by Sergey Lepilov
06/10/2011 | 02:18 PM
In the first months following the announcement of a new graphics card we can usually buy only a copy of the reference sample with stickers and accessories peculiar to the particular brand. This was also the case with the Radeon HD 6970 and HD 6950 cards released by AMD in late 2010. However, there was already a variety of custom-made products as soon as the spring of 2011. They feature more effective or less noisy coolers, custom PCBs with reinforced power circuitry, pre-overclocked GPU and memory frequencies, and gorgeous accessories with various extras. Among them we can name such products as Gigabyte GV-R697OC2-2GD, ASUS EAH6970 DCII/2DI4S/2GD5, HIS 6970 IceQ Turbo, PowerColor PCS+ HD6970 as well as the two products we are going to discuss today: MSI Radeon R6970 Lightning and VTX3D Radeon HD 6970 X-Edition.
The MSI Radeon R6970 Lightning comes in a large and colorful box that bears a picture of a supersonic jet caught in a lightning.
You can find a description of the product’s features and differences from the reference card on the spread of the box cover.
Bundled with the MSI Radeon R6970 Lightning are the following accessories:
It’s rather odd that such an expensive graphics card comes without one or two recently released games which would make it even more appealing. It is priced at about $390 or some 40$ more than the recommended price of Radeon HD 6970 2GB. This product comes with a 3-year warranty on the territory of the USA and Canada and mostly 2-year warranties elsewhere.
MSI’s Lightning is impressive for its sheer size. Considering the size of the cooler casing, the card is 35 millimeters longer and 28 millimeters wider than the reference Radeon HD 6970. You should keep this in mind if you want to install an alternative cooler, especially a liquid cooling system, on it.
The card matches the reference sample in its height, so its dimensions are 308 x 125 x 41 millimeters. This is one of the largest single-processor graphics cards of our time.
MSI doesn’t indulge us when it comes to the card’s connectivity options. We have two DVIs, one HDMI and two mini-DisplayPorts here.
There is gold sputtering on the video connectors. Another external difference from the reference Radeon HD 6970 is that there are two 8-pin power connectors here instead of one 8-pin and one 6-pin connector. The R6970 Lightning also has two standard CrossFireX connectors.
The place for a BIOS switch is empty. There is another switch nearby that selects the operation mode for the graphics card and its cooler.
According to the user manual, setting it into the Performance position extends the adjustment range for the graphics card’s clock rates and voltages (in fact, the card just boots up using a second BIOS). However, this switch didn’t help us in overclocking. It can also be set into a third position but the card does nothing then. It doesn’t even boot up with the switch in that position.
There are also three connectors for measuring main voltages. They are going to be useful for serious overclockers.
Now let’s take a look at the MSI Radeon R6970 2GB Lightning without its cooler and the plate that covers its power components:
We can see that the power circuit is reinforced:
Instead of the reference Radeon HD 6970’s 9-phase circuit (6+2+1), the MSI card features as many as 18 power phases: 14 for the GPU, 3 for the memory chips and 1 for VDDCI. Coupled with high-quality Japanese capacitors, ultracapacitors from NEC and 10-layer PCB, this allows MSI to claim unprecedented stability and high overclocking potential. The PCB is indeed unique in this respect. No other maker has ever introduced anything like that although there is a lot of custom-made Radeon HD 6970s available on the market now.
The Cayman processor installed on our sample of the MSI card was manufactured on the 47th week of 2010 in Taiwan.
It is clocked at 940 MHz at a voltage of 1.175 V in 3D mode, which is 60 MHz (or 6.8%) higher than the GPU clock rate of the reference Radeon HD 6970. The frequency and voltage are lowered to 250 MHz and 0.9 volts in 2D mode. To remind you the Cayman specs, the GPU incorporates 1536 shader processors, 96 texture-mapping units, and 32 raster back-ends. The GPU die is 389 sq. mm large.
The MSI Radeon R6970 Lightning is equipped with 2 gigabytes of GDDR5 memory in eight FBGA chips from Hynix Semiconductor located on the face side of the PCB.
The memory chips are marked as H5GQ2H24MFR R0C. They have a voltage of 1.5 V and a frequency of 5500 MHz (but they are rated for 6000 MHz). There is no difference from the reference Radeon HD 6970 here. The memory clock rate is lowered to 600 MHz in 2D mode. The memory bus is 256 bits wide.
Here is what the GPU-Z utility reports about the MSI card:
The Radeon R6970 Lightning is fitted with a Twin Frozr III cooler which is claimed to cool the GPU by 26°C more effectively and produce 12 dBA less noise in comparison with the reference Radeon HD 6970’s cooler. Being all nickel-plated, the Twin Frozr III is going to look cool for a very long time. It consists of a copper base, five heat pipes that go through that base (two of the pipes are 8 millimeters in diameter and three, 6 millimeters), and slim aluminum fins of the heatsink.
The contact between the details of the cooler is established through soldering although some fins are somewhat loose on the pipes. A steel plate with thermal pads serves to cool the power circuit components and memory chips.
The whole arrangement is cooled with two 92mm impellers secured in special frames with three screws.
The actual maker of the fans is Power Logic (it is the PLA09215B12H model with ball bearings).
The speed of the fans is PWM-controlled automatically within a range of 1000 to 3750 RPM. The fans’ startup voltage was measured to be 4.1 volts. Their peak power consumption was 9.7 watts.
To check out how hot the MSI Radeon R6970 Lightning (and the other cards in this review) is, we ran the benchmark from Aliens vs. Predator (2010) in five cycles with maximum graphics quality settings at 2560x1600 with 16x AF and 4x FSAA. There was also a second test: a Bitcoin-client poclbm-qui generating a high load comparable to that of the latest FurMark but having more practical value. MSI Afterburner 2.2.0 Beta 3 and GPU-Z version 0.5.3 were used as monitoring tools. These tests were carried out in a closed system case at an ambient temperature of 23°C.
First, let’s see how hot the MSI card is with its fans working in the automatic mode.
Aliens vs. Predator (2010)
When running a 3D game, the GPU got as hot as 68°C and the fan speed was increased to 2090 RPM. The GPU temperature only increased to 70°C in the Bitcoin mode but the fan speed was 2220 RPM then. It is clear that the MSI card with Twin Frozr III is colder than the reference Radeon HD 6970 which may have a GPU temperature up to 90°C and a fan speed up to 2400 RPM.
The graphics card is even colder at the maximum speed of its fans:
Aliens vs. Predator
Of course, 3700 RPM is far from comfortable, but we will return to the problem of noise later on. Right now let’s check out the overclocking potential of this graphics card.
Without increasing the GPU voltage, we managed to lift the GPU clock rate of our MSI Radeon R6970 Lightning up to 970 MHz. This is only 30 MHz above the card’s default clock rate but as much as 90 MHz higher than the GPU clock rate of the reference Radeon HD 6970. Setting the GPU voltage at 1.225 volts helped us increase the clock rate to 995 MHz but that’s hardly worth putting the card under such a stress. So, we preferred to leave the default GPU voltage and limit our overclocking to 970 MHz. The memory chips could be overclocked to 6000 MHz without losing anything in stability or image quality:
When overclocked, the card got 4°C hotter while running the 3D game: with its fans being controlled automatically, the GPU temperature was 72°C.
The speed of the fans increased but a little, to 2280 RPM. Thus, the MSI Radeon R6970 Lightning is much colder and quieter than the reference Radeon HD 6970 even when overclocked.
The second product in this review is marketed by the not very well-known yet VTX3D which was founded in 2009 and specializes in graphics cards with AMD processors.
The company offers a completeRadeon HD 6xxx product range, the VTX3D HD 6970 2GB GDDR5 X-Edition being perhaps the most exciting model in it.
As opposed to MSI, VTX3D ships its graphics card in a smaller and less informative box.
There is enough of information on the box to help the customer with his shopping choice, though.
Included with the card are a 6-pin->8-pin power adapter, one mini-DisplayPort->DisplayPort adapter, a CrossFireX bridge, a CD with drivers, and an installation guide.
There are no bonus CDs or games in the kit. The warranty period is 2 years. This card sells for about $380 at retail shops.
The VTX3D Radeon HD 6970 X-Edition is a very beautiful, even juicy, graphics card.
The bright-red PCB matches the gaudy coloring of the cooler casing while the 8mm heat pipes going out of the cooler’s base make the card look serious and even aggressive. The PCB is 267 millimeters long and the cooler doesn’t go beyond its dimensions. The card is 125 millimeters wide and 39 millimeters high.
The Radeon HD 6970 X-Edition offers a standard selection of video outputs: two DVIs, one HDMI and two mini-DisplayPorts.
You will find the power and CrossFireX connectors as well as the BIOS switch in their habitual locations.
This is indirect evidence that the Radeon HD 6970 X-Edition is based on the reference PCB design. We can make sure of this by removing the card’s cooler:
We can only note the aluminum heatsinks painted like copper on the power circuit components. The VTX3D card has a newer GPU than the MSI. It was manufactured on the first week of 2011.
Interestingly, its clock rate is also 940 MHz or 6.8% above the GPU clock rate of the reference Radeon HD 6970. The memory chips are the same as we’ve seen on the MSI card above.
The memory frequency of the VTX3D card is 5700 MHz, which is 200 MHz (3.6%) higher than the reference card’s.
Here is a summary of the graphics card specs:
For the card to be cooled properly at minimum noise, VTX3D equips it with an original cooler that consists of a copper base with four copper heat pipes (8 mm in diameter), an aluminum heatsink, and a plastic casing with fans.
A proper contact between the pipes and heatsink is ensured through soldering. The slim 0.25mm fins of the heatsink are long and placed 1 millimeter apart from each other. Their total surface area is almost twice as large as that of the reference card’s cooler.
The plastic frame contains two 92mm impellers that are a mere 10-12 millimeters high. Judging by the sticker on the motor, the fans are manufactured by Rotek Electronics.
The fans both have the same speed which is varied automatically within a certain range. We could not determine that range with any monitoring tool but our controller reported that the fans could work at a speed of 1140 to 3570 RPM. The startup voltage of the fans is 2.7 volts; their peak power consumption is 7.5 watts.
The temperature of the card turned out to be similar to that of the MSI. Its GPU was as hot as 70°C in the automatic fan control mode when running a 3D game. When we launched the Bitcoin client, we temperature grew up to 73°C.
Aliens vs. Predator (2010)
At the maximum speed of the fans the GPU temperature wasn’t higher than 62°C when running Aliens vs. Predator (2010) and no higher than 66°C when running the Bitcoin client.
Aliens vs. Predator (2010)
Thus, VTX3D’s cooler proves to be even more effective than the cooler of the MSI Radeon R6970 Lightning. But how would it compare with, for example, an Arctic Cooling AcceleroXTREME 5870 which is actually the best air-based cooler for the Radeon HD 58xx and 69xx series? We could not install it on the MSI card due to the heat-spreading plate but it fitted nicely on the VTX3D Radeon HD 6970 X-Edition:
Here are the results in the Bitcoin mode:
3 x 1000 RPM
3 x 1490 RPM
3 x 1870 RPM
The AcceleroXTREME 5870 is expectedly superior to the VTX3D cooler (and, obviously to the cooler of the MSI Radeon D6970 Lightning) but its advantage is not very large, so the original coolers of the two Radeon HD 6970s we are discussing today are very high quality.
Now let’s get back to the VTX3D card to check out its overclocking potential (with its default cooler, of course). Its GPU was able to work at 975 MHz without our raising its voltage. The memory chips worked at 5880 MHz.
By increasing the GPU voltage from 1.175 to 1.225 volts, we managed to add 30 MHz to the GPU frequency, reaching as high as 1005 MHz, but this is not as large a benefit as to justify our running the card in such a stressful mode. When overclocked to 975/5880 MHz, the VTX3D Radeon HD 6970 X-Edition got a mere 2°C hotter under max load.
All graphics cards were benchmarked in a closed system case with the following configuration:
In order to lower the dependence of the graphics cards performance on the overall platform speed, I overclocked our 32 nm six-core CPU with the multiplier set at 24x and “Load-Line Calibration” (Level 2) enabled to 4.5GHz. The processor Vcore was increased to 1.475V in the mainboard BIOS:
The 6 GB of system DDR3 memory worked at 1.5 GHz frequency with 7-7-7-16_1T timings and 1.64V voltage. Turbo Boost and Hyper-Threading technologies were disabled during our test session.
The test session started on May 17, 2011. All tests were performed in Microsoft Windows 7 Ultimate x64 SP1 with all critical updates as of that date and the following drivers:
Taking into account the class of our today’s testing participants, we performed all tests in two resolutions: 1920x1080 and 2560x1600 pixels. The tests were performed in two image quality modes: “HighQuality + AF16x” – highest texturing quality with enabled 16x anisotropic filtering and “HighQuality + AF16x + AA8x(4x)” with enabled 16x anisotropic filtering and full screen anti-aliasing (MSAA) 8x or 4x if the average framerate was high enough for comfortable gaming experience. We enabled anisotropic filtering and full-screen anti-aliasing from the game settings or configuration files. If the corresponding options were missing, we changed these settings in the Control Panel of Catalyst drivers. There were no other changes in the AMD Catalyst driver settings.
Since we have already tested AMD Radeon HD 6970 and HD 6950 graphics cards before, we decided to cut down the list of benchmarks so that only the most interesting and resource-hungry games remained. Besides the recently added Warhammer 40,000 Dawn of War II: Retribution and Crysis 2, we also included two new titles: Total War: Shogun 2 and DiRT 3. As usual, all games and applications were updated with new patches available at the time of tests. As a result, the list this time included one popular semi-synthetic benchmarking suite, one technical demo and 13 games. Here is the complete list of tests used with the settings (all games listed in their release order):
If the game allowed recording the minimal fps readings, they were also added to the charts. We ran each game test or benchmark twice and took the best result for the diagrams, but only if the difference between them didn’t exceed 1%. If it did exceed 1%, we ran the tests at least one more time to achieve repeatability of results.
The goal of our today’s tests is to benchmark the two custom-made Radeon HD 6970s we have described above. We will also benchmark the VTX3D card, which proved to have the higher overclocking potential, at its maximum clock rates (975/5880 MHz). Finally, we will also benchmark a CrossFireX configuration built out of these two cards at 940/5700 MHz (we’ve done such tests before, but this time around we’ve got a new Catalyst driver and some new games on our list).
The graphics cards will be shown in the diagrams in the order of descending performance. The results of the CrossFireX configuration with two Radeon HD 6970s are purple. The results of the VTX3D Radeon HD 6970 X-Edition at its default and overclocked frequencies are red. The MSI Radeon HD 6970 Lightning is dark-gray and the reference AMD Radeon HD 6970 is blue in the diagrams. Each of these cards has 2 gigabytes of onboard memory, so this number is not indicated.
The difference between the pre-overclocked cards from MSI and VTX3D on one part and the reference AMD Radeon HD 6970 on the other amounts to 2-4% in 3DMark 2011, which doesn’t look like much. The pair of pre-overclocked HD 6970s shows their best in CrossFireX mode, enjoying a 76 to 96% advantage over one such card.
We’ve got the same picture here as in the previous test except that the CrossFireX configuration has an even larger advantage that amounts to 95-97%!
Here, the pre-overclocked cards increase the average frame rate by 2-4% in the lower-quality mode and by 5-8% in the 4x MSAA mode. The CrossFireX configuration can only show its huge potential at 2560x1600 in the MSAA mode when it is 90% faster than the single graphics card.
The graphics cards from MSI and VTX3D are faster than the reference AMD Radeon HD 6970 by 4-7% in this game, too. CrossFireX technology isn’t very efficient here, adding 21 to 56% to the performance of the single card.
The gap between the pre-overclocked and reference Radeon HD 6970s is very narrow even in such a resource-consuming game as Metro 2033. CrossFireX technology is more efficient here than in Borderlands, boosting the average frame rate by 71-90%.
CrossFireX gets even more efficient in Just Cause 2, providing a performance boost of 92-96%! The gap between the pre-overclocked cards from MSI and VTX3D on one side and the reference Radeon HD 6970 on the other is no larger than 4%, which means a mere 1 or 2 frames per second.
The MSI and VTX3D cards again have but a very small advantage over the reference card from AMD. More interestingly, CrossFireX technology allows the tandem to beat the single card by over 100%!
The pre-overclocked cards from MSI and VTX3D are 4-7% faster than the reference card from AMD, yet this only translates into one or two frames per second, the frame rates being rather low in this game. The performance benefits from CrossFireX technology vary from 72 to 85% depending on the resolution and graphics quality settings.
AMD-based solutions are generally inferior to Nvidia-based ones in this game. If you enable full-screen antialiasing, even top-end cards with AMD GPUs will prove to be too slow. CrossFireX technology is not as efficient as in other games, either.
We’ve got the same picture here as in the other games.
CrossFireX technology should be given credit for boosting the frame rate by 84 to 97% but the three graphics cards that perform singly do not differ much in performance.
The new patch and the new Catalyst driver make CrossFireX work in Crysis 2, but the performance growth is no higher than 50% and the bottom speed of the CrossFireX configuration is just as low as that of the single cards.
Our Radeon HD 6970 CrossFireX configuration doesn’t show its full potential in Retribution, being limited by the performance of the CPU and the testbed at large. The diagram makes it clear that a single Radeon HD 6970 is going to be enough for playing this game. The only problem is that the bottom frame rate is too low despite our running every game from a RAID0 built out of two SSDs and having a 4.5GHz six-core CPU in our testbed. Hopefully, the situation will improve with driver updates or game patches.
We used the maximum graphics quality settings for this game:
The test scene was heavy and very beautiful.
Here are the results:
The new test from Total War: Shogun 2 is so heavy that even the two pre-overclocked Radeon HD 6970s in CrossFireX mode do not feel at ease. A single Radeon HD 6970 may fail to deliver a playable frame rate if you enable the highest settings. Two such cards in CrossFireX mode will double both the minimum and average frame rates. So, AMD’s multi-GPU technology is most efficient in this game.
The newest game on our list was tested with AMD Catalyst 11.5bRC5 dated 05.25.2011. We used the highest graphics quality settings:
The visuals and level of detail of the new racing sim are most impressive:
Here are the results:
CrossFireX is impressively efficient in this game, boosting the frame rate by 90 to 97% compared to the single card.
You can also check out the table below with full test results:
The first pair of summary diagrams shows the advantage of the VTX3D Radeon HD 6970 X-Edition overclocked to 940/5700 MHz over the reference AMD Radeon HD 6970 (880/5500 MHz):
With the GPU and memory clock rates increased by 6.8% and 3.6%, respectively, over those of the reference Radeon HD 6970, the VXT3D card proves to be 4 to 4.5% faster in the lower-quality mode and 3.4 to 4% faster in the FSAA mode. The MSI card overclocked to the same GPU frequency delivered almost the same results through most of the tests because memory frequency is less important for such top-end graphics cards.
The second pair of summary diagrams illustrates the efficiency of CrossFireX technology. We compare a CrossFireX tandem built out of two Radeon HD 6970s clocked at 940/5700 MHz to a single such card clocked at the same frequencies.
CrossFireX is not very efficient or the CrossFireX tandem is limited by the CPU in such games as Borderlands, Warhammer 40 000 Dawn of War II: Retribution, and in the FSAA-less mode of S.T.A.L.K.E.R.: Call of Pripyat and StarCraft 2: Wings of Liberty. But in games like Aliens vs. Predator (2010), Total War: Shogun 2, Just Cause 2, Tom Clancy’s H.A.W.X. 2 and DiRT 3 CrossFireX proves to be as efficient as 90% and higher. The average across the games we’ve tested it in is 67-75% in the lower-quality mode and 78-81% in the 8x/4x MSAA mode.
We measured the power consumption of systems with different graphics cards (without the monitor) using a specially modified power supply. A gaming load was emulated by running Aliens vs. Predator (2010) at 2560x1600 with 16x anisotropic filtering and 4x MSAA. For the maximum load we launched FurMark 1.9.0 in the stability check mode at 2560x1600 with 16x AF and 4x MSAA together with Linpack x64 (LinX 0.6.4, 4750 MB, 5 threads). These two programs load heavily the graphics card and CPU, respectively, so we can determine the peak power draw of the whole system and see what power supply will suffice for it. You can see the results in the diagram:
A 550-watt power supply is going to be sufficient for a system with one Radeon HD 6970 under typical gaming loads. A CrossFireX tandem built out of two such cards will call for a 750-watt or better power supply. And if you are going to put your graphics subsystem under stress loads, the power supply requirements get even stricter: 650 watts for the single card and 900 watts for the CrossFireX tandem. Comparing the three different Radeon HD 6970s, the system with our MSI Radeon R6970 Lightning has the highest power consumption whereas the system with the VTX3D Radeon HD 6970 X-Edition, despite its higher clock rates, needs somewhat less power and is comparable to the system with the reference 6970 in this respect.
We also measured the level of noise using an electronic noise-level meter CENTER-321 in a closed room about 20 sq. meters large. The noise-level meter was set on a tripod at a distance of 15 centimeters from the graphics card which was installed on an open testbed. The mainboard with the graphics card was placed at an edge of the desk on a foam-rubber tray. The instrument’s bottom measurement limit is 29.8 dBA. When measured from this distance, a noise level of 36 dBA seems to be subjectively comfortable. The speed of the fans was adjusted by changing their voltage with a step of 0.5 volts using our in-house controller.
Here are the results:
The MSI Radeon R6970 Lightning is about as noisy as the Arctic Cooler AcceleroXTREME 5870 which is one of the best GPU coolers available. The cooler of the VTX3D Radeon HD 6970 X-Edition is not so quiet. Moreover, the fans of the VTX3D cooler are rather unstable, meaning that their noise does not grow up proportionally to their speed. This makes the card somewhat uncomfortable when you launch 3D applications. On the other hand, both alternative coolers are much quieter than the reference cooler at the maximum speed of the fans.
The MSI Radeon R6970 Lightning and VTX3D Radeon HD 6970 X-Edition graphics cards are surely highly interesting top-end products. Besides the increased GPU frequency (and memory frequency too, in the case of the VTX3D card), they boast high-performance coolers which work quieter than the reference Radeon HD 6970’s cooler. As a matter of fact, the factory overclocking of these cards only increases their performance by an average of 4%. This can hardly be a strong argument in favor of a solution with an alternative design. So, it is the 20°C lower temperature of these cards and the lower level of noise that make them preferable over the reference Radeon HD 6970. The MSI Radeon R6970 Lightning is especially good in this respect as its cooler is almost as quiet as the best alternative GPU cooler from Arctic Cooling.
The VTX 3DRadeon HD 6970 X-Edition, in its turn, is more compact, consumes somewhat less power and does not need two 8-pin power connectors from your PSU. MSI’s solution can respond to this with its gold-plated connectors, 18-phase power circuit and special connectors for measuring voltages. VTX3D can then mention the higher memory clock rate and the eye-catching exterior of its product, so this argument might go on infinitely. But the only serious downside we can find about these cards is that their accessories are rather too scanty. A modern game like DiRT 3 would make them even more appealing.