|PowerColor PCS+ HD6950 Vortex II|
|Reviews - Featured Reviews: Video Cards|
|Written by David Ramsey|
|Monday, 23 May 2011|
Page 13 of 17
PCS+ HD6950 Temperatures
Adding an enhanced cooling solution is the first change any vendor will make to a reference video card. By their nature, reference coolers are designed to keep the GPU at a temperature that NVIDIA and AMD deem safe at reasonable ambient temperatures. With noise and cost as major considerations for the stock cooler, these "safe" temperatures often leave little headroom for overclocking.
FurMark is the application to use if you want to drive video card temperatures as high as possible. Still, this is complicated by the extra controls AMD's built into its latest generation GPUs and drivers, which will automatically throttle themselves down if power draw or temperature exceed certain limits. In this respect they're similar to Intel's Sandy Bridge processors, which will also aggressively throttle themselves to prevent damage.
AMD Overdrive, an overclocking feature built into AMD's Catalyst Control Center, has a "Power Control Settings" slider you can adjust to allow up to 20% extra power draw by the GPU, and I adjusted this control to the maximum 20% for temperature, power, and overclocking. I let FurMark run until the reported temperature was stable for 5 minutes, and recorded the load temperatures with the fans on "Auto" as well as 100% speeds.
When I removed the cooler from the card for photography (after this test), I was concerned by two things: the mounting system didn't seem to apply much pressure to clamp the heat sink to the GPU, and the excessive amount of thermal paste. But obviously neither of these factors mattered much, because this cooler keeps the card really, really cool. These temperatures are 1°C higher for load and 3° cooler under load with 100% fans than the similar cooler on the PowerColor PCS++ HD6950 card Benchmark Reviews tested previously.
The automatic fan control kept the fans at about 28-30% speed under load, at which speed they were almost inaudible. Manually ramping the fans up to full speed decreases temperatures dramatically, but at a significant sonic cost. I don't see any reason to take the fans off auto control when running at stock speeds, since the card simply doesn't get that hot, even under heavy loads.
The overclocked temperatures I recorded with FurMark were identical to the stock temperatures, as were the frame rates reported by the utility. This is because AMD's drivers detect when certain benchmarking applications like FurMark are being used, and throttle the card accordingly. I did see load temperatures up to 74°C (auto fans) during some gaming benchmarks, so FurMark's definitely being throttled here. With older versions of AMD's drivers, simply renaming FurMark was enough to defeat this protection, but that doesn't work with the current drivers from either ATI or NVIDIA, which will limit FurMark's continued use as a stress testing tool. AMD and NVIDIA both claim that utilities like FurMark aren't realistic in that no real-world use will ever apply the continuous loads these test utilities do, and that the detection of these utilities prevents possible damage to their video cards.