Sapphire HD 6970 dual fan edition review

Sapphire has recently announced the new HD6970 dual fan edition, and thanks to Sapphire SA, Mygaming was lucky enough to get hands on one of the first cards to land in South Africa.

The new dual fan edition of the graphics card has a host of new features over the original HD6970 which was released released nearly a year ago; but is it worth upgrading? Read on to find out.

New features

Before we get into the performance of the card, the new 6970 does bring with it a few noteworthy features.

The first is a new dual bios system that separates the normal operation bios from a high performance bios optimised for overclocking. When the former is active, the card runs at default clock speeds and makes use of a low speed fan profile to keep temperatures in check.

If you select the performance bios, the core voltage of the GPU is raised to 1.2v to assist overclocking, and more aggressive fan speed settings are applied. In addition to this, max GPU and memory parameters in overclocking software such as AMD Overdrive or Sapphire TriXX are raised to 1500MHz on the core, and 2000MHz for the memory.

While these speeds are unlikely to be achieved using conventional air cooling, having them available to you at the flick of a switch is a nice feature to have.

The 6970 also has a very comprehensive output configuration, which includes two DVI ports, one HDMI port and two mini DisplayPort connectors. The card can also support a VGA monitor with the dongle supplied. Using the DVI or HDMI outputs together with the two DisplayPorts enables this card to support multiple monitors in Eyefinity (SLS) mode.

Heat and noise

The new HD6970 has a dual fan cooling solution featuring twin 90mm fans that make use of thin impeller blades to moved more air at a lower rotation speed. This results in a quieter card than the reference design, while cooling ability is on par. A bonus of using thin impeller blades is that even at high speeds (when the performance BIOS is selected for example), the card remains surprisingly quiet.

The only issue with this design is that it dumps hot air directly into the chassis, rather than directing the air out of the rear expansion slots. Users with high airflow cases won’t be affected, though if you’re using a case with poor airflow (such as the Coolermaster Cosmos S) you might see higher than expected temperatures.

We installed the HD6970 in a high airflow chassis (NZXT Phantom) and tested idle and load GPU temperatures.

For idle, we booted up the PC and navigated around Windows Explorer, worked in Microsoft Office and copied over folders between hard drives. After measuring the temperature we then did the same for a Gainward GTX570 Phantom edition card.

To simulate real world load, we played through a level of Metro 2033 with all graphics settings maxed out, at a resolution of 1920 x 1200. The results were as follows:

HD6970
•    Idle temperature: 31°C
•    Load temperature: 80°C

GTX570
•    Idle temperature: 31°C
•    Load temperature: 78°C

The cards performed identically in idle conditions, while the GTX570 is just ahead under load temperatures. The new dual fan cooler does a better job cooling the HD6970 than the original reference cooler, and is noticeably quieter especially when the fan speed is increased.

Performance

Unigine heaven

First up for the Synthetic benchmarks is Unigine heaven, a DirectX 11 benchmark that focuses on a variety of methods to stress modern graphics cards. We made use of the HWbot Xtreme DX11 pre-configured settings during testing, and the benchmark was run three times on each card, with the upper and lower samples disregarded.

Scores:
•    GTX570 – X1257.93
•    HD6970 – X885.89

As a result of the massive tessellation calculation abilities of the GTX570, it seems to eclipse the 6970 in this benchmark.

3D Mark Vantage

3D Mark Vantage is still one of the best synthetic benchmarks to compare current generation graphics cards. While there are still a host of DirectX 10 titles on the market, performance in this segment will remain relevant.

You’ll notice that there are two overall scores for the GTX570 in the below table. One of these scores was achieved with Physics (PPU) turned off, the general practise when comparing AMD and ATI graphics cards.

However, I feel the Physics abilities of the GTX570 are relevant in real world scenarios where gamers would make use of this feature, and so scores with and without Physics have been posted.

Scores:
•    GTX570 – 25481
•    GTX570 – 20719 physics off
•    HD6970 – 20630

With Physics on, the GTX570 has a clear advantage in this benchmark, though as one turns it off you can see just how evenly the two graphics cards are matched. This is also clear in the 3D Mark Vantage GPU score, with each card scoring within a few hundred points of the other.

3D Mark Vantage GPU

Scores:
•    GTX570 – 21661
•    GTX570 – 21657 physics off
•    HD6970 – 21511

Trackmania

The now outdated DirectX 9.0c standard is often ignored in modern benchmarks, and for good reason. The standard is almost completely irrelevant for modern gamers, though as a quick and easy indication of real world performance for those who (like me) still enjoy the occasional classic title, so it was added in for good measure.

Tests were run at 1920 x 1200, with all graphics settings maxed out, and AA set to 16x.

Scores:
•    GTX570 – 87.3fps
•    HD6970 – 65.8fps

Metro 2033

Known as the new Crysis, Metro 2033 is currently one of the most demanding games on your hardware.

We played through a set level three times, and recorded the average frame rate of the level using Fraps. These were then averaged to get the result below. All graphics settings were maxed, while tessellation was turned off. The resolution was 1920 x 1200.

Scores:
•    GTX570 – 24.5fps average
•    HD6970 – 25.9fps average

The HD6970 inches ahead of the GTX570 in this benchmark, though this is without tessellation enabled. If we had to redo the tests with tessellation on, the GTX570 would probably nudge into the lead, however the game would become completely unplayable thanks to the poor framerates from both cards.

Unfortunately, due to time constraints we couldn’t perform other gaming tests thoroughly enough to generate meaningful data. However, judging by the fact that the new HD6970 runs the same clock speeds as the original version, we can surmise from our own benchmarks and others from around the web, that on average the GTX570 and HD6970 are evenly matched cards, being within a few percent of each other across a variety of benchmarks.

On average the cards are able to trade blows extremely well, and only when a game makes use of a rendering technique that the specific card excels at (such as tessellation for the GTX570) do the performance figures become noticeably higher or lower.

Pricing

At the time of publication, no local recommended retail price was available for the card. When using the local price of the standard Sapphire 6970 2GB graphics card from a local retailer as a reference point, this is how the HD6970 dual fan edition stacked up to the competition:

Conclusion

The HD6970 dual fan edition is a difficult card to review for two reasons. Firstly the card is an amazing package; it runs cool, the fans are quiet, there are a host of useful features such as the twin bios setting and the display connectivity ports, and stock performance is really impressive. For this reason, the 6970 deserves a great score.

However, the GTX570 is nearly as good as the HD6970 in terms of overall performance; it runs just as quietly, just as cool, and costs nearly 19% less.

With all things said and done, the HD6970 dual fan edition is a great card that might come in at a price point just too high to fend off competition from Nvidia.

Score

Performance: 8
Cooling: 9
Features: 8
Overall: 85%

Pros: Feature Rich, impressive cooler, high end performance
Cons: Price

Sapphire HD 6970 dual fan edition review << Comments and views