All you need to know about your products!
 |
 |
|
3DNews Vendor Reference English Resource - All you need to know about your products! |
|
||||
|
|
|
||||
|
|
|
||||
|
|
|
||||
|
||||||
GeXcube Radeon 9800XT ReviewAuthor: Aleksandr MitrofanovDate: 27/04/2004
GeXcube Radeon 9800XTUntil recent times, the demand for high-end video cards have not been essentially high. The thing is most 3D games presented on the market offer primitive graphics which can be easily "digested" by mid-end cards. Various promises of game developers to release a super cool game (HalfLife2, Doom3 etc. in practice turned out to be just another deferral of release dates :). But things radically changed with the release of FarCry game which immediately amazed all with the outstanding graphics quality (albeit some contradictory gameplay :). Many owners of mid-end cards (as well as high-end cards of the previous generation) were surprised to realize that their system is simply "weak" for the game played at the maximum possible image quality. And this is experienced even with the antialiasing disabled, the sky displayed in static textures (which is a serious loss to the ancient Unreal :), and exploded cars smash into 4-5 dark debris (whereas this number should be a hundred times more :). Anyway, it is understandable for the FarCry game developers (as well as other developers): the number of owners of high-end cards (as well as powerful platforms with speedy processors and abundant RAM capacities) are negligible and amount to no more than 5% as per some estimates. As a result, the "graphic niceties" are cut down at the root to make the game playable at mid-end systems. In any case, games like FarCry is a big step forward: purchasing a high-end card gives an immediate return, and in many ways becomes a must for decent gameplay. Therefore, we are reviewing a typical high-end card built on the ATI Radeon9800XT chip: it is GeXCube Radeon9800XT (for details of this company, read the review GeXcube Radeon 9600XT Extreme). The video card is packaged in a mid-sized cardboard box.  Apart from the card, there are the following components in the card:
 Of special interest is the coupon to acquire Half-Life 2 for free.  It has a serial number closed with an opaque erasable layer. Once the game is officially released, you can order a free shipment (if that option is available for your country) with this coupon or get a link to download the game. Surely, it's better not to open the number before the time. Otherwise you might be in for trouble in upgrading the video card (judging by the ever deferred release dates for HL2, the probability of upgrade for Radeon9800XT is rather high :). Now let's look in to the card itself. Like all the high-end video cards, GeXCube Radeon9800XT require external power and have an additional Molex-connector. Once I unpacked the card, I found a rather long splitter cable already plugged in to the connector (therefore, the number of free power connectors from the PSU is not reduced). Albeit a trifle, but how nice :)  The card is a precise replica of the ATI Radeon9800XT reference design. This means that the interface of the card is AGP 4x/8x, the core runs at 412 MHz, and the DDR SDRAM capacity is equal to 256 MB.  More details about the memory. On the video card, there are eight BGA chips HY5DU573222 made by Hynix located on the from and rear sides the PCB.  The memory access time is 2.5 ns, which is equivalent to 400(800) MHz. But according to the reference design, the memory runs at 365(730) MHz, with the memory bandwidth = 256 bit. That is, the memory has some margin for frequency increase (or overclocking potential :). The operating frequency of the chip = 412 MHz.  It makes sense to mention about the "Overdrive" mode that switches the card to the "dynamic overclocking" mode. Upon the start-up of a 3D application, the driver increases the video chip frequency. The maximum possible frequency is 432 MHz; there is also an interim stage where the chip frequency equals 418 MHz.  The decision as to the stage is made by the driver on the base of the information about the chip temperature. Below 52 C, operation is possible at the maximum frequency (432 MHz), and within 52 to 65 C the chip runs at 418 MHz, and at temperatures over 65 C the frequency is re-set to the "default" (412 MHz). As a result, the "Overdrive" is a very simple and useful feature which allows to gain performance boost provided the housing is properly fanned. Of course, we ran all the performance tests with the "Overdrive" mode disabled.
Overclocking and heat emissionOnce I received the video card, I ran a series of overclocking experiments. As a result, we get the following maximum frequencies: the chip clock speed = 429 MHz, memory frequency = 391(782) MHz.  Further increase of frequencies resulted in the uprise of artifacts. The board heated up immensely, which tells to the need to modify the cooling system and proceed with the cooling. And there is definitely enough space to proceed: albeit with slight artifacts, the board kept on running at speeds up to 450-460 MHz (for the chip) and 400(800) MHz (for the memory frequency).  The video card was overclocked with the super popular Riva Tuner software which can hardly leave the Release Candidate stage (which is no surprise though, considering the flaws in the user interface :). It is worth noting that the overclocking results is a purely subjective thing, and for each specific card the maximum frequencies may be varied. Anyway, the scatter of readings is rarely too large. This holds true not only for video cards of a single manufacturer, but for the vast majority of cards built on the same chip (in this case, it is ATI Radeon 9800XT). It rarely happens when a manufacturer departs from the reference design requirements and either fits a faster memory or changes the operating voltage (as is known, the higher chip and memory voltages allow attaining higher operating frequencies). On the other hand, this allows the use not to worry much about the fate of the video card: a nominal cooler guarantees excellent cooling at both standard and increased frequencies. By the way, let's look at the cooler.  Engineers at GeXCube decided not to rack their brains at developing their own cooler but inherited the design from the reference. In practice, the cooler proved to offer a very quiet operation. Also note that the cooling system is very compact and does not block the adjoining PCI slot.  To overclock to 400/450 MHz, we used additional fans The cooler is made up of a copper base plate, a fan and a plastic lid which guides the air flow. Of special note is that the fan is positioned not right above the core, but with some displacement. This is done to eliminate the "dead spot" and maximize the efficiency of using the air flow.  The copper plate is positioned on the core and is fastened with four pins. The plate is also intended for cooling the memory chips. In respective points on the plate there are extruded pads which provide tight contact with the memory. Between the memory and the plate there is a layer of quite dubious thermal interface.  But they were overly generous with the thermal interface (a different type of paste :) for the 9800XT video chip.  Finally, the memory on the reverse side of the board is cooled by a copper plate where there is thermal interface of the third type :).  There was nothing to complain about this component: during the tests, the rear plate was heating immensely, which tells to a good heat transfer from the memory chips. By the way, the quality of assembly did appeal to me: the plate ideally fits to both the video chip and each of the memory chip. The same can be said about the rear plate. But the processing quality of the thermal interface of the front plate leaves much to be desired: on replacing the thermal paste with Zalman thermal paste, I was able to raise the operating frequencies of the chip from 429 to 440 MHz. Note that the cooler fan is plugged in via a 3-pin connector. This implies the rotational speed can be adjusted. But it is inaccessible to the user - all is done automatically. Reminding it once again that the card heats up immensely (even with the open test system). Therefore, for the summer period I would recommend fitting additional cooling, especially if the card is installed in a closed housing.
PerformanceTo test the performance, we assembled the most powerful system using available components. In particular, it is the Abit IC7-MAX3 (Intel 875P) motherboard. Then, we installed a Prescott Pentium4 processor and set the clock speed = 3.6GHz (225õ16). In so doing, the memory (PC3200 Corsair TwinX) was set to run in the asynchronous mode at 180 MHz with the memory latency timings set to 2-3-6-3. Clearly, the more powerful the system is, the better. This allows us to maximize the load upon the video card and estimate its true potential. Now regarding the tests. The first what interests us is the operating speed of the GeXCube Radeon9800XT relative to the similar product of a different manufacturer (we chose the Asus Radeon 9800XT video card). Then we were interested in performance drop upon the maximum increase of 3D image quality (antialiasing and anisotropy). And lastly, we looked at the video card's performance at increased frequencies. We used the following set of benchmarks:
Let's first take a look at the results of synthetic benchmarks.   3D Mark 2001 is already outdated, and we used it simply as a tribute to tradition :). The performance in the first three tests ("terminator", "dragotic" and "matrix" :) is already much more than enough (the fps rate exceeds 300-400), so we restricted to only the last test - "Mother Nature".           And the last synthetic (or semi-synthetic :) benchmark. We will be using real-world gaming applications.  A special comment is regarding the Comanche game: this application puts more load to the processor, rather than the video card. In particular, the use of the Northwood processor gives an essential performance boost (the average fps goes over 70 !) For details, read our review about the Prescott core.  But at CodeCreatures, the speed of the platform does not play any part. The performance at this test depends solely on the video card used.
 And - "the spice of the show" - here goes FarCry. As we can see, at the maximum settings the speed of the game is definitely not high. Even with this powerful platform where a Pentium4 3.6 GHz processor is used. Do not flatter yourself with the >60 fps figure. This result is "averaged", which in its turn means a possible performance drop during hot battle scenes (or "gore" :)).  Since in the GunMetal test the AA mode is enabled by default (AA2x), we decided to run the tests at AA 4x from the start.   Now look at the performance in the OpenGL application. While the classical Quake3 is an ideal testing environment for processors, memory and motherboards, it does not work for testing video cards because the graphics is too simplified. That is why we used Return to Castle which, albeit based on the Q3 engine, offers a higher quality graphics.    Another, demanding and hard for video cards, benchmark using DirectX 9.0. Remarkably, nVidia products lose seriously at this benchmark to ATI chips. The lag is so substantial that programmers at nVidia were unable to eliminate it even through the use "cheats". As a result, pressed by nVidia, the game developers had to disable the benchmark feature in the latest patch.  There is one more interesting benchmark, indicative in terms of nVidia vs. ATI opposition. Since the game offers no integrated benchmark, programmers at nVidia did not do any cheats for that (like, e.g., for Unreal Tournament). However, Alexandr Kondratyuk, a developer of the BenchemAll, found a way to enable the benchmarking feature (for which he deserves special respects :), and the program showed that nVidia's new drivers absolutely don't raise performance in this game.   The last, fairly hard test, albeit not using DirectX 9.0 to the full, is able putting a serious load upon your video card. Especially once the antialiasing is enabled. As regards the board's performance in the overclocking mode, the boost is about ~10% on the average. But it is important to note here that operation at 450/400 without artifacts is possible only after serious revamping of the cooling system. But if you don't touch the cooler (thus preserve the guarantee for this very expensive card), then the maximum possible frequencies attained are 429-432 MHz for the chip and 380MHz for the memory. In this case, the performance boost will be a bit more than 5%. The only what I can say regarding the maximum quality mode (antialiasing and anisotropy) is that it makes sense to enable it only with games whose speed exceeds 100 fps. In this case, the speed will definitely drop (by 30-50%), but this change will not affect the gameplay quality. Final WordsTo date, video cards based on the Radeon 9800XT chip are among the fastest video accelerators. They are practically free of any serious shortcomings, except the price which is over $450. It's just the price which is a serious obstacle for their wide spread occurrence. However, in the nearest future the situation may radically change. First of all, the price may quickly drop because of the active introduction of the new PCI Express bus. Reminding it, that early this summer the new Intel platform - Socket LGA775 - will be released, for which the i915 and i925 chipsets are meant and which offer no support for the AGP bus. Therefore, it makes sense to give up the idea of buying Radeon 9800XT boards until the price drops to $300. Only in this case, video cards of this type may be serious rivals to boards built on the Radeon 9800 and Radeon 9800Pro chips which to date offer the best "price/performance" ratio. The same advice I would give to the overclocker: be patient and wait a bit. Since Radeon 9800XT (R360) appears to be an extremely overclocked version of the R350 core (and thus the R300 core), then the overclocking potentials are not very high. So the owners of revamped (i.e. enabled pipelines) Radeon9500 and Radeon 9800SE video cards can calm down for the time being: at speed, their video cards are not much inferior to Radeon9800XT. The same applies to owners of overclocked Radeon 9800 (Pro :) video cards. Now, specific findings regarding the GeXCube Radeon 9800XT video card. This video cards is a very well made product built on the reference design. As a result, it is at least on par with other boards built on the Radeon 9800XT chip at both performance and the overclocking potentials. In fact, for hardcore overclocking you would have to improve the cooling system which proved to be a joy due its practically noiseless operation in the nominal modes. Nor there were any issues regarding the package bundle. As regards the price, to date it is about $462 (i.e. one of the cheapest among Radeon 9800XT boards). So if you have firmly decided to buy just this board based on the Radeon 9800XT chip, then GeXCube could be regarded as one of the possible candidates. Read more on this topic
|
|
|
| |||||
