by Alexey Stepin , Yaroslav Lyssenko
12/19/2006 | 03:39 PM
The ATI Radeon X1950 Pro graphics card has easily won the title of the best product at $199, leaving its immediate market rival Nvidia GeForce 7900 GX far behind in almost each and every test. ATI can price the new card that low because it uses the RV570 chip (for details see our article called Midheavyweights: ATI Radeon X1950 Pro against Nvidia GeForce 7900 GS), developed specifically for the mainstream sector, and a rather simple PCB.
The ATI RV570 was designed in accordance with the 3-to-1 concept, having 36 shader processors, 12 TMUs and 12 ROPs. Unlike the Radeon X1900 GT, the Radeon X1950 Pro model doesn’t have non-functioning subunits. Coupled with the 0.08-micron tech process, this has helped reduce the chip’s power consumption a little, yet it is still rather high at 65W in 3D mode, which is much higher than that of the GeForce 7900 GS (45 watts) and even higher than the power draw of the more advanced GeForce 7950 GT (about 60 watts). The large amount of dissipated heat didn’t allow ATI to install a simple cooling system, like Nvidia does on its GeForce 7900 GS, GeForce 7900 GT and GeForce 7950 GT. So, ATI developed a new copper cooler especially for the new card. It is as large and heavy as the cooler of the Radeon X1800 GTO/X1900 GT, but has better thermal and ergonomic properties.
Although we couldn’t find any fault with the new cooler that showed itself an efficient and quiet solution, some of ATI’s partners decided different. Particularly, PowerColor has released two models with a more advanced cooling system, besides the regular PowerColor X1950 Pro 256MB equipped with the reference cooler. The original models are called PowerColor X1950 Pro Extreme 512MB and PowerColor X1950 Pro Extreme 256MB and it’s about the model with 256 megabytes of graphics memory on board that we’ll be talking today. The two models don’t differ in anything else besides the amount of memory, so we’ll be referring to the card as to PowerColor X1950 Pro Extreme throughout this review.
This is a product selling in retail, so its packaging and accessories need a description.
This simple design solution isn’t anything new for PowerColor. We’ve seen it in the design of the packages of the PowerColor X1900 GT, PowerColor X1900 XT, and PowerColor X1800 GTO. This simplicity and unification are appropriate here because this design doesn’t irritate one’s eyes and creates a unified image of the company’s product line-up. Besides some standard captions, there is a “Silently Cooled” sticker on the box. It informs you that the graphics card is equipped with a cooler from Arctic Cooling.
Inside the glossy wrapper there is a white cardboard box with the graphics card in an antistatic pack. The accessories can be found under the card. Here they are:
This is a standard set of accessories PowerColor cards come with, but there are a few points that need to be specifically mentioned. First, we were somewhat surprised to see only one DVI-I → D-Sub adapter since the card is equipped with two DVI connectors. Second, a flexible bridge to build a CrossFire configuration is missing. However, PowerColor assured us that all their graphics cards would be shipping with a CrossFire cable and in case the cable is not in the box (which we discovered in the very first packages) they will ship one to the users directly upon request.
As you know, RV570-based devices do not need a special Master card since their graphics core already contains all the necessary logic and the two cards in a CrossFire configuration are joined together by means of two flexible cables.
We don’t know if all the owners of the packages with the absent cable will figure out that they have to claim it from the manufacturer, but it was missing in the box with our card and we wouldn’t have been able to build a CrossFire tandem out of two PowerColor X1950 Pro Extreme. In our opinion, the connecting bridge must be included with every card by default, the integrated frame compositing engine being one of the key features of the Radeon X1950 Pro. Otherwise, there is no sense in that integration because it may be very difficult to find and purchase such a bridge separately.
The brief manual is in fact a poster in several languages that gives you some basic info about how to install the card into your system. The information about configuring a CrossFire tandem doesn’t refer to this model since it describes configurations with a Master card and those that transfer data through the PCI Express bus. If you want a pair of Radeon X1950 Pro cards to work in CrossFire mode, you have to connect them with two flexible bridges and then enable the necessary mode in the Catalyst Control Center.
The previous version of the manual having been a rather thick brochure that covered most of the installation and usage issues, the new manual from PowerColor looks like a step back, especially as we didn’t find a full version of the manual on the CD with drivers.
Despite the mentioned drawbacks, this set of accessories is sufficient for using the PowerColor X1950 Pro Extreme. Not all users want to pay extra for such bonuses as additional games, CD cases or web-cameras.
The graphics card from PowerColor differs from the ordinary Radeon X1950 Pro with its cooling system design only. They are identical otherwise.
Even with a new PCB, the Radeon X1950 Pro is still quite long and may not fit into small system cases. The position of the power connector is questionable. It is placed in the middle and near the edge of the PCB, which is less convenient than the usual position in the top left corner.
The PowerColor X1950 Pro Extreme processor works at 595MHz, although the manufacturer’s website declares 600MHz for all the X1950 Pro series models, including the model that matches ATI’s reference card. We don’t know the reason for this erroneous information.
The graphics card carries eight 256Mb Samsung K4J55323QG-BC14 chips (8Mx32) for a total of 256 megabytes. The X1950 Pro Extreme 512MB model seems to be equipped with Samsung K4J52324QC chips that have a twice larger capacity. The memory frequency of the PowerColor X1950 Pro Extreme is 700 (1400) MHz. This is the rating frequency of 1.4ns chips and is somewhat higher than that of the reference card, which is 690 (1380) MHz.
Traditionally for ATI’s Radeon X1800, X1900 and X1950 series, the card carries a Rage Theater chip that endows it with VIVO functionality – not a very useful feature nowadays, yet a nice additional bonus that the competitor’s graphics cards do not offer. The PowerColor X1950 Pro Extreme has a standard configuration of connectors that a majority of today’s top-end graphics cards have: two DVI-I ports and one universal video output. Since this card supports VIVO, the video output is a 9- rather than 7-pin connector. It is not compatible with the standard S-Video connector and requires an adapter (like the one included with the card) to connect a TV-set to this interface.
The PowerColor X1950 Pro models with the word Extreme in the name are equipped with the Accelero X2 cooler developed by Arctic Cooling specifically for ATI Radeon X1800/X1900/X1950 cards instead of the single-slot cooler from ATI we described in an earlier review. Let’s take a closer look at this cooler.
The Arctic Cooling Accelero series is absolutely different from the Silencer series, although uses the same configuration of fan blades. The Silencer exhausts hot air through slits in the graphics card’s mounting bracket, but the Accelero drives the stream of air perpendicularly to the mainboard’s PCB. This is a queer and, in our opinion, questionable solution. The hot air from the graphics card won’t harm the mainboard much, but won’t do it any good, either. Note that the heatsink and casing of the Accelero X2 are shaped in such a way that some of the air stream goes out at the side and is blown at the graphics card’s power circuit elements.
In spite of the radical difference in the principle of operation, the Accelero heatsink can be viewed as a development of the Silencer: there is a copper base with a heatsink consisting of aluminum plates. It is also equipped with three heat pipes for a more uniform distribution of heat. Every connection in the cooler that might affect the cooling efficiency is made very carefully, so we didn’t find anything to cavil about.
The heatsink is covered with a black casing in which a 60mm Arctic Cooling fan (a max speed of 2000rpm) is installed. For comparison, the Silencer series used a 72mm fan. The fan runs on a fluid bearing that reduces noise and increases the fan lifetime. Arctic Cooling declares an MTBF of 400,000 hours, shipping its Accelero coolers with a 6-year warranty. The fan uses a 3-wire connection with a standard connector for ATI cards.
The fan is fastened in a way we’ve never seen before. In most other coolers the fan is fixed firmly, but the Accelero uses a rubber membrane the fan motor is installed on. The flexible hanger prevents vibration to pass from the fan to the cooler’s casing, thus eliminating any extra tones in the spectrum of noise it generates. Coupled with the fluid bearing, this should make the Accelero X2 a near-silent cooler. We’ll check this out soon.
The thermal interface is provided by the cooler developer, too. There are elastic 3M thermal pads between the cooler and the memory chips, and the copper sole contacts with the GPU surface through a layer of high-quality Arctic Cooling MX-1 grease. It doesn’t conduct electricity, so as not to close any contacts on the GPU wafer, and hardens in 200 hours after application, preventing evaporation or leakage that might worsen the quality of heat transfer.
According to Arctic Cooling, a layer of MX-1, once applied, can serve effectively for 8 years.
The Arctic Cooling Accelero X2 is intended to cool hottest graphics cards from ATI Technologies, so it is easy for it to cope with the not-very-hot Radeon X1950 Pro. Perhaps we’ll get better results at overclocking. Let’s check this out and see how quiet this cooling system is.
We measured the level of noise produced by the cooling system installed on the PowerColor X1950 Pro Extreme with a digital sound-level meter Velleman DVM1326 (0.1dB resolution) 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 a working testbed with a passively cooled graphics card inside was 40dBA. We got the following results:
The PowerColor X1950 Pro Extreme is somewhat better than the reference Radeon X1950 Pro in its noise characteristics. Basing on the readings of the sound-level meter, we might say that the graphics cards produce the same noise as it’s impossible to catch such a small difference by ear. But besides the numbers in decibel, there are certain nuances you can indeed catch without any tools. The noise spectrum of the Accelero X2 doesn’t have any plastic tone caused by the resonating casing because the fan is fastened through a vibration-damping membrane. The fan is always rotating at 2000rpm, so the PowerColor X1950 Pro Extreme produces the same noise all the time. Although the card is not exactly silent, it may be considered as such because its noise isn’t audible against the noise from other system components like hard drives, CPU cooler, and system/PSU fans.
Alas, the PowerColor X1950 Pro failed in our overclockability tests in spite of the advanced cooling system. The core of our card, which is pre-overclocked by the manufacture, showed some frequency growth like 15-20MHz at first, but began to display image artifacts after a while and we had to roll back to 595MHz. The memory refused to get overclocked at all. There were flickering bands and other signs of over-overclocking after we had tried to increased its frequency even by 5MHz.
Perhaps there was poor contact between the cooler’s sole and the GPU die, but we think the real reason was zero overclockability of our sample of the PowerColor X1950 Pro Extreme. You may be luckier with another sample of the same graphics card model. Since we didn’t achieve a tangible growth of GPU and memory frequencies, we decided to benchmark our PowerColor X1950 Pro Extreme 256MB only at its default clock rates of 595/700 (1400) MHz.
During our comparative testing of the Radeon X1950 Pro and GeForce 7900 GS we used the following hardware platforms:
The graphics card drivers were set up in such a way as to provide the highest possible quality of texture filtering.
We selected the highest possible graphics quality level in each game. We didn’t modify the games’ configuration files. Performance was measured with the games’ own tools or, if not available, manually with Fraps utility. We also measured the minimum speed of the cards where possible.
We tested the cards in three standard resolutions according to our testing methodology: 1280x1024, 1600x1200 and 1920x1200. Since PowerColor Radeon X1950 Pro Extreme belongs to high-end graphics cards, we only used FSAA 4x + Aniso 16x mode. We enabled FSAA and anisotropic filtering from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia ForceWare.
We ran the tests with disabled FSAA only for those games that do not support FSAA due to the specifics of their engine or use HDR (FP16). The thing is that the GeForce 7 family cannot perform FSAA together with floating-point HDR.
Besides the PowerColor Radeon X1950 Pro Extreme 256MB, we included the following graphics cards into this review:
We used the following games and benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters:
The three Radeon X19x0 models included into this review have ranked up in order of ascending specs. The PowerColor has a tiny advantage over the reference card from ATI, less than 5%, notwithstanding the word Extreme in its name. The GeForce 7900 GS is considerably slower than its opponents. Despite the generally high level of performance, the Nvidia card is 25% slower than the PowerColor in 1600x1200.
Battlefield 2142 is a much heavier application than its predecessor, but the graphics cards on ATI chips feel at ease here, at least in the resolution of 1280x1024 pixels. The PowerColor has a tiny advantage over the reference card, which amounts to only 5% in 1600x1200.
The increased GPU clock rate helps the PowerColor X1950 Pro Extreme take first place, a little ahead of the GeForce 7900 GS. But this is not enough to provide a comfortable performance with enabled antialiasing: a frame rate like 30fps is required for that, not 3fps!
There is no big difference between the PowerColor X1950 Pro Extreme and the GeForce 7900 GS in 1280x1024, but the Nvidia card can barely give out a comfortable frame rate in 1600x1200. It is ATI’s Radeons that dominate in 1920x1200 as they have a higher GPU clock rate and a more efficient memory controller.
The GeForce 7900 GS is on the losing side from the beginning on the Research map because its 20 pixel processors working at 450MHz cannot match the 36 processors of the Radeon X1950 Pro and Radeon X1900 GT. The overclocked Radeon X1950 Pro isn’t much faster than ATI’s reference card – they differ by no more than 3fps, although the speed of processing pixel shaders, which depends directly on the GPU clock rate, is a crucial factor in this test.
Note that the resolution of 1920x1200 is unavailable for owners of GeForce 7900 GS, not to mention GeForce 7600 GT, due to low performance provided by these cards.
The game’s FP HDR mode can be used in practice on Radeon X1950 Pro and PowerColor X1950 Pro Extreme graphics cards, but only in resolutions not higher than 1280x1024. The speed would be too low in the higher display modes.
This doesn’t refer to the GeForce 7900 GS, which provides a near-comfortable frame rate on the Research map, but fails to cope with the increased load on the Pier level that is all about open spaces with lots of vegetation and water.
The PowerColor card turns in the best result here, but doesn’t reach 60fps which is the comfortable speed for first-person shooters. Well, you can try to play the game as the speed is never lower than 25fps, yet we’d recommend you to disable FSAA on your $199-class graphics card to have some reserve of performance for the hardest game scenes.
As you could have learned from our review, F.E.A.R. Extraction Point differs from the original only in making wider use of complex visual effects. That’s why it’s not surprising to see the same picture of performance here as in F.E.A.R. Moreover, the average performance of every graphics card is somewhat higher than in the original game.
The deferred rendering technique employed by this game makes it incompatible with full-screen antialiasing, so the cards were benchmarked with anisotropic filtering only.
The increased GPU clock rate allows the PowerColor X1950 Pro to successfully contend with the GeForce 7900 GS here, but the game engine is so resource-consuming that the frame rate of 60fps is unachievable. On the other hand, there are no slowdowns to below 25fps which is the minimum you need for smooth gameplay.
The PowerColor X1950 Pro Extreme is 4-5% ahead of the ordinary Radeon X1950 Pro in every resolution. This is not much, so 1600x1200 pixels is the highest resolution you can play this game with comfort on either card in 4x FSAA + 16x Aniso mode. This should satisfy a majority of gamers because monitors that support higher resolutions aren’t widespread. And owners of such monitors shouldn’t find it a problem to buy a more advanced graphics card, we guess.
The GeForce 7900 GS leaves the Radeon X1950 Pro behind in this OpenGL application, but the overclocked version of the latter card offered by PowerColor overtakes the leader. There is in fact little difference, like 1-2fps, between these three solutions. None of today’s mainstream graphics card from the $199 category can yield 60fps with enabled 4x FSAA and at the highest graphics quality settings in this game.
The cards take the same places as in the previous test, but deliver higher performance in general. Almost all the included cards allow playing with comfort in 1280x1024 with enabled 4x FSAA. The GeForce 7600 GT with only 12 pixel processors, 8 ROPs and a 128-bit memory bus is an exception as it belongs to a lower product category than the rest of the tested graphics cards.
You need a graphics card with 512MB of memory on board to get acceptable performance in Serious Sam 2 at the highest graphics quality settings. The graphics cards in this test all have 256MB of memory and their performance is low. Note, however, that the GeForce 7900 GS has the best minimum of speed but also the worst average frame rate due to its 20 TMUs. It’s because the game makes wide use of shaders with multiple texture lookups.
The PowerColor X1950 Pro Extreme is successfully competing with the GeForce 7900 GS in 1280x1024 and has no rivals in the higher resolutions, but its minimum frame rate is considerably lower than that of the Nvidia card. So, the GeForce 7900 GS should be considered a better choice for Hitman: Blood Money due to the bigger reserve of performance it provides. If you want to have a high speed in this game on your $199 graphics card, you should turn FSAA off.
In this game, it is all about how fast the graphics card can execute pixel shaders. That’s why it is expectable to see the PowerColor X1950 Pro Extreme in the lead. Third-person shooters requiring a lower average frame rate to run smoothly than first-person ones, the Radeon X1950 Pro is fast enough for you to play in the widescreen 1920x1200 resolution with 4x FSAA and 16x anisotropic filtering.
As we wrote in our review of the GeForce 8800 GTX model, the Next Generation Content mode is only applicable on new-generation graphics cards like Nvidia’s G80-based ones. We shouldn’t expect an acceptable performance from mainstream $199 graphics cards, especially with enabled full-screen antialiasing. Note, however, that the Radeon X1950 Pro and its overclocked version from PowerColor still deliver the best results among the tested products.
The current version of Gothic 3 doesn’t support FSAA, so we benchmarked the cards with anisotropic filtering only.
Version 1.09 of the game is too heavy for the graphics subsystem and none of the $199 cards can provide even a minimum of comfort here. We should wait for patches and hope they’ll improve this situation.
This game looks best when you enable HDR, so we benchmarked the card in that mode. Moreover, the Nvidia GeForce 7 series cannot use both FP HDR and full-screen antialiasing simultaneously.
Due to time limits, we had to use the same version of the game as in our GeForce 8800 GTX review. It has a number of drawbacks, doesn’t support refraction and reflection effects on ATI Radeon X1000 cards, and runs too slow. As you can see, none of the cards can give you more than 20fps.
The Nvidia GeForce 7 architecture doesn’t allow using FP HDR along with FSAA, so we benchmarked the cards in TES IV: Oblivion with anisotropic filtering only. The game loses much of its visual appeal without HDR, although this is an arguable point.
The GeForce 7900 GS performs somewhat better than the PowerColor X1950 Pro Extreme does, but these results are of little practical interest due to low performance. The resolution of 1280x1024 is available for normal play, but it is the PowerColor card that has better average and min frame rates then.
In open scenes the R580 and RV570-based cards provide a much higher minimum of speed than the GeForce 7900 GS does, except for 1920x1200. So, it is the solutions on ATI’s chips that are the best among $199 products for playing TES IV: Oblivion .
Nvidia’s GeForce 7 cards have low performance in this game, and the GeForce 7900 GS can’t compete with the Radeon X1950 Pro. The PowerColor X1950 Pro Extreme has a tiny advantage over the reference card, just like in most other tests. It is comfortable to play this game even at an average speed of 30-40fps, but we guess it would be wise to disable FSAA to have enough performance in action-heavy scenes.
The ATI Radeon X1000 series doesn’t support vertex texturing and cannot render the water surface with the best possible quality in this game. This is only available on Nvidia’s GeForce 7 and 8 cards.
The GeForce 7900 GS is much faster than the PowerColor X1950 Pro Extreme and allows playing in 1600x1200. ATI’s graphics cards cannot do the same because the game engine is optimized for Nvidia’s solutions.
It’s quite the opposite of the previous test because X3: Reunion prefers graphics card on ATI Technologies’ chips. Each of the Radeon X19x0 models included into this review provides a comfortable level of performance with enabled 4x FSAA whereas the GeForce 7900 GS can’t do that even in 1280x1024.
Behaving alike to the Radeon X1950 Pro, the PowerColor X1950 Pro Extreme outperforms the GeForce 7900 GS in two out of three resolutions and loses to it in 1920x1200. There’s little difference between the PowerColor and the reference card from ATI.
This game abounds in cutting-edge technologies and special effects, allowing both versions of Radeon X1950 Pro to have a higher average frame rate than that of the GeForce 7900 GS. The Nvidia card provides a higher minimum speed in two out of three resolutions, though. Moreover, it’s clear that you can’t play this game on $199 graphics cards with 4x FSAA due to terrible slowdowns in some scenes. In order to ensure an acceptable level of comfort, you’ll want to give up antialiasing for this game.
The PowerColor X1950 Pro Extreme enjoys a small advantage over the ordinary Radeon X1950 Pro, but it grows ever smaller in higher resolutions. It is the memory subsystem performance that becomes a crucial factor in FSAA modes and it is almost identical between these two graphics cards.
The relatively small advantage in terms of GPU frequency helps the PowerColor X1950 Pro Extreme reach the 10,000 points mark, scoring 404 points more than the standard Radeon X1950 Pro does. This is expectable since 3DMark05 defaults to 1024x768 resolution without FSAA, and the memory subsystem performance isn’t very important under such conditions.
We can see the same trend in every test: the PowerColor card has the biggest advantage over the reference card in 1280x1024. In the higher resolutions it is limited by the memory frequency. The GeForce 7900 GS is always slower than the PowerColor, but differently in each test. The lag is the biggest in the second test that requires high speed of execution of shaders, both vertex and pixel, but doesn’t need a high fill rate.
The PowerColor X1950 Pro Extreme wins in 3DMark06, too, but doesn’t look as impressive as in 3DMark05: it scores a mere 662 points more than the GeForce 7900 GS.
The GeForce 7900 GS isn’t far behind the leader in the first group of graphics tests because its lower core frequency is partially compensated by the large number of TMUs. The PowerColor X1950 Pro Extreme scores 226 points more, though.
The models of the Radeon X19x0 series look much better than the GeForce 7900 GS in the SM3.0/HDR tests due to their higher computing power and more advanced architecture.
Everything remains the same in the SM2.0 tests if we take them separately. The first large-scale scene requires a high fill rate, so the GeForce 7900 GS looks competitive against its opponents. The second test demands high performance of pixel and shader processors, and the ATI Radeon X1900/X1950 family wins easily. The PowerColor X1950 Pro Extreme enjoys an almost twofold advantage over the GeForce 7900 GS in the latter case.
So, the overall scores of the cards look well-deserved, making allowances for the use of FSAA in the separate graphics tests.
PowerColor’s version of the ATI Radeon X1950 Pro graphics card has showed itself a well-made product. Its key feature is the unique cooling system developed by Arctic Cooling. This made the PowerColor X1950 Pro Extreme larger, but also almost silent, notwithstanding its increased GPU and memory frequencies.
Alas, this Extreme card from PowerColor is not extremely faster in comparison with the reference Radeon X1950 Pro. There is tiny difference between them in most applications, 5% at best, while the overclocking potential of the PowerColor card is near zero (this may refer to our sample of the card only; other samples may be better at overclocking).
The PowerColor X1950 Pro Extreme has exhibited all of the strong aspects of the standard Radeon X1950 Pro. That is, it beat the competing GeForce 7900 GS in almost each and every test we tried them in. In some cases, particularly in Ghost Recon Advanced Warfighter, Prey and Quake 4 , the small performance increase over the reference card helped the PowerColor catch up with the Nvidia card, thus becoming an overall winner even in those applications where the GeForce 7900 GS performed on its home turf.
It’s about the accessories that we’ve got some complaints to express. First, there is only one DVI-I → D-Sub adapter included whereas the card has two DVI connectors. Second, the lack of a flexible connecting bridge may make it impossible to build a CrossFire configuration in case you purchase a second card in the future because it’s next to impossible to find this bridge selling separately. So make sure, that you request this cable from PowerColor, as they assured us that they would be supplying all the users who received no cable with their card for free. The user manual might also be more informative and detailed, especially as we didn’t find a full electronic version of the manual on the CD with drivers. But besides these things, the accessories to the PowerColor X1950 Pro Extreme 256MB are sufficient and do not include any extras that would make the product more expensive.
So, we’ve got positive feelings about the PowerColor X1950 Pro Extreme 256MB. A relatively small sum of money buys you a high-performance mainstream solution, a little faster than the standard Radeon X1950 Pro. It provides a high-quality anisotropic filtering, unachievable on same-class solutions from Nvidia, and comes with an efficient Arctic Cooling Accelero X2 cooler. Unfortunately, it’s going to be difficult to built CrossFire configurations out of two such cards because PowerColor doesn’t think it necessary to include appropriate connectors.