by Alexey Stepin , Yaroslav Lyssenko
08/25/2006 | 09:27 AM
Historically, the Radeon X1900 XTX was released by ATI Technologies in response to Nvidia’s GeForce 7800 GTX 512 which had earlier proved its superiority over the ATI Radeon X1800 XT. ATI took an original approach to increasing the performance of the flagship product in the Radeon X1000 series which has come to be known as the 3-to-1 concept. The point of this concept is explained in our review of the Radeon X1900 XTX called The Fast and Furious: ATI Radeon X1900 XTX Review, but we want to repeat it in brief here. ATI holds that this ratio of arithmetic to textural operations is optimal and promises the highest possible performance in modern games since there is a clear tendency among game developers to maker wider use of math1ematically complex shaders to create various special effects.
This is very true nowadays when quite a lot of projects are developed for several platforms, for the PC and for a variety of gaming consoles (Microsoft’s Xbox and Xbox 360 and Sony’s PlayStation 2 and 3). Even last-generation consoles have a small amount of memory in comparison with modern PCs, so the developers have to do without high-resolutions textures that require a lot of memory to be stored. One and the most popular way to make a game’s visuals appealing is to employ pixel shaders with complex math1ematical computations. Such shaders don’t require too much memory, but use the GPU’s execution units to the full, thus providing spectacular visual effects at a rather low cost. Shader-based effects in combination with a rather low resolution the game consoles work with allow using lower-quality textures. From this point of view, ATI’s putting more pixel shader processors into its recent GPUs looks justifiable and compatible with game developers’ plans.
The ATI R580 GPU was the result of this strategy. It inherited all the features of the R520, but was endowed with 48 rather than 16 pixel processors. The pure computational power of the new GPU thus grew by 300%, although it only had 20% more transistors than the older chip. The new GPU also acquired a large Z-buffer, a few more general-purpose registers, and a Fetch4 feature to accelerate the rendering of shadows. All this helped the Radeon X1900 XTX become the fastest top-end graphics solution with a good reserve for the future and with high performance in existing games. In our review the new flagship product from ATI beat the GeForce 7800 GTX 512 in almost each and every gaming test.
There is a downside to each compromise, though. The realities of the PC world are different from those of the console realm. High resolutions and full-screen antialiasing are widespread here, and low quality of textures strikes the gamer’s eye immediately. Texturing speed is the crucial parameter in such modes and it directly depends on the number of texture-mapping units in the graphics processor. Moreover, even when a complex realistically-looking material is rendered by means of pixel shaders, such shaders have to contain a lot of texture lookups and thus put a considerable load on the graphics card’s TMUs and memory subsystem.
In other words, the quantity of TMUs is still a decisive factor in many cases, and the release of the G71 chip and the GeForce 7900 graphics card series confirmed that point. The senior member of the series GeForce 7900 GTX didn’t differ architecturally from the GeForce 7800 GTX 512, but had a higher core clock rate thanks to the transition to the 0.09-micron tech process: 650MHz against the predecessor’s 550MHz. This was enough to increase the efficiency of the GeForce 7900 GTX to the level of the Radeon X1900 XTX and even higher in some cases.
So, why was it so simple for Nvidia to restore the balance? Because game developers have to account for the huge amount of graphics cards currently in use that are not as good as the Radeon X1900 series at math1ematical calculations. So, if they used too many math1ematics-heavy shaders, they would leave quite a lot of users out of play, and this means lost profits. Besides that, the use of high resolutions and full-screen antialiasing brings about new requirements to the graphics subsystem as we’ve said above.
As a result, the combination of 24 TMUs with 24 PSUs (Pixel Shader Units) have proved to be almost as efficient as the combination of 16 TMUs with 48 PSUs. Yes, the GeForce 7900 GTX sometimes provides a smaller reserve of speed, i.e. has lower minimum performance, but in average performance it is at least no slower than the Radeon X1900 XTX and also has much better thermal and electrical characteristics.
This problem became even more urgent for ATI after the arrival of the GeForce 7950 GX2 (for details see our article called Two for One: Nvidia's Dual-Chip GeForce 7950 GX2 Reviewed) because the company lost its competitive edge in the sector of top-end single-card graphics solutions. The fact that the GeForce 7950 GX2 is actually a SLI tandem didn’t make life easier for ATI. They had to find a way to improve the performance of top-end models in the Radeon X1000 series. ATI didn’t have time to equip its R580 chip with additional TMUs notwithstanding the flexibility of the Radeon X1000 architecture which allows making a chip from ready-made functional units as if from Lego pieces. This approach might have been fruitful, but would have also meant projecting a new chip almost from scratch. Although this could have been made faster than with a classic graphics architecture, it didn’t make sense in view of the imminent arrival of the next generation of GPUs. But where to find new resources then?
Actually, such resources were built into the Radeon X1000 architecture originally. We mean the memory controller. As we said above, using high resolutions along with FSAA raises the requirements bar not only in terms of texturing speed but also in terms of graphics memory bandwidth. So if there is no way to improve the former parameter, the other may be tinkered with. That’s exactly what they did to create the Radeon X1950 XTX!
The memory controller in the Radeon X1000 series had originally been designed to support GDDR4, so ATI decided to use this new memory type on the new card as having a much higher frequency potential than GDDR3 while having comparable electrical and thermal characteristics. ATI Technologies played a key role in the development of GDDR4, so the company had had a chance to try the new memory under real conditions before the release of graphics cards on next-generation processors where GDDR4 is expected to be used.
Mass production of the new memory was launched by Samsung in June of this year. The slowest chip in the GDDR4 series has an access time of 0.91 nanoseconds and is capable of working at 1100 (2200) MHz frequency whereas the fastest GDDR3 chip is only rated for 900 (1800) MHz. If clocked at the same frequency of 1000 (2000) MHz, a GDDR4 chip would consume 45% less power than a GDDR3 chip. It means that the sky-high operating frequency of the new memory type does not lead to higher heat dissipation and does not make some special cooling systems necessary.
After BIOS-based reprogramming of the memory controller the R580 chip becomes R580+ and in that capacity is installed on a new family of graphics cards called Radeon X1950. The minor addition to the chip name and to the graphics card model name looks reasonable considering that the product characteristics have changed so little.
So what are the details of the Radeon X1950 XTX which is meant to beat the GeForce 7900 GTX and be competitive with the GeForce 7950 GX2? Here is its specification.
The new model doesn’t differ from the older one in anything except the graphics memory bandwidth, but the growth of the latter parameter by 14.4GB/s is not to be underestimated. For comparison, the memory bandwidth of the GeForce 7600 GS is a mere 12.8GB/s! Today, the bandwidth of 64GB/s is a record for 3D graphics hardware and is 9.4GB/s higher than the previous record set by the GeForce 7800 GTX 512 which clocked its memory at 850 (1700) MHz.
It’s clear that such a high memory bandwidth is not called for in low resolutions, so the Radeon X1950 XTX is unlikely to have any advantage over the GeForce 7900 GTX in such modes. But top-end graphics cards are usually bought to play in high resolutions and at highest graphics quality settings, usually with enabled 4x or even 6x/8x full-screen antialiasing. It’s under such conditions that the Radeon X1950 XTX will have an opportunity to show the advantages of the new memory type and be victorious over its rivals. The rest of the characteristics of the new graphics card from ATI Technologies are fully identical to the previous model’s. It has the same capabilities, including the ability to work with FSAA and HDR (FP16) simultaneously, support for adaptive antialiasing and for high-quality anisotropic filtering that doesn’t depend on the angle of view. Taking the baton from the R580, the R580+ features the highest functionality among consumer-class graphics processors and will probably remain such until the new generation of GPUs arrives.
The Radeon X1950 series consists of two models with identical technical characteristics: Radeon X1950 XTX and Radeon X1950 CrossFire. The latter has a Compositing Engine that is responsible for sewing together parts of the frame in CrossFire mode. For details about this engine and about CrossFire technology at large, refer to our review of the Radeon X1800 XT CrossFire called Caught in the Cross-Fire: ATI RADEON X1800 XT CrossFire Edition Review .
The new cards aren’t very expensive. The MSRP of both models is only $449. This is a serious blow at the GeForce 7900 GTX which has the same suggested price.
With prices like that, ATI says a highest-performance CrossFire tandem will only cost you $898 (without the cost of an appropriate mainboard). For comparison: an Nvidia quad-SLI complex consisting of two GeForce 7950 GX2 cards would cost you almost $1300 (without a mainboard). As soon as we have a chance to build a Radeon X1950 CrossFire platform in our labs, we’ll tell you how competitive the mentioned multi-GPU platforms are.
Unfortunately, the new graphics card won’t be available for purchase on the announcement day (August 23, 2006). It will begin shipping in mass quantities at the beginning of September. We don’t know exactly why the shipments of the new Radeon are delayed, but this is probably due to a shortage of GDDR4 memory chips.
The PCB of the Radeon X1950 XTX is a slightly modified PCB of the Radeon X1900 XTX, which is not a surprise considering that both these cards are based around the same GPU.
There seems to be no difference between the cards, but you can spot it on closer inspection. The power circuitries are different in the first place. A Volterra VT1103 controller is responsible for power on the Radeon X1800 and X1900, but the Radeon X1950 XTX comes with a Volterra VT1105 that has a rectangular rather than square packaging. So, the wiring of this part of the PCB was changed accordingly. This in its turn provoked other changes that concerned the placement of various smaller elements, capacitors, coils, etc. Some power circuit elements are not even soldered on the new card due to the reduced power consumption of GDDR4. Particularly, one power MOSFET with an accompanying inductance coil Pulse PA0511 is missing at the bottom of the PCB and a number of other elements are absent, too. You can see the same on the reverse side of the mainboard: some elements that used to be included into the memory power circuit are not installed now. Moreover, the fan connector, a 4-pin one now, has moved to the back part of the PCB.
There are fewer differences in that part of the PCB where the GPU resides, yet you can spot some: the contact pad with an unclear purpose (perhaps to install an additional TMDS transmitter) that was present in the top left corner of all Radeon X1800 and X1900 cards is now missing. There are more points of difference on the reverse side of the PCB, but all of them are due to lack of certain elements that used to be installed on the Radeon X1900 XTX. The Rage Theater chip that endows the card with VIVO functionality is in its customary place. There’s not much use from it today, but extra functionality is always welcome.
The marking on the GPU chip remained intact: there is no “plus” in the core code-name. The R580 chip on our sample of the card was manufactured on the 21st week of 2006, i.e. at the end of May. The core is clocked at 650MHz in 3D mode, just as on the Radeon X1900 XTX. This frequency is lowered to 500MHz in 2D mode. The die packaging is equipped with a metal frame for protection. This is an appropriate measure considering the large die area and the massive cooling system.
It was difficult to read the marking on the memory chips as it had been virtually erased by the heat-conductive pads, yet we managed to do that. There are eight Samsung K4U52324QE-BC09 chips of GDDR4 memory on the card. According to the specification, these 512Mb chips are designed as 16Mx32 and work at 1.8V voltage.
The BC09 suffix denotes an access time of 0.91 nanoseconds. This access time means that the chip is rated to work at a frequency of 1100 (2200) MHz, yet its consumption current is only 1130mA at the maximum (in Burst mode), which gives a power consumption of about 2W per chip. The fastest GDDR3 memory chip (K4J5324QC-BJ11), which works at 2.0V voltage and at 900 (1800) MHz frequency, consumes a current of 1180mA and has a power consumption of 2.36W. So, GDDR4 is obviously better than the older memory type in terms of power consumption, especially considering its higher operating frequency. Moreover, the BC09-suffixed chip is the slowest in the K4U52324QE series whereas the fastest BC07-suffixed chip is capable of working at 1400 (2800) MHz, consuming this same 2.36W.
The memory frequency is somewhat reduced below the rated one on the Radeon X1950 XTX card and is 1000 (2000) MHz. Accessed across a 256-bit memory bus, it provides a memory bandwidth of 64GB/s, which is the new record for consumer 3D graphics hardware. The memory frequency is lowered to 600 (1200) MHz in 2D mode.
The design of the cooling system ATI installed on its Radeon X1800 XT, Radeon X1900 XT and Radeon X1900 XTX cards was criticized in our as well as other resources’ reviews due to obvious reasons. The main error was made back at the times of the Radeon X850 XT. When designing the cooling system of that card, they put the blower behind the GPU heatsink, making it suck hot air into itself and throw it out of the PC case. It’s good to have the hot air exhausted, especially when it comes to top-end solutions with high heat dissipation, but sucking air in from a heatsink is always less efficient than blowing at it because the speed of the air stream is much higher in the latter case. That old cooler design was adapted for the increased heat dissipation of the Radeon X1800 and X1900 families by adding heat pipes and enlarging the heatsink ribbing, but the efficiency of the improved cooler was still far from perfect. Moreover, the plastic casing turned to be a good resonator and would sing in tune with the fan when the latter was working at an increased speed at high GPU temperatures. This characteristic and rather irritating “plastic” sound is familiar to all people who have ever dealt with ATI’s flagship products from Radeon X850 XT to Radeon X1900 XTX equipped with the reference coolers.
But after all ATI decided to use a new original cooling system on the Radeon X1950 XT which seems to be free from the deficiencies of the older one. This time there are no dubious innovations here: the cooler is designed in a classic way and resembles the popular Arctic Cooling Silencer series, but with a smaller diameter of the fan and certain improvements for higher cooling efficiency.
The cooler of the Radeon X1950 XTX is designed in such a way that the air stream from the fan cools the heatsink and is then exhausted out of the case through the slits in the graphics card’s mounting bracket. The GPU die has contact with a thick copper base a long thin-ribbed copper heatsink is bonded to. To distribute the GPU heat uniformly, the heatsink is also connected to the base with a heat pipe. The whole arrangement is covered with a casing which is made of translucent dark-red plastic. An X-shaped silvery piece with a relief ATI logo is on the front of the casing. The combination of red and silver looks quite impressive and stylish.
The cooler’s fan is 75 millimeters in diameter and is equipped with a stabilizing rim. The shape of the blades may be not as optimal as in Arctic Cooling fans, but this shouldn’t have any great effect on cooling efficiency. We don’t know the power rating of the fan and the type of its bearing. We guess a frictionless bearing is used as having a longer life cycle and better reliability but somewhat worse noise characteristics in comparison with sliding bearings. The 4-wire fan connection is indicative of possible improvements in the fan speed management system.
Notwithstanding its considerable size and weight, the cooler is secured on the PCB with only four screws, but the threaded poles are shaped in such a way as to prevent skewed installation whereas the metal spring-loaded frame doesn’t allow the PCB to bend. Coupled with the protective frame around the GPU die, this helps avoid damaging the device.
The memory chips are cooled by an additional copper heatsink milled out of a solid piece of copper. It has curiously shaped ribs and is fastened to the PCB with four screws separately from the main cooler. Special juts that are aligned with the holes in the PCB make sure that the cooler is installed exactly upright. The heatsink on the memory regulator’s MOSFETs is made from copper, too. There are special slits in the cooler casing for the fan to take more air in, ensuring additional cooling of the memory chips and power circuit.
Traditional dark-gray thermal paste, very thick and with a low thermal resistance coefficient, is used as a thermal interface between the cooler’s base and the GPU die. The memory chips have contact with the heatsink through blue-colored plastic pads.
The new cooler from ATI seems to be free from obvious drawbacks, being an implementation of a time-tested design. There’s only one thing left to check out – how quiet it is.
We measured the level of noise produced by the ATI Radeon X1950 XTX’s cooler 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 new cooler from ATI is almost noiseless and is only 0.8dBA louder than the passive solution most of the time. It’s only after long working in 3D mode that the fan begins to work at a higher speed, and the level of noise grows to 47.1dBA, which is still somewhat quieter than the GeForce 7950 GX2. At a distance of 5 centimeter the soft rumble of the fan bearing is the main source of noise.
The cooler is indeed silent subjectively and only reminds of itself when the GPU temperature reaches a certain threshold value. The spectrum of its noise is different from the Radeon X1900 XTX’s cooler and without that characteristic “plastic” tone. The noise seems quieter and softer despite the reading of the sound-level meter. So, ATI had succeeded in developing a very quiet and efficient cooler. There’s no need to replace it unless you plan to equip your gaming platform with a liquid or cryogen cooling system.
Then we measured the power consumption of the new graphics card on the following testbed:
The mainboard in this testbed was specially modified: we connected measurement shunts into the power lines of the PCI Express x16 slot and equipped them with connectors to attach measuring instruments. We also added such a shunt to a 2xMolex → PCI Express adapter. The measurements were performed with a Velleman DVM850BL multimeter (0.5% accuracy).
We loaded the GPU by launching the first SM3.0/HDR graphics test from 3DMark06 and running it in a loop at 1600x1200 resolution and with enabled 16x anisotropic filtering. The Peak 2D load was created by means of the 2D Transparent Windows test from Futuremark’s PCMark05 benchmarking suite. The results follow below:
The power consumption of the new Radeon X1950 XTX is nearly that of its predecessor, being only 4W higher under Max 3D load and 3.5W higher in Peak 2D mode. But the use of GDDR4 and the changes in the power circuit design have led to a different distribution of load across the lines:
Click to enlarge
It has become somewhat more balanced, although it is still the external +12V line that is under the highest load in 3D mode. Note also the increased load on the +3.3V line which used to supply 5.5W previously, but now supplies 6.5W.
Unfortunately, we can’t directly compare the Radeon X1950 XTX and the GeForce 7950 GX2 in this test because the latter refuses to work on the mainboard installed in our testbed. These cards should have comparable power consumption, though. By our estimate, the dual-chip solution from Nvidia consumes about 110-120W. Not yet having any accurate data about the power consumption of the GeForce 7950 GX2 we have to name the Radeon X1950 XTX the most voracious premium-class graphics card of today. Anyway, one such card can be easily fed by any high-quality 450W ATX 2.0 power supply. For building Radeon X1950-based CrossFire systems ATI recommends using a 550W or higher power supply. This rather moderate requirement is partially due to the low consumption of the CrossFire Xpress 3200 chipset.
There was no reason why the Radeon X1950 XTX could be good at overclocking since it uses the ordinary R580 core that seldom achieves frequencies above 670-680MHz without extreme overclocking methods like volt-modding. The current version of RivaTuner doesn’t support the Radeon X1950 XTX, so we had to use a special utility provided by ATI Technologies. Our apprehensions came true: using standard overclocking methods, we only managed to increase the frequency of the 384-million-transistor chip to 670MHz. This is normal considering that the R580 dissipates more heat than many CPUs.
The memory chips refused to be clocked higher than 1100 (2200) MHz: the Desktop would all go blinking squares and the system hung up, but the mentioned frequency was stable. For higher operating frequencies the wiring of the PCB needs optimization, we guess. So, the highest frequencies at which the overclocked Radeon X1950 XTX was stable were 670/1100MHz, and it would be unwise to demand anything more from a premium-class product. Graphics cards in this category usually work at the limit even at their default frequencies. Better results may be achieved through extreme overclocking methods, but this is beyond the scope of this review.
Focusing on speed in modern games in the first place, many hardware reviewers pay less or no attention to the quality of graphics, particularly to the quality of texture filtering and antialiasing. This approach became the norm about three years ago when ATI Technologies improved greatly the quality of anisotropic filtering on its Radeon 9700/9800 whereas Nvidia reduced that quality in its GeForce FX. As a result, the picture delivered by the competing products was almost the same quality, so speed became the major concern, especially because there was a definite lack of sheer speed in some games like Far Cry .
It is different today, however. The GeForce 7 series uses even simpler anisotropic and tri-linear filtering algorithms than those employed in the GeForce FX whereas the Radeon X1000 chips have acquired support of “high-quality AF”. So, the two competitors may differ in the quality of anisotropic filtering and the user must be aware that one graphics card may be slower but deliver a better-looking image and vice versa.
Our readers have reported they have seen “noise” in textures on Nvidia GeForce 7 graphics cards and have requested that we use a better-quality mode when testing products from that series in comparison with ATI’s Radeon X1000. Yet each time we tried to find any difference in static screenshots we couldn’t sport any great discrepancy in the quality of texture filtering on ATI and Nvidia cards. But since that noise shows up most conspicuously in motion, we decided to check it out by moving along the same trajectories in popular games and recording this into video clips by means of the Fraps utility.
Having watched a lot of records made in different games we’ve come to the conclusion that although the anisotropic filtering methods employed by ATI and Nvidia are far from perfect, products from the former company offer a much better tri-linear filtering, which results in the lack of eye-irritating texture noise in some cases.
For you to see it with your own eyes, we offer you a few video clips for download:
As you can see, the texture noise is obvious in some cases and can be put up with in others, like in Elder Scrolls: Oblivion . We can’t offer you a general recommendation, but each time we see a small advantage of a GeForce over a Radeon, which is not due to advantages or shortcomings of this or that architecture, we should be aware that this may be the consequence of the lower rendering quality.
We measured the performance of the ATI Radeon X1950 XTX on a testbed that was configured like follows:
ATI’s and Nvidia’s drivers were set up as follows.
We select the highest graphics quality settings in each game, identical for graphics cards from ATI and Nvidia. We do not modify the games’ configuration files and do not use the drivers’ profiles optimized for specific games. The frame rate is measured with the game’s own tools or, if not available, with the Fraps utility. We also measure a minimum frame rate where possible.
Besides the two standard resolutions of 1280x1024 and 1600x1200 pixels, we also used 1920x1200 resolution (with an aspect ratio of 16:10) in games that support widescreen modes. The reviewed card belonging to the top-end product category, we didn’t test it without full-screen antialiasing. We turned on FSAA and anisotropic filtering from the game menu. If such options were unavailable, we forced FSAA and AF through the appropriate options of the ATI Catalyst and Nvidia ForceWare driver.
Besides the Radeon X1950 XTX, the following graphics cards took part in the tests:
The cards were tested in these games and benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
The increased memory bandwidth of the Radeon X1950 XTX affects its fill rate positively when working with the Z-buffer only or when mapping one texture. In other cases the number of TMUs becomes a limiting factor and the GeForce 7900 GTX has three times more TMUs, its core frequency being the same as the Radeon X1950 XTX’s.
The new card from ATI doesn’t have anything new to show us when it comes to processing pixel shaders. The memory frequency isn’t important here while the core frequencies of the Radeon X1900 XTX and the Radeon X1950 XTX are identical at 650MHz. As a result, the cards deliver identical performance, except for the shader that emulates per-pixel lighting and accesses memory often. The Radeon X1950 XTX is considerably faster than its predecessor at executing that shader.
There’s no wonder that the Radeon X1950 XTX delivers results comparable to those of the Radeon X1900 XTX. Their GPUs contain the same number of pixel processors and are clocked at the same frequency of 650MHz. There are improvements, however. You can observe them in shaders that work with textures (Cook Torrance + Texture + Fresnel, 27-Pass Fur or NPR (hatch)) as well as in shaders that contain intensive math1ematic calculations (Metal + Phong, Wood). We’ll only see in real-life situations, i.e. in gaming tests, what effect this is going to have on the new card’s real performance.
Among other things we can note that the Radeon X19x0 family is still invincible when executing shaders with dynamic branching. Note also that the GeForce 7950 GX2 isn’t far better than the GeForce 7900 GTX in a majority of shaders included in this synthetic benchmark except for shaders with dynamic branching: the dual-chip solution from Nvidia improves its results here relative to the single-chip flagship thanks to the double number of pixel processors.
Just as we had expected, the Radeon X1950 XTX has a 10% advantage over the Radeon X1900 XTX in 1600x1200 and none in 1280x1024. This advantage is not big enough to catch up with the GeForce 7950 GX2, but the gap is only 10% and the average speed of 113fps is more than enough for comfortable play in the highest resolution supported by this game with enabled 4x FSAA. If you want to have an even better-quality picture, switch into 6x FSAA mode. The Radeon X1950 XTX is strong enough to permit that.
There is almost no gain from the use of GDDR4 in this game even in 1920x1200 resolution. The reason is that the game runs on a modified, yet still obsolete engine from Quake 3. The modification mostly comes down to hanging the engine over with numerous shader-based special effects. Those effects do not probably access the graphics memory too often, and the advantage of the new graphics card is just a few frames per second at best.
So, the new Radeon cannot challenge the GeForce 7950 GX2 in this test, but it keeps up with the GeForce 7900 GTX. Despite the old engine, the game is a heavy application and you can use FSAA on single-chip top-end graphics cards in resolutions not higher than 1280x1024.
Showing a large scene, the Pier demo is dependent on the CPU speed. This can be observed with premium-class graphics cards even when full-screen antialiasing is in use. It’s only in the resolution of 1920x1200 pixels, which is still a luxury for many gamers, that the Radeon X1950 XTX offers a few extra frames per second relative to its GDDR3-equipped predecessor.
The Research demo doesn’t rely so heavily on the CPU, and the Radeon X1950 XTX delivers a nice performance boost in high resolutions that varies from 10% to 15%. This helps it keep close to the GeForce 7950 GX2, although having only one GPU with 16 texture-mapping units. On the other hand, this test doesn’t need a high fill rate since the action goes on in the closed and rather small environment of an underground base.
There is a performance boost in the HDR (FP16) modes, but it is now in the highest resolution of 1920x1200 that the speed gain is the smallest. As a result, the Radeon X1950 XTX is capable of delivering the same speed as the GeForce 7900 GTX, even though HDR support for ATI’s cards is not implemented optimally in Far Cry.
The 1000 (2000) MHz memory on board the new Radeon doesn’t affect its performance much in F.E.A.R., but it is anyway ahead of the GeForce 7900 GTX in 1600x1200. The average speed of the new card is close to the comfortable 55-60fps.
This game employs the deferred rendering technique which is incompatible with full-screen antialiasing, so we only publish “pure speed” results here.
Even with disabled FSAA and with rather simple textures in use there is a clear effect from the super-fast memory of the new Radeon. It is especially conspicuous in the resolution of 1280x1024, amounting to 10%. Moreover, the new card has lifted the minimum speed limit up thus preventing the game from slowing down to below comfortable level in the most graphically complex scenes. But if you want to use resolutions higher than 1280x1024, you have to buy a dual-GPU solution: a GeForce 7950 GX2, Nvidia SLI or ATI CrossFire.
The Radeon X1950 XTX is less than 10% ahead of the Radeon X1900 XTX but this is just enough for it to catch up with the GeForce 7900 GTX in resolutions higher than 1280x1024. It is possible to play with enabled 4x FSAA even in 1920x1200. The more expensive GeForce 7950 GX is not actually needed for that, although it ensures a higher min frame rate.
Prey runs on the Doom 3 engine and uses OpenGL, but the Radeon X1950 XTX isn’t any worse than the GeForce 7900 GTX here and is 15% ahead of the Radeon X1900 XTX (and this difference in speed is only due to the difference in the memory clock rate of the two cards: 1000 (2000) MHz against 775 (1550) MHz). The GeForce 7950 GX2 defaults from this race, betraying one of the downsides of SLI technology: it needs support for the particular game in the ForceWare driver to work correctly.
Quake 4 is an OpenGL application, too, yet the Radeon X1950 XTX performs even better than in the previous test! It is a mere 3fps behind the GeForce 7900 GTX in 1280x1024 and is only second to the GeForce 7950 GX2 in 1600x1200, thus turning in the best result among today’s single-chip premium-class graphics cards. Owners of large monitors with a diagonal of 23-24” won’t be disappointed: the average speed of the new card is 65fps with enabled full-screen antialiasing.
Serious Sam 2 uses shaders with multiple texture lookups and it looks like it’s possible to speed it up not only by means of a lot of TMUs but also by increasing the memory bandwidth as is indicated by the results of the Radeon X1950 XTX. The new card is no farther than 5-6% behind the traditional leader GeForce 7900 GTX but has a twice higher minimum speed! With this performance, the Radeon X1950 XTX becomes the first card that allows playing Serious Sam 2 normally in the “eye candy” mode in a resolution higher than 1024x768. It looks like winning an away match!
The 20% performance boost lifts the Radeon X1950 XTX up to the level of the GeForce 7900 GTX, but its minimum speed is still lower than the Nvidia card’s. This is probably due to the fewer TMUs (high texturing speed helps render complex scenes faster) or to the game engine’s specifics.
Anyway, you don’t have a comfortable speed on any single-chip graphics card in resolutions above 1280x1024 if you enable FSAA, even considering that the comfortable speed bar is lower for third-person shooters than for first-person ones.
The use of the new high-performance memory type has a positive effect on the performance of the Radeon X1950 XTX in Splinter Cell: Chaos Theory. It is about 15% behind the GeForce 7950 GX2 in 1280x1024 and 10% behind in the higher resolutions. That’s an excellent result considering the lower price and the better noise characteristics of the new Radeon.
The super-fast GDDR4 memory on board the Radeon X1950 XTX is of little use in the special effects rich Tomb Raider: Legend, although we launched the game in the Next Generation Content mode in which Shader Model 3.0 capabilities are employed. The GeForce 7900 GTX is just a little faster than the two flagship Radeons and, like them, allows playing in resolutions no higher than 1280x1024 when full-screen antialiasing is enabled.
The Nvidia GeForce 7 architecture doesn’t allow using HDR (FP16) along with FSAA. That’s why we tested the cards in TES: Oblivion at the “pure speed” settings only: without HDR the game loses much of its visual appeal. The speed of the cards was measured with the Fraps utility, so the numbers may be somewhat inaccurate.
Fast execution of pixel shaders is important in closed scenes, so the Radeon X1950 XTX is just a little ahead of the Radeon X1900 XTX. The new memory should show its best in open scenes of the Tamriel world. We guess that the lower minimum speed of the new Radeon in 1920x1200 is most likely due to an error during the manual benchmarking. Note also that despite having a total of 48 pixel processors, the GeForce 7950 GX2 is slower than the single-chip cards from ATI in terms of minimum speed in resolutions up to 1600x1200.
The performance of the Radeon X1950 XTX has really improved relative to its predecessor in open environments of this game, but the difference isn’t as big as we had anticipated. Perhaps the 16 TMUs of the new card limit its performance somewhat. Anyway, the new Radeon takes the first place among single-chip cards, once again confirming the fact that ATI R580-based products are best for playing Oblivion.
Fast memory doesn’t have a big effect on performance in Titan Quest, either. In this Diablo clone you only need an average speed of 30-40fps for comfortable play. It is also desirable that the speed be always higher than 25fps. All modern graphics cards from the premium and high-end categories equipped with 512 megabytes of graphics memory can give you that speed. The GeForce 7950 GX2 acts up in 1280x1024 for some unclear reason, but regains its strength in higher resolutions.
Graphics cards from the Radeon X1000 family do not support vertex texturing and cannot use SM3.0 to render the water surface with the highest quality in this game. This is only available on Nvidia’s GeForce 7 cards.
We can’t expect an exceptional performance from the Radeon X1950 XTX because ATI’s solutions are limited here by the specifics of the game engine which is optimized for Nvidia’s architectures. The new Radeon allows playing Pacific Fighters with comfort in 1280x1024 with enabled 4x FSAA and 16x AF.
The fast memory of the Radeon X1950 XTX only leads to higher min frame rates in resolutions below 1600x1200. But in 1920x1200 the new card delivers a nearly 20% performance boost, making it possible to play with comfort in this resolution with enabled full-screen antialiasing. Moreover, this lifts the card up to the level of the dual-chip GeForce 7950 GX2, which is an achievement, too.
The Radeon X1950 XTX beats each of its opponents, including the GeForce 7950 GX2, enjoying the biggest advantage in low resolutions. The new card gives you more than enough speed even in 1920x1200: the frame rate is never lower than 40fps, so you’ll have accurate control over your army in the toughest mass battles. So, that’s yet another victory for ATI Technologies.
This test produces most curious results: the Radeon X1950 XTX is for some reason worse than the Radeon X1900 XTX in average performance but better in minimum speed. It must be the driver’s fault because there can’t be any other explanation considering that these two cards are identical in everything except the frequency and type of their memory chips.
By default, 3DMark05 uses 1024x768 resolution without full-screen antialiasing. The Radeon X1950 XTX cannot show its best under such conditions because its fast memory is of little use then.
The first test depends heavily on the speed of processing textures and, accordingly, on the graphics memory bandwidth. Since we also enable full-screen antialiasing in our tests, the Radeon X1950 XTX has something to show its muscle on and it does so, leaving the Radeon X1900 XTX from 15% to 20% behind.
The second test renders a smaller scene, and the Radeon X1950 XTX only enjoys a 10% performance boost. This is enough to make it the leader among single-chip cards, though.
The new card from ATI is about 10% ahead of the Radeon X1900 XTX across all the resolutions in the third test, too. It doesn’t reach the level of the GeForce 7950 GX2, but it’s incorrect to compare these two cards directly. Nvidia’s solution is a dual-chip device and is in fact a SLI tandem, although occupies one PCI Express x16 slot.
So, the new Radeon X1950 XTX really has a bigger potential than the Radeon X1900 XTX but only reveals it in high resolutions and/or with enabled full-screen antialiasing.
The Radeon X1950 XTX has a bigger advantage over its predecessor in 3DMark06 than in 3DMark05 if we compare their overall scores because the newer 3DMark defaults to 1280x1024 resolution. The gap is 231 points.
ATI’s new card cannot catch up with the GeForce 7900 GTX, which has 24 TMUs, in the SM2.0 tests. The difference is only 141 points, though.
But the new graphics card from ATI meets no competition in the SM3.0/HDR tests. It is faster by 216 points than the GeForce 7900 GTX which is a big enough difference, the overall scores being about 2500 points. Let’s now see how the Radeon X1950 XTX does in particular tests.
The first SM2.0 test is similar to 3DMark05’s first test, but the Radeon X1950 XTX has a smaller advantage here, less than 10% in high resolutions. Anyway, the new card is in the lead.
The second test produces a picture similar to what we have seen in Rise of Nations: Rise of Legends. The Radeon X1900 XTX is faster than the Radeon X1950 XTX in the standard resolutions. This is most likely a fault of the current version of the Catalyst driver.
None of Nvidia’s graphics cards can simultaneously work with HDR and FSAA, so the Radeon X1950 XTX can be compared only with its predecessor in the SM3.0/HDR tests. As in the previous 3DMark06 tests, the new card isn’t far better than the older one; the difference is no bigger than 10%.
It’s all generally the same in the second SM3.0/HDR test, but the gap is the biggest in 1920x1200 resolution, which is logical because the load on the memory subsystem is the biggest, too. So, the Radeon X1950 XTX is not far better than the Radeon X1900 XTX even if full-screen antialiasing is used.
ATI’s new Radeon X1950 XTX is a worthy successor to the Radeon X1900 XTX, the previous flagship product in the R580-based series. Equipped with new-type memory, capable of working at an unprecedented frequency of 1000 (2000) MHz, the card has got faster and better than the GeForce 7900 GTX where the Radeon X1900 XTX used to lag behind. As might have been expected, the biggest performance growth is observed when full-screen antialiasing is used along with high resolutions. The value of this growth varies from 5% to 20%, but is zero in some cases, e.g. in Call of Duty 2, F.E.A.R., and Tomb Raider: Legend. Anyway, this is an excellent achievement considering that the new card’s specs differ from the older model’s in only one point.
We also want to single out the fact that having the same or higher performance than the GeForce 7900 GTX, the Radeon X1950 XTX provides a much higher quality of anisotropic filtering. This is not too conspicuous in the heat of a battle, but the difference is striking in some cases. The Radeon X1950 XTX is undoubtedly preferable in this respect. Besides that, like any other member of the Radeon X1000 family, the new card from ATI permits to use FSAA and enhanced floating-point color representation (HDR FP) simultaneously which greatly improves the visuals in some games, e.g. in TES IV: Oblivion. Having two G71 chips under the hood, the Nvidia GeForce 7950 GX2 cannot offer anything like that with regards to image quality, but is ahead of the Radeon X1950 XTX in performance across a number of applications.
Our praises also go to the designers of the new cooling system the Radeon X1950 XTX is equipped with. It is a quiet, yet highly efficient cooler which is no worse than alternatives from Artic Cooling or Zalman and is far better than the older cooler that used to be installed on the Radeon X1900 XTX. Silence-seeking users won’t have to buy a third-party cooler for this card. And it looks impressive, too!
The only disappointing thing about the new card from ATI is that the company couldn’t begin mass shipments of it on the day of the announcement as it had done with the Radeon X1900 XTX. This is due to a certain deficit of GDDR4, it being a new memory type, but we hope this all will have been adjusted by September when the Radeon X1950 XTX is going to become widely available. At a suggested price of $449 and with excellent performance the new graphics card from ATI has a chance to get highly appreciated by those people who want to have not only a highest frame rate possible but also high image quality in games they play.