radrokIt all depends on what you think it is inefficient, if it brings 50% more performance for 40% more power consumption it isn't inefficient by my standard, it's still a power hog but it delivers.

timmyisme221.21 gigawatts indeed, punani.

Aquinus...but how does it scale? How much more power can you get out of it? The 7970 is power efficient and a little more voltage goes a very long way. If Kepler runs hot, what good is over-clocking going to do if you don't have the thermal headroom and you're chip is already consuming so much power.

I'm afraid I have to agree with m1dg3t since efficiency determines how a platform scales with voltage, clocks, shader counts, and temperature.

Aquinus...but how does it scale? How much more power can you get out of it? The 7970 is power efficient and a little more voltage goes a very long way. If Kepler runs hot, what good is over-clocking going to do if you don't have the thermal headroom and you're chip is already consuming so much power.

I'm afraid I have to agree with m1dg3t since efficiency determines how a platform scales with voltage, clocks, shader counts, and temperature.

m1dg3tI always factor efficiency into everything, really how good is something if it is inefficient?

To me inefficiency = poor design

newtekie1And to me people made way to big of a deal out of the "inefficiency" of nVidia cards. I mean if you take the HD6970 and GTX570, two cards that performed virtually identically, the difference in performance per watt was a whole 4%. Not something to complain about at all really, IMO. And when you look at actual power consumption between the two, the GTX570 consumes a whole 3w more when idle and a whole 9w more under load. Again, nothing to complain about really.

Yes the GTX580 was more inefficent, about 10% worse, but it also performed about 13% better, so it is actually doing slightly better than linear when it comes to raw performance and performance per watt, which is actually good.

nvidiaintelftwTypically cards don't even use all the power supplied to them though.

newtekie1Very good point, like the GTX580 has an 8+6 pin design good for 300w and peaks in the 225w range and the GTX570 it 6+6 good for 225w and it peaks in the 190w range. So if the GK104 samples are popping up with 6+6 I would guess peak power consumption around 200w as well.

And frankly the number of power connectors on a card is a terrible way to judge power consumption, the GTX560 has a 6+6 pin design as well and peaks under 150w...

btarunrTo begin with, the picture confirms that the board will have 5 NVVDD phases, and up to three miscellaneous power domains.

newtekie1Very good point, like the GTX580 has an 8+6 pin design good for 300w and peaks in the 225w range and the GTX570 it 6+6 good for 225w and it peaks in the 190w range. So if the GK104 samples are popping up with 6+6 I would guess peak power consumption around 200w as well.

And frankly the number of power connectors on a card is a terrible way to judge power consumption, the GTX560 has a 6+6 pin design as well and peaks under 150w...

DarkOCeantpucdn.com/reviews/Powercolor/HD_7950_PCS_Plus/images/power_maximum.gif
Wizz uses crisys 2 for power consumption not furmark so this is max real power consumption .

radrokWho cares about efficiency on high end cards? Buy midrange if you want good perf/watt.
High end should bring loads of performance and IF possible efficiency, not necessarily anyway.

AquinusOr you can use Kombustor (Furmark) that fully taxes your GPU and not your CPU. It is built to stress every part of your GPU and I can tell you this. My 6870 gets a lot hotter running Kombustor than running Crysis 2. Unless you have numbers to prove the difference, I wouldn't make such a claim...

Prima.VeraEasy to say when your dad pays for the bills. ;)

radrokWho are you to make such insinuations? For the record I went green with solar energy some years ago, my electricity bills are 0.
I'd avoid such blind statements in the future if I were you, it just isn't polite.