- ASUS HD 7970 DirectCU II 28nm Graphics Card Review
- A Closer Look - DirectCU II
- Features - GPU Technologies
- Features - Software and Utilities
- Testing - Temps, Voltages, Acoustics, and Overclocking
- Benchmarks - Batman: Arkham City
- Benchmarks - Dirt 3
- Benchmarks - Aliens vs. Predator
- Benchmarks - Battlefield 3
- Benchmarks - 3DMark 11
- Benchmarks - Unigine Heaven
- Benchmarks - WQHD Performance
- Final Thoughts
- All Pages
Features - GPU Technologies
AMD's 7970 GPU is based upon their "Graphics Core Next" (GCN) architecture, and is manufactured using a 28nm process. The HD 7970 is ready for 4K resolutions and support for stereo 3D, bezel compensation, and flexible display configurations to make the most of how you want to use EyeFinity. The reference GPU core clock for the HD 7970 is 925MHz while the memory is clocked at 5500 MHz (1375 MHz actual), which are the same speeds as the ASUS 7970-DC2-3GD5 we've got on the bench today . The card comes with 3GB of GDDR5 memory on a 384-bit memory bus.
Under the Heatsink - Non-Reference Design
The direct-contact Copper heatpipe base is the foundation for the "DirectCU" brand name. You might also be surprised how relatively thin the actual heatsink is underneath the huge bezel, but the fans account for much of the depth. The way the entire assembly comes together, the fans require their own dedicated cross-braces/mounting supports, and the bezel, PCB, and backplate all connect to the integrated support frame, which also acts as the VRM heatsink, a great 2-for-1 design solution. This multi-layered construction creates a very rigid overall design that completely eliminates any noticeable card sag, which also helps provide greater longevity to the PCB.
In addition to the obviously non-reference cooling design in DirectCUII, ASUS has built their HD 7970 DCUII from the ground up as a non-reference card. One thing that's apparent is that the black PCB has a matte finish on it, which looks much better in my opinion than the glossy black PCBs. Rounded corners on the PCB are a nice touch, and give the card a sleeker look, although not something that will immediately jump out at you.
The Tahiti GPU is surprisingly small, and it's polished to a mirror smooth finish which was revealed after some very careful and high-dexterity removal of the thermal paste (shown in the gallery above). The circuit traces are very clean and neat, and the PCB trace layouts are another non-reference re-design by ASUS that goes deeper than the surface specifications, but will ultimately add to the overall reliability and stability of the card. Surrounding the GPU are Hynix GDDR5 chips providing the 3GB of onboard video memory.
Super Alloy Power (SAP) and Digi+ VRM
Another non-reference design ASUS proudly touts is its power delivery system, comprised of their Digital VRM design, Digi+ and their " Super Alloy Power," components abbreviated as "SAP." The VRM, which stands for "Voltage Regulator Module," is present on every graphics card and is responsible for converting the +12V provided by the power supply to the lower voltages needed by the GPU. This step down in voltage has a side effect of producing heat, and since the performance of capacitors and inductors which are responsible for providing the needed voltage to the power supply are heavily dependent upon temperature, these components (and their associated cooling) can be critical to stability and overclocking headroom.
The VRM consists of a PWM, capacitors, chokes, and MOSFETs which steps down and stabilizes the +12V DC power from the PSU to the ~1V DC power neeeded by the graphics card. The PWM (Pulse Width Modulator) is the controller which controls power delivery, and can be digital (more recent implementions) or analog (now becoming uncommon). ASUS' Digi+ design refers to the digital PWM which allows for greater control over power delivery options, lower power response latency, and greater precision with voltage adjustments. The Digital PWM is also what enables OS-side adjustments through a more intuitive and less-intimidatig UI, such as GPU tweak.
The SAP chokes, capacitors, and MOSFETs are made from metals which are especially magnetic, heat-resistant, and anti-corrosive - translating into longer lifetimes and more reliable operation across the board. The MOSFETs are essentially the "transformers" which step down the voltage, and the inductors (chokes) and capacitors act as reservoirs/compensators to smooth out ripple or fluctuations to provide the GPU with "cleaner/filtered" voltage. The more efficient components also allow them to be downsized, which enables ASUS greater capability to optimize trace layouts on the PCB. The heat reistance (and additional cooling provided by the integrated support frame/heatsink) will reduce thermal performance reduction/instability of the capactitors and inductors which ultimately leads to lower ripple in the GPU and memory components, which is often the lowest common denominator for overclocking performance.
The SAP capacitors themselves also have much higher capacity than reference caps, which becomes absolutely crucial for stable power delivery when the card becomes heavily loaded or experiences very sudden peaks in power draw, especially when water cooling or other exotic cooling methods are being used which allow for high power draw. In addition to greater capacity and durability, ASUS claims that the production process required to produce their SAP components also allows for a wider thermal threshold which means your components will have higher tolerance to temperature fluctuation. Similar philosophies have also been validated by their employment in ASUS' best-selling motherboards.
Another important factor is the number of phases used by the power delivery system. You can think of phases as parallel "rails" for delivering power to the various graphics card components, and the more rails you have, the more the power demand is spread out amongst the comonents, which means you have less demand on any given phase. The PWM controls power delivery across the phases by staggering power pulse peaks between the phases. ASUS' design uses an 8+2+2 phase design to deliver a total of 360A. The reference design is a 5+1+1 phase with a total of 300A. A comparative table demonstrating PCB differences is shown below:
|PCB Layout||Power Phase||Current Capacity|
|ASUS HD7970 DCII||8+2+2||
45A x 8 +
45A x 2 +
45A x 2
|AMD Reference 7970||5+1+1||
60A x 5 +
60A + 60A
The Super Alloy Power logo is neatly stamped on the chokes, and the integrated support frame acts as the heatsink which runs across the VRM components. The MOSFETs are the hottest part of the VRM along with the drivers, especially when bumping up the voltage for overclocking. Many non-reference cards don't have heatsinks on the MOSFETs and Drivers, and this can induce instability which cripples overclock potential. ASUS implemented an integrated approach where the support frame uses thermal rubber to also act as a heatsink for the power components.
Other subtleties in the PCB layout are these capacitors (which ASUS calls the SAP Caps) which provide the GPU with power and helps it handle sudden changes in power demands. The longer a wire or circuit is, the more resistance, power loss, heat generation, and potential instability it may have. To combat this and have these critical capacitors as close to the GPU as possible, they're connected directly behind the GPU on the opposite side of the PCB.