(06-30-2016, 08:50 AM)ocre Wrote:(06-30-2016, 04:45 AM)BoFox Wrote: Hell, with that aluminum heatsink, it's like having a stock Intel heatsink and trying to overclock the thing to 4.5GHz.
Those who really want more performance from overclocking should just avoid the stock cooler like the plague and wait for custom coolers by AIB's (especially if there's a version that allows for increased voltage). Temperatures would drop by maybe 25-30C, reducing power leakage by quite a bit (perhaps 20W or more), and allow for much higher overclocking (perhaps as high as 1500MHz, but I think more than 1450MHz would be hard to accomplish with just 1 6-pin PCIE power connector. By the time MSI or Gigabyte designs a board with an 8-pin PCIE power plug, AMD would probably have launched Vega by then, heh.
The power consumption would not drastically change. Amd took a page from Nvidia and tried their best to keep the Temps in the low to mid 80s. This is the same spot Nvidia has been capping their cards at. It's not running 95 degrees.
Look at Nvidia cards reference cards like the 1070, 970, 980, etc. Their target was low 80s and their performance per watt was optimal. Custom cards that run cooler don't show radically improved efficiency. Actually, the custom cards tend to not be better at all when it comes to efficiently. I don't ever remember seeing that, not even once. custom cards with better cooling do drop Temps for maxwell and pascal, but the overall consumption tends to be higher than Nvidia reference designs running in the low 80s.
See, custom cards have more components and are more robust but those things come with added power draw. The improved cooling doesn't seem to lower power consumption and with pascal at least, they haven't brought major leaps in overclocking capabilities.
I think the 1600mhz Polaris over clocks were just bologna, just as the 2500mhz pascal rumors. Just inflated bs hype.
So, better cooling on the 480 wouldn't be bad but I seriously doubt there would or could be such a radical drop in power consumption. They seem to stay around 80c already, Nvidia has the same temp targets and I trust they have aimed their cards in the ideal range. It's not like their blowers are the issue, not so much, this is just their fan target speeds. If a few degrees lower was so much more efficient, they would be targeting below 81c. Their cards hold 70 degree Temps with just a minor change in fan curve. So i am pretty sure that the low 80s is the ideal zone for temps, power, and efficiency...at least this is what Nvidia seems to think.
You might be right - it seems that power leakage isn't as bad as it was at 40-45nm. Some chips made at 45nm would generally leak as much as 40W extra per 20C increase (when over-volted really high in order to overclock as high as possible). Now, the voltage is far lower, so perhaps leakage is hardly noticeable at all.
I'd beg to differ with you, though - 28nm cards with custom cooling seemed to show lower power consumption figures than those with stock cooling. Just look at the HD 280X or 7970 GHz Edition compared to the 7970. Most review sites received custom-cooled cards for review, which ran at far lower temperatures along with lower noise, and the power consumption was also somewhat lower in order to compare better against GTX 680/770.
I hope a hardware site would give a real good look at power leakage vs temperature with 14/16nm cards compared to 28nm (and also 40nm GPUs). Anandtech did a real great analysis on 45nm Intel cpus (yeah, so many years ago, when power leakage was getting way worse than 65nm and 90nm, but Intel promised that it would get better after 45nm).