by Sergey Lepilov
11/16/2009 | 10:00 AM
It is no secret that all ATI Radeon HD 58xx series cards available today are nothing but clones of the two reference designs: Radeon HD 5870 and HD 5850. After a while, there will of course be versions with modified PCBs, faster memory chips (on the HD 5850) or alternative cooling systems but at the current moment AMD does not allow any liberty on the partners’ side and produces identical products. Therefore we called this review "Attach of the Clones" as it covers CrossFireX tandems built out of Radeon HD 5870 and Radeon HD 5850 cards discussed in detail in our previous article.
Although the technology for uniting multiple Radeon cards into one graphics subsystem has been around for quite a while already, under the names of CrossFire and, later, CrossFireX, such configurations are still considered a luxury. You can find one in a system of a wealthy gamer or hardware enthusiast. Besides the need to pay for two rather than one graphics card, the popularity of CrossFireX is restrained by its persistent problems. CrossFireX does not work on some game engines and the AFR rendering mode does not always ensure comfortable gameplay.
Anyway, this won’t prevent me from checking out the performance benefits you can get by joining two Radeon HD 58xx series cards into a single graphics subsystem. I will also compare them with dual-processor graphics cards and tandems and will perform some power consumption measurements as well. First, I will give you a brief description of two graphics cards from HIS.
Products from Hightech Information System Ltd. should be familiar to our readers. Graphics cards with this brand are often equipped with effective and low-noise coolers, have increased frequencies, original accessories and competitive pricing. This is not the case with these two models, though. Both cards are copies of the reference sample, so I will only focus on what differences they have from the latter. I will try to cover some things that were not discussed in our previous Radeon HD 5870 and HD 5850 reviews.
The two cards come in upright, medium-size boxes embellished with a picture of a sword piercing water.
That’s beautiful. Besides the key features of the card, there is a small “Dirt 2” sticker on the face side of each box. It means that the box contains a coupon for a free download of the new game after its release. Besides, each card comes with the following accessories:
The CrossFireX bridge pack is empty because the bridge was installed on the cards at that moment.
There is a sticker with a sword, the card’s model name and the manufacturer’s name on the cooler’s casing (here and elsewhere the Radeon HD 5870 is on the left, the Radeon HD 5850 is on the right).
Each card is installed into a PCI-E slot and has two DVI-I ports, one HDMI and one DisplayPort, two 6-pin connectors for additional power supply, and a CrossFireX connector.
Here are photographs of the connectors area:
And here are the graphics cards’ 40nm Cypress (RV870) graphics processors.
And the next photo shows Samsung’s GDDR5 memory chips (0.40ns):
The GPUs and memory are the same as in my previous review, so I won’t discuss them here.
These two HIS cards have the same specs as the respective reference samples:
The HIS Radeon HD 5870 is equipped with the same blower as the reference Radeon HD 5870. The blower of the HIS Radeon HD 5850 differs from the reference one with a sticker only.
Judging by the sticker, the BASA0425R2U fan is made by AVC and has a fluid dynamic bearing whereas the ATI Radeon HD 5850 has a blower from NTK (HK) Limited. Despite this difference, the fans have identical specs, including speed range and noise level.
In my recent report on the Radeon HD 58xx series I checked out the temperature of the cards when the fan speed was set at automatic mode (31% full speed) as well as at 50% and 100% speed. And here I will check out how much noise the reference coolers produce at a subjectively moderate level of noise when the fan speed is manually set at 25-27% (1520-1660rpm). The test method remains the same: a room temperature of 23°C, a closed system case, FurMark 1.7.0 running at 1680x1050, and MSI Afterburner 1.2.0 with GPU-Z for monitoring frequencies and temperatures.
As you can see, the trick did not work out in both cases. Each GPU quickly grew as hot as 90°C and the card’s protection circuitry reduced the frequencies and voltages to save it from danger. So, if you want to have a quiet system case, you may want to replace the default cooling system of your Radeon HD 5870 or HD 5850 with an alternative cooler. But don’t forget that the voltage regulator chips need cooling, too!
I also used MSI Afterburner andAMD GPU Clock Tool to check out the frequency potential of the HIS cards at 50% speed of their reference coolers. The cards ran FurMark and 3DMark 2006 (the Firefly Forest scene). I managed to increase the GPU frequency of the HIS Radeon HD 5870 to 905MHz (+6.5%) at its default voltage of 1.149V. At a voltage of 1.25V I reached a GPU frequency of 960MHz (+12.9%). This is not much, really. The GPU of the HIS Radeon HD 5850, like the GPU of the reference card I had tested earlier, proved to have high overclocking potential. At a default voltage of 1.087V, its GPU frequency was increased to 900MHz (+24.1%). At a voltage of 1.2V it could be increased to 975MHz (+34.5%). That’s a second time in my practice that the GPU of a Radeon HD 5850 has lower overclockability than the GPU of a Radeon HD 5850.
The highest memory frequency was 5000MHz with the HIS Radeon HD 5870 and 5080MHz with the HIS Radeon HD 5850.
The hardware configuration of our testbed remained exactly the same as the one used in the Radeon HD 5850 review that is why we are not going to repeat it here again. You can always go back to the previous review to read about it. We used identical ATI Radeon HD 58xx and HIA Radeon HDF 58xx models for our CrossFireX configurations. Besides CrossFireX we also tested Radeon HD 5870 and GeForce GTX 275 installed for PhysX technology and physical effects calculations:
|
|
To ensure that Radeon and GeForce graphics cards could work fine together we had to use a relatively old version of the driver for the latter – ForceWare 185.85 - together with a special patch, because we couldn’t activate PhysX with the latest 190.xx drivers even with the help of a patch.
Besides CrossFireX configurations of two Radeon HD 5870 and two Radeon HD 5850 cards as well as a combination of a Radeon HD 5870 and a GeForce GTX 275, we also included the results for two GeForce GTX 285 cards in an SLI configuration.
Moreover, we also added the results for dual-processor Radeon HD 4870 X2 and GeForce GTX 295 from the previous review. The operating system, drivers, all games, benchmarks and image quality settings are almost fully identical to those used in our article called "Hidden Threat": ATI Radeon HD 5850 Graphics Card Review. The changes have only been applied to Tom Clancy’s H.A.W.X. game and resolutions. Namely, we used only one of them – 1920x1200, which is the maximum for our monitor and at the same time the least CPU dependent one. So, that said, let’s go over to the test results and their analysis.
Graphics cards with Radeon processors are marked with red in the diagrams. Graphics cards with Nvidia’s processors are marked green. The combination of a Radeon HD 5870 and a GeForce GTX 275 is blue.
Let’s start with synthetic benchmarks.
The 2006 benchmark from 3DMark is not a resource-consuming 3D application anymore as the results in the default test mode indicate: it is the speed of the platform that limits the performance of the top-end graphics cards or multi-GPI configurations thereof. CrossFireX technology can show its best when we turn on full-screen antialiasing and anisotropic filtering and switch into a high resolution. This configuration beats the dual-processor Radeon HD 4870 X2 and GeForce GTX 295 as well as the SLI tandem of two GeForce GTX 285. CrossFireX brings about a performance growth of 60% for the Radeon HD 5870 and 72% for the Radeon HD 5850. As expected, the combination of a Radeon HD 5870 and a GeForce GTX 275 is ineffective because 3DMark 2006 does not support PhysX.
The more modern and resource-consuming 3DMark Vantage produces a different picture. The CrossFireX configurations show a good performance growth (about 70% for each tandem), the Radeon HD 5870 tandem winning the test. The Radeon HD 5850 configuration is challenged by the two GeForce GTX 285. When a GeForce GTX 275 is added to a Radeon HD 5870, the CPU score grows twofold (from 25 to 55 thousand) but this fact affects the overall score in the Performance mode only.
This is the last synthetic benchmark and the CrossFireX configurations of Radeon HD 5870 and HD 5850 cards boost the frame rate by 86-87% relative to the respective single cards. Nvidia’s solutions only win this test because they work in DirectX 9 mode. If the Radeons worked in DirectX 9 mode too, their performance would be 80-90% higher.
When there is no FSAA and AF, the two Radeon HD 5870 cards are limited by the CPU’s performance in the test from World in Conflict even at a resolution of 1920x1200. It is only when FSAA and AF are turned on that the Radeon HD 5870 CrossFireX is faster than its opponents, although the gap is small. When CrossFireX is not limited by the speed of the platform, it delivers a scalability of 84%. Although not record-breaking, this result is impressive nonetheless, considering the lack of any lags or jerks in gameplay. Take note of the confident performance of the GeForce GTX 285 SLI tandem which is competing successfully with the two Radeon HD 5850 cards.
The CrossFireX configurations are not much faster than the respective single cards in Crysis. For example, the two Radeon HD 5870 are only 38-40% ahead of the single such card, and the Radeon HD 5850 pair is only 45-50% faster than the single HD 5850. SLI technology is more effective as the GeForce GTX 285 SLI delivers a performance growth of 86-89%! The CrossFireX configurations are not limited by the speed of the platform (I will explain this at length in my upcoming report on the CPU dependence of the Radeon HD 5870). So what is the reason for the clumsy performance of CrossFireX in Crysis? I guess it is the driver’s fault but we will only see that after the next version (or even a few next versions) of Catalyst is released.
Both CrossFireX pairs offer a performance growth of 62% to 80%, depending on the test mode, in Unreal Tournament 3. That’s good, but the two GeForce GTX 285 are faster than the single such card by 75-89%. Therefore, the SLI configuration is competitive to the Radeon HD 5850 tandem in the low-quality mode and beats the two Radeon HD 5870 in the FSAA+AF mode.
Adding a second Radeon HD 5870 and HD 5850 into the graphics subsystem boosts the frame rate by 76-80% in Lost Planet: Colonies, which is quite a lot. But the pair of GeForce GTX 285 cards is just incredible: the SLI tandem is 108-110% faster than the single card! I rechecked these results and confirmed them. By the way, Lost Planet: Colonies is not the only game in this test session in which a second graphics card brings a 100% or higher performance growth over a single such card. The combined Radeon & GeForce configuration does not show any benefits.
Being friendly towards ATI’s GPUs, S.T.A.L.K.E.R. eagerly welcomes a second graphics card into the system, increasing the average frame rate by 90-100%. The two Radeon HD 5850 cards easily outperform the pair of GeForce GTX 285 with their 85% performance growth, which is quite high, too. The tandem of two fastest single-processor graphics cards has no opponents here.
It is similar to S.T.A.L.K.E.R. except that the GeForce GTX 285 SLI feels very confident in the FSAA + AF mode, being only inferior to the Radeon HD 5870 tandem. The lack of PhysX in this game makes the GeForce GTX 275 useless ballast for the Radeon HD 5870.
Like Unreal Tournament 3 or 3DMark 2006, this game is not a difficult task for the CrossFireX and SLI systems I am benchmarking today. The performance growth is modest here and the average frame rate is limited by the speed of the whole platform.
Left 4 Dead is even easier for the graphics cards. Adding a second graphics card only makes sense for playing at the highest settings or resolutions,
CrossFireX is useless for Cryostasis: Sleep of Reason. It is better to buy an Nvidia card to add to a Radeon HD 58xx. Thanks to this addition, the Radeon HD 5870 proves to be faster than two HD 5870 in CrossFireX mode and is only inferior to the SLI tandem of two GeForce GTX 285 cards.
CrossFireX technology makes its only slip in this test session in Warhammer 40000: Dawn of War II. As you can see, the pairs of Radeons work slower here than the single cards. It should be noted that SLI is not perfect, either. The SLI tandem is only 22 to 43% faster than the single card. Real-time strategies are generally CPU-dependent and this should be taken into account. Unfortunately, I did not have a faster processor than the Core i7-920 overclocked to 4.1GHz at the time of my tests.
CrossFireX wins, showing a performance growth of 80-86% over the respective single cards.
CrossFireX is even more confident in Stormrise. Although SLI delivers a higher performance growth (97-100%) than CrossFireX (89-91%), both pairs of Radeons are far ahead of the GeForce team.
As opposed to the previous testing in this game, every card runs in DirectX 10 mode now.
The two GeForce GTX 285 cards can only compete with the Radeon HD 4870 X2. The CrossFireX tandems of single-processor cards of the new generation have no rivals in this test.
This game produces a picture similar to the previous test.
Graphics cards with Nvidia’s GPUs are preferable than their opponents in Wolfenstein even though the CrossFireX and SLI technologies have about the same scalability in the modes unaffected by the CPU speed.
Batman: Arkham Asylum is a second game in this review where the mixed tandem of two graphics cards with completely different GPUs turns in a wonderful performance, beating not only the dual-processor GeForce GTX 295 but also the GeForce GTX 285-based SLI tandem.
Despite the 60% performance growth in CrossFireX mode, the two GeForce GTX 285 in SLI mode are in the lead as they are 85-87% faster than the single such card.
The first two diagrams compare the performance of the Radeon HD 5850 CrossFireX and the GeForce GTX 285 SLI where the latter is taken as the reference point.
The two pairs of graphics cards are comparable in price and comparable in performance. The GeForce GTX 285 SLI is better in four games, namely Crysis, Cryostasis: Sleep of Reason, Batman: Arkham Asylum and Resident Evil 5. The two Radeon HD 5850 cards beat their opponent in eight games: World in Conflict, S.T.A.L.K.E.R., Far Cry 2, Call of Duty 5: World at War, BattleForge, Stormrise, Tom Clancy's H.A.W.X. and Call of Juarez: Bound in Blood. In the remaining five games there is no clear winner. These are Lost Planet: Colonies, Warhammer 40 000: Dawn of War II, Wolfenstein, Unreal Tournament 3 and Left 4 Dead. In some games the gap is negligible. In others it is as large as 76%. It is not easy to choose between these tandems. Besides such a high variation in performance depending on the specific game, there are such factors to be considered as DirectX 10, PhysX, antialiasing quality, power consumption and noise level.
The next pair of diagrams shows the difference between CrossFire tandems built out of two Radeon HD 5870 and two Radeon HD 5850, the latter being the reference point.
The difference is actually the same as between the respective single cards. The senior models are 12.7% ahead in the low-quality image mode and 13.1% ahead in the FSAA + AF mode.
The third pair of diagrams shows the difference between two CrossFireX tandems and a single GeForce GTX 295 taken as the reference point.
The CrossFireX configurations easily beat the fastest dual-processor card from Nvidia in most of tests, but any CrossFireX tandem with Radeon HD 58xx cards is going to cost more than a GeForce GTX 295. So, this comparison does not make much practical sense.
And finally, the last two diagrams show the performance growth of CrossFireX and SLI configurations relative to the respective single cards.
The Radeon HD 5870 tandem delivers an average 53% performance growth in the low-quality mode and 59% in the FSAA+AF mode. For the Radeon HD 5850 CrossFireX these numbers are 57% and 64%. For the GeForce GTX 285 SLI, the performance growth is 70 and 77%, respectively. Why does the Radeon HD 5850 tandem show more efficiency under the same conditions than the Radeon HD 5870 tandem? I guess the more advanced configuration is just more limited by the CPU.
I carried out my power consumption measurements with a specially modified power supply. To create maximum load I launched FurMark 1.7.0 in stability check mode at 1920x1200 and Linpack x64 (LinX 0.6.3, 3500MB, 7 threads). These two programs load heavily the graphics card and CPU, respectively, so we can see the peak power draw of the whole system. I also found out the most resource-consuming game – Lost Planet: Colonies – and measured the power consumption of the systems while running it, too.
You can see the results in the diagram:
.png)
Two GeForce GTX 285 cards working together are the most voracious graphics subsystem. You need a 750W and higher power supply for it. Next goes the system with two Radeon HD 5870 in CrossFireX mode. Interestingly, the system with two Radeon HD 5850 consumes less than the system with a dual-processor GeForce GTX 295 and Radeon HD 4870 X2.
The Radeon HD 5870 and HD 5850 both left a highly positive impression on me, but it is not so clear with the CrossFireX configurations based on them. Such configurations indeed offer the highest performance available today but in some games two GeForce GTX 285 in SLI mode beat the Radeon HD 5850 tandem and, occasionally, the Radeon HD 5870 pair, too. A Radeon HD 58xx-based CrossFireX configuration needs a fast platform that would not limit the speed of the graphics subsystem. It also needs support on the side of the game without which purchasing two such cards won’t make any sense.
Another drawback of such powerful configurations is the lack of PhysX support as was indicated by the results of the Radeon HD 5870 + GeForce GTX 275 subsystem in Cryostasis: Sleep of Reason and Batman: Arkham Asylum. And these are only two games that I used in my tests – you can benchmark graphics cards in them without FRAPS. A full list of PhysX-supporting games can be found here. I don’t mean that the lack of PhysX should prevent you from buying a Radeon HD 58xx, but you should be aware of this fact anyway.
As for building CrossFireX subsystems, I could not correctly install the Catalyst driver for both cards at a first attempt. The system would not see the second Radeon HD 5870. So, I first installed one card, rebooted the system a couple of times and added another one. And then, after a third reboot, I successfully enabled CrossFireX mode. There were occasional jerks in Lost Planet: Colonies that I mentioned in me review of the single card. In Call of Duty 5: World at War the menu would be replaced with a black square for a couple of seconds (but there were no problems in the game proper). Besides the games included into this test session I also checked CrossFireX out in Borderlands, Painkiller: Resurrection and Left 4 Dead 2 (demo) and had no problems playing them.
