I do seem to recall some slides showing lower power, non-HT mobile quads, which I think Intel definitely needs.
I think that's not enough. In some ways it might turn out to be a downgrade from the dual core, and end up exactly like the Nehalem mobiles, which were 10-15% faster than the duals due to thermal limits heavily limiting clocks.
By the way, there is a reason to upgrade, because my 2600K desktop is surely lot more responsive than the Core i7 3517U Ultrabook. Being more "responsive" is clearly the only thing I need, and likely for lot of people. I just noticed in Bench that the reason I did not feel 2600K being an improvement over i5 661 is because the single thread improvement is only 30%.
Minimum for an upgrade: 50-60%(single thread), which means with current 5%/year improvement, takes 10 Ticks and Tocks from Ivy Bridge, which turns out to be about 2025. Really, this is probably my acceptable bar for a Desktop that will cost me $500 for an upgrade.
To justify spending on a $1000+ laptop though, it should be 100%, or I guess about 15 generations. I guess in about 2030 I'll think of changing it.
Later this year 15W Skylake-U GT3e will be able to offer better CPU + graphics performance than the 17W Core i7 3517U + 15W Geforce GT620M (GF117) inside my ASUS ultrabook. Battery life gain should be great too.
You are different, but for me I really can't see why you'd spend that much for an upgrade that'll end up mostly on graphics that'll merely be acceptable. The point of going integrated graphics should have been to save significant amounts of money and get battery life advantage. The only real advantage of the GT3e U is for those that goes for portability as #1, and the graphics is still barely acceptable.
Since the U chips can't offer #1 portability and the systems are outrageously expensive, the GPU performance should be GT4e Quad level. GT3e would be ok at $599-699 devices.
It doesn't matter whether Moore's Law is at an end, or they are having real trouble with next process generation, or that ILP limits are reached. Whatever. As a customer it needs to be drastically better.
Sweepr said:
I hope Skylake-Y's Core M improved long-term performance. 17% CPU & 41% iGPU bump (same number of EUs as Broadwell-Y) is more than I expected, probably the result of higher base clocks + higher sustained Turbo clocks under load
I hope that its lot better than this. Let's extrapolate. Original claims of Intel said 2.5 points in Cinebench R11.5 for Core M, and most devices turned out to be 1.8-2.0. 17% faster means 2.1 to 2.3. While the Chi is quite exceptional in terms of performance and price, most are not. Best case 2.3 points for average devices mean 2016 Skylake Y is still slower than a Core i5 from 2012, and other benchmarks reflect that. That's quite sad. Review also point out that Broadwell U is the more efficient part in terms of battery life.
Sustained performance
Meh: 2.1 to 2.3(Extrapolated Skylake Y based on current systems)
Acceptable: 2.5-2.7(Original claim of Broadwell Y)
Very Good: 1.17 x 2.5-2.7