by Alexey Stepin , Yaroslav Lyssenko
04/22/2009 | 07:11 PM
The recent launch of the Radeon HD 4890 has changed the standings in the sector of single-chip graphics cards since AMD has got a product capable of rivaling the Nvidia GeForce GTX 285. Yet notwithstanding AMD’s strategy of filling the top market segment with dual-processor solutions, AMD’s graphics department is in no hurry to introduce a Radeon HD 4890 X2. Perhaps the company’s partners will be offered the opportunity to create such a product but AMD does not yet agree to it, and the Radeon HD series is still topped by the Radeon HD 4870 X2 which is somewhat inferior to the GeForce GTX 295 in performance (see this news story for details).
However, the arrival of the Radeon HD 4890 has provided the opportunity to build CrossFireX subsystems out of two, three or even four such cards. Even a dual-processor tandem should be no worse than a GeForce GTX 285 SLI, let alone a single GeForce GTX 295. And it is going to be cheaper because the recommended price of the Radeon HD 4890 is within $229-249. So, theoretically, a graphics subsystem made out of two such cards is going to cost you no more than $500. For comparison, the GeForce GTX 285 sells in retail for $330-340 and more, and a couple of such cards are going to cost more than $600.
The GeForce GTX 295 can be found selling at $500-550 but it will surely be inferior to a Radeon HD 4890 CrossFireX tandem, particularly because it has less of graphics memory available for 3D applications (896MB against 1024MB). This also refers to the GeForce GTX 275 SLI tandem that has higher clock rates but only 896MB of effective graphics memory and a cut-down GPU configuration with 28 raster back-ends per a core as opposed to the GeForce GTX 285 SLI’s 32 RBEs per a core. Thus, the most interesting thing is to compare the Radeon HD 4890 CrossFireX and GeForce GTX 285 SLI. The battle is going to be exciting if we put aside the factor of price and focus on gaming performance only.
Despite the global crisis, gamers are still interested in advanced and expensive multi-GPU configurations, and we will cater to our readers by performing a comparative test of Radeon HD 4890 CrossFireX and GeForce GTX 285 SLI tandems, especially as we’ve got four graphics cards necessary for such a test just in time. The “green” camp will be represented by EVGA GeForce GTS 285 FTW and Zotac GeForce GTX 285 AMP! cards. They will be opposed in this heavyweight fight by ASUS’s EAH4890 and EAH4890 TOP. Let’s now take a closer look at each of the fighters.
EVGA uses a unified design for its product packaging. It appeals to the customer with simplicity and neatness, differing from model to model with but a few details such as the color and texture of the strip at the top of the box and the typeface of the product’s model name. The design of the packaging of the EVGA GeForce GTX 285 FTW is the same as that of the EVGA GeForce GTX 285 SSC we described in an earlier review:
The single difference is the FTW Edition sticker you can see below the orange strip. The memory type is indicated as DDR3 rather than GDDR3, which is a common error among graphics card makers. There is a window in the back of the box for you to compare the serial numbers on the graphics card’s PCB and packaging without taking the card out of the box. This is necessary if you care about the manufacturer’s warranty as well as about the EVGA Step-Up program. The box is compact and will easily fit into any bag.
There is a transparent plastic blister pack inside the box. The graphics card is fixed within it and is secured against any hazards during its transportation and storage. The card is accompanied with the following accessories:
As you can see, there are almost no differences from the EVGA GeForce GTX 285 SSC save for the coupon that provides a 10% discount if you purchase up to 5 games or applications at the nZone.com shop. The EVGA card comes with the fewest accessories among the reviewed graphics cards, yet you do get everything necessary to run the card normally. Not all users appreciate extra accessories, after all. So, this is not a drawback in our eyes.
Besides Fraps, the drivers disc contains a copy of the EVGA Precision utility we have mentioned in our previous reviews of EVGA products. It is a simple and handy tool for overclocking and monitoring the graphics card’s temperatures. It is somewhat superior to ASUS SmartDoctor in functionality but cannot control the GPU voltage.
Well, the latter function depends not only on the software but also on the voltage regulator’s controller chip. Alas, the reference design of GeForce GTX 285 uses a controller that does not support software control of the GPU voltage. So far, EVGA Voltage Tuner can only work with graphics cards equipped with Volterra VT1165 controllers, and the GeForce GTX 285 is one of them.
The packaging and accessories of the EVGA GeForce GTX 285 FTW have not changed since the previous EVGA card we tested, and our opinion about it has not changed, either. The box design is good while the accessories are just sufficient. We don’t view the latter as a drawback, but cannot add more points to the product for its accessories, either.
This product comes in Zotac’s classic packaging: an upright box made from thick cardboard with a translucent window in the front panel. The box is painted the corporate black-yellow-orange using a cell pattern. It looks cure enough. Besides the window, there is a picture of a dragon on the face side of the box:
Besides, there is basic information about the product’s characteristics, particularly the amount of graphics memory and factory overclocking. The stickers indicate that the box contains copies of the racing simulator Race Driver: GRID and the popular benchmarking suite Futuremark 3DMark Vantage Advanced. This version of 3DMark Vantage only differs from Professional Edition in the licensing conditions that do not allow its commercial usage. There is one more sticker on the right side of the box that lists a detailed specification of the graphics card except for the frequency of its shader processors. This seems to be a trifle, yet it is most helpful for the customer. It is sad that many other manufacturers neglect such trifles.
The quality and protective properties of the packaging are very good. The interior of the box is occupied by a foam-rubber tray with cutouts for the graphics card and its accessories. The tray is additionally covered with a transparent plastic cap from above. The Zotac GeForce GTX 285 AMP! comes with the following set of accessories:
The accessories are up to the product class. Besides the required necessities, the box includes a nice bonus: Race Driver: GRID and 3DMark Vantage.
Graphics card makers have begun to supply exclusive overclocking tools for their products. Zotac offers one, too. Besides the drivers, the included disc contains the Firestorm utility that is similar in purpose to EVGA’s and ASUS’s overclocking tools. Its interface looks like follows:
Besides controlling the frequencies of the GPU and memory, the utility can control the speed of the fan, create profiles with settings and keep track of the GPU temperature. There is nothing extraordinary in this tool, just like in most other such programs from other manufacturers, but it does its job well enough. Some people may find it even easier and simpler to use than the more advanced and multifunctional, but also more difficult, RivaTuner.
The first impression about the Zotac GeForce GTX 285 AMP! is all positive. Its packaging looks nice and protects the card well. The accessories include nice bonuses: Race Driver: GRID and 3DMark Vantage.
The boxes these two cards ship in look almost the same. They differ only by the graphics adapter model name and a few insignificant details. Both are designed in red-orange color scheme with a knight on a horse raised on its hind legs. The knight and some other design elements are glossy and embossed while the rest of the box is matt. It looks pretty attractive, although a little commonplace.
The front of both boxes contains only the basic info about the product, such as type and size of installed memory (the type is actually marked as DDR5 and GDDR5 at the same time). As we have already said, this funny mistake is pretty popular among graphics card makers. But while DDR3 and GDDR3 are really two different memory types and mentioning one of them instead of the other can really confuse an inexperienced user, then DDR5 simply doesn’t exist.
The internal structure of the packaging is pretty interesting. If you remove the bright but pretty common-looking external cover, you will see a black box made of thick rigged cardboard with golden embossed lettering.
In my opinion, this box would look much more appealing and rich on the store shelves. Inside there is a rich-looking folder with some accessories, such as disks and a user manual. There is a tray made of polyurethane foam beneath it that holds the graphics card very securely. And the bundled cables and adapters are all placed in a small side section of it. Both cards come with the same exact accessories. Here they are:
I have to say that this is a pretty rich accessories set for sub-$250 graphics adapters. The bundle has everything you might ever need, except the second 2x4-pin Molex → 1x6-pin PCIe adapter. However, most contemporary power supply units are equipped with at least one 6-pin graphics card connector anyway, so the owners of the new Radeon HD 4890 are very unlikely to ever need a second adapter. We were a little surprised to see a full set of adapters supporting outdated analogue video output standards, however, some users may actually need them. So, we should definitely give Asus due credit for being so farsighted.
One of the bonuses for the potential customers is the well-known S.T.A.L.K.E.R.: Clear Sky game, which is currently one of the games used to test contemporary graphics cards performance. Another bonus deserves a special mention. It is a mouse pad made of brown imitation leather on porous rubber base that is stitched along the perimeter with coarse thread. It is very nice to the touch. Our Logitech Revolution MX mouse worked perfectly fine on it. Although it is not a serious gaming mouse pad also because it is pretty small in size, it will fit perfectly for everyday use. Besides, it looks very nice.
The disk with the drivers contains Asus GamerOSD and SmartDoctor utilities. The former is very similar to a well-known Fraps, but besides monitoring the performance, taking screenshots and recording gaming videos, GamerOSD allows overclocking GPUs and adjusting brightness and contrast on the fly, without quitting the game.
Another utility called Asus SmartDoctor provides more advanced overclocking options:
On EAX4890 graphics cards it allows not only adjusting GPU and memory frequencies, but also increasing the graphics processor voltage. The maximum voltage you can set is 1.45V, which should be more than enough for most overclockers. However, even this official software voltmod doesn’t eliminate the possibility of the graphics card failure in overclocked mode. Therefore, you should be very careful with this SmartDoctor function.
It is hard to tell who the winner is in this first round, but if we take into account only the package and accessories bundle, Asus EAX4890 graphics cards seem to be a little more attractive, in our opinion, than the solutions from EVGA and Zotac. But this is just a preliminary conclusion on one specific aspect. Let’s move on to the PCB design and features.
Since both these graphics cards use reference PCB design and a cooling solution developed by Nvidia, we won’t go too much into details about them now, because we have already discussed the reference layout in our GeForce GTX 285 review before. In fact, the only distinguishing feature of these two solutions is the brand name logo sticker on the cooler casing. Zotac’s sticker bears a fire-spouting dragon, while EVGA’s sticker depicts a barrel of some sci-fi weapon:
The PCB of both cards is 27cm long, which makes them unfit some certain system cases with shorter base and specific HDD chassis design. However, those who will be building a gaming SLI platform will definitely make sure that their case is big enough for a system like that.
EVGA GeForce GTX 285 FTW
Zotac GeForce GTX 285 AMP!
Unlike GeForce GTX 280, all GeForce GTX 285 memory chips are located on the front of the PCB. As a result, the PCB layout as well as the cooler is made way simpler.
The voltage regulator circuitry includes a six-phase regulator for the GPU managed by Intersil ISL6327 controller and a two-phase regulator for the video memory managed by an unknown chip of microscopic size. On EVGA GeForce GTX 285 FTW this chip is marked as BR=AL U07, and on Zotac GeForce GTX 285 AMP! - as BR=AG E1V.
Although both solutions have increased GPU and memory frequencies and one of the spots on the PCB is designed for 8-pin PCIe 2.0 connector with up to 150W load capacity, the cards are equipped with a pair of regular 6-pin PCIe 1.0 connectors designed for 75W load. Since G200b has lower power consumption than G200, it should be more than enough, although we know that during overclocking these connectors may be receiving up to 70W of power, which is rather close to their maximum capacity. A small two-pin connector should be used for S/PDIF to HDMI translation to the sound card. As we have already mentioned, there is a special cable for that bundled with both cards. This solution is not as convenient as the one from ATI that integrated a sound core with HDMI support into their graphics processors. However, there is nothing you could actually do about it: Nvidia solutions do not know yet how to work with sound streams on their own.
Both graphics cards have 16 512Mbit GDDR3 memory chips onboard (16Mx32) that form a 1024MB bank with 512-bit access. According to Nvidia’s official specifications the reference memory frequency for GeForce GTX 285 is 1242 (2484) MHz. However, the video memory on Zotac and EVGA solutions we are discussing today has been pre-overclocked by the manufacturers:
EVGA GeForce GTX 285 FTW
Zotac GeForce GTX 285 AMP!
As you can see, EVGA GeForce GTX 285 FTW was overclocked somewhat more aggressively – to 1323 (2646) MHz, while Zotac GeForce GTX 285 AMP! was overclocked to 1296 (2592) MHz. As a result, the peak memory subsystem bandwidth is a little higher: 169.3GB/s vs. 165.9GB/s respectively. The difference in percents is only 2%, so you will barely notice it in real gaming conditions.
The GPUs have also been overclocked. The official default frequencies for the main and shader GeForce GTX 285 core domains are 648 and 1476MHz respectively. However, EVGA GeForce GTX 285 FTW core works at 702/1584MHz and Zotac GeForce GTX 285 AMP! Core – at 702/1512MHz. So, EVGA graphics card has an advantage in shader processors speed. It is interesting that EVGA GeForce GTX 285 SSC we have tested before worked at the same GPU and memory frequencies, but the official company web-site has no mention of this product anymore. The only pre-overclocked GeForce GTX 285 models listed there are GeForce GTX 285 FTW and GeForce GTX 285 Superclocked. Looks like EVGA GeForce GTX 285 SSC may have got a new suffix in the model name. As for Zotac GeForce GTX 285 AMP!, it is, on the contrary, not a flagship solution in the product lineup – there is Zotac GeForce GTX 285 Infinity above it overclocked to 722/1548 MHz. However, it needs to be included into the existing liquid-cooling contour, so we can’t call it a fully independent solution.
Since both graphics cards are an exact copy of Nvidia’s reference design, they have all standard connectors, including a pair of dual-channel DVI-I ports, seven-pin mini-DIN for the analogue video out and two MIO interface connectors providing the cards with Triple-SLI support. There is a plastic light pipe on the card’s mounting bracket next to the mini-DIN connector. It comes from the green and red LEDs indicating normal functioning (green) or reporting any existing power supply issues (red).
As we have already said, GeForce GTX 285 has a simpler cooler than GeForce GTX 280. The new cooler has a smaller heatsink and extremely simple cooling system for the voltage regulator components: they are simply cooler by the fan blowing air through the slits in the heatsink base.
The heat from the memory chips moves through fibrous pads soaked in white thermal paste. The base of the heatsink on top of the GPU itself is covered with a layer of thick dark-gray thermal interface.
We can’t say that the GeForce GTX 285 reference cooler is simple in design. However, it is very thoroughly made and boasts good cooling efficiency with pretty comfortable though not record-breaking acoustics.
The users will be more than pleased with it for the most part and will not need to search for alternatives of any kind. Only those who need a completely silent solution or extreme overclocking enthusiasts in hunt for a new 3DMark world record may want to replace the default cooler with a liquid- or even cryogen-system.
As we have already said in our earlier Radeon HD 4890 Review, the first batch of graphics cards based on the new ATI graphics core will use the reference PCB layout and cooling system design, and only a little later the market will welcome a few uniquely designed solutions. These two Asus EAH4890 graphics cards that will be racing against the GeForce GTX 285 SLI tandem in our today’s test session belong to the first generation of graphics cards, so they are designed exactly the same as the recently discussed PowerColor HD4890 Plus graphics adapter – only the brand name logo stickers are different. As for the exterior differences between Asus EAH4890 and EAH4890 TOP, there are none at all.
It is almost impossible to tell Radeon HD 4890 from Radeon HD 4870 without removing the cooling system, but nevertheless, it uses a new PCB, because RV790 is not pin-compatible with RV770. The voltage regulator is designed as “5+2”, where 5 phases are responsible for powering the GPU and 2 phases – for powering the memory. There are two Volterra VT1165 PWM controllers providing software support for power management. This technology on Asus EAH4890 cards is called “Voltage Tweak”.
Radeon HD 4890 PCB design implies the use of one 6-pin and one 8-pin power connector. However, just like with GeForce GTX 285, both graphics cards are equipped with two 6-pin PCIe 1.0 power connectors capable of handling up to 75W of power. Our regular readers already know that despite our expectations, the new card consumed a little less power than Radeon HD 4870 in our previous test session. As a result, the connectors load does not exceed 50W each even by Radeon HD 4890 OC, so there is nothing to worry about here and no real need to put 8-pin PCI1 2.0 power connectors on the new Radeon solutions.
The cards are equipped with 8 1 Гбит GDDR5 chips (32Mx32) providing the total of 1024MB with 256-bit access. According to the official ATI specifications, the memory should work at 975 (3900) MHz. and it does work at this particular frequency by both cards. Taking into account that the equivalent frequency is 3900MHz, we get 124.8GB/s bandwidth, which is way lower than by GeForce GTX 285. However, practice shows that it is not enough to have high memory bandwidth: you should also be able to really take advantage of it. Theoretically, Nvidia solutions have a certain advantage here, especially in 2560x1600, however, only gaming benchmarks will really answer this question once and for all.
Asus EAH4890 TOP
The only difference in technical specifications between Asus EAH4890 and EAH4890 TOP is the GPU frequencies: in the former case the core works at 850MHz, while in the latter – at 900MHz. The core has maximum configuration in both cases and consists of 160 superscalar processors (5 ALU in each, 800 ALU total), 40 texturing units and 16 RBE. So, if we compare Radeon HD 4890 CrossFireX with GeForce GTX 285 SLI tandem, it will only win in the number of ALU (1600 vs. 480), but will yield in the number of TMU and RBE (160 vs. 80 and 64 vs. 32, respectively). However, things we have just said about the memory subsystem are absolutely true here, too. It doesn’t make sense to compare the number of functional units in contemporary GPUs without taking into consideration their architectural peculiarities. Besides, software optimizations have pretty significant effect on their performance in real life situations.
Both Asus EAH4890 models are equipped with two DVI-I connectors supporting dual-channel mode, a universal analogue video output and two CrossFireX connectors. Unlike the competitor technology that allows tying together three graphics cards maximum, CrossFireX can accommodate up to four cards in multi-GPU mode. Although we have to keep in mind that not every mainboard supporting CrossFireX will fit a configuration like that, especially, with the cards of double-height, like Radeon HD 4890.
As for sound-over-HDMI support, it is implemented in RV790 and doesn’t require a sound card connection. There is an additional sound device in the system marked as Digital Output Device (HDMI) in the Control Panel and all you need to do to activate it, is either to select this sound device as a default sound card for Windows or to select it on the configuration options page of your video player. In the latter case only the soundtrack of your movie will be played through HDMI. Of course, you will need an appropriate adapter included with Asus EAH4890.
Just like the above described EVGA and Zotac, both Asus EAH4890 models use a reference cooler. ATI and Nvidia reference coolers are very similar from the ideological standpoint, because they use an aluminum heatsink cooled with a radial turbine and oust hot air outside the system case.
Overall, ATI’s cooler design is a little less effective, because it uses a heatsink with smaller effective cooling surface and a noisier fan, which crackling has become a distinguishing feature of Radeon HD graphics cards. However, it also has an advantage: it removes the heat from the power components of the voltage regulator circuitry via a special lug in the heatsink base and a greenish fibrous thermal pad. The memory chips transfer their heat to multi-layer elastic pads, and the GPU heat-spreader contacts the heatsink base through a layer of traditional dark-gray thermal interface that is used in most contemporary graphics solution coolers. Despite our concerns, this cooler performed pretty well in terms of cooling efficiency as well as acoustics, at least if you do not manually set the fan rotation speed to the maximum.
All four graphics cards participating in our today’s test session belong to pretty high-performance solutions and three of them are pre-overclocked by the manufacturers. Nevertheless, we decided to find out if we could squeeze a little more out of them and undertook an additional overclocking session. The software power management technology available on Asus EAH4890 cards was off, so that all the testing participants could be in equal testing conditions. GeForce GTX 285 doesn’t support any technologies like that yet and may never acquire one for the voltage regulator circuitry with Intersil ISL6327 controller.
Either way, we managed to obtain the following results:
EVGA GeForce GTX 285 FTW
Taking into account the already existing overclocking, it is a pretty good achievement for the GPU and quite mediocre for the memory.
Zotac GeForce GTX 285 AMP!
Zotac solution proved a little more confident during overclocking having outperformed EVGA graphics card in memory frequency and main GPU domain frequency. However, it couldn’t surpass the 1600MHz bar for the shader speed. It is really hard to draw any final conclusions: both products overclock pretty similarly. In fact, they overclock similarly poor, considering the factory overclocking and extreme complexity of the G200b.
Now let’s move on to Asus EAH4890:
Asus EAH4890 couldn’t repeat the success of PowerColor HD4890 Plus and conquer the 1GHz barrier. Memory overclocked even less – only to 1050 (4100) MHz vs. 1200 (4800) MHz on PowerColor solution.
Asus EAH4890 TOP
As for Asus EAH4890 TOP, it did better: it was only 20MHz short of the sacred 1Ghz barrier, while the memory worked stably at 1150 (4600) MHz. Nevertheless, PowerColor HD4890 Plus remains an undefeated overclocking leader among ATI Radeon HD 4800 models.
Since our today’s review is devoted to the performance comparison between Radeon HD 4890 CrossFireX and GeForce GTX 285 SLI configurations, we decided not to test single graphics cards in overclocked mode, because we have already performed these tests for both: GeForce GTX 285 as well as Radeon HD 4890, and the results of these tests are very well familiar to our readers.
As for the power consumption, our current testbed doesn’t allow us to measure this parameter for two cards working in multi-GPU mode. Therefore we will only provide the numbers obtained as a sum of power consumption measurements for single graphics cards in 3D mode, because in 2D mode one of the cores is idle.
Actual numbers may be slightly different from the ones mentioned above, because even in 2D mode the power consumption of the secondary card never gets down to 0. Nevertheless, it will give you an idea of the power consumption levels in contemporary multi-GPU systems.
Of course, the leader is the triple-GPU Radeon HD 4870 3-way CrossFireX system. It is far ahead. However, the Radeon HD 4890 CrossFireX tandem consumes even a little less than a single Radeon HD 4870 X2. GeForce GTX 285 SLI is pretty power-hungry, but even despite the transition to 5nm process we didn’t expect anything else from a pair of G200b working in a maximum configuration at 648/1476MHz frequencies. You absolutely must have a powerful high-quality PSU in a system like that, although it doesn’t make sense to hunt for 1kW power supplies, even if you have Radeon HD 4870 3-way CrossFireX configuration. At this point we would like to wind up our theoretical part and move on to the practical performance tests.
We are going to investigate the performance of GeForce GTX 285 SLI and Radeon HD 4890 CrossFireX tandems in contemporary games using the following testbed:
The graphics card drivers were configured in the same way as before: to provide the highest possible quality of texture filtering and to minimize the effect of default software optimizations. As a result, our ATI catalyst and Nvidia GeForce driver settings looked as follows:
The list of benchmarks includes the following gaming titles and synthetic tests:
First-Person 3D Shooters
Third-Person 3D Shooters
We selected the highest possible level of detail in each game using standard tools provided by the game itself from the gaming menu. The games configuration files weren’t modified in any way, because the ordinary user doesn’t have to know how to do it. We made a few exceptions for selected games if that was necessary. We are going to specifically dwell on each exception like that later on in our article.
Besides the above discussed multi-GPU configurations, we also inclouded the following mupti-processor solutions in our today’s test session:
We ran our tests in the following resolutions: 1280x1024, 1680x1050, 1920x1200 and 2560x1600. Everywhere, where it was possible we added MSAA 4x antialiasing to the standard anisotropic filtering 16x. We enabled antialiasing from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia GeForce drivers.
Performance was measured with the games’ own tools and the original demos were recorded if possible. We measured not only the average speed, but also the minimum speed of the cards where possible. Otherwise, the performance was measured manually with Fraps utility version 2.9.8. In the latter case we ran the test three times and took the average of the three for the performance charts.
Starting from version 1.3 we use the game’s integrated benchmarking options together with a custom demo record. Unfortunately, this method does not report the bottom frame rate.
The two heavyweights engage into a hot fight right from the start. The GeForce GTX 285 SLI is 7 to 15% slower than its opponent at resolutions up to 2560x1600 pixels but wins the latter resolution by almost 8%. The multi-GPU subsystems both deliver excellent speed irrespective of the display mode.
But the highest speed in this game can be achieved on the Radeon HD 4870 3-way CrossFireX configuration which is, however, bulky, noisy, uneconomical, and not easy to use. When it comes to such consumer properties, the GeForce GTX 295 is the overall winner. It is not much slower than the leaders but is quiet, occupies only one PCIe x16 slot and does not require a mainboard with support for multi-GPU technologies.
It is far more interesting to watch the titans fight in Crysis Warhead because they are trying to deliver a comfortable frame rate at resolutions above 1280x1024. The winner is the GeForce GTX 285 SLI whereas its opponent, having almost the same average performance at 1680x1050, cannot keep the bottom speed above 25fps.
Take note that the dual-processor Radeon HD 4890 CrossFireX is far more effective than the triple-processor Radeon HD 4870 3-way CrossFireX at every resolution. Judging by the results, the overhead for synchronizing the three GPUs has eaten all the performance growth. Perhaps the Catalyst driver is not optimized well for this game.
We disabled the integrated frame rate limiter in the game console for the sake of comparing the cards. The game’s built-in benchmarking options do not provide information about the bottom speed, so there is no such info in the diagrams.
The GeForce GTX 285 SLI is in the lead at resolutions up to 1920x1200 inclusive, outperforming the Radeon HD 4890 CrossFireX by 2-9%. The gap is shrinking as the resolution grows up. ATI’s solution wins at 2560x1600 but all this winning has little practical wroth since the graphics subsystems are both fast enough for every display mode.
The Radeon HD 4870 3-way CrossFireX platform is again imperfect, taking first place at 2560x1600 only. It is a mere 11fps ahead of the GeForce GTX 295 while the average frame rates of both solutions are 100-115fps. Thus, the GTX 295 shows itself as the best single graphics card available today but we are expecting ATI partners’ answer in the way of Radeon HD 4890 X2.
Nvidia enjoys an all-around triumph with its GeForce GTX 285 SLI tandem outperforming the Radeon HD 4890 CrossFireX by 12-29%. However, the bottom speed of both platforms was comparable at resolutions up to 2560x1600 and both were equals in terms of gamer’s comfort. ATI’s solution maintains a bottom speed of over 25fps at 2560x1600 but falls far behind the GeForce GTX 285 SLI. This additional reserve of bottom speed may come in handy in certain scenes.
As for the GeForce GTX 295, it can successfully compete with the Radeon HD 4890 CrossFireX even at the highest resolution. If Nvidia’s solution has a comparable or lower retail price, we’d recommend it to Far Cry 2 players because it is easier to use.
Here, ATI’s solution is in the lead at the first three resolutions although the advantage of the Radeon HD 4890 CrossFireX over the GeForce GTX 285 SLI is negligible and its bottom speed is considerably lower. Despite its three GPUs the Radeon HD 4870 3-way CrossFireX configuration delivers the same performance at a much higher level of power consumption.
The GeForce GTX 285 SLI subsystem is only 3% ahead of its opponent at 2560x1600 in terms of average performance. The GeForce GTX 295 is still the best solution that delivers a frame rate of 60fps at 2560x1600 but is compatible with nearly every mainboard that has at least one PCI Express x16 slot.
The game runs on the Source engine and has an integrated benchmark, but the latter does not report the bottom speed information.
The two multi-GPU tandems are roughly equal but Nvidia’s solution gains the upper hand at 2560x1600. The Radeon HD 4890 CrossFireX is challenged by the GeForce GTX 295 here. The latter card can make an optimal purchase for a gamer who does not care much about his computer budget but does not want to bother about multiple cards, SLI connectors, etc. But if you are willing to put up with the drawbacks typical of discrete multi-GPU solutions, we can recommend you to buy two GeForce GTX 285 cards and an appropriate mainboard.
To achieve a playable speed in this game we disabled FSAA and such resource-consuming options as Sun rays, Wet surfaces and Volumetric Smoke. We use the Enhanced full dynamic lighting (DX10) mode for our test and additionally enable the DirectX 10.1 mode for the ATI cards.
This game, besides Crysis Warhead, is among the most demanding applications on our list of benchmarks despite the relaxed settings described above. Therefore it is highly interesting to watch top-end hardware run it.
This time every test participant with two or more GPUs can deliver a playable speed at the resolution of 2560x1600. The best bottom speed can be achieved with the Radeon HD 4890 CrossFireX tandem thanks to its increased clock rates and DirectX 10.1 support whereas the GeForce GTX 285 SLI is somewhat inferior to it in terms of both average and bottom speed. We can also note that Nvidia’s solutions feel more confident at lower display resolutions here.
Today’s multi-GPU solutions are overkill for this game. Even the single Radeon HD 4890 allows playing at 2560x1600 comfortably with excellent average and bottom speed showings. The Radeon HD 4890 CrossFireX does best at 1680x1050 and 1920x1200 whereas the GeForce GTX 285 SLI is ahead of it at 1280x1024 and 2560x1600.
The 3-way CrossFireX configuration boasts the best result in the two top resolutions but, as we’ve said above, this speed is not necessary for Devil May Cry 4 as it provides no practical benefits.
Although it is Nvidia’s solutions that are usually the best performers in this test, the Radeon HD 4870 3-way CrossFireX takes top place in every display mode, although its bottom speed is comparable to that of the GeForce GTX 285 SLI or even lower.
The bottom speed remark refers to the Radeon HD 4890 CrossFireX tandem, too. The difference is insignificant, though. The performance of this solution at 2560x1600 is quite comfortable although the GeForce GTX 285 SLI ensures a higher reserve of speed that may come in handy in some especially dynamic scenes.
This round is a tie, actually. Both titans of 3D graphics cope with delivering comfortable performance at every resolution but the Radeon HD 4890 CrossFireX system is somewhat slower than its opponent in terms of bottom speed at 2560x1600.
Judging by the results, the multi-GPU technologies are not really called for in this game because it runs fast enough even on a Radeon HD 4890 or GeForce GTX 285 at the highest display resolution.
The Radeon HD 4890 CrossFireX tandem is barely ahead of the ordinary Radeon HD 4870 X2 notwithstanding the higher GPU and memory frequencies, but provides somewhat more comfort at 1920x1200 thanks to the higher bottom speed. The latter is 25fps, which means the lack of any reserve, as opposed to the GeForce GTX 285 SLI subsystem that has a bottom speed of nearly 40fps. The GeForce GTX 295 is somewhat worse in speed but better in usage, so this card is actually the best choice for playing Mass Effect.
None of the multi-GPU solutions can provide a playable bottom speed at 2560x1600 but the GeForce GTX 285 SLI is close to doing that. Subjectively, the game is playable on that platform with a reduction in gameplay smoothness in the most action-heavy scenes.
A single GeForce GTX 285 is enough for playing this game even at 2560x1600, although this card has the worst result among the tested solutions. The Radeon HD 4890 CrossFireX equals the GeForce GTX 285 SLI because the small advantage of AMD’s solution does not show up in practice.
The 3-way CrossFireX solution fails yet another test, being only ahead of the Radeon HD 4890 tandem at 2560x1600. Again, this gives no practical benefits to the owner of such a noisy and uneconomical configuration considering the overall high level of performance in this game.
We use the in-game benchmarking tools that do not allow to measure the bottom frame rate. We also enable DirectX 10.1 support for ATI’s solutions.
ATI’s single cards do not look very good in this game but CrossFireX technology changes the picture: two Radeon HD 4890 cards in one graphics subsystem show excellent scalability and superb performance at resolutions up to 1920x1200 inclusive. The GeForce GTX 285 does not scale up well in SLI mode, and the appropriate SLI configuration is 37% slower than its opponent at 1920x1200.
This is also the only game on our list where the Radeon HD 4870 3-way CrossFireX platform provides a practical benefit as it makes the resolution of 2560x1600 playable with comfort. The GeForce GTX 295 has a performance slump due to the lack of local graphics memory (it has 896MB as opposed to the SLI pair’s 1024MB).
The game has a frame rate limiter fixed at 30fps. We could not disable it.
Despite the dramatic improvement of average performance in SLI mode, Nvidia’s solutions still cannot boast a good bottom speed if you enable full-screen antialiasing in this game. ATI’s solutions are better in this respect: the single Radeon HD 4890 is never slower than 22fps even at 1920x1200.
The Radeon HD 4890 CrossFireX tandem wins the 2560x1600 resolution but its bottom speed is no higher than 20fps. It means that the gameplay may lose its smoothness in some situations, preventing you from playing a complex combination with your units that would require a quick and combined action.
The recently released add-on to the original game does not introduce any technical innovations but contains a new plotline that allows you to play for the USSR.
In this strategy game the GeForce GTX 285 SLI comes to the fore. It is the only solution capable of maintaining a bottom speed of over 25fps at 2560x1600 pixels. The bottom speed of the Radeon HD 4890 CrossFireX is lower and its playability is limited to 1920x1200.
The results of the Radeon HD 4870 3-way CrossFireX are almost the same as those of the Radeon HD 4890 CrossFireX. Building such a configuration does not make much sense, except for individual games like Tom Clancy’s H.A.W.X.
3DMark06 suggests that there is not much point in assembling a Radeon HD 4870 3-way CrossFireX subsystem because its overall score is only 1156 points higher than that of the Radeon HD 4890 CrossFireX but the latter consumes far less power and is quieter. Well, we don’t think that a hypothetical Radeon HD 4890 3-way CrossFireX would be much better in this old version of 3DMark that defaults to 1280x1024.
Otherwise ATI’s solutions – both single- and multi-GPU ones – are quite fast.
The individual groups of tests provide far more interesting results: the Radeon HD 4890 CrossFireX tandem is even ahead of the Radeon HD 4870 3-way CrossFireX in the SM2.0 tests whereas the latter is ahead in the SM3.0/HDR tests, scoring over 11,000 points. The other multi-GPU solutions can only score about 9500 points.
We minimize the CPU’s influence by using the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering). We also publish the results of the individual tests across all display resolutions to provide a full picture.
This test supports hardware PhysX acceleration on Nvidia’s GPUs, so the SLI systems, from the GeForce GTX 295 to the GeForce GTX 285 pair, take top places. The Radeon HD 4890 CrossFireX cannot score even 9000 points while its opponents easily overtake that barrier. The Radeon HD 4870 3-way CrossFireX is competitive, yet cannot beat the GeForce GTX 285 SLI.
The GeForce GTX 285 SLI tandem is ahead in the first three resolutions of the first test, but the Radeon HD 4870 3-way CrossFireX wins at 2560x1600. The Radeon HD 4890 CrossFireX is always slower than the GeForce GTX 295 but the gap shrinks to naught at 2560x1600.
Nvidia is not so victorious in the second test: the GeForce GTX 285 SLI wins at the first two resolutions only and loses to ATI’s triple-GPU platform. The Radeon HD 4890 CrossFireX is ahead of the GeForce GTX 295 at high resolutions, starting from 1920x1200.
The fight is over and the heavyweights are leaving the ring. Who is the winner then? Strictly speaking, the Radeon HD 4890 CrossFireX and GeForce GTX 285 SLI belong to somewhat different weights because the latter is considerably more expensive. However, they did fight and we should examine the results in more detail to see if this comparison makes sense.
The relatively inexpensive configuration based on the new RV790 processors from ATI takes a good start, being a mere 3% slower than the more expensive opponent on average. Moreover, this configuration is either faster or comparable to the GeForce GTX 285 SLI! It is far slower in three games only: Far Cry 2, S.T.A.L.K.E.R.: Clear Sky and World in Conflict: Soviet Assault. The advantage of the Radeon HD 4890 CrossFire over the GeForce GTX 285 SLI is higher than 20% in such games as F.E.A.R. 2: Project Origin and Tom Clancy’s H.A.W.X.
We also found out that the dual-processor RV790-based configuration is no worse or even better than the triple-RV770 one, being only 1% slower on average. The exceptions are Prince of Persia and Tom Clancy’s H.A.W.X. but the Radeon HD 4870 3-way CrossFireX does not have any practical advantage at 1280x1024 anyway.
At the resolution of 1680x1050, which seems to be the most popular among modern gamers on the PC platform, the advantage of the GeForce GTX 285 SLI tandem over the Radeon HD 4890 CrossFireX is even lower at an average 2%. The biggest gap is in Tom Clancy’s H.A.W.X., but the next update of the GeForce driver is going to solve this problem. The situation is overall more equal in comparison with the resolution of 1280x1024.
Comparing the Radeon HD 4890 CrossFireX and the Radeon HD 4870 3-way CrossFireX, the latter is still ahead in Prince of Persia and Tom Clancy’s H.A.W.X. but the performnance growth is not worth the money and trouble associated with the high noise, heat dissipation and power consumption of such a subsystem.
Notwithstanding the different price categories, the fighters equal each other at 1920x1200: the GeForce GTX 285 SLI has an average advantage of less than 1%. It is 32% ahead of the Radeon HD 4890 CrossFireX in Mass Effect but 37% behind it in Tom Clancy’s H.A.W.X. In most other tests the difference is no bigger than 2-4% to either side, excepting Call of Duty: World at War, Far Cry 2 and World in Conflict: Soviet Assault. Thus, both platforms are suitable for the resolution of 1920x1200, and you should just take into account what games you are going to play. Alas, we can’t test them subsystems in all the games available on the market.
At the highest resolution there is no gap between the two multi-GPU solutions at all. Nvidia wins 8 and loses 7 tests but has a serious advantage in Far Cry 2 only (17%). There is also only one serious defeat – by 20% in Crysis Warhead. That’s rather funny considering the kinship of the two games.
People who love flight simulators can find a good use for a Radeon HD 4870 3-way CrossFireX subsystem together with a 30-inch monitor since it delivers comfortable performance in Tom Clancy’s H.A.W.X. even at 2560x1600 at the highest graphics quality settings plus full-screen antialiasing. Or perhaps you should wait for a Radeon HD 4890 X2?
Anyway, the Radeon HD 4890 CrossFire has proved to be a worthy opponent to the GeForce GTX 285 SLI configuration. The latter is generally faster at low resolutions, but ATI’s configuration secures a draw at 1920x1200 and higher resolutions. Therefore, we can expect it to be faster than the GeForce GTX 285 SLI tandem from the same price category. We will check this out practically in an upcoming review.
As for practical recommendations for gamers, it would not be wise for us to make them basing on our list of 15 games only, especially as you should take into account the display resolution you are going to play at. Still, we can note that if you want maximum performance in Crysis Warhead or Far Cry 2, you should consider building a GeForce GTX 285 SLI, but a Radeon HD 4890 CrossFireX configuration would be better for F.E.A.R. 2: Project Origin or Tom Clancy’s H.A.W.X.
Both these cards are copies of Nvidia’s reference design and feature factory overclocking. Therefore, we can draw similar conclusions about them, at least when it comes to technical specifications and performance. EVGA solution works at slightly higher memory and GPU frequencies, however, they are still not high enough to ensure a visible performance advantage over Zotac GeForce GTX 285 AMP! card in contemporary games. On the other hand, the latter comes bundled with a more attractive set of accessories that includes Race Driver: GRID game and 3DMark Vantage suite.
The choice in this case depends solely on the retail price of these graphics accelerators, so the cheaper card will be a better buy.
The representatives of Asus EAH4890 lineup, just like the previously reviewed PowerColor HD4890 Plus, belong to the first generation of Radeon HD 4890 solutions using reference PCB design and cooling system. This way, the only way to win the competition between the first-round players is to offer the best accessories bundle and the highest overclocking potential. Asus EAH4890 and EAH 4890 TOP do not leave PowerColor HD4890 Plus a single chance in the first nomination. However, when it comes to overclocking potential, the PowerColor solution easily outperforms both Asus contestants easily reaching 1GHz GPU and 1200 (4800) MHz memory frequencies. Asus EAH4890 TOP is just a little behind, so PowerColor HD4890 Plus remains a formal winner.
As for Asus EAH4890, it may be a preferable choice for those who are into overclocking, especially, if it retails at a much lower price than EAH4890 TOP.