ASUS HD6950 2GB DirectCU II Review
Temperature & Overclocking
As you can imagine, overclocking was an absolute breeze on this card. The main obstacle is the lack of BIOS switch and possibility for unlocking so for the HD6950 to compete against even the reference cards, it will have to be able to outperform a HD6970 once overclocked. The ASUS HD6950 DirectCU II comes pre-overclocked at 810/5000MHz, 10MHz more than the reference cards on the core. Despite the cooler, ASUS did not provide a healthier overclock leaving users the joy of overclocking. The card operates at 1.0V at idle when it clocks down to 450/5000MHz in dual screen mode and 175/300MHz with just one monitor. Under load, it operates at 1.1V.
Because of the custom design of the card, SmartDoctor is the only application capable of tweaking the voltages using the SHE controller. Initially, the application limited the maximum core clock to 950MHz but the headroom limit can be removed in the settings. In the end a final clock speed of 984/5480MHz was reached with a voltage of 1.25V on the core. This represents a 24% boost on the core speed and a 10% increase in memory clocks. The clocks were tested to be stable by undergoing gaming sessions and the benchmark suite without problems. In this case, voltage tweaking also has its disadvantages. As expected, increasing the voltage also raises power consumption so improved headroom is inhibited by PowerTune’s maximum TDP limit. Getting optimal performance without TDP reducing the core speed dynamically requires a fine balance of lower voltages and stable clock speeds. The power control setting was set to 20% throughout testing. Thankfully, the improved power circuitry helps in reducing power consumption and the maximum TDP has been raised to match the HD6970s. When overclocked, the core idles at 500MHz instead of 400MHz.
The overclock provides a significant boost to the HD6950’s specification, increasing the fill rates and shader operation by 24%. Similarly, the memory bandwidth is also increased by 10% making it close to that of the HD6970’s. The overclocked HD6950 manages to surpass the specifications of the HD6970 despite its 1408 stream processors.
In real world performance, the 24% overclock resulted in an average 13.9% boost, nowhere as linear as what we would expect but nonetheless free performance. It is possible that PowerTune limited the scalability of the overclock. Still this overclock is enough to match the performance of a stock clock HD6970.
Temperature & Noise
The highlight of the cooler is without a doubt the cooler which not only dominates its look but also the card’s appeal. The temperature results do the premium feel of the card justice with some of the lowest temperatures we have recorded. At stock speed, the temperature delta under load was a mere 49C, lower than everything bar the DirectCU based HD5850. More impressively, it did so at a measly 25% fan speed which was inaudible as you would expect. At idle, the fan operated at 13%.
Even when overclocked, the load temperature only increased by 4C for a delta of 53C, 14C lower than the reference cooler. Under load, the fan only ramped up to a maximum 28% speed to compensate for the higher heat output. At idle, the fan speed only went up by 1% compared to previously. To make sure that the fan was indeed alive, I increased the fan speed to see at which point it became audible. At 30%, a gentle hum could be heard and increasing to 40% still yielded somewhat quiet noise levels. It wasn’t till the fans reached 55% fan speed that it became intrusive but again this was easily drowned by games. Even at 100%, the fans never reached the blow dryer noise levels of the reference vapour chamber cooler. Kudos to ASUS!
As you can imagine, overclocking was an absolute breeze on this card. The main obstacle is the lack of BIOS switch and possibility for unlocking so for the HD6950 to compete against even the reference cards, it will have to be able to outperform a HD6970 once overclocked. The ASUS HD6950 DirectCU II comes pre-overclocked at 810/5000MHz, 10MHz more than the reference cards on the core. Despite the cooler, ASUS did not provide a healthier overclock leaving users the joy of overclocking. The card operates at 1.0V at idle when it clocks down to 450/5000MHz in dual screen mode and 175/300MHz with just one monitor. Under load, it operates at 1.1V.
Because of the custom design of the card, SmartDoctor is the only application capable of tweaking the voltages using the SHE controller. Initially, the application limited the maximum core clock to 950MHz but the headroom limit can be removed in the settings. In the end a final clock speed of 984/5480MHz was reached with a voltage of 1.25V on the core. This represents a 24% boost on the core speed and a 10% increase in memory clocks. The clocks were tested to be stable by undergoing gaming sessions and the benchmark suite without problems. In this case, voltage tweaking also has its disadvantages. As expected, increasing the voltage also raises power consumption so improved headroom is inhibited by PowerTune’s maximum TDP limit. Getting optimal performance without TDP reducing the core speed dynamically requires a fine balance of lower voltages and stable clock speeds. The power control setting was set to 20% throughout testing. Thankfully, the improved power circuitry helps in reducing power consumption and the maximum TDP has been raised to match the HD6970s. When overclocked, the core idles at 500MHz instead of 400MHz.
The overclock provides a significant boost to the HD6950’s specification, increasing the fill rates and shader operation by 24%. Similarly, the memory bandwidth is also increased by 10% making it close to that of the HD6970’s. The overclocked HD6950 manages to surpass the specifications of the HD6970 despite its 1408 stream processors.
In real world performance, the 24% overclock resulted in an average 13.9% boost, nowhere as linear as what we would expect but nonetheless free performance. It is possible that PowerTune limited the scalability of the overclock. Still this overclock is enough to match the performance of a stock clock HD6970.
Temperature & Noise
The highlight of the cooler is without a doubt the cooler which not only dominates its look but also the card’s appeal. The temperature results do the premium feel of the card justice with some of the lowest temperatures we have recorded. At stock speed, the temperature delta under load was a mere 49C, lower than everything bar the DirectCU based HD5850. More impressively, it did so at a measly 25% fan speed which was inaudible as you would expect. At idle, the fan operated at 13%.
Even when overclocked, the load temperature only increased by 4C for a delta of 53C, 14C lower than the reference cooler. Under load, the fan only ramped up to a maximum 28% speed to compensate for the higher heat output. At idle, the fan speed only went up by 1% compared to previously. To make sure that the fan was indeed alive, I increased the fan speed to see at which point it became audible. At 30%, a gentle hum could be heard and increasing to 40% still yielded somewhat quiet noise levels. It wasn’t till the fans reached 55% fan speed that it became intrusive but again this was easily drowned by games. Even at 100%, the fans never reached the blow dryer noise levels of the reference vapour chamber cooler. Kudos to ASUS!
