Thermal Performance

To thermally test the cards we recorded GPU temperatures for each individual core using RivaTuner. Idle readings were taken at the desktop in Windows Vista 64-bit Home Premium with Windows Aero enabled, while the load reading taken while the GPU configuration was running our Crysis benchmark at 1,920 x 1,200 with 0xAA.

This is one of the most demanding tests in our benchmarking suite, and is sure to get those GPUs sweating. We didn't go higher because we felt that if we increased the load too much, some of the slower cards wouldn't deliver realistic numbers because they'd be bottlenecked in other ways.

All temperatures were taken with the cards cooled using only their standard stock coolers and running on our open air test benches, which have no additional airflow other than that present in the room.


Thanks to the reduced die size from switching to the 55nm manufacturing process, the GTX 295 is able to match the single GPU GTX 280 for thermal performance at idle, despite the use of two GPUs rather than one. However, it's still far far cooler than the RV770 card, all of which idle much hotter.

What's interesting however is difference in GPU temperature between the multi-GPU Nvidia and ATI cards. While the GTX 295's dual GPUs (and quad GPUs in the SLI configuration) all idle at roughly the same temperature, the Radeon HD 4870 X2's are incrementally cooler - the hottest is GPU0 and then GPU1, GPU2 and GPU3 are not as warm.


When at load the situation is much the same, with the GTX 295 roughly matching the GTX 280 for operating temperature be it in single or multi GPU configuration, and every GPU operating at roughly the same temperature - close to 80°C, which is fairly acceptable. The 4870 X2 however once again is incrementally cooler on subsequent GPUs (although still hotter than the GTX 295) and in CrossFire even under heavy load the second card is much cooler, indicating that not all the processing power available in being fully utilised.