http://www.notebookcheck.net/AMD-A-...r-Specifications-and-Benchmarks.144218.0.html
Code:
15w 35w ?
in % 100 233
3d mark 11 (P) 1764 2359 2181
in % 100 133 (123,6)
cpu slightly above kaveri a10-7400p
200% would be double the power, i.e. Carrizo requires more than double the power (233%) to reach far less than the half the performance (33%) more. This is absolutely remarkable.
And makes it clear immediately why AMD claims Carrizo to be (relatively) more efficient at 15 watts. It means the currency of "electricity" is increasingly loosing value above of 15 watts. Such a thing is otherwise called runaway hyper-inflation.
It is not explicitly specified which thermal design power budget of the cpu was used to perform the test. Since these tests always want to determine the maximum performance it is plausible, if it is assumed that the tests were conducted under conditions, under which the CPU can deliver their maximum performance (regardless of the efficiency) namely at 35W TDP. A lower configured TDP should be explicitly declared. Otherwise, the whole test basically is worthless.
On the other hand there are three results to 3D Mark 11 at various TDP figures, two results with specified TDP details (see above) and a test result with unknown TDP (see below). If we assume a performance increase linear to the TDP and vice versa, we can calculate the unknown TDP simply as follows:
35 - 15 = 20
2359 - 1764 = 595
595 / 20 = 29.75 Points per tdp-watt
2181 - 1764 = 417
417 / 29.75 = 14.02
15 + 14.02 = 29.02 Watt
Therefore, the test was performed with a TDP configured at approximately 30 watts. Taking into account the ratio of the increases of TDP and performance, so the thing is as follows:
100 / 233 = 100 / 133
233 / 133 = 1.75
100 / 1764 * 2181 = 216.5%
15 Watt / 100% * 216.5% = 32.475 Watt
Therefore, the test was performed with a configured TDP of between 29 Watts and 32 watts.
Code:
octane v2 8324 core-m 5y10: +11%
cineb r15 single 69 core i3 4030u: +2%
cineb r15 multi 194 core-m 5y71: +0%
mozilla kraken 3488 core-m 5y10: -4%
geekbench3 32b multi 4775 core-m 5y70: -6%
geekbench3 32b single 2006 core-m 5y10: +0%
geekbench3 64b multi 5105 core-m 5y71: -9%
geekbench3 64b single 2170 core-m 5y71: +24%
3d firestrike physics 1920x1080 2685 core-m 5y70: -2%
3d cloud gate physics 1280x720 2007 core i3 4030u: -3%
3d ice storm unlimited 1280x720 22946 core i3 3217u: -6%
3d ice storm extreme 1920x1080 22721 core-m 5y71: +4%
3d ice storm standard 1280x720 22740 core-m 5y70: +0%
sunspider 164.3 core-m 5y10: -1%
truecrypt serpent mean 0.2 core i5 4210h: -4%
truecrypt twofish mean 0.3 core i5 5257u: -1%
truecrypt aes mean 1.5 core-m 5y71: +2%
x264 hd 4.0 pass 2 15.2 core i5 3437u: 0%
x264 hd 4.0 pass 1 74.4 core-m 5y31: -3%
winrar 1328 core i3 3110m: -1%
----------------------
3d 11 physics 1280x720 2181 core-m 5y10: -1%
----------------------
3d vantage cpu no physics 6585 core i3 4030u: +2%
3d 06 cpu 2843 core i5 3317u: +1%
r11.5 cpu single 64b 0.9 core i5 3339y: +1%
r11.5 cpu multi 64b 2.4 core i3 5010u: 0%
r10 rendering single cpu 64b 3264 core i3 3227u: 0%
r10 rendering multi cpu 64b 8949 core-m 5y70: 0%
r10 rendering single 32b 2343 core i3 3217u: 0%
r10 rendering multi cpu 32b 6443 core i3 4158U: -1%
super pi mod 1.5 xs 1m 19.9 core i3 4030u: 0%
super pi mod 1.5 xs 2m 44.4 core i3 4030u: 0%
super pi mod 1.5 xs 32m 965.5 core i3 4158u: -2%
All in all Carrizo (at about 30W) seems to reach a performance level of the (older) Broadwell Core M (6W or so).
How is that possible? Then Skylake Core M (+ 17% CPU, GPU + 41%) will likely burn off Carrizo?
One got an idea, how:
http://forums.anandtech.com/showpost.php?p=37585572&postcount=1902
"Intel chips are underspecced"
That's why lean too fast and stylish notebooks with Intel CPUs always explode, and why their battery runtimes tend to zero? And if they do not overheat and explode and its batteries last longer than you think, it is a large-scale conspiracy.
The OEMs receive the real TDP specification that is kept secret to the public, and so also opportunity to incorporate larger batteries, whose Spezifikation is also forged from the public.
Is this your theory as to why Intel can run longer and faster, much faster and with much more "bella figura" than AMD?