I have asked Notebookcheck and they confirmed me this device is running 2x4 GB
Dualchannel. Their article is not yet finished.
Not looking too good, as expected AMDs 3dmark marketing slides are misleading. Compared with HD 5500:
http://www.notebookcheck.com/Intel-HD-Graphics-5500.125585.0.html
(make sure you are comparing the better dualchannel results)
There is only a big difference in 3dmark. Basically all other benchmarks from Cinebench OpenGL, gfxbench and most gaming benchmarks is a close match. In some it is even losing to HD5500. And this is only Broadwell.
Firstly, I wouldn't call that a close match. Against HD6000, yes it trades blows, but then you'd need to find an R7 to compare to instead.
As for the score discrepency with AMD's slides.. well, have you actually looked at the results on NBC for the core U? If so, have you not noticed the delta between the many different implementations? it's as much as 100% on gaming tests! ,far far more than LP single ch vs dual ch alone can account for. It's clearly very difficult to get anywhere close to consistent results across machines with any low power, dynamically clocked CPU/APU. particularly in gaming results due to the CPU+GPU loading
Just look at the Core M debacle (Which wouldn't be a debacle at all if they wern't trying to push it into such a low TDP/SDP which it's really not suitable for)
In a nutshell , don't write off Carrizo based on a single test result.. especially one that's just ~10% off the 'ideal' Reference system.
It's pretty obvious it's running 12-15w TDP, given it aligns closest to the expected results floating around, and given that, even it's CPU performance isn't that far off the mark. Much like how BW GT2 isn't far off the mark in it's graphics performance.
I think BOTH these facts are a surprise, and I tip my hat to both companies really in each respect.