I swapped the noisy stock cooler on my Sapphire HD4830 for a Zalman VF900-CU and I've been trying to stress-test it to see how hot it will go. I'm glad today is the coolest day we've had in a few weeks.
So I fire up Furmark and go for a stress test, and I'm surprised as the temps rise into the mid 90's C with the cooler running at full speed (and noisy it is too.) The heatsink doesn't even feel all that warm to the touch, though plenty of warm air is coming away from it. The temperature graph shows a nice gentle curve up to temperature, and I've reseated the heatsink with decent TIM without change - it just looks like the cooler isn't up to cooling the card.
Everything is running at stock speed. Case cooling isn't an issue as it's running in an open environment on my desk.
So I try a different program to stress it, falling back to trusty RTHDRIBL. GPUZ reports 98% GPU load with it, good enough for me. Temps don't rise above 60, and I can dial down the fan to a whispery 5V for a load temp of 70C.
So is furmark pointless for GPU stress testing? I keep reading folks say nothing ever pushes a graphics card as hard as Furmark. I used to think no games pushed a card as hard as RTHDRIBL's shiny spheres. I guess I should try some actual games and see what temps I get.
Basically should I panic my GPU is going to melt just because Furmark eats it? Do I 'need' a better GPU cooler? Does it look like I've mounted it wrong? (I don't see how I could have when other stressing programs are kept reasonably cool.)
So I fire up Furmark and go for a stress test, and I'm surprised as the temps rise into the mid 90's C with the cooler running at full speed (and noisy it is too.) The heatsink doesn't even feel all that warm to the touch, though plenty of warm air is coming away from it. The temperature graph shows a nice gentle curve up to temperature, and I've reseated the heatsink with decent TIM without change - it just looks like the cooler isn't up to cooling the card.
Everything is running at stock speed. Case cooling isn't an issue as it's running in an open environment on my desk.
So I try a different program to stress it, falling back to trusty RTHDRIBL. GPUZ reports 98% GPU load with it, good enough for me. Temps don't rise above 60, and I can dial down the fan to a whispery 5V for a load temp of 70C.
So is furmark pointless for GPU stress testing? I keep reading folks say nothing ever pushes a graphics card as hard as Furmark. I used to think no games pushed a card as hard as RTHDRIBL's shiny spheres. I guess I should try some actual games and see what temps I get.
Basically should I panic my GPU is going to melt just because Furmark eats it? Do I 'need' a better GPU cooler? Does it look like I've mounted it wrong? (I don't see how I could have when other stressing programs are kept reasonably cool.)