AMD Radeon HD 6970 and HD 6950: Cayman GPUs are There!

Radeon HD 6970

AMD has released today the new Radeon HD 6900 Series with the Radeon HD 6970 and Radeon HD 6950. Cayman GPU is the first GPU designed by AMD that is able to rasterize 2 triangles per clock (all previous GPUs, Barts included, can only rasterize one triangle par clock). With AMD’s Radeon HD 6970 and HD 6950, there is a new term: PowerTune and a new name for FurMark and other GPU intensive apps: outlier application 😉

Let’s see what the web says about AMD’s new babies.

Radeon HD 6970

From left to right: Radeon HD 6970, 6950, 6870, 6850 – source

1 – AMD Radeon HD 6970

• GPU: Cayman XT @ 880MHz, TSMC 40nm
• Stream processors: 1536 (or 24 SIMD, one SIMD == 64SP)
• Memory: 2048MB GDDR5 @ 1375MHz real speed, 256-bit
• Texture units: 96 (read: 96 filtered texels per clock)
• ROPs: 32 (read: 32 pixels par clock)
• Z-stencil: 128
• 3D APIs: OpenGL 4.1 and Direct3D 11
• GPU computing: OpenCL 1.1, DirectCompute
• TDP: 250W
• Power connectors: one 8-pin and one 6-pin
• Price: around USD 370$

2 – AMD Radeon HD 6950

• GPU: Cayman XT @ 800MHz, TSMC 40nm
• Stream processors: 1408 (or 22 SIMD, one SIMD == 64SP)
• Memory: 2048MB GDDR5 @ 1250MHz real speed, 256-bit
• Texture units: 88
• ROPs / Z-stencil: 32 / 128
• 3D APIs: OpenGL 4.1 and Direct3D 11
• GPU computing: OpenCL 1.1, DirectCompute
• TDP: 200W
• Power connectors: one 8-pin and one 6-pin
• Price: around USD 300$

3 – AMD Radeon HD 6970 and HD 6950 OpenGL performances

ShaderToyMark OpenGL 2 GPU test – source

TessMark OpenGL 4 tessellation test – source

TessMark OpenGL 4 tessellation test – source

TessMark tests resulted in rather horrid results for Cayman cards, but this is clearly a software issue related to older Catalyst releases and it should not be taken seriously. AMD claims TessMark performance will be vastly improved in new driver updates, scheduled for early 2011. Despite this, we chose to include the results in our review, just to annoy fanatic fanboys who have already issued several fatwas calling four our untimely deaths.

TessMark OpenGL 4 tessellation test – source

Hmm. TessMark uses OpenGL rather than Direct3D to access the GPU, and apparently AMD’s OpenGL drivers aren’t yet fully aware of Cayman’s expanded geometry processing capabilities. Frustrating.

4 – AMD Radeon HD 6970 and HD 6950 Direct3D performances

3DMark11 Direct3D 11 gaming test – source

DiRT2 Direct3D 11 gaming test – source

5 – AMD Radeon HD 6970 and HD 6950 power consumption and temperatures

More details about AMD PowerTune:

• Redefining TDP With PowerTune @ anandtech.com

PowerTune throttling table – source

In the case of Metro the average clockspeed was 850MHz; Metro spent 95% of the time running at 880MHz, and only at a couple of points did the core clock drop to around 700MHz. Conversely FurMark, a known outlier, drove the average core clock down to 600MHz for a 30% reduction in the core clock. So while PowerTune definitely had an impact on FurMark performance it did almost nothing to Metro, never mind any other game/application.

PowerTune throttling table – source

As expected, power and temperature both increase with FurMark with PowerTune at 300W. At this point FurMark is no longer constrained by PowerTune and our 6970 runs at 880MHz throughout the test. Overall our power consumption measured at the wall increased by 60W, while the core clock for FurMark is 46.6% faster. It was under this scenario that we also “uncapped” PowerTune for Metro, when we found that even though Metro was being throttled at times, the performance impact was impossibly small.

PowerTune throttling table – source

Right off the bat everything is lower. FurMark is now at 292W, and quite surprisingly Crysis is also at 292W. This plays off of the fact that most games don’t cause a card to approach its limit in the first place, so bringing the ceiling down will bring the power consumption of more power hungry games and applications down to the same power consumption levels as lesser games/applications.

The above charts show the AMD Radeon HD 6970 graphics dynamically changes engine clock speed to effectively manage TDP, without drastically affecting performance. Conversely, the GTX 580 is shown to simply maintain a static engine clock – source

FurMark OpenGL 2 stress test – source

FurMark OpenGL 2 stress test – source

6 – AMD Radeon HD 6970 and HD 6950 Reviews Round-up

• AMD’s Radeon HD 6970 and Radeon HD 6950: Paving The Future For AMD @ anandtech.com
• AMD’s Radeon HD 6950 and 6970 graphics processors @ techreport.com
• AMD Radeon HD 6950 and 6970 CrossFire Video Card Reviews @ legitreviews.com
• XFX HD6970 3-way CFx and HD6950 CFx review @ legitreviews.com @ kitguru.net
• XFX Radeon HD 6970 and HD 6950 tested @ fudzilla.com
• HIS Radeon HD 6970 2 GB @ techpowerup.com
• HIS Radeon HD 6950 2 GB @ techpowerup.com

TechPowerUP – Radeon HD 6970:
AMD’s new Radeon HD 6900 series is a disappointment to users who expected HD 5970 or GTX 580 performance levels from the new series. The new VLIW4 shader design can not make a huge performance impact as some have guessed, but it opens up a lot of room for die size and transistor optimization which effectively makes the GPU cheaper to produce, the cards cheaper to buy – more performance for your hard earned Dollar. Unfortunately AMD’s Radeon HD 6970 can not meet that price target. With performance comparable to GeForce GTX 570, but a price that is $50 higher it is difficult to justify the investment.