by Alexey Stepin
09/14/2005 | 01:58 PM
As you know from our earlier reports, the ATI RADEON X700 PRO graphics card is slower than the GeForce 6600 GT in nearly all current games due to the difference in their clock rates: 420/864MHz against 500/1000MHz. The ATI card has more memory on board (256 megabytes), but the larger amount of memory doesn’t help it much in low resolutions without full-screen antialiasing (as you can learn from our review, the amount of memory doesn’t affect the performance of a graphics card much in such display modes). The 6 vertex processors don’t give the ATI card an advantage, either, since it is the pixel shader performance (directly dependent on the GPU frequency) that affects the speed of modern games the most.
In a lower price category, however, NVIDIA’s GeForce 6600 meets ATI’s RADEON X700 which has much better technical characteristics. The default clock rates of the RADEON X700 are 400/700MHz (GPU/memory) and you can compare this with the GeForce 6600’s 300/500MHz (GPU/memory). The difference in clock rates does show up in real applications as our full-scale testing of today’s graphics hardware showed: the RADEON X700 is really almost always faster than the GeForce 6600.
High-performance graphics cards from different manufacturers use the reference PCB design as a rule and often differ from each other only in the color of the lacquer coating of the PCB and the design of the cooler, and sometimes in the clock rates. It’s quite different at the bottom of the market where inexpensive solutions use a variety of PCBs, memory types, cooling systems, etc.
Today we’re going to talk about PowerColor’s version of ATI RADEON X700. Will the PowerColor Bravo X700 graphics card get our “bravo!” for its performance at the end of the review? Let’s see.
The package of the PowerColor Bravo X700 graphics card should be already known to our readers from our other reviews of PowerColor’s products – its design hasn’t changed since. The text on the front of the box indicates that we deal with a version of RADEON X700 with 256MB of GDDR2 memory, two DVI connectors and with support of HDTV display devices. The picture in the center also promises a silent cooling system on heat pipes.
We opened the box and found the following:
The user’s manual is comprehensive, but is somewhat out-dated. It describes the process of connecting HDTV devices to the card, but the driver settings are described only for the Catalyst Control Panel rather than for the more up-to-date Control Center. The manual is in English only. The software bundle isn’t very generous as it includes only one and not very fresh game, although the PowerColor Bravo X700 belongs to the GameFX series. The CyberLink software isn’t quite appropriate here, either. The card is not equipped with a Rage Theater chip and cannot capture video from external sources, so the only really useful program on that CD (with this particular graphics card) is the software DVD player.
The quality of the S-Video and RCA cables might be better, too. On the other hand, if you really want to use these video interfaces, you can separately purchase high-quality and longer cables.
So, the accessories to the PowerColor Bravo X700 are on the whole sufficient for using the product, but the software bundle might be better. But anyway, it’s the card that we’re actually interested in, so let’s have a closer look at it.
The PowerColor Bravo X700 seems to resemble the RADEON X700 PRO at first sight, but there are still some important differences. Particularly, it has two DVI connectors and a SiI1162 transmitter from Silicon Image that is responsible for the second DVI output. The elements in the power circuit are placed in a different way and the circuit itself includes more electrolytic capacitors.
The complexity of the power circuit is justifiable as this graphics card uses GDDR2 memory that consumes more power than GDDR3. Despite the number “2”, GDDR2 is in fact a version of ordinary DDR and, like DDR, works at VDD = 2.5V, but has a lower VDDQ (1.8V). GDDR2 memory works at higher frequencies than typical PC3200 modules, so its power consumption and heat dissipation are much higher. And these are the reasons why GDDR2 is rarely employed on modern graphics cards.
We know about three cases when GDDR2 was ever used: on ATI RADEON 9800 PRO 256MB, on NVIDIA GeForce FX 5800 Ultra and on GeForce FX 5700 Ultra (in the latter case the manufacturers soon gave up this memory type and replaced it with the more economical and cold GDDR3). GDDR3 traces its origin to DDR2, works at a 1.8-2.0V voltage (both VDD and VDDQ) and generates much less heat. We have no idea why PowerColor used GDDR2 in the reviewed product, but it is a fact that the memory they put on the PowerColor Bravo X700 is not of the best type possible.
The placement of the memory chips is unique, too. At least we haven’t seen a similar PCB layout on any other card. Like the RADEON X700 PRO, the PowerColor Bravo X700 carries eight 256Mbit chips on board that yield a total of 256 megabytes of graphics memory. The chips are placed in pairs on both sides of the PCB crosswise rather than along the edges, as usual. Again, it’s the first time we see this way of placing memory chips and we don’t know the reasons for using it.
Samsung’s memory chips are marked as K4N56163QF-GC37 and this gives us some thinking matter, too. The company does not manufacture such GDDR2 chips anymore, offering K4N56163QF-GC36 chips with a higher rated frequency instead – 275(550) MHz against 266(533) MHz of the older version.
PowerColor must have employed GDDR2 because it had such obsolete chips in stock or because they managed to get a big lot of such chips at a discount. It also means that the PowerColor Bravo X700 graphics card has a reduced memory frequency as compared with the reference RADEON X700 that works at 350 (700) MHz. The PowerStrip utility agrees with us, showing that the memory of the PowerColor Bravo X700 really works at 266 (533) MHz, i.e. at its rated frequency. This Samsung memory isn’t very overclocker-friendly, so we don’t have much hope to reach even 350 (700) MHz – we’ll check this in the appropriate section of the review. As for the graphics processor, it works at 400MHz, like on the reference RADEON X700.
It’s not the first time we meet products from PowerColor with down-clocked frequencies – see our article called Mainstream PCI Express Graphics Cards Review: PowerColor X600 XT vs. Albatron Trinity PCX 5750 , for example. We don’t approve of this practice since the customer is misled, even though unintentionally. Yet you shouldn’t forget that many versions of the RADEON X700 come even with TSOP-packaged DDR memory clocked at 250 (500) MHz frequency. The PowerColor Bravo X700 looks preferable to such products, of course.
The cooling system deployed on the PowerColor Bravo X700 closely resembles the passive Thermaltake CL-G0003 cooler (for details see our article called Thermaltake CL-G0003 Passive VGA Cooler Review ) and even has the same number of heat pipes. The heatsinks are smaller here, however, so the total heat dissipation area is smaller, too. Yet it is big enough for cooling the RV410 chip manufactured on 0.11-micron tech process and working at a relatively low frequency. If your system case is small or badly ventilated, you can install the enclosed 60mm fan (or any other fan of an appropriate size) on the reverse-side heatsink. You should do so if you are going to overclock the card or are just not content with the efficiency of the passive cooler.
The heatsinks are rather massive, and there are special rubber pads between them and the card’s PCB to prevent the heatsinks from shifting aside and damaging the device. Against our expectations we found no thermal interface between the memory chips and the heatsinks. If there’s poor ventilation inside the system case, the lack of such interface may lead to overheat. The GPU die contacts the heatsink base through a layer of pink-colored thermoplastic material. There is no protecting frame on the GPU. The heatsink isn’t polished well – you don’t only see the traces of the milling machine, but can even feel them with your fingers. As for the spots where the heat pipes are fastened, they are all generously covered with silicone thermal paste. So, the Bravo X700 is equipped with a classic system on heat pipes. Such solutions have long proved their efficiency.
The cooling system of this graphics card is completely silent in the passive mode. If the additional fan is installed, there is some noise, of course, but the fan is not too loud, at least no louder than the other system components. And you also have an opportunity to replace it with another 60mm fan with better acoustic characteristics. So, the PowerColor Bravo X700 is blameless as concerns noise.
In our overclocking experiments, however, the card simply failed. We met visual artifacts in 3DMark03 and Unreal Tournament 2004 as soon as we increased the memory frequency to 300 (600DDR) MHz. The artifacts didn’t vanish even at the default frequency of 266 (533DDR) MHz, and we only got rid of them by setting the memory frequency at 250 (500DDR) MHz. So, we kind of got a card with negative overclockability for the first time in our tests. The memory chips couldn’t overheat because we set an additional 120mm fan to blow at the card, so we must have had a defective sample. The graphics processor could overclock to 420MHz; some visual artifacts appeared at 430MHz and at 440MHz the card would hang up in 3DMark03. So, we have to confess that our sample of the PowerColor Bravo X700 was no good for overclocking at all.
The image quality in 2D applications this card provides is quite high. The image remained sharp, without blurring or shadows, on our Dell P1100 monitor even in 1600x1200@85Hz and 1800x1440@75Hz display modes.
We used the games’ integrated benchmarks if possible. If the benchmark could produce the minimal fps rate besides the average one, we put it into the table of results and marked it in white. Games without built-in benchmarking tools were tested using the FRAPS utility and the minimal fps rates are also indicated in the diagrams.
We tested graphics cards with the PCI Express x16 interface on the following testbed:
Sticking to our standard testing procedure, we set up the drivers from ATI and NVIDIA in the following way:
ATI CATALYST 5.7:
NVIDIA ForceWare 77.72:
We turn on full-screen antialiasing and anisotropic filtering from the game menu, if possible. Otherwise, we force the necessary mode from the driver. We do not edit the games’ configuration files. We select the highest graphics quality settings in each game, the same for graphics cards from ATI and NVIDIA, but choose the rendering mode depending on the graphics card’s capabilities, i.e. Shader Model 2.0 and 2.0b for graphics cards on ATI’s GPUs and Shader Model 3.0 for products on NVIDIA’s GPUs.
Besides the PowerColor Bravo X700, we tested the following graphics cards:
The GeForce 6600 and RADEON X700 graphics card models were emulated by down-clocking the GeForce 6600 GT and RADEON X700 PRO, respectively. We didn’t test the PowerColor card in the overclocked mode since it turned to have a very poor overclocking potential.
The list of games and benchmarks used follows below.
First Person 3D Shooters:
Third Person 3D Shooters:
As expected, the PowerColor Bravo X700 is somewhat slower than the ordinary RADEON X700 with 400/700MHz clock rates. The gap is about 10% and doesn’t depend on the test mode and resolution. The GeForce 6600 family don’t show their best in this game, the junior model even failing to outperform the long-obsolete RADEON X600 XT.
The RADEON X700 can’t beat the GeForce 6600 in the “pure speed” mode as the game engine that actively uses stencil shadows is better suited for the GeForce 6/7 architectures. In the “eye candy” mode with enabled full-screen antialiasing and anisotropic filtering the standard model overtakes the competitor, but the PowerColor card can’t do the same due to its slow memory. It’s only in 1600x1200 that the PowerColor Bravo X700 has the same speed as the GeForce 6600.
The reduced memory frequency of the PowerColor card leads to its being just a little step ahead of the RADEON X600 XT in the “pure speed” mode. The situation improves somewhat in the “eye candy” mode, but the Bravo X700 still resides on the last but one place. The gap between the PowerColor and the standard RADEON X700 is about 20% in the highest resolution! We should also note that this game is optimized for the GeForce 6/7 architecture and uses the OpenGL API, so among ATI’s cards with their less efficient OpenGL driver the RADEON X700 PRO is the only one to ensure a comfortable frame rate in Doom 3 , and only in 1024x768.
It’s the same on the d3dm4 map, but the RADEON X700 manages to overtake the GeForce 6600 in high resolutions of the “eye candy” mode. The PowerColor Bravo X700 is just a little slower than the entry-level solution from NVIDIA. It’s important to have fast memory after all, and 500MHz is not sufficient for playing modern games.
The PowerColor is almost as fast as the GeForce 6600 in the “pure speed” mode, the difference between them no bigger than 3fps. However, when full-screen antialiasing and anisotropic filtering are enabled, the Bravo X700 uses its efficient memory controller to gain an advantage of up to 20% in 1600x1200 over its opponent. Meanwhile, the RADEON X700 is even faster than that, so the Bravo is still a loser.
Unlike on the Pier map, the PowerColor Bravo X700 can’t beat the GeForce 6600 at the “eye candy” settings – the memory subsystem speed doesn’t have a big effect on the card’s performance in scenes with closed environments. Of course, the RADEON X700 is still on top, about 20% ahead of the PowerColor.
The GeForce 6600 GT is the only participating graphics card to yield more than 30fps in this game. The Bravo X700 is 20% and 25% ahead of the GeForce 6600 in the “pure speed” and “eye candy” mode, respectively.
Not as demanding as F.E.A.R. , Half-Life 2 runs comfortably fast on all the participating graphics cards even in 1600x1200 at the “pure speed” settings. The PowerColor Bravo X700 is slower than the RADEON X700 in both test modes, though. In the “eye candy” mode it delivers the same performance as the GeForce 6600, excepting 1600x1200 resolution. Well, there are only two graphics cards that ensure any comfort in this resolution – the RADEON X700 PRO and the GeForce 6600 GT.
The scene we recorded on the d3_c17_02 map is different, but produces the same results: the PowerColor Bravo X700 is faster than the GeForce 6600 but slower than the standard RADEON X700 in all resolutions. The gap between the two versions of RADEON X700 amounts to 20% again.
The sequel to Painkiller isn’t a very resource-consuming application, and the PowerColor card gives out a playable frame rate even in 1600x1200 with enabled full-screen antialiasing. It also keeps closer to the RADEON X700 here than in the other cases – the load on the graphics memory subsystem must be low in this game.
The Bravo X700 is no more than 10% slower than the reference RADEON X700 in Pariah , too. This game doesn’t support full-screen antialiasing at the best visual quality settings, just like many other third-person shooters that use the Bloom effect. The PowerColor is also slower than the GeForce 6600 in high resolutions, even though it wins in 1024x768.
There are no high-resolution textures in Project: Snowblind and the memory load is low in the “pure speed” mode. That’s why the Bravo X700 is almost as fast as the RADEON X700 and leaves the GeForce 6600 behind thanks to the higher GPU frequency. When the memory load becomes higher in the “eye candy” mode, there’s a bigger gap between the two RADEON X700 incarnations. On the other hand, you can only play this game comfortably on an entry-level graphics card in resolutions no higher than 1024x768.
Unreal Tournament 2004 is a rather easy test, but it features complex and beautiful textures. Considering that the Torlan map is a huge open environment, the importance of fast memory grows even higher. The results agree with theory: the Bravo X700 is quite far behind the RADEON X700 and cannot outperform the GeForce 6600 like the RADEON X700 does. In the “eye candy” mode, however, the efficient memory controller keeps the Bravo X700 on the same level with the GeForce 6600.
The gap between the PowerColor Bravo X700 and the RADEON X700 is small on the Metallurgy map – the action goes on indoors thus putting a smaller load on the graphics memory subsystem. In the “eye candy” mode, however, the reference RADEON X700 proves to be faster again, the Bravo X700 equalling the GeForce 6600.
Price of Persia: Warrior Within owes its beauty to pixel shaders mostly, so the frequency of the graphics processor becomes the main performance-determining factor in this test. And still we can see how the Bravo X700 suffers from having slow memory – even though it ensures a comfortable frame rate even in 1600x1200, it is about 8-12% behind the standard RADEON X700.
Although the GeForce 6600 works in the Shader Model 3.0 mode of this game, it competes with the PowerColor Bravo X700 not only at the “pure speed” but also at the “eye candy” settings. The gap isn’t bigger than 10% in the hardest test mode, but the game looks much better on the GeForce. Note, however, that even the reference RADEON X700 cannot ensure a comfortable frame rate even in 1024x768 at the maximum graphics quality settings, so you’ll need something like GeForce 6600 GT or RADEON X800 GT to play this game.
The difference between the PowerColor’s and ATI’s versions of RADEON X700 is very small even with enabled FSAA – the GPU frequency and the number of pixel pipelines are the main performance-determining factors in this game.
Strangely enough, the PowerColor Bravo X700 isn’t much worse than the RADEON X700 in Lock On , even though flight simulators generally put a high load on the graphics memory subsystem.
Graphics cards with ATI’s GPUs find it hard to compete with NVIDIA’s solutions in flight simulators from Maddox Games. These games use OpenGL by default and the support of this API is still deficient in ATI’s driver. So, it is no wonder the humble GeForce 6600 easily beats the dangerous opponents from the ATI camp. The PowerColor Bravo X700 is about 10% slower than the reference version in both test modes.
The GeForce 6600 GT is the only graphics card in this test to offer a playable speed in this resource-consuming strategy game. Even the RADEON X700 PRO can hardly notch 20fps, not to mention the less advanced products. The PowerColor Bravo X700 is ahead of the GeForce 6600 in low resolutions, but their speeds become similar in higher display modes. These data are of no practical use, however, as you cannot play a game at a speed of 5-15fps.
The gap between the RADEON X700 and the PowerColor Bravo X700 diminishes in higher resolutions, although it should be otherwise (high resolutions mean a higher memory load). The entry-level cards deliver playable frame rates only in 1024x768 in the “pure speed” mode. Full-screen antialiasing is for products of higher categories here.
There are few complex shaders in Aquamark3 , but the test scene contains open environments, so fast memory would be welcome. The PowerColor Bravo X700 has slow memory and can only compete with the GeForce 6600. The PowerColor is a little faster than its opponent in the “eye candy” mode, but the standard RADEON X700 is still about 20% ahead.
This test depends greatly on the performance of the vertex processors and, accordingly, on the GPU clock rate. Both versions of RADEON X700 have the same GPU frequency, so there’s a minimal difference between them. The GeForce 6600 is on the losing side because of a lower GPU clock rate.
The PowerColor card scores about 1100 points less than the standard RADEON X700 and about 400 points less than the GeForce 6600. That’s the result of putting slow memory on this card.
The difference between memory frequencies of 700MHz and 533MHz can be seen well in the first test: it’s only in the “eye candy” mode that the PowerColor competes with the GeForce 6600.
The PowerColor Bravo X700 is still slower than the GeForce 6600 in the second test, even though the gap is very small. The latter features the UltraShadow II technology which helps a lot to render the test scene with its stencil shadows.
It’s different in the third test: a considerable load on the vertex processors is added to the dynamically generated stencil shadows. But even 6 vertex processors of the Bravo X700 don’t help it beat the GeForce 6600 which has only 3.
The fourth test is the most difficult one as it uses version 2.0 pixel shaders and is complex in terms of geometry. The high frequency of the graphics core and the six vertex processors of the PowerColor Bravo X700 help it surpass the GeForce 6600 in the “pure speed” mode, and keep on the same level in the “eye candy” mode. So, the 3DMark03 tests agree that slow memory is the main performance-limiting factor of the PowerColor Bravo X700.
3DMark05 sees it in a slightly different way: the Bravo X700 is still worse than the RADEON X700, but the GeForce 6600 has an even smaller score despite its support of Shader Model 3.0.
The Bravo X700 is still the same 20% behind the RADEON X700 in the first, shooter-like test. Yet the PowerColor card successfully competes with the GeForce 6600, except in the “eye candy” mode where even its more efficient memory controller can’t help. The PowerColor comes with 256MB of graphics memory, so it can pass the test in all resolutions with enabled full-screen antialiasing, while the GeForce 6600 cannot.
The second, “role-playing” test puts a high load on the vertex processors, but the graphics memory subsystem influences the results strongly, too: the PowerColor Bravo X700 is again 20-22% slower than the RADEON X700.
It is difficult to classify the genre of the third test, but one thing is certain – it is the most difficult of the three. The higher GPU performance of the PowerColor Bravo X700 helps it keep on the same level with the NVIDIA GeForce 6600. In the “eye candy” mode the PowerColor card doesn’t slow down, but comes ahead of the GeForce 6600 even. So, the overall scores of the cards naturally come from their performance in separate tests: the PowerColor Bravo X700 is a little faster than the GeForce 6600 in average.
The PowerColor Bravo X700 is an arguable solution. On one hand, the silent cooling system on heat pipes is a big plus, but on the other hand, the low quality of manufacture and the use of out-dated GDDR2 memory rated for 266(533) MHz frequency spoil the overall impression from the product. It possible we got a defective sample of the Bravo X700, but the failure of the memory chips at the first (and rather discreet) overclocking attempt was not a pleasant fact at all. The use of GDDR2 didn’t make the PowerColor card an overclocker-friendly product. On the contrary, the overclocking potential of this card is close to zero.
Moreover, the standard RADEON X700 as described by ATI Technologies’ recommendations clocks its memory at 350 (700) MHz and the use of slower memory inevitably reduced the overall performance of the Bravo X700. The PowerColor is 5% to 20% slower than the standard RADEON X700, depending on the particular game and mode. Thus, the PowerColor Bravo X700 lacks the main advantage of the RADEON X700 over the more functional GeForce 6600 – speed!
So, we don’t recommend the PowerColor Bravo X700 to overclockers or to people who choose between the RADEON X700 and the GeForce 6600. They should better consider more expensive, but faster models that use FC-BGA-packaged memory with an access time of 2.8 nanoseconds and a frequency of 350 (700) MHz.
We think the PowerColor Bravo X700 may be interesting for people who are building a silent computer, who don’t care about playing games in high resolutions, and who have a limited PC budget (this card is a very affordable product after all). The passive cooler also makes this product suitable for home multimedia centers, especially since this graphics card fully supports the HDTV standard. The card also supports two DVI interfaces, which might be a useful feature at work.
So, in brief, the PowerColor Bravo X700 is a good choice for non-gaming systems, including multimedia and office computers, for which silent operation is a priority.