by Alexey Stepin
11/27/2005 | 07:22 AM
ATI Technologies has already struck a blow on its main market rival NVIDIA in the sector of mainstream solutions by announcing the RADEON X800 GT graphics card. The card was tested in our labs and you could see that the NVIDIA GeForce 6600 GT which had earlier had no competitors in its price sector was now facing a very dangerous rival. What’s more, the RADEON X800 GT could even challenge the GeForce 6800 notwithstanding the latter’s 12 pixel pipelines. The strong points of that graphics card from ATI were high clock rates and a 256-bit memory bus which helped it greatly in high resolutions and with enabled full-screen antialiasing.
As a matter of fact, ATI Technologies seems to have large stores of last-generation GPUs in stock, but the users’ interest towards high-performance graphics cards on such GPUs is declining because they are technically less advanced than GeForce 6/7 or RADEON X1000 series devices. So in order to avoid some heavy loss, the company has to throw away the stored chips to the market and this problem has been partially solved with the release of the above-mentioned RADEON X800 GT which can come with either an R480 or an older R423 chip on board. Being a 0.11-micron, but poorly overclockable chip, the R430 also became out-dated and shared the fate of its 0.13-micron mates the same instant the RADEON X1600 came out. In other words, these chips are now like cumbersome ballast for ATI Technologies.
Meanwhile, in consequence of the recent aggressive price reduction on NVIDIA’s part, the GeForce 6800 graphics card based on the very successful NV42 chip has come to cost $199 or less and has thus become a very appealing buy. Unlike the ATI R430, the NVIDIA chip is overclocker-friendly – we can recall our overclocking the GPU of our sample of the MSI NX6800-TD128E from the default 325MHz to an impressive 460MHz and enjoying an appropriately hefty performance boost. As said above, the RADEON X800 GT delivers about the same performance, yet the GeForce 6800 is better in some cases as it has more pixel pipelines.
ATI Technologies had to react to NVIDIA’s actions and the reaction came quickly in the form of the September announcement of the ATI RADEON X800 GTO graphics card. The new card emerged under the same “Putting Old GPUs to Good Use” motto. It differed from the RADEON X800 GT in having 12 active pixel pipelines and a lower core clock rate (400MHz against 475MHz), so the company could use not only R480/R423 chips, but also R430 which was unsuitable for the RADEON X800 GT due to its limited frequency potential (420-430MHz at best). Costing about $200, the new graphics card is a direct market opponent to the GeForce 6800 and its faster memory makes it the more dangerous, especially in high resolutions and/or with enabled full-screen antialiasing. And now we are going to see with our own eyes which of these two graphics cards is better by taking a PowerColor X800 GTO VIVO model for our tests.
The box of the new product from PowerColor has the new design that we already described in our review of the PowerColor X800 GT graphics card.
We think this new design is better than the older one. It looks somewhat more modern and, accordingly, more appealing for the today’s user. The main color of the cover is red, not black, and the winged cyber-creature is looking in the opposite direction – these are the only points of difference from the box of the PowerColor X800 GT. The product description on the face side of the paper cover will immediately disappoint all overclockers because it has no mention of 0.11-micron tech process. It can mean one thing only – that this sample of the RADEON X800 GTO is based on the ATI R430 chip which has poor overclockability.
Although the PowerColor X800 GTO VIVO belongs to the GameFX series, it is oriented at multimedia enthusiasts rather than at gamers. At least, this thought is suggested by the accessories included:
You get all the things necessary for processing video here, but no games (for example, the PowerColor X800 GT was shipped with a copy of the popular flight simulator Pacific Fighters ). One more drawback is apparent at once: you can’t use the video input and a HDTV display device simultaneously because the card has only one VIVO connector and you can attach either the VIVO splitter or the YPbPr adapter to it, but not both at once. An I/O unit like the one we described in our ASUS Extreme N7800GTX TOP review might be better in this respect (well, we are not absolutely sure if the ASUS I/O unit permits to use an YPbPr output and an S-Video or Composite input at the same time). So, the accessories to the reviewed graphics card are not quite gorgeous, but you do get everything necessary to use it.
The PowerColor X800 GTO VIVO uses the reference PCB design of the ATI RADEON X800 XL, but has a different cooling system and sports a Rage Theater chip – the one that supports the VIVO functionality.
You can see for yourself that there’s no difference from the RADEON X800 XL. The cards coincide to a minutest detail and quite understandably so – there is not much sense in developing a completely new PCB design for a product that is destined for a rather short lifecycle.
Although there is a Rage Theater processor on board, the card is not equipped with an additional yellow-colored onboard header to connect to a composite video input that can be found on the front panel of some system cases. An additional power connector is missing, too, because even the 16-pipelined RADEON X800 XL can be powered up through the PCI Express x16 slot alone. Of course, the slot will power up the 12-pipelined RADEON X800 GTO which uses the same R430 GPU as well.
The cooling system is the same as was deployed on the PowerColor X800 GT, differing only in the color of the sticker. This simple all-aluminum design with an axial fan can well cope with a RADEON X800 GT or an X800 GTO, and without much noise, too. The GPU die touches the heatsink’s sole through a layer of thick dark-gray thermal paste whereas the memory chips have no contact with the heatsink whatever. It is not a big problem as GDDR3 chips working at frequencies up to 500 (1000) MHz do not heat up much, but the “heat bag” effect may occur, especially if your system case is not ventilated well.
The graphics processor is clocked at 400MHz, exactly as described in the RADEON X800 GTO specification, while the 2.0ns K4J55323QF-GC20 memory chips from Samsung are clocked at 490 (980) MHz. We shouldn’t expect overclocking wonders from this graphics card since the R430 is not a very overclocker-friendly chip. Some memory frequency growth may be achieved, but the frequency will hardly get higher than 550 (1100) MHz. Well, let’s better check it out right away.
The PowerColor X800 GTO VIVO carries the same standard cooling system as installed on the PowerColor X800 GT, so its noise characteristics are exactly the same. The noise is not very low, but remains in a comfortable range all the time – the speed control system never increased the speed of the fan during our tests. It means the GPU was effectively cooled and its temperature was always within the acceptable range.
In our attempts to overclock the PowerColor X800 GTO VIVO we at first increased the GPU frequency to an impressive 480MHz (almost a record for the R430 chip!), but the graphics card soon began to hang up in tests. By reducing the GPU frequency to 450MHz we made the card stable and it normally passed all of our trials. We call this an excellent result anyway since an overwhelming majority of ATI R430 chips can only work at 420-430MHz at best. The memory performed up to our expectations and was stable at 550 (1100) MHz – there were visual artifacts at higher frequencies. So we didn’t witness a marvel this time around like we did with the GeForce 6800 GS.
The 2D image quality the graphics card provided was high and up to today’s standards, but with one reservation: only one DVI output of our sample of the card worked. When attached to the other one, the monitor reported lack of signal. A PowerColor representative said to us that it was not a defect of our particular sample. That was a problem of incompatibility with some mainboards. Unfortunately, the Intel Desktop Board D925XCV and ASUS A8N-SLI Premium mainboard models that our testbeds are based around are in the incompatibles list, too. The problem may be solved in newer BIOS versions, so the owner of a PowerColor X800 GTO VIVO and people who want to buy this card may want to keep track of the BIOS updates on the manufacturer’s website.
We installed the PowerColor X800 GTO VIVO graphics card into our testbed computer:
We set up the ATI and NVIDIA drivers in the following way:
ATI CATALYST 5.9:
NVIDIA ForceWare 81.87:
We select the highest graphics quality settings in each game, identical for graphics cards from ATI and NVIDIA. If possible, we use the games’ integrated benchmarking tools (to record and reproduce a demo and then measure the reproduction speed in frames per second). Otherwise we measure the frame rate with the FRAPS utility. If it is possible, we measure minimal as well as average fps rates to give you a fuller picture.
We turn on 4x full-screen antialiasing and 16x anisotropic filtering in the “eye candy” test mode from the game’s own menu if possible. Otherwise we force the necessary mode from the driver. We don’t test the “eye candy” mode if the game engine doesn’t support FSAA.
Besides the PowerColor X800 GTO VIVO, the following graphics cards took part in this test session:
These games and applications were used as benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
The PowerColor X800 GTO is of course weaker than the GeForce 6800 GS because its GPU works at a lower frequency, but it is confidently ahead of the GeForce 6800 and ensures a playable frame rate even in 1280x1024 with enabled FSAA and anisotropic filtering. In the hardest video mode the new graphics card even beats the GeForce 6800 GS owing to the more efficient memory controller, but that mode is practically unplayable as the speed may go down below comfortable level.
As it could be expected, the inefficient OpenGL driver doesn’t permit the PowerColor card to show its best in this game. The user is in fact limited to 1024x768 resolution. Full-screen antialiasing is out of the question altogether.
The performance of the PowerColor X800 GTO in Doom 3 can be characterized in the same words as above. Leaving the RADEON X1600 XT behind at the “pure speed” settings, the PowerColor card only overtakes the GeForce 6800 in 1600x1200 resolution of the “eye candy” mode where there’s no talking about comfortable play.
Version 2.0 pixel shaders are executed very quickly on the RADEON X8 architecture, so the PowerColor X800 GTO feels at ease in Far Cry. Its performance only goes below playable level in 1600x1200 resolution of the “eye candy” mode. The GeForce 6800 can’t compete at all with the described product, mostly due to the slow memory, but NVIDIA’s newer GeForce 6800 GS card is only second to the 16-pipelined RADEON X800 XL in this test.
Graphics cards with Shader Model 3.0 support have an advantage on the Research map where per-pixel lighting is created by means of such shaders. The gap between the PowerColor X800 GTO and the GeForce 6800 is smaller as a consequence. The difference amounts to 2fps only in 1600x1200 resolution.
The two competing cards do not differ much in F.E.A.R., either. The GeForce 6800 and the PowerColor X800 GTO can be said to have the same speed in this game, except higher resolutions of the “eye candy” mode where none of the two can provide a playable frame rate. Well, comfortable play is only possible in 1024x768 resolution of the “pure speed” mode – and even then the speed will occasionally bottom out to 20fps.
Half-Life 2 being a not very demanding application, the PowerColor X800 GTO offers enough performance for playing it comfortably with enabled 4x FSAA and 16x anisotropic filtering in all resolutions. The RADEON X1600 XT and the GeForce 6800 are limited to 1280x1024 resolution in this case.
Unlike in the previous scene, the GeForce 6800 GS is not faster than the PowerColor X800 GTO in the battle scene we recorded on the d3_c17_02 map. The difference is 0.2fps. By the way, this scene is more complex than the previous one, so if you want to play Half-Life 2 on the RADEON X800 GTO with full-screen antialiasing enabled, you may want to restrain yourself to 1280x1024 resolution to have some reserve of speed.
Judging by the numbers, Project: Snowblind is indifferent to the number of pixel pipelines. The results of the PowerColor X800 GTO and the RADEON X800 XL differ little even in high resolutions. The PowerColor is considerably slower than the GeForce 6800 GS, but fast enough to provide a playable frame rate in almost all video modes, except the hardest ones.
Quake 4 produces a very curious picture: the PowerColor X800 GTO starts out at the same speed as the GeForce 6800 GS, but then falls behind it in higher resolutions. When we turn on FSAA and anisotropic filtering, we see the opposite thing happen and the cards differ by no more than 2-3fps in 1600x1200 resolution of the “eye candy” mode. Both cards make the whole range of resolutions playable, but the speed in 1600x1200 with enabled 4x FSAA and 16x anisotropic filtering may go down below comfortable level at times.
The RADEON X800 architecture lacks Shader Model 3.0 support and suffers a crushing defeat in Serious Sam 2. The PowerColor X800 GTO does not suit for playing this game at all. You can of course reduce the level of detail to arrive at a more or less comfortable frame rate, but you’ll lose some of the fun by doing so.
The PowerColor X800 GTO yields 60+ fps in the hardest video mode here, but the GeForce 6800 follows just 5-6fps behind, despite the worse technical characteristics.
The standings are the same on the Metallurgy map, but the gap between the GeForce 6800 and the PowerColor X800 GTO is even smaller, no more than 3fps. Of course, the GeForce 6800 GS is much faster, but NVIDIA’s GPUs have a higher fps/MHz ratio than ATI’s in general and in this case the PowerColor card even has a lower GPU clock rate than GeForce 6800 GS.
The average performance of the PowerColor X800 GTO in this game is 65fps and the card ensures 55fps even in the hardest scenes. The GeForce 6800 and the RADEON X800 XL do about the same, while the GeForce 6800 GS is much faster – it is 40% ahead of the PowerColor card in 1600x1200!
The PowerColor X800 GTO outperforms the GeForce 6800 in the “pure speed” mode and almost overtakes the GeForce 6800 GS at the “eye candy” settings. These achievements are, however, negated by the fact that the PowerColor works in the game’s Shader Model 2.0 rendering mode, i.e. under easier conditions than the Shader Model 3.0-supporting graphics cards, which ensure a higher image quality. The game is very demanding, so you will only find low resolutions playable on the PowerColor X800 GTO.
This racing simulator with simple pixel shaders used to run very fast on the RADEON X8 architecture. The PowerColor X800 GTO confirms the point by leaving the GeForce 6800 behind in all the test modes, but it is still slower than the GeForce 6800 GS that only meets competition from the 16-pipelined RADEON X800 XL. The game isn’t very hard, but the performance of the PowerColor X800 GTO is insufficient for normal play in 1600x1200 resolution with enabled FSAA and anisotropic filtering.
The PowerColor X800 GTO seems to give out a playable frame rate in 1024x768 in Pacific Fighters, but one glance upon the min speed numbers tells you clearly enough that the speed may bottom out far below comfortable level even in that resolution. The reason is the same as with Doom 3 and The Chronicles of Riddick – a very slow OpenGL driver from ATI.
The PowerColor X800 GTO seems to be not much worse than the GeForce 6800 GS, especially in higher resolutions, but do not forget that the RADEON X8 architecture does not support Shader Model 3.0. In other words, the speed of the cards being similar, the GeForce 6800 GS displays a much nicer-looking image. Well, comfortable play is only possible in 1024x768 resolution anyway.
The PowerColor X800 GTO is confidently ahead of the GeForce 6800 at first, notwithstanding the latter’s support of UltraShadow II technology which accelerates processing of stencil shadows, but the cards change their places as soon as we switch to the “eye candy” mode. This is rather strange, considering the faster memory and the more efficient memory controller of the PowerColor card. Yet it is a fact that you can play this game in low resolutions on the GeForce 6800 when FSAA and anisotropic filtering are enabled. The RADEON X800 GTO doesn’t allow you to do so because the speed sometimes bottoms out to 15fps and lower.
Aquamark3 is not a very hard trial, and graphics cards do not have to argue which is better at executing complex pixel shaders here. More important for this benchmark is the ability to remove invisible surfaces and the performance of the GPU’s vertex processors. The PowerColor X800 GTO VIVO is everywhere a little slower than the GeForce 6800 GS because the GPU of the latter card has only 5 vertex processors against 6 in the R430. The RADEON X1600 XT performs but poorly. Although ATI improved the invisible surface removal algorithm in the RV530, they also equipped that GPU with only four texture-mapping units which become a bottleneck in many applications. Moreover, the RADEON X1000 architecture is generally oriented at executing complex version 3.0 pixel shaders which do not occur in Aquamark3.
The numbers do not have much to say to a hardware tester, yet we can observe that the benchmark doesn’t load the pixel pipelines much: the PowerColor X800 GTO is not much slower than the RADEON X800 XL which has 4 such processors more.
Scoring over 9500 points is a nice result for a 12-pipelined graphics card based on the RADEON X8 architecture. When overclocked, it even leaves the 16-pipelined RADEON X800 XL behind. The GeForce 6800 GS remains unrivalled, however, because three out of four 3DMark03 tests run better on GeForce 6/7 cards.
The GeForce 6800 GS turns in an easy win, while the GeForce 6800 outperforms the PowerColor X800 GTO in the “pure speed” mode only. In the “eye candy” mode the GeForce 6800 is slowed down by its low GPU and memory clock rates.
It’s all vice versa in the second test: the GeForce 6800 is only slower than the PowerColor card when we do not use full-screen antialiasing and anisotropic filtering. Otherwise the difference in performance is no more than 1 fps.
The third test has nothing new to tell us, although its geometry is more complex than that of the second test. The PowerColor X800 GTO just seems to be a little more confident in the “eye candy” mode.
Thanks to its high speed of execution of version 2.0 pixel shaders, the PowerColor X800 GTO manages to have the same “pure speed” as the GeForce 6800 GS has, but the NVIDIA-based card is ahead as soon as we turn on FSAA and anisotropic filtering, and the PowerColor can overtake it only through overclocking.
So, it’s all correct in 3DMark03: the GeForce 6800 GS is much faster than the PowerColor X800 GTO in three tests out of four and has the same performance in the fourth one, so it deserves the overall victory. The hero of this review has to content itself with the third place, right behind the RADEON X800 XL.
The PowerColor X800 GTO VIVO scores just over 4500 points in the newer 3DMark, while the GeForce 6800 GS scored over 5000 and the RADEON X1600 XT, whose architecture is optimized for efficient processing of very complex shaders, reached the 5200 points mark. On the other hand, the GeForce 6800, the direct rival to the RADEON X800 GTO, could not score even 4000 points. But let’s examine the situation in more detail.
The PowerColor X800 GTO VIVO is far ahead of the GeForce 6800 in the “pure speed” mode, but sinks to its level at the “eye candy” settings, despite its more advanced memory subsystem. Unfortunately, we couldn’t test the GeForce 6800 in high resolutions because the benchmark requires 256MB of graphics memory to launch in high display modes with enabled FSAA and anisotropic filtering. The GeForce 6800 GS only meets some competition from the RADEON X1600 XT in 3DMark05.
The second test scene is somewhat limited if compared with the scenes from the first or the third tests. The PowerColor X800 GTO feels confident here and is not so far behind the GeForce 6800 GS as in the first test – the gap is no more than 10% in the “pure speed” mode and 20% in the “eye candy” mode.
The micro-architecture of the PowerColor X800 GTO is not optimized to execute long pixel shaders the scene abounds in, hence, the particular results are lower compared to competing products because of relatively low clock-speed as well as only 12 pixel processors. As you can easily guess, the GeForce 6800 GS holds the top place for itself in all the resolutions, while the RADEON X800 XL and RADEON X1600 XT share the second position. The former makes up for the architectural deficiencies with its 16 pixel pipelines, while the latter features Shader Model 3.0 support and is optimized to execute pixel shaders of the maximum degree of complexity.
So, the overall scores are well-deserved, too, although the GeForce 6800 GS is generally faster than the RADEON X1600 XT in high resolutions and in the “eye candy” mode. As for the RADEON X800 GTO, it is impeded by its lack of Shader Model 3.0 support and by its 12 pixel pipelines. For example, the 16-pipelined RADEON X800 XL is no worse than the GeForce 6800 GS in the 3DMark05 tests.
Having tested one version of the RADEON X800 GTO, we can claim that it is as dangerous a rival to the GeForce 6800 as the RADEON X800 GT is to the GeForce 6600 GT.
The ATI RADEON X800 GTO is generally faster – and much faster at times – than the NVIDIA GeForce 6800. Not a big surprise, actually, considering its higher clock rates and bigger amount of graphics memory. The exceptions are few: the shooters Doom 3 and The Chronicles of Riddick and flight sims from Maddox Games. These are all OpenGL applications, and ATI Technologies doesn’t seem to have a really efficient OpenGL driver as yet. This is not a very big problem, however, because there are very few OpenGL games around and some of them can work with Direct3D, too.
So, the bigger problem for ATI is that the NVIDIA GeForce 6800 GS can be already bought for less than the officially recommended $249, and that graphics card easily beats the RADEON X800 GTO, supports Shader Model 3.0 and HDR and overclocks well. Lacking the modern technologies and not overclocker-friendly because of the R430 chip, the RADEON X800 GTO looks pale against the new product from NVIDIA, even though surpasses the RADEON X1600 XT with its bottleneck of four TMUs, in almost all games.
Talking about the PowerColor X800 GTO VIVO in particular, the overall impression about this device was somewhat marred by the lack of any games in the package as well as by the compatibility problems with some mainboards. Those problems may be solved in future BIOS updates, but you must agree it is not a pleasant situation to buy a new graphics card only to find that it refuses to work normally on your mainboard.
Otherwise, it is a typical PowerColor product, with no gorgeous accessories and a simple cooling system. It is a reliable and inexpensive graphics card based on the time-tested RADEON X800 architecture for people who do not need high performance in OpenGL applications and Shader Model 3.0 in those few games that use it. As a kind of bonus, PowerColor offers the VIVO functions for people who capture video from analog sources.
The product won’t interest overclockers. There’s a tiny, negligible chance to unblock the extra pipelines, while the overclocking potential of the employed R430 chip is very low, usually not more than 20-30MHz. We think we were just lucky to speed up the GPU of our sample of the card to 450MHz, even though the extra 50MHz helped the PowerColor X800 GTO VIVO to overtake the RADEON X800 XL in some tests.
And still, if you are choosing between a PowerColor X800 GTO VIVO and a GeForce 6800, the decision is obvious, and not in the favor of the GeForce.