by Alexey Stepin , Yaroslav Lyssenko
10/19/2007 | 09:16 AM
The ATI Radeon HD 2600 XT was meant to enter the market in two flavors, with GDDR3 memory and with high-frequency and faster GDDR4. However, even the latter version couldn’t surpass the Nvidia GeForce 8600 GT due to the bottlenecks in the architecture of the new RV630 graphics processor. Being superior to the opposing G84 core in sheer computing power, the RV630 is greatly limited by having only eight texture and four raster processors, which affects its performance at high resolutions and with enabled full-screen antialiasing.
<%BANNER[article]%>The next model in AMD/ATI’s graphics product line-up is ATI Radeon HD 2900 XT, but the technical characteristics of this graphics card are far superior to those of the RV630-based solutions. The gap between the two families has diminished somewhat after the release of the ATI Radeon HD 2900 Pro, yet it’s still quite large because the new card is selling at over $200.
This vacuum is to be filled by the dual-chip Radeon HD 2600 X2 but all previous attempts to promote such solutions on the market have been unsuccessful. They were regarded as rare exotics in the best case. Otherwise they would be completely forgotten like the Nvidia GeForce 7950 GX2. The reasons are many, particularly the high manufacturing cost due to the use of a complex PCB, two GPUs and a double amount of memory chips. There are compatibility issues, too. While classic single-chip graphics cards guarantee maximum performance in every game, a dual-chip solution depends heavily on the support of multi-GPU technology in the driver and on the specifics of the particular game engine.
So, until the arrival of new GPUs from ATI that would make a good foundation for building classic single-chip solutions capable of competing with the GeForce 8600 GTS and with Nvidia’s upcoming mainstream solutions, AMD/ATI’s partners can do nothing but release overclocked versions of Radeon HD 2600 XT. Considering the support for GDDR4 memory which can be clocked at very high frequencies and the good frequency potential of the RV630 core, this can indeed lead to appealing products. One of them, produced by MicroStar International, is going to be reviewed in this article. Let’s see if it can match the performance of the GeForce 8600 GTS.
As usual, the card came to us in its retail packaging with all the accessories the manufacturer thought it proper to put in. The medium-size box has a handy handle and looks like that:
There is nothing exceptional about its design that follows MSI’s traditional style. The left part of the box with the company’s name and logo imitates polished metal. There is a row of labels with technical information on the right. As opposed to the NX8600GTS-T2D256E-HD-OC whose package had a picture of a fantasy warrior, this box is embellished with a cyborg girl, which is quite a hackneyed figure, too.
Inside the wrapper there is the main box that contains the card, firmly fixed in a foam-plastic tray, and accessories:
So, there is everything you need to make full use of your RX2600XT except for a flexible connector necessary for building a CrossFire configuration. Mainstream graphics card, like the described product from MSI, are but seldom used in such tandems, so that’s not a great problem. The user manual is a traditional poster, which is a rather inconvenient format, but there is additionally a brochure with all the information the user may need.
The packaging of the RX2600XT Diamond 512 deserves a good word for its quality and practicality if not for its design. There are no special bonuses among the accessories, yet the kit contains everything you want to use the card in a modern multimedia system with HDMI support.
The PCB of the RX2600XT Diamond 512 follows the reference design developed by ATI for Radeon HD 2600 XT GDDR4, but MSI used a red rather than black solder mask, which endowed the device with such an imposing appearance. Exclusive products are usually colored like that.
Like the reference card, the RX2600XT Diamond 512 features a complex power circuit based on an uP6201AQ controller. Some of the elements are not installed, which means that the circuit can be scaled up for newer, more power-hungry, processors. This is also indicated by the fact that the PCB allows installing a standard 6-pin connector for additional power supply. Here, this connector is missing and its seat is occupied by an electrolytic capacitor. Having a power draw much lower than 75W, the graphics card receives all the power it needs from the PCI Express slot. The power circuit is sharply separated from the main part of the PCB, meaning that ATI follows the modular design principle not only when developing GPUs but also when developing PCBs.
Since this version of Radeon HD 2600 XT GDDR4 is equipped with 512 megabytes of graphics memory, it has four memory chips on the reverse side of the PCB, which you don’t often see on modern graphics cards. The card carries a total of eight GDDR4 chips (Samsung K4U52324QE-BC09, 16Mbx32, 1.8V). The chips have a rated frequency of 1100 (2200) MHz but are clocked at 1150 (2300) MHz by the card. Although each chip has a 32-bit interface, the total width of the memory bus is 128 bits, the standard value for the RV630 chip.
The graphics processor has a standard configuration with 120 ALUs grouped into 24 universal shader processors, 2 texture and 1 raster processor which are equivalent to 8 TMUs and 4 ROPs. The RV630 is weaker in this respect than the G84 that has 16 TMUs and 8 ROPs. In fact, the superiority of the GeForce 8600 GTS over the Radeon HD 2600 XT GDDR4 comes from this advantage whereas in sheer computing power the Nvidia card is far inferior to the ATI one. The GPU of the RX2600XT Diamond 512 is clocked at 850MHz which is 50MHz higher than the frequency of the standard Radeon HD 2600 XT.
Save for the described peculiarities, the RX2600XT Diamond 512 is identical to the reference ATI Radeon HD 2600 XT GDDR4. It has two DVI-I ports with dual-link mode, a universal VIVO port, and CrossFire connectors. The VIVO functionality is not supported, though. The appropriate chip is not installed on the PCB, although the latter has a seat for it. As opposed to some other versions of Radeon HD 2600, the reference PCB design doesn’t provide for the installation of a HDMI connector instead of one DVI.
The increased clock rates mean that the MSI card consumes more power and produces more heat that the reference card. Even though this is not critical for the RV630 chip, MSI decided to equip it with a rather advanced cooling system.
The cooler represents the classic dual-slot design with a heat pipe. Exhausting the hot air out of the system case, it resembles the cooler from the ATI Radeon X1950 XTX, but is somewhat smaller and has a fan with a peculiarly shaped impeller.
The base and heatsink of the cooler are aluminum. The only copper element is the nickel-plated heat pipe. The photo shows the thick dark-gray thermal grease typical of top-performance graphics cards and rubber spacers on the fastening poles. The rather small heatsink consists of thin profiled plates that are fastened to each other with the jutting pieces. It is mounted on the base and connected to it with the heat pipe that ensures a uniform distribution of heat to increase the overall efficiency of the cooler. The memory chips on the front side of the PCB are cooled with an L-shaped ribbed plate. The chips on the reverse side of the PCB are cooled by means of natural convection.
Judging by the fan marking, its max power is 2.76W. The nine blades have a queer S-like shape and are getting thicker towards the ends. We haven’t seen such fans before. The fan has a 3-wire connection with an output of the velocity sensor. The whole arrangement is covered with a profiled casing made from translucent red plastic that drives the air through the heatsink to the slits in the mounting plate and out of the system case. The cooler is fastened to the PCB by means of four nuts and an X-shaped back-plate. The heatsink on the memory chips is fastened with three screws.
Representing a time-tested design, this cooler should deliver high performance at low noise. It may even increase the overclocking potential of the MSI RX2600XT Diamond 512. The only downside is the height of the cooler that may prevent you from installing the card into a compact system case. On the other hand, the ATI Radeon HD 2600 XT GDDR4 has a too large PCB to start with, which makes it incompatible with small system cases.
We measured the level of noise produced by the MSI card with a digital sound-level meter Velleman DVM1326 using A-curve weighing. At the time of our tests the level of ambient noise in our lab was 36dBA and the level of noise at a distance of 1 meter from the working testbed with a passively cooled graphics card inside was 43dBA. We got the following results:
The cooling system installed on the RX2600XT is not just powerful, it is even redundant for such a modest card as ATI Radeon HD 2600 XT. That’s why its fan can rotate at a small speed to cool the card successfully. As a result, the cooler is virtually silent, matching the best reference coolers from ATI and Nvidia.
Unfortunately, our attempt to overclock the MSI RX2600XT Diamond 512 was not much of a success. The memory with a rated frequency of 1100 (2200) MHz but clocked at 1150 (2300) MHz proved to have little potential. The card was only stable at a memory frequency of 1179 (2358) MHz. The graphics core had been overclocked by the manufacturer from 800MHz to 850MHz and the driver reported a frequency growth to 857MHz only. Such a modest growth of frequencies cannot result in a considerable performance gain. That’s why we decided not to benchmark the RX2600XT Diamond 512 at the overclocked frequencies.
To test the performance of MSI RX2600XT Diamond 512 we assembled the following standard test platform:
Since we believe that the use of tri-linear filtering optimizations is not justified in this case, the AMD and Nvidia graphics card drivers were set up to provide the highest possible quality of tri-linear and anisotropic texture filtering. Also, to ensure maximum image quality, we enabled transparent texture filtering - Adaptive antialiasing for AMD Catalyst and Transparency antialiasing (multisampling) for Nvidia ForceWare. As a result, our AMD and Nvidia driver settings looked as follows:
ATI Catalyst:
Nvidia ForceWare:
We selected the highest possible graphics quality level in each game using standard tools provided by the game itself. The games configuration files weren’t modified in any way. Performance was measured with the games’ own tools or, if not available, manually with Fraps utility version 2.9.1. We also measured the minimum speed of the cards where possible.
With a few exceptions the tests were performed in standard resolutions: 1280x1024/960, 1600x1200 and 1920x1200/1440. We used “eye candy” mode everywhere, where it was possible without disabling the HDR or Shader Model 3.0/4.0. Namely, we ran the tests with enabled anisotropic filtering as well as MSAA 4x. We enabled them from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of Catalyst and ForceWare.
MSI RX2600XT Diamond 512 will be competing against the following graphics accelerators participating in our test session:
For our tests we used the following games and benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
RPG
Simulators
Strategies
Synthetic Benchmarks
The MSI RX2600XT is faster than the reference card by 8-10% depending on the resolution. However, this performance growth is insufficient to catch up with the Nvidia GeForce 8600 GT in the playable resolutions whereas in the resolutions above 1280x1024 pixels this growth is unimportant due to a too low speed.
This game has a speed limiter set at 30fps – that’s the rate at which the physical model is being updated on the server. The only mainstream graphics card to achieve this level is the GeForce 8600 GTS. The game engine doesn’t need a high computing capacity from the graphics card, but it favors Nvidia’s GeForce 6/7/8 architectures with their UltraShadow technology. The MSI RX2600XT Diamond 512 has a very small advantage over the ATI Radeon HD 2600 XT GDDR4 in this test. It is less than 5%.
It’s like in the previous test, but Nvidia’s mainstream solutions are on the losing side as they have a smaller computing capacity than ATI Radeon HD 2600 cards.
Anyway, this is a very demanding application and the best result the RX2600XT Diamond 512 can deliver is only 27fps at 1280x1024. This is far from being comfortable. To have a better result you’ll need at least a GeForce 8600 GTS 320MB, but even its performance may drop below 25fps in scenes with lots of objects, characters and special effects.
Talking about the upcoming DirectX 10 version of Call of Juarez, even Nvidia’s GeForce 8800 GTS doesn’t look good in it. We didn’t test the resolutions above 1280x1024 as it doesn’t make sense considering that you have a frame rate of only 4-17fps in it. We should note, however, that the MSI product is almost 20% ahead of the reference card.
The MSI card is hardly any faster than the reference sample but far slower than the Nvidia GeForce 8600 GTS. Its average speed is close to the allowable minimum in open scenes, meaning that you don’t have any reserve of speed.
The RX2600XT Diamond feels much better when you enable HDR but turn FSAA off. Its advantage over the reference card from ATI is obvious, and its average speed is high enough for you to play comfortably at 1280x1024 or even at 1600x1200 pixels. When the latter resolution is used, the MSI card is only 5% and 10% slower than the GeForce 8600 GTS on the Pier and Research maps, respectively.
The Radeon HD 2600 based solutions have a very low minimum of speed in this test. It is below 10fps even with the overclocked MSI card while the Nvidia GeForce 8600 GT easily provides 17fps and the GeForce 8600 GTS, 25 fps, which ensures normal gaming conditions considering the average speed of about 44fps. If you want to play at resolutions higher than 1280x1024, you need a graphics card like GeForce 8800 GTS 320MB or better.
Both versions of ATI Radeon HD 2600 XT GDDR4 deliver comfortable performance at a resolution of 1280x1024 pixels. They are somewhat faster than the Nvidia GeForce 8600 GTS, but the latter regains its leadership at the higher resolutions. This must be due to the Nvidia card’s having two times as many raster operators as the ATI card has, and the card’s rasterization performance is important for using full-screen antialiasing.
Prey runs on a modified Doom 3 engine, and the Radeon HD 2600 XT GDDR4 is slower than the Nvidia GeForce 8600 GTS just it was in Quake Wars. For the latter card, this is just a formal victory, as its average speed of 35fps is not far better than 28-29fps the ATI cards provide. There is no playing comfort in both cases since the speed will inevitably fall below comfortable level in the most complex scenes. If you want to play Prey with comfort, you need at least an ATI Radeon X1950 Pro, which often proves the best solution among inexpensive mainstream graphics cards despite the arrival of DirectX 10 compatibles.
The game doesn’t support FSAA when you enable the dynamic lighting model, but loses much of its visual appeal with the static model. This is the reason why we benchmarked the cards in S.T.A.L.K.E.R. using anisotropic filtering only.
The ATI Radeon HD 2000 cards are still slow in this game, but the Nvidia GeForce 8600 GTS is no winner, either. Its average speed of about 20fps equals the ATI Radeon X1950 Pro’s minimum speed. The game will only feel really comfortable on a GeForce 8800 GTS 320MB or a more advanced graphics card.
Save for the GeForce 8800 GTS 320MB, none of the tested graphics cards could show acceptable results in this test. The GeForce 8600 GTS is the closest to that while both versions of ATI Radeon HD 2600 XT GDDR4 took last place with an average speed of below 20fps and a minimum speed of less than 10fps. In other words, the inexpensive mainstream graphics cards don’t suit for playing this game unless you lower the graphics quality settings.
We try to get the highest possible image quality from each game, and choosing between FSAA and HDR we preferred the latter. That’s why the game was tested with anisotropic filtering only.
The GPUs with a unified architecture show their best when high shader-processing performance is called for. The RV630 is superior to the G84 in this respect, and the Radeon HD 2600 XT GDDR4 is competitive against the GeForce 8600 GTS at 1280x1024. The MSI RX2600XT Diamond 512 adds a few percent more to the reference card’s results. The resolution of 1600x1200 still doesn’t seem playable on either version of the card due to the low minimum of speed.
The current version of the game doesn’t support FSAA, so we performed the test with anisotropic filtering only.
Judging by the results, the current generation of inexpensive mainstream graphics cards can’t provide a playable speed in Gothic 3. The MSI RX2600XT Diamond 512 is the closest to doing this, but the real comfort can only be achieved with an Nvidia GeForce 8800 GTS 320MB or better graphics card.
The game offers FSAA starting with version 1.04, but its support for HDR is still deficient, so we are benchmarking the cards in the eye candy mode only.
What we’ve said about the previous test is also true for Neverwinter Nights 2. Only two out of the six cards, except for the Nvidia GeForce 8800 GTS 320MB, can deliver 15fps or something at 1280x1024, and their performance bottoms out to below this minimum playable level in the most complex scenes. As for the GeForce 8800 GTS 320MB, this graphics card makes the game playable at the highest graphics quality settings at 1600x1200 with 4x FSAA.
The game loses much of its visual appeal without HDR. Although some gamers argue that point, we think TES IV looks best with enabled FP HDR and test it in this mode.
The MSI RX2600XT Diamond 512 outperforms the Nvidia GeForce 8600 GTS in both minimum and average speed, and provides a good level of comfort at resolutions below 1600x1200, at least in closed environments.
Each graphics card is naturally slower in open scenes of the game, but the MSI card isn’t much worse than the GeForce 8600 GTS even then. Its minimum speed, and that of the other mainstream graphics cards, is below 20fps at 1600x1200, so we recommend you to limit the resolution to 1280x1024.
The gap between the MSI and the reference ATI card is getting smaller as the resolution grows: from 10% at 1280x1024 to 5% at 1920x1200. This proves to be enough to catch up with the Nvidia GeForce 8600 GTS, yet the speed is too low for playing comfortably at 1280x1024 with maximum graphics quality settings and 4x FSAA. Note that the Radeon X1950 Pro has the best minimum of speed here.
This game having a frame rate limiter, you should compare the minimum frame rates in the first place because it is the minimum speed that determines your playing comfort in Command & Conquer 3.
Only three out of six cards can yield over 20fps, and one of them, the Radeon X1950 Pro, is based on an outdated architecture and one more, the GeForce 8800 GTS 320MB, belongs to a higher class than the ATI Radeon HD 2600 XT GDDR4 and GeForce 8600 GT/GTS. In other words, if you want to have a really comfortable play, you should spend $150-200 and more for the graphics card, and this is true not only for C&C 3.
If we don’t count in the rather expensive Nvidia GeForce 8800 GTS 320MB, it is the ATI Radeon X1950 Pro that feels the best in this game. The GeForce 8600 GTS is at the allowable minimum of performance while the ATI Radeon HD 2600 XT GDDR4 has the worst result of all the tested graphics cards and the overclocking doesn’t help it much.
Starting with version 1.71, this game has no problems with FSAA, so we test it with enabled FSAA.
With enabled anti-aliasing MSI RX2600XT Diamond 512, just like any other graphics card on ATI Radeon HD 2600, can hardly demonstrate great performance. As a result, we would recommend you to disable FSAA completely or to play in 1280x1024 resolution.
The tests in DirectX 10 mode prove again that this API hasn’t yet come for real despite the arrival of the new-generation GPUs. Only owners of premium graphics cards can enjoy the capabilities of DirectX 10 and not in every game even.
The increased GPU and memory frequencies help the MSI RX2600XT Diamond 512 overcome the 11.000 points mark, but it scores a mere 269 points more than the ordinary ATI Radeon HD 2600 XT GDDR4. If expressed in percent, the performance growth is negligible.
The MSI card behaved just as you could expect in the first and third tests but the double amount of graphics memory (in comparison with the reference card) shows its worth in the second test, which it did not in the gaming tests. Anyway, the MSI card didn’t reach the level of the GeForce 8600 GTS.
The gap between the ordinary and overclocked versions of Radeon HD 2600 XT GDDR4 is even smaller in 3DMark06.
The gap is larger in the SM3.0/HDR tests, but barely noticeable in the SM2.0 tests.
Quite habitually, the ATI Radeon HD 2600 XT loses to the GeForce 8600 GT in the first test but equals it in the second test. As opposed to the second test from 3DMark05, there is no performance growth from the 512 megabytes of graphics memory installed on the MSI card.
The strong points of ATI Radeon HD 2000 architecture dealing with complex calculations show their best in the SM3.0/HDR tests: the MSI solution competes successfully with GeForce 8800 GTS in the first test and yields just a little bit to the rival in the second. Moreover, GPU overclocking results into a more significant performance improvement than in SM2.0 tests, originating most likely from the complex maths shaders.
The MSI RX2600XT Diamond 512 did well in our tests, leaving a positive impression overall. Its dual-slot cooler can be viewed as a kind of drawback, but it makes the card silent despite the increased GPU and memory frequencies. This card doesn’t suit for compact system cases, but multimedia PCs are quite often assembled in rather large cases, which can accommodate a dual-slot graphics card. The accessories to the RX2600XT Diamond 512 provoke no complaints, either.
As for pure gaming performance, the MSI product ensures an average performance gain up to 10% over the reference ATI Radeon HD 2600 XT GDDR4, and up to 20% in a few cases. Of course, this is not enough for this card to be competitive against the Nvidia GeForce 8600 GTS. The moderate frequency growth must be not enough to make up for the driver-related problems, 24x VIW execution processors, and possible problems with the rasterization units or memory controller of the ATI RV630 chip.
We have to acknowledge once again that graphics cards like ATI Radeon HD 2600 XT GDDR4 and Nvidia GeForce 8600 GT/GTS are often unable to deliver high performance in modern games, at least when the gamer doesn’t want to compromise and reduce the level of detail or when he wants to play in resolutions higher than 1280x1024. Our tests show that such cards often turn to be slower than previous-generation solutions such as ATI Radeon X1950 Pro.
The Nvidia GeForce 8800 GTS 320MB, on its part, provides a high enough performance for you to play comfortably at the maximum level of detail and with 4x FSAA at a resolution of 1600x1200 and, in some games, even 1920x1200. This is obviously the minimum everyone who is interested in modern games should go for. The same should be true for the new ATI Radeon HD 2900 Pro as well.
So, it looks like the only application left for the ATI Radeon HD 2600 XT and Nvidia GeForce 8600 GT/GTS is multimedia PCs, especially as they suit that purpose even better than the top-performance products. You should only buy one for games if you don’t play games seriously or if you can’t afford a more advanced graphics card.
If you need a quiet and economical graphics card, and games is not your priority, you should definitely consider the MSI RX2600XT Diamond 512 as one of the best options, but you should look up in the higher class if you want to play games with high-quality visuals. This is going to require more money, but you won’t be disappointed with the card’s performance.
Highs:
Lows: