Testing (continued):
Track Mania Nations Forever:
Track Mania Nations Forever is a fun (and free) 3D car racing game that also includes a built-in benchmark. This game may only be DirectX 9.0c capable, but it can be configured for high graphic detail and makes for an interesting test. With a resolution of 1280x1024, 8x anti-aliasing, 8x anisotropic filtering, and high settings, it was executed on all three cards. The results presented below are the average frame rate values achieved, and higher is better. What we see here is that the HD4870 handles the high level of detail the best, and takes the victory in this round.
Thermal Testing:
Now that performance testing was out of the way, the next step was to analyze the range of temperatures that the GPU core operated in. This was accomplished by running
CPUID's HWMonitor 1.10 (64-Bit) within Windows Vista Home Premium (64-Bit). The idle condition was established by recording the temperature after letting the system sit at the Window's desktop for at least 30 minutes with no other applications running. The load condition was established by running 3DMark Vantage two times back-to-back. HWMonitor logs current, minimum, and maximum temperature, so once both rounds of testing were completed, the maximum temperature was recorded. All testing was completed on an open testbed in a room with an ambient temperature of 21C.
The results for the HD4870 were actually provided earlier in the review, but here we see confirmation that no matter what the card is asked to do, there is a very tight band of temperatures that the GPU will see (79C-85C). As a point of reference we see that the HD3870 and the HD3870 X2 idle about 25C lower, and that when stressed they don't get quite as hot as the HD4870. The HD4870 does have a larger die size, which could mean more power and more heat, and in my opinion the cooler should be reconfigured to deal with this heat better. The HD4870 makes about as much noise as other cards I have used (which isn't to say it is silent, but it isn't all that loud), and I feel a little extra fan speed wouldn't hurt the noise output too much, but could greatly improve the cooling performance. For those that don't want to wait for a factory fix, an aftermarket cooler or installing
ATITool to run with the stock cooler may be good ideas.
Power Consumption Testing:
The final phase of testing was to monitor the total system power consumption, and compare the results between the different cards. A Seasonic Power Angel power monitoring device was plugged inline between the power supply and wall out, and the Wattage screen was monitored. Since all other hardware remained the same, any change in this value would be the result of the video card's demands. Idle conditions were established by letting the system sit at the Windows desktop until a stable reading was achieved (which never took more than a few minutes after booting up). For load testing I found that DirectX 9 applications did a much better job than DirectX 10 applications, and I chose to keep an eye on the Power Angel's display while running the DirectX 9 version of the Devil May Cry benchmark.
What we see is that the new generation of ATI graphics cards doesn't do anything for energy efficiency, as the HD4870 uses about 55% more power than the HD3870 at idle, and about 35% more power under a load. The HD3870 X2 is the most power hungry under a load, but it does have two GPUs to feed, and at least it is more efficient than running two separate HD3870 cores. The HD4870's larger die size plays a part in this, but the results are the opposite of what we have come to expect when new generations of products are released.
