance of the model X1800 XT as well as the GeForce 7800 GTX!
This should probably be attributed to much higher frequencies as well as availability of 512MB frame-buffer on the latest RADEON. The RADEON X1800 XL cannot boast with speed higher than that of the RADEON X850 XT PE product.
We can observe the same picture in another demo sequence recorded in Doom III. While the numbers are generally higher, the gap between the top-end offerings has shrunk for sure. The RADEON X1800 XL, unfortunately, cannot manage to compete really well the GeForce 7800 GT.
We should note that virtually all high-end graphics cards today can play Doom III at comfortable speeds at up to 1280x1024 with all eye candy enabled. The GeForce 7800 GTX – the unconquered champ – can even allow to play at up to 1600x1200.
FarCry used to be a rather demanding game in the past, but today even with FSAA 4x and aniso set to 16x, nearly all high-end graphics cards can achieve 50 – 60 fps here.
Due to high dependence of our demo from the Pier level, every high-end graphics board demonstrates equal results, unless the antialiasing and anisotropy are enabled.
Once eye-candy features are enabled, all the graphics cards except the RADEON X1800 XT begin to quickly drop their speed. Performance hit on the RADEON X1800 XT, by contrast, is negligible – thanks to extreme memory speed, it is 34% faster than the rival in the most complex case. The RADEON X1800 XL could match allegedly more powerful GeForce 7800 GTX in our case.
The general situation in the Research demo of the Far Cry game is pretty surprising: despite of the fact that both ATI and NVIDIA hardware works using Shader Model 3.0 code-path, NVIDIA’s products manage to grab the lead away from ATI.
Given that even the RADEON X850 XT PE succeeds in leaving the RADEON X1800 XL behind, we believe there is something going on with the drivers for the RADEON X1800-series, but as for now, NVIDIA’s execution of Shader Model 3.0 code-path in Far Cry seems to be more efficient compared to ATI’s.
The eye-candy mode changes the situation by putting the RADEON X1800 XT impressively ahead of the GeForce 7800 GTX due to tremendous memory clock-speeds. The model X1800 XL cannot beat the model 7800 GT, even despite of allegedly more efficient memory controller.
Half-Life 2 game, just like the Far Cry, used to demand a high-performance graphics card. However, driver developers managed to boost the speed of their products and today all the high-end offerings run Half-Life 2 absolutely smoothly
There is hardly a point to talk about performance in cases with no image quality enhancements implemented, as all the graphics cards demonstrate about 100fps.
Surprisingly, in eye-candy mode the GeForce 7800 GTX manages to outperform the newcomer from ATI, which may either indicate that the 7800 GTX has much more raw pixel shader 2.0 power, or ATI’s RADEON X1800-series still needs better drivers.
Another demo of the Half-Life 2 showcases the same results as the first one: all the graphics cards are equally good.
But in the demo with highly-intensive action on the streets of the City 17 demonstrates that that the RADEON X1800 XT can, in fact, match the GeForce 7800 GTX in Half-Life 2. Just like GeForce 7800 GT can be matched by the RADEON X1800 XL.
The Painkiller game runs smoothly on all graphics cards of today.
Even with FSAA and anisotropic filtering, all high-end GPUs score more than 150 fps in Painkiller .
In F.E.A.R. multiplier beta demo, which is probably among the most demanding applications today, RADEON X1800 XT surprisingly cannot beat the GeForce 7800 GTX, possibly because the fact that the latter is more efficient when it comes to heavy usage of pixel shaders. It seems like the same is correct about the RADEON X1800 XL and the GeForce 7800 GT.
In the eye-candy mode the RADEON X1800 XT manages to grab the lead because of extra memory bandwidth, whereas the XL model cannot be that fast.
Pariah uses heavily-modified Unreal Engine and sports some blurring effects as well as bump-mapping; that’s is not a state-of-the-art technology today, but Pariah still can offer a thrilling gameplay.
All the graphics cards deliver awesome results, but as the resolution grows, the GeForce 7800 GTX take the lead. Perhaps, there is a problem with the ATI’s drivers for the new RADEON X1800-series, perhaps, computational power of the 7800 GTX is simply unbeatable?
Project Snowblind is a yet another sci-fi first-person shooter that greatly uses pixel shader 2.0 effects.
It has so far been considered that ATI’s RADEON X1800-series “sacrificed” Shader Model 2.0 performance efficiency in favour of some other capabilities, and, in fact, it does seem so, based on the results in Project Snowblind title. The RADEON X1800 XL cannot achieve results similar to those of the GeForce 7800 GT, whereas the RADEON X1800 XT seems to beat the GeForce 7800 GTX in terms of performance seemingly because of extreme clock-speeds of core and memory.
With full-scene antialiasing and anisotropic filtering activated, the RADEON X1800 XT continues to greatly outperform the GeForce 7800 GTX, nevertheless, the RADEON X1800 XL does not really seem to be better than the GeForce 7800 GT.
Game engine used in the Unreal Tournament 2004 was developed about three years ago and it does not push graphics accelerators to the boundaries. Basically speaking, all the modern high-end GPUs demonstrate pretty close results in the Torlan demo.
In the eye-candy mode we see that the GeForce 7800-series outperform the RADEON X1800-series in this demo. The reason is unknown, but, possibly the drivers, of the new ATI’s products is the matter.
Metallurgy seemingly puts higher load on the GPUs compared to the Torlan demo. Nevertheless, the GeForce 7800-series graphics processors greatly outperform their competitors among the RADEON X1800 family of products.
This game is pretty rich in textures and not very complex pixel shaders, which are used to create different visual effects, such as fog, halos, etc.
RADEON X1800 XT doesn’t have any opportunity to reveal its remarkable features that is why it is just a little bit ahead of the GeForce 7800 GTX. Note that despite that both graphics cards show similar minimal performance level.
In resolutions above 1024x768 RADEON X1800 XL fails to outperform the previous generation solution, namely RADEON X850 XT PE, not to mention GeForce 7800 GT. So, I would say that the performance of ATI newcomers in Prince of Persia: Warrior Within isn’t really that impressive.
This shooter is based on the modified Unreal engine. After some additional improvements and changes, the game started to support normal maps, per-pixel lighting and other contemporary technologies. It also supports Shader Model 3.0.
RADEON X1800 XT is not that much faster than GeForce 7800 GTX here, however ATI’s solution boasts much greater “safety factor”, due to significantly higher minimum fps rate. The younger model, RADEON X1800 XL, also competes with GeForce 7800 GTX quite successfully. As for RADEON X850 XT PE, it works in lighter conditions, because it doesn’t support Shader Model 3.0 – this is exactly why its results are so high in this test.
In eye candy mode RADEON X1800 solutions dash forward. RADEON X1800 XL owes this success to its efficient memory controller, and RADEON X1800 XT – to the memory controller, of course, and to incredibly high working frequencies.
The well-known race simulator called Colin McRae Rally 2005 uses a tremendous amount of relatively simple (for today) pixel shaders to create various effects, such as smoke, dust coming from under the tires, highlights, etc. The performance in this game depends a lot on the number of pixel processors and their computational power.
In this case there is parity between RADEON X1800 XT and GeForce 7800 GTX: the first one reaches these heights due to high GPU clock and memory frequencies, while the second one owes its high great results to 24 pixel processors. The high working frequency of the RADEON X1800 XL, however, fails to make up for the fewer pixel processors: 500MHz is definitely not enough to successfully oppose the 20 pixel processors of the GeForce 7800 GT working at 400MHz (at least with the memory working at the same speed). Moreover, here the XL solution yields even to the RADEON X850 XT Platinum Edition, which may indicate that the drivers could use some extra work.
Once we switch to the test mode with enabled FSAA and anisotropic filtering, the situation changes radically. RADEON X1800 XT gets far ahead of everyone else due to its memory working at 750 (1500) MHz and communicating with the GPU via a highly efficient controller. RADEON X1800 XL also outperforms GeForce 7800 GT, and in 1280x1024 even catches up with GeForce 7800 GTX.
Graphics adapters based on GPUs from ATI have never been very successful in flight simulators from Maddox games. And the primary reason for that was the inefficient OpenGL driver, because these games use this particular API by default. This API provides the best image quality for them.
Well, the new RADEON based graphics cards failed to work wonders here. RADEON X1800 XT and RADEON X1800 XL performed equally fast in low resolutions, which points at a definite issue with the OpenGL driver. However, in high resolutions RADEON X1800 didn’t show anything remarkable. Only in 1600x1200 with enabled anti-aliasing, the top ATI RADEON model manages to perform at least close to GeForce 6800 Ultra. Here I would like to point out that despite their evident defeat, ATI RADEON X1800 XT/XL based graphics cards provide acceptable gaming performance in all resolutions except only 1600x1200 with enabled FSAA and anisotropic filtering.
Lock On has often surprised us with pretty unexpected results, but this time there were no surprises: driver issues prevented ATI RADEON X1800 from reaching even the 28-30fps rate, which is the maximum any known graphics adapter has ever reached in this game. The indirect evidence that we should blame the driver again is the identical performance of the RADEON X1800 XT and XL models with up to 0.7fps precision. GeForce 7800 GTX/GT behave practically the same in pure gaming mode, however, unlike their ATI competitors they manage to reach the maximum performance rate.
This strategy game uses all versions of pixel and vertex shaders up to 2.0. The game features dynamic landscape, and all military vehicles and buildings cast very realistic shadows. As a result, the game requires a lot from the graphics subsystem, especially from the pixel shader performance.
When ATI designed its new graphics architecture, they focused primarily on increasing the pixel shader performance as much as possible, that is why RADEON X1800 XT and RADEON X1800 XL work so well in this game. The top model is so fast that you can enjoy this game even with enabled full-screen anti-aliasing. RADEON X1800 XL, of course, performs slower, but it is still enough to surpass GeForce 7800 GTX. The only exception is 1600x1200 resolution with enabled FSAA: here RADEON X1800 XL seems to be slowed down by its relatively slow memory. However, due to efficient the Ring Bus architecture it only loses 1fps to GeForce 7800 GTX.
Dawn of War is a classical real-time strategy, it doesn’t have any of those graphics “crotchets” the previous game, Perimeter, boasts. However, there can still be quite many military units on the screen at the same time, which imposes pretty heavy workload on the vertex processors.
Geometry processing speed is one of the strengths of the new RADEON X1800, however, it still yields to G70 based solutions in lower resolutions. As the resolution increases RADEON X1800 XT manages to catch up with the GeForce 7800 GT, but not the GTX: this guy stays beyond any competition. RADEON X1800 XL perform almost as fast as the RADEON X850 XT Platinum Edition.
The scenes in Aquamark3 the graphics adapter to be able to remove hidden surfaces efficiently enough: the overdraw coefficient in this benchmark is very high. Also, the performance of the vertex processors is very important in this test, because some scenes feature complex geometry.
RADEON X1800 XT is the leader here, but even though its hidden pixels removal techniques have been significantly improved, the gap between the leader and its closest rival, GeForce 7800 GTX, is very small: only 3-5fps, while the overall performance level is around 60-75fps. It seems to me that the newcomer wins only due to its higher core working frequency. The slower RADEON X1800 XL competes with GeForce 6800 Ultra in pure speed mode and with GeForce 7800 GT in eye candy mode with enabled FSAA and anisotropic filtering.
The Final Fantasy XI test has very heavy texturing workload. So the results in this benchmark depends on the ability of the testing participants to cope with this type of workload.
RADEON X1800 XT features only 16 pixel processors, while the top GeForce 7 has 24 of them, so even 625MHz graphics chip frequency cannot help ATI solution here. It loses, although not dramatically. Moreover, even RADEPN X850 XT PE appears faster than RADEON X1800 XT. It gave us some food for thought that maybe there are some serious issues with the current version of the CATALYST driver. We don’t know if this benchmark uses fixed TnL functions, but if it does, then the problem might be connected with the inefficient TnL emulation.
As is known, the Game 1 test from 3DMark03 testing suite uses only fixed TnL functions, and the other two tests use dynamic stencil shadows, which puts GeForce 6/7 into more favorable conditions. Only high working frequencies help RADEON X1800 XT to perform close to GeForce 7800 GTX. But as for RADEON X1800 XL, it cannot boast the same weapon, so it loses about 1000 points to the 20-pipeline GeForce 7800 GT. Now let’s see how the new ATI solutions behave in each of the 3DMark03 tests.
The new graphics card family doesn’t show its best in the applications using the fixed TnL functions: if in 1024x768 RADEON X1800 XT manages to retain its leading position, then in higher resolutions its performance drops down to the level of GeForce 7800 GTX. This is only true for modes with disabled FSAA. Once FSAA is enabled, the newcomer catches up with GeForce 7800 GTX due to the new memory controller with Ring Bus architecture. As for the RADEON X1800 XL, throughout the entire test session it yields not only to GeForce 7800 GT, but also to GeForce 6800 Ultra.
It is evident that the new ATI graphics solutions do not feel quite at home in applications using pixel shaders 2.0 and up. I assume, however, that the situation might improve once the new CATALYST driver version is out.
Game 2 benchmark uses normal maps and dynamic stencil shadows. The maximum shader versions it supports are 1.4 and 1.1 for pixel and vertex shaders respectively. RADEON X1800 XT again wins only thanks to super high clock frequencies: this is the conclusion we make after checking out the RADEON X1800 XL results.
Nevertheless, the result close to GeForce 7800 GT is not such a bad result for the XL model, I should say. It has only 16 pixel pipelines against 20 by the NVIDIA solution.
In eye candy mode with enabled FSAA and anisotropic filtering both RADEON X1800 solutions feel much better already: for example in 1600x1200 RADEON X1800 XL runs as fast as GeForce 7800 GTX. Nothing to be surprised at, actually: work modes with enabled anti-aliasing have always been one of RADEON’s trumps, and now ATI’s solutions feel even more confident in these modes due to the new advanced memory controller.
In Game 3 test the scenes acquire much more complex geometry: the Game 3 scene consists of 280,000 polygons against 150,000 in Game 2 test. Additional polygons are used to create realistic hairs on the main character’s head. Even though 8 vertex processors of the RADEON X1800 XT work at 625MHz and this frequency should provide the newcomer with a pretty significant performance advantage, it works slower than in Game 2 test and rests at the level of GeForce 7800 GTX. When we enabled FSAA and anisotropic filtering, RADEON X1800 XT makes it to the head of the race, but the gap between the leader and GeForce 7800 GTX is again much smaller than in the second gaming test. RADEON X1800 XL yields to GeForce 7800 GT in pure speed mode, and in eye candy mode with enabled FSAA outperforms the competitor only in high resolutions.
Game 4 test is a much more complex benchmark than the Game 2, Game 3 and especially Game 1, because it uses Shader Model 2.0, and besides uses something like HDR. The total number of polygons per scene reaches 800,000. It might seem like pretty favorable conditions for the RADEON X1000 architecture. However, the results of RADEON X1800 XL do not prove this point: it works not any faster than RADEON X850 XT Platinum Edition in the pure speed mode, even though it boasts enhanced pixel processors and 8 vertex processors.
When we enable FSAA, it manages to outperform the previous generation ATI solution, although the performance advantage of the newcomer doesn’t exceed 10%. Moreover, GeForce 7800 GT with a less advanced memory controller runs faster in this mode, probably due to higher computational power. RADEON X1800 XT again stays ahead due to its high working frequencies, however, it doesn’t leave the GeForce 7800 GTX too far behind in eye candy mode. It corresponds perfectly well with the results of our theoretical benchmarks where both RADEON X1800 didn’t perform that well with shaders using complex math1ematical calculations. As for the multi-pass shaders and branching shaders, there are none of those in 3DMark03, that is why RADEON X1800 XT/XL cannot really show off here.
The results of individual tests proved the total result we discussed in the beginning of this chapter. However, we shouldn’t forget that 3DMark03 has already become somewhat outdated, so the performance picture we get for contemporary graphics accelerators from this testing suite is not very objective any more. 3DMark05 should do the job much better, so let’s move on to it now.
3DMark05 is better balanced than 3DMark03: there are no tests favoring ATI or NVIDIA architecture here. Besides, this benchmarking set is more up-to-date from the technology standpoint as it uses a lot of complex pixel and vertex shaders and other contemporary features. Since RADEON X1000 architecture was designed to satisfy the today’s needs as well as to meet the standards of tomorrow, both RADEON X1800 models perform very well in 3DMark05. RADEON X1800 XT almost hits the 9000 point bar, while RADEON X1800 XL almost reaches the 7000 point, which is a today’s record for single ultra high-end and high-end graphics cards working at their nominal speeds. Now let’s find out what determined this great victory of RADEON X1800 in 3DMark05.
The “shooter” test contains a scene typical of any science fiction 3D shooter game. There are dynamic shadows generated with high resolution depth maps (2048x2048). The GPU is mostly loaded by numerous soldier models and complex lighting: there are 8 light sources.
RADEON X1800 shows its architectural advantages right away: the top model provides pretty acceptable fps rate even in 1024x768 resolution with enabled FSAA and anisotropic filtering. The younger model also performs quite well keeping up with GeForce 7800 GT in all modes and resolutions.
The second test loads mostly the vertex processors. Texturing workload is not that significant, as the action takes place in a night forest and you cannot see far. The test uses realistic light dissipation and dynamic shadows. In this case RADEON X1800 XL starts to get ahead of GeForce 7800 GT because of the higher working frequency of its vertex processors (500MHz against 440MHz by GeForce) and their higher capacity. The performance advantage is not very big: it doesn’t exceed 5-8%. Of course, RADEON X1800 XT is undefeated here: its working frequencies are a way higher than those of GeForce 7800 GTX.
The most complicated benchmark in the entire 3DMark05 package loads any GPU to the maximum. Canyon walls and water surface are generated with the most complex pixel shaders using Shader Model 2.0. For instance, the rock material is created with two color maps, two normal maps and diffuse shading. The water surface is rendered in six passes to ensure its highest realism. Pixel processors are loaded very heavily by the math1ematical calculations as well as texturing tasks. RADEON X1800 solutions win in pure speed mode, although the XL model cannot catch up with GeForce 7800 GT in 1600x1200.
The results in eye candy mode with enabled FSAA demonstrate lower computational power of RADEON X1800 XT compared against GeForce 7800 GTX. As the resolution grows, NVIDIA solution starts dashing forward even though the rival from the new RADEON family boasts higher working frequencies and intellectual memory controller. The performance of RADEON X1800 XL also drops faster in this mode than the performance of GeForce 7800 GT, so that their results level out in 1600x1200.
All in all, the results of the total performance analysis are quite logical: 3DMark05 uses 1024x768 resolution by default, and in this resolution RADEON X1800 XT is the best. However, RADEON X1800 XT may be not that impressive any more in higher resolutions, and sometimes with enabled FSAA, although it will still be faster than GeForce 7800 GTX. The same would be true for RADEON X1800 XL, which wins an indisputable victory over GeForce 7800 GT only in lower resolutions.
Today we observed performance of the most technologically advanced graphics card to date – ATI RADEON X1800 XT 512MB – as well as its less speedy brother, the RADEON X1800 XL. Based on the benchmark results we have obtained, we can say for sure that the new high-end lineup from ATI delivers incredible performance in all types of games popular today.
However, we cannot say that ATI’s new flagship product, the RADEON X1800 XT 512MB, is the absolute champion across the board, nor we can say that the XL model is a king of high-end, as both graphics accelerators are noticeably left behind by competing GeForce 7800 GTX and GT in games like Doom III, Chronicles of Riddick, Pacific Fighters and some other OpenGL applications, for example. Nevertheless, when it comes to modern games with intensive use of shaders, such as Battlefield 2, Colin McRae Rally, F.E.A.R., Perimeter, Splinter Cell: Chaos Theory and some others, the new RADEON X1800-series takes the lead.
Performance advantage over the competing GeForce 7800 GTX comes to the RADEON X1800 XT 512MB at the cost of tremendous clock-speeds as well as never-before-seen power consumption. However, we should keep in mind that high consumption of power seems to be a result of installation of 512MB of GDDR3 memory onboard.
In fact, 90nm process technology allowed ATI to keep power consumption of its RADEON X1800 XL graphics chip in the same envelope as that of the RADEON X850 XT PE, which has two times less transistors. Furthermore, die sizes of the RADEON X1800-series seem to match those on the RADEON X800 family, which means that the cost per die, provided that model X1800 and X800 yields are equal, remains approximately the same as with the previous generation products.
Currently there are several strong points the RADEON X1800 (or the R520, if you prefer) has: very good Shader Model 2.0 performance (at least, when it comes to games, generally the RADEON X1800 is not behind the GeForce 7800), the most efficient Shader Model 3.0 implementation on the PC to date (as theoretical tests showcase), FSAA support with HDR, Avivo video engine and hardware H.264 decoding support, higher-quality anisotropic filtering, great memory clock-speed of the model X1800 XT.
Unfortunately, the CATALYST drivers for the RADEON X1800-series still seem to need certain performance optimizations. Stability-wise, though, everything seems to be okay. This means that customers who get the RADEON X1800-series products today are advised to update their drivers later on.
While generally speaking the whole RADEON X1800-series family deserves a credit for nice performance and feature-set, it comes more than a quarter after the GeForce 7800 GTX, which offered similar speed and capabilities. Unless ATI manages to flood the market with the models X1800 XT and X1800 XL, success of the whole family will be under question mark. To say in short, ATI needs volume availability as soon as possible and, as a result, price drops, of the whole RADEON X1800-series in order to fight high-end market back from NVIDIA.
So, let us summarize everything up about both graphics cards we tested today.
ATI RADEON X1800 XT 512MB
Highs:
Lows:
ATI RADEON X1800 XL
Highs:
Lows: